[clang] [llvm] [AMDGPU] Extend permlane16, permlanex16 and permlane64 intrinsic lowering for generic types (PR #92725)

Vikram Hegde via cfe-commits cfe-commits at lists.llvm.org
Tue Jun 25 04:40:10 PDT 2024


https://github.com/vikramRH updated https://github.com/llvm/llvm-project/pull/92725

>From b00271123e45c8e1fb66e8ec777eb6bb1ee6b311 Mon Sep 17 00:00:00 2001
From: Vikram <Vikram.Hegde at amd.com>
Date: Wed, 10 Apr 2024 11:53:16 +0000
Subject: [PATCH 1/9] [AMDGPU] Extend readlane, writelane and readfirstlane
 intrinsic lowering for generic types

---
 clang/lib/CodeGen/CGBuiltin.cpp               |    4 +
 clang/test/CodeGenOpenCL/builtins-amdgcn.cl   |    4 +-
 llvm/include/llvm/IR/IntrinsicsAMDGPU.td      |   15 +-
 .../Target/AMDGPU/AMDGPUAtomicOptimizer.cpp   |   10 +-
 llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp |    3 +
 llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h   |    4 +
 llvm/lib/Target/AMDGPU/AMDGPUInstrInfo.td     |   29 +
 .../lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp |  190 +
 llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.h  |    3 +
 llvm/lib/Target/AMDGPU/SIISelLowering.cpp     |   60 +
 llvm/lib/Target/AMDGPU/SIInstructions.td      |    2 +-
 llvm/lib/Target/AMDGPU/VOP1Instructions.td    |   11 +-
 llvm/lib/Target/AMDGPU/VOP2Instructions.td    |   16 +-
 .../UniformityAnalysis/AMDGPU/intrinsics.ll   |    6 +-
 .../atomic_optimizations_mul_one.ll           |   13 +-
 .../atomic_optimization_split_dt_update.ll    |    2 +-
 .../test/CodeGen/AMDGPU/global-atomic-scan.ll |   48 +-
 .../AMDGPU/global_atomic_optimizer_fp_rtn.ll  |  120 +-
 .../AMDGPU/global_atomics_iterative_scan.ll   |    2 +-
 .../global_atomics_iterative_scan_fp.ll       |    8 +-
 .../global_atomics_optimizer_fp_no_rtn.ll     |   24 +-
 .../AMDGPU/llvm.amdgcn.readfirstlane.ll       |  654 +++-
 .../AMDGPU/llvm.amdgcn.readfirstlane.ptr.ll   |   92 +
 .../CodeGen/AMDGPU/llvm.amdgcn.readlane.ll    |  960 +++++-
 .../AMDGPU/llvm.amdgcn.readlane.ptr.ll        |  105 +
 .../CodeGen/AMDGPU/llvm.amdgcn.writelane.ll   | 3041 ++++++++++++++++-
 .../AMDGPU/llvm.amdgcn.writelane.ptr.ll       |  292 ++
 .../InstCombine/AMDGPU/amdgcn-intrinsics.ll   |   32 +-
 28 files changed, 5485 insertions(+), 265 deletions(-)
 create mode 100644 llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ptr.ll
 create mode 100644 llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ptr.ll
 create mode 100644 llvm/test/CodeGen/AMDGPU/llvm.amdgcn.writelane.ptr.ll

diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index f9ee93049b12d..a275ada586d13 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -18479,6 +18479,10 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID,
         CGM.getIntrinsic(Intrinsic::amdgcn_update_dpp, Args[0]->getType());
     return Builder.CreateCall(F, Args);
   }
+  case AMDGPU::BI__builtin_amdgcn_readlane:
+    return emitBinaryBuiltin(*this, E, Intrinsic::amdgcn_readlane);
+  case AMDGPU::BI__builtin_amdgcn_readfirstlane:
+    return emitUnaryBuiltin(*this, E, Intrinsic::amdgcn_readfirstlane);
   case AMDGPU::BI__builtin_amdgcn_div_fixup:
   case AMDGPU::BI__builtin_amdgcn_div_fixupf:
   case AMDGPU::BI__builtin_amdgcn_div_fixuph:
diff --git a/clang/test/CodeGenOpenCL/builtins-amdgcn.cl b/clang/test/CodeGenOpenCL/builtins-amdgcn.cl
index c2ef9ea947e93..83b75f9e2d618 100644
--- a/clang/test/CodeGenOpenCL/builtins-amdgcn.cl
+++ b/clang/test/CodeGenOpenCL/builtins-amdgcn.cl
@@ -306,14 +306,14 @@ void test_ds_bpermute(global int* out, int a, int b)
 }
 
 // CHECK-LABEL: @test_readfirstlane
-// CHECK: call i32 @llvm.amdgcn.readfirstlane(i32 %a)
+// CHECK: call i32 @llvm.amdgcn.readfirstlane.i32(i32 %a)
 void test_readfirstlane(global int* out, int a)
 {
   *out = __builtin_amdgcn_readfirstlane(a);
 }
 
 // CHECK-LABEL: @test_readlane
-// CHECK: call i32 @llvm.amdgcn.readlane(i32 %a, i32 %b)
+// CHECK: call i32 @llvm.amdgcn.readlane.i32(i32 %a, i32 %b)
 void test_readlane(global int* out, int a, int b)
 {
   *out = __builtin_amdgcn_readlane(a, b);
diff --git a/llvm/include/llvm/IR/IntrinsicsAMDGPU.td b/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
index be8048ca2459c..457566944069e 100644
--- a/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
+++ b/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
@@ -2176,26 +2176,23 @@ def int_amdgcn_wave_reduce_umin : AMDGPUWaveReduce;
 def int_amdgcn_wave_reduce_umax : AMDGPUWaveReduce;
 
 def int_amdgcn_readfirstlane :
-  ClangBuiltin<"__builtin_amdgcn_readfirstlane">,
-  Intrinsic<[llvm_i32_ty], [llvm_i32_ty],
+  Intrinsic<[llvm_any_ty], [LLVMMatchType<0>],
             [IntrNoMem, IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
 
 // The lane argument must be uniform across the currently active threads of the
 // current wave. Otherwise, the result is undefined.
 def int_amdgcn_readlane :
-  ClangBuiltin<"__builtin_amdgcn_readlane">,
-  Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty],
+  Intrinsic<[llvm_any_ty], [LLVMMatchType<0>, llvm_i32_ty],
             [IntrNoMem, IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
 
 // The value to write and lane select arguments must be uniform across the
 // currently active threads of the current wave. Otherwise, the result is
 // undefined.
 def int_amdgcn_writelane :
-  ClangBuiltin<"__builtin_amdgcn_writelane">,
-  Intrinsic<[llvm_i32_ty], [
-    llvm_i32_ty,    // uniform value to write: returned by the selected lane
-    llvm_i32_ty,    // uniform lane select
-    llvm_i32_ty     // returned by all lanes other than the selected one
+  Intrinsic<[llvm_any_ty], [
+    LLVMMatchType<0>,   // uniform value to write: returned by the selected lane
+    llvm_i32_ty,        // uniform lane select
+    LLVMMatchType<0>    // returned by all lanes other than the selected one
   ],
   [IntrNoMem, IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]
 >;
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUAtomicOptimizer.cpp b/llvm/lib/Target/AMDGPU/AMDGPUAtomicOptimizer.cpp
index 1d645002b1fe6..9ac0a52f63fd8 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUAtomicOptimizer.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUAtomicOptimizer.cpp
@@ -433,7 +433,7 @@ Value *AMDGPUAtomicOptimizerImpl::buildReduction(IRBuilder<> &B,
   // Pick an arbitrary lane from 0..31 and an arbitrary lane from 32..63 and
   // combine them with a scalar operation.
   Function *ReadLane =
-      Intrinsic::getDeclaration(M, Intrinsic::amdgcn_readlane, {});
+      Intrinsic::getDeclaration(M, Intrinsic::amdgcn_readlane, B.getInt32Ty());
   V = B.CreateBitCast(V, IntNTy);
   Value *Lane0 = B.CreateCall(ReadLane, {V, B.getInt32(0)});
   Value *Lane32 = B.CreateCall(ReadLane, {V, B.getInt32(32)});
@@ -523,10 +523,10 @@ Value *AMDGPUAtomicOptimizerImpl::buildShiftRight(IRBuilder<> &B, Value *V,
                      {Identity, V, B.getInt32(DPP::WAVE_SHR1), B.getInt32(0xf),
                       B.getInt32(0xf), B.getFalse()});
   } else {
-    Function *ReadLane =
-        Intrinsic::getDeclaration(M, Intrinsic::amdgcn_readlane, {});
-    Function *WriteLane =
-        Intrinsic::getDeclaration(M, Intrinsic::amdgcn_writelane, {});
+    Function *ReadLane = Intrinsic::getDeclaration(
+        M, Intrinsic::amdgcn_readlane, B.getInt32Ty());
+    Function *WriteLane = Intrinsic::getDeclaration(
+        M, Intrinsic::amdgcn_writelane, B.getInt32Ty());
 
     // On GFX10 all DPP operations are confined to a single row. To get cross-
     // row operations we have to use permlane or readlane.
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
index d35a022ad6806..9cd7496ba9f96 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
@@ -5496,6 +5496,9 @@ const char* AMDGPUTargetLowering::getTargetNodeName(unsigned Opcode) const {
   NODE_NAME_CASE(LDS)
   NODE_NAME_CASE(FPTRUNC_ROUND_UPWARD)
   NODE_NAME_CASE(FPTRUNC_ROUND_DOWNWARD)
+  NODE_NAME_CASE(READLANE)
+  NODE_NAME_CASE(READFIRSTLANE)
+  NODE_NAME_CASE(WRITELANE)
   NODE_NAME_CASE(DUMMY_CHAIN)
   case AMDGPUISD::FIRST_MEM_OPCODE_NUMBER: break;
   NODE_NAME_CASE(LOAD_D16_HI)
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h
index 3814b56a4d56a..02c3dcf39e554 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h
+++ b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h
@@ -558,6 +558,10 @@ enum NodeType : unsigned {
   FPTRUNC_ROUND_UPWARD,
   FPTRUNC_ROUND_DOWNWARD,
 
+  READLANE,
+  READFIRSTLANE,
+  WRITELANE,
+
   DUMMY_CHAIN,
   FIRST_MEM_OPCODE_NUMBER = ISD::FIRST_TARGET_MEMORY_OPCODE,
   LOAD_D16_HI,
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUInstrInfo.td b/llvm/lib/Target/AMDGPU/AMDGPUInstrInfo.td
index 702f6e67c5527..e4f329b200c86 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUInstrInfo.td
+++ b/llvm/lib/Target/AMDGPU/AMDGPUInstrInfo.td
@@ -342,6 +342,22 @@ def AMDGPUfdot2_impl : SDNode<"AMDGPUISD::FDOT2",
 
 def AMDGPUperm_impl : SDNode<"AMDGPUISD::PERM", AMDGPUDTIntTernaryOp, []>;
 
+def AMDGPUReadfirstlaneOp : SDTypeProfile<1, 1, [
+  SDTCisSameAs<0, 1>
+]>;
+
+def AMDGPUReadlaneOp : SDTypeProfile<1, 2, [
+  SDTCisSameAs<0, 1>, SDTCisInt<2>
+]>;
+
+def AMDGPUDWritelaneOp : SDTypeProfile<1, 3, [
+  SDTCisSameAs<0, 1>, SDTCisInt<2>, SDTCisSameAs<0, 3>
+]>;
+
+def AMDGPUreadlane_impl : SDNode<"AMDGPUISD::READLANE", AMDGPUReadlaneOp>;
+def AMDGPUreadfirstlane_impl : SDNode<"AMDGPUISD::READFIRSTLANE", AMDGPUReadfirstlaneOp>;
+def AMDGPUwritelane_impl : SDNode<"AMDGPUISD::WRITELANE", AMDGPUDWritelaneOp>;
+
 // SI+ export
 def AMDGPUExportOp : SDTypeProfile<0, 8, [
   SDTCisInt<0>,       // i8 tgt
@@ -506,3 +522,16 @@ def AMDGPUdiv_fmas : PatFrags<(ops node:$src0, node:$src1, node:$src2, node:$vcc
 def AMDGPUperm : PatFrags<(ops node:$src0, node:$src1, node:$src2),
   [(int_amdgcn_perm node:$src0, node:$src1, node:$src2),
    (AMDGPUperm_impl node:$src0, node:$src1, node:$src2)]>;
+
+def AMDGPUreadlane : PatFrags<(ops node:$src0, node:$src1),
+  [(int_amdgcn_readlane node:$src0, node:$src1),
+   (AMDGPUreadlane_impl node:$src0, node:$src1)]>;
+
+def AMDGPUreadfirstlane : PatFrags<(ops node:$src),
+  [(int_amdgcn_readfirstlane node:$src),
+   (AMDGPUreadfirstlane_impl node:$src)]>;
+
+def AMDGPUwritelane : PatFrags<(ops node:$src0, node:$src1, node:$src2),
+  [(int_amdgcn_writelane node:$src0, node:$src1, node:$src2),
+   (AMDGPUwritelane_impl node:$src0, node:$src1, node:$src2)]>;
+
diff --git a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
index bd7bf78c4c0bd..6ffc8a20f76fa 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
@@ -5387,6 +5387,192 @@ bool AMDGPULegalizerInfo::legalizeDSAtomicFPIntrinsic(LegalizerHelper &Helper,
   return true;
 }
 
+bool AMDGPULegalizerInfo::legalizeLaneOp(LegalizerHelper &Helper,
+                                         MachineInstr &MI,
+                                         Intrinsic::ID IID) const {
+
+  MachineIRBuilder &B = Helper.MIRBuilder;
+  MachineRegisterInfo &MRI = *B.getMRI();
+
+  Register DstReg = MI.getOperand(0).getReg();
+  Register Src0 = MI.getOperand(2).getReg();
+
+  auto createLaneOp = [&](Register Src0, Register Src1,
+                          Register Src2) -> Register {
+    auto LaneOp = B.buildIntrinsic(IID, {S32}).addUse(Src0);
+    switch (IID) {
+    case Intrinsic::amdgcn_readfirstlane:
+      return LaneOp.getReg(0);
+    case Intrinsic::amdgcn_readlane:
+      return LaneOp.addUse(Src1).getReg(0);
+    case Intrinsic::amdgcn_writelane:
+      return LaneOp.addUse(Src1).addUse(Src2).getReg(0);
+    default:
+      llvm_unreachable("unhandled lane op");
+    }
+  };
+
+  Register Src1, Src2;
+  if (IID == Intrinsic::amdgcn_readlane || IID == Intrinsic::amdgcn_writelane) {
+    Src1 = MI.getOperand(3).getReg();
+    if (IID == Intrinsic::amdgcn_writelane) {
+      Src2 = MI.getOperand(4).getReg();
+    }
+  }
+
+  LLT Ty = MRI.getType(DstReg);
+  unsigned Size = Ty.getSizeInBits();
+
+  if (Size == 32) {
+    // Already legal
+    return true;
+  }
+
+  if (Size < 32) {
+    Register Src0Cast = MRI.getType(Src0).isScalar()
+                            ? Src0
+                            : B.buildBitcast(LLT::scalar(Size), Src0).getReg(0);
+    Src0 = B.buildAnyExt(S32, Src0Cast).getReg(0);
+    if (Src2.isValid()) {
+      Register Src2Cast =
+          MRI.getType(Src2).isScalar()
+              ? Src2
+              : B.buildBitcast(LLT::scalar(Size), Src2).getReg(0);
+      Src2 = B.buildAnyExt(LLT::scalar(32), Src2Cast).getReg(0);
+    }
+
+    Register LaneOpDst = createLaneOp(Src0, Src1, Src2);
+    if (Ty.isScalar())
+      B.buildTrunc(DstReg, LaneOpDst);
+    else {
+      auto Trunc = B.buildTrunc(LLT::scalar(Size), LaneOpDst);
+      B.buildBitcast(DstReg, Trunc);
+    }
+
+    MI.eraseFromParent();
+    return true;
+  }
+
+  if ((Size % 32) == 0) {
+    SmallVector<Register, 2> PartialRes;
+    unsigned NumParts = Size / 32;
+    auto IsS16Vec = Ty.isVector() && Ty.getElementType() == S16;
+    MachineInstrBuilder Src0Parts;
+
+    if (Ty.isPointer()) {
+      auto PtrToInt = B.buildPtrToInt(LLT::scalar(Size), Src0);
+      Src0Parts = B.buildUnmerge(S32, PtrToInt);
+    } else if (Ty.isPointerVector()) {
+      LLT IntVecTy = Ty.changeElementType(
+          LLT::scalar(Ty.getElementType().getSizeInBits()));
+      auto PtrToInt = B.buildPtrToInt(IntVecTy, Src0);
+      Src0Parts = B.buildUnmerge(S32, PtrToInt);
+    } else
+      Src0Parts =
+          IsS16Vec ? B.buildUnmerge(V2S16, Src0) : B.buildUnmerge(S32, Src0);
+
+    switch (IID) {
+    case Intrinsic::amdgcn_readlane: {
+      Register Src1 = MI.getOperand(3).getReg();
+      for (unsigned i = 0; i < NumParts; ++i) {
+        Src0 = IsS16Vec ? B.buildBitcast(S32, Src0Parts.getReg(i)).getReg(0)
+                        : Src0Parts.getReg(i);
+        PartialRes.push_back(
+            (B.buildIntrinsic(Intrinsic::amdgcn_readlane, {S32})
+                 .addUse(Src0)
+                 .addUse(Src1))
+                .getReg(0));
+      }
+      break;
+    }
+    case Intrinsic::amdgcn_readfirstlane: {
+      for (unsigned i = 0; i < NumParts; ++i) {
+        Src0 = IsS16Vec ? B.buildBitcast(S32, Src0Parts.getReg(i)).getReg(0)
+                        : Src0Parts.getReg(i);
+        PartialRes.push_back(
+            (B.buildIntrinsic(Intrinsic::amdgcn_readfirstlane, {S32})
+                 .addUse(Src0)
+                 .getReg(0)));
+      }
+
+      break;
+    }
+    case Intrinsic::amdgcn_writelane: {
+      Register Src1 = MI.getOperand(3).getReg();
+      Register Src2 = MI.getOperand(4).getReg();
+      MachineInstrBuilder Src2Parts;
+
+      if (Ty.isPointer()) {
+        auto PtrToInt = B.buildPtrToInt(S64, Src2);
+        Src2Parts = B.buildUnmerge(S32, PtrToInt);
+      } else if (Ty.isPointerVector()) {
+        LLT IntVecTy = Ty.changeElementType(
+            LLT::scalar(Ty.getElementType().getSizeInBits()));
+        auto PtrToInt = B.buildPtrToInt(IntVecTy, Src2);
+        Src2Parts = B.buildUnmerge(S32, PtrToInt);
+      } else
+        Src2Parts =
+            IsS16Vec ? B.buildUnmerge(V2S16, Src2) : B.buildUnmerge(S32, Src2);
+
+      for (unsigned i = 0; i < NumParts; ++i) {
+        Src0 = IsS16Vec ? B.buildBitcast(S32, Src0Parts.getReg(i)).getReg(0)
+                        : Src0Parts.getReg(i);
+        Src2 = IsS16Vec ? B.buildBitcast(S32, Src2Parts.getReg(i)).getReg(0)
+                        : Src2Parts.getReg(i);
+        PartialRes.push_back(
+            (B.buildIntrinsic(Intrinsic::amdgcn_writelane, {S32})
+                 .addUse(Src0)
+                 .addUse(Src1)
+                 .addUse(Src2))
+                .getReg(0));
+      }
+
+      break;
+    }
+    }
+
+    if (Ty.isPointerVector()) {
+      unsigned PtrSize = Ty.getElementType().getSizeInBits();
+      SmallVector<Register, 2> PtrElements;
+      if (PtrSize == 32) {
+        // Handle 32 bit pointers
+        for (unsigned i = 0; i < NumParts; i++)
+          PtrElements.push_back(
+              B.buildIntToPtr(Ty.getElementType(), PartialRes[i]).getReg(0));
+      } else {
+        // Handle legalization of <? x [pointer type bigger than 32 bits]>
+        SmallVector<Register, 2> PtrParts;
+        unsigned NumS32Parts = PtrSize / 32;
+        unsigned PartIdx = 0;
+        for (unsigned i = 0, j = 1; i < NumParts; i += NumS32Parts, j++) {
+          // Merge S32 components of a pointer element first.
+          for (; PartIdx < (j * NumS32Parts); PartIdx++)
+            PtrParts.push_back(PartialRes[PartIdx]);
+
+          auto MergedPtr =
+              B.buildMergeLikeInstr(LLT::scalar(PtrSize), PtrParts);
+          PtrElements.push_back(
+              B.buildIntToPtr(Ty.getElementType(), MergedPtr).getReg(0));
+          PtrParts.clear();
+        }
+      }
+
+      B.buildMergeLikeInstr(DstReg, PtrElements);
+    } else {
+      if (IsS16Vec) {
+        for (unsigned i = 0; i < NumParts; i++)
+          PartialRes[i] = B.buildBitcast(V2S16, PartialRes[i]).getReg(0);
+      }
+      B.buildMergeLikeInstr(DstReg, PartialRes);
+    }
+
+    MI.eraseFromParent();
+    return true;
+  }
+
+  return false;
+}
+
 bool AMDGPULegalizerInfo::getImplicitArgPtr(Register DstReg,
                                             MachineRegisterInfo &MRI,
                                             MachineIRBuilder &B) const {
@@ -7330,6 +7516,10 @@ bool AMDGPULegalizerInfo::legalizeIntrinsic(LegalizerHelper &Helper,
     Observer.changedInstr(MI);
     return true;
   }
+  case Intrinsic::amdgcn_readlane:
+  case Intrinsic::amdgcn_writelane:
+  case Intrinsic::amdgcn_readfirstlane:
+    return legalizeLaneOp(Helper, MI, IntrID);
   default: {
     if (const AMDGPU::ImageDimIntrinsicInfo *ImageDimIntr =
             AMDGPU::getImageDimIntrinsicInfo(IntrID))
diff --git a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.h b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.h
index e5ba84a74a0f8..40e056154527f 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.h
+++ b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.h
@@ -208,6 +208,9 @@ class AMDGPULegalizerInfo final : public LegalizerInfo {
   bool legalizeBufferAtomic(MachineInstr &MI, MachineIRBuilder &B,
                             Intrinsic::ID IID) const;
 
+  bool legalizeLaneOp(LegalizerHelper &Helper, MachineInstr &MI,
+                      Intrinsic::ID IID) const;
+
   bool legalizeBVHIntrinsic(MachineInstr &MI, MachineIRBuilder &B) const;
 
   bool legalizeFPTruncRound(MachineInstr &MI, MachineIRBuilder &B) const;
diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index 0a3a56e9b3a0b..6c23bdf09974b 100644
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -6086,6 +6086,62 @@ static SDValue lowerBALLOTIntrinsic(const SITargetLowering &TLI, SDNode *N,
       DAG.getConstant(0, SL, MVT::i32), DAG.getCondCode(ISD::SETNE));
 }
 
+static SDValue lowerLaneOp(const SITargetLowering &TLI, SDNode *N,
+                           SelectionDAG &DAG) {
+  EVT VT = N->getValueType(0);
+  unsigned ValSize = VT.getSizeInBits();
+  unsigned IntrinsicID = N->getConstantOperandVal(0);
+  SDValue Src0 = N->getOperand(1);
+  SDLoc SL(N);
+  MVT IntVT = MVT::getIntegerVT(ValSize);
+
+  auto createLaneOp = [&DAG, &SL](SDValue Src0, SDValue Src1, SDValue Src2,
+                                  MVT VT) -> SDValue {
+    return (Src2 ? DAG.getNode(AMDGPUISD::WRITELANE, SL, VT, {Src0, Src1, Src2})
+            : Src1 ? DAG.getNode(AMDGPUISD::READLANE, SL, VT, {Src0, Src1})
+                   : DAG.getNode(AMDGPUISD::READFIRSTLANE, SL, VT, {Src0}));
+  };
+
+  SDValue Src1, Src2;
+  if (IntrinsicID == Intrinsic::amdgcn_readlane ||
+      IntrinsicID == Intrinsic::amdgcn_writelane) {
+    Src1 = N->getOperand(2);
+    if (IntrinsicID == Intrinsic::amdgcn_writelane)
+      Src2 = N->getOperand(3);
+  }
+
+  if (ValSize == 32) {
+    // Already legal
+    return SDValue();
+  }
+
+  if (ValSize < 32) {
+    SDValue InitBitCast = DAG.getBitcast(IntVT, Src0);
+    Src0 = DAG.getAnyExtOrTrunc(InitBitCast, SL, MVT::i32);
+    if (Src2.getNode()) {
+      SDValue Src2Cast = DAG.getBitcast(IntVT, Src2);
+      Src2 = DAG.getAnyExtOrTrunc(Src2Cast, SL, MVT::i32);
+    }
+    SDValue LaneOp = createLaneOp(Src0, Src1, Src2, MVT::i32);
+    SDValue Trunc = DAG.getAnyExtOrTrunc(LaneOp, SL, IntVT);
+    return DAG.getBitcast(VT, Trunc);
+  }
+
+  if ((ValSize % 32) == 0) {
+    MVT VecVT = MVT::getVectorVT(MVT::i32, ValSize / 32);
+    Src0 = DAG.getBitcast(VecVT, Src0);
+
+    if (Src2.getNode())
+      Src2 = DAG.getBitcast(VecVT, Src2);
+
+    SDValue LaneOp = createLaneOp(Src0, Src1, Src2, VecVT);
+    SDValue UnrolledLaneOp = DAG.UnrollVectorOp(LaneOp.getNode());
+    return DAG.getBitcast(VT, UnrolledLaneOp);
+  }
+
+  return SDValue();
+}
+
 void SITargetLowering::ReplaceNodeResults(SDNode *N,
                                           SmallVectorImpl<SDValue> &Results,
                                           SelectionDAG &DAG) const {
@@ -8553,6 +8609,10 @@ SDValue SITargetLowering::LowerINTRINSIC_WO_CHAIN(SDValue Op,
   }
   case Intrinsic::amdgcn_addrspacecast_nonnull:
     return lowerADDRSPACECAST(Op, DAG);
+  case Intrinsic::amdgcn_readlane:
+  case Intrinsic::amdgcn_readfirstlane:
+  case Intrinsic::amdgcn_writelane:
+    return lowerLaneOp(*this, Op.getNode(), DAG);
   default:
     if (const AMDGPU::ImageDimIntrinsicInfo *ImageDimIntr =
             AMDGPU::getImageDimIntrinsicInfo(IntrinsicID))
diff --git a/llvm/lib/Target/AMDGPU/SIInstructions.td b/llvm/lib/Target/AMDGPU/SIInstructions.td
index f9e811f54d05e..628fa438f696a 100644
--- a/llvm/lib/Target/AMDGPU/SIInstructions.td
+++ b/llvm/lib/Target/AMDGPU/SIInstructions.td
@@ -3400,7 +3400,7 @@ def : GCNPat<
 // FIXME: Should also do this for readlane, but tablegen crashes on
 // the ignored src1.
 def : GCNPat<
-  (int_amdgcn_readfirstlane (i32 imm:$src)),
+  (i32 (AMDGPUreadfirstlane (i32 imm:$src))),
   (S_MOV_B32 SReg_32:$src)
 >;
 
diff --git a/llvm/lib/Target/AMDGPU/VOP1Instructions.td b/llvm/lib/Target/AMDGPU/VOP1Instructions.td
index 4a56fad0cd603..9c5d6a7bf6d0b 100644
--- a/llvm/lib/Target/AMDGPU/VOP1Instructions.td
+++ b/llvm/lib/Target/AMDGPU/VOP1Instructions.td
@@ -112,7 +112,7 @@ class getVOP1Pat <SDPatternOperator node, VOPProfile P> : LetDummies {
         !if(P.HasOMod,
             [(set P.DstVT:$vdst, (node (P.Src0VT (VOP3OMods P.Src0VT:$src0,
                                                   i1:$clamp, i32:$omod))))],
-            [(set P.DstVT:$vdst, (node P.Src0RC32:$src0))]
+            [(set P.DstVT:$vdst, (node (P.Src0VT P.Src0RC32:$src0)))]
         )
     );
 }
@@ -243,11 +243,16 @@ def VOP_READFIRSTLANE : VOPProfile <[i32, i32, untyped, untyped]> {
 // FIXME: Specify SchedRW for READFIRSTLANE_B32
 // TODO: There is VOP3 encoding also
 def V_READFIRSTLANE_B32 : VOP1_Pseudo <"v_readfirstlane_b32", VOP_READFIRSTLANE,
-                                       getVOP1Pat<int_amdgcn_readfirstlane,
-                                                  VOP_READFIRSTLANE>.ret, 1> {
+                                       [], 1> {
   let isConvergent = 1;
 }
 
+foreach vt = Reg32Types.types in {
+  def : GCNPat<(vt (AMDGPUreadfirstlane (vt VRegOrLdsSrc_32:$src0))),
+        (V_READFIRSTLANE_B32 (vt VRegOrLdsSrc_32:$src0))
+  >;
+}
+
 let isReMaterializable = 1 in {
 let SchedRW = [WriteDoubleCvt] in {
 // OMod clears exceptions when set in this instruction
diff --git a/llvm/lib/Target/AMDGPU/VOP2Instructions.td b/llvm/lib/Target/AMDGPU/VOP2Instructions.td
index d2af1753d5503..b1df57320cfdd 100644
--- a/llvm/lib/Target/AMDGPU/VOP2Instructions.td
+++ b/llvm/lib/Target/AMDGPU/VOP2Instructions.td
@@ -780,14 +780,22 @@ defm V_SUBREV_U32 : VOP2Inst <"v_subrev_u32", VOP_I32_I32_I32_ARITH, null_frag,
 
 // These are special and do not read the exec mask.
 let isConvergent = 1, Uses = []<Register> in {
-def V_READLANE_B32 : VOP2_Pseudo<"v_readlane_b32", VOP_READLANE,
-  [(set i32:$vdst, (int_amdgcn_readlane i32:$src0, i32:$src1))]>;
+def V_READLANE_B32 : VOP2_Pseudo<"v_readlane_b32", VOP_READLANE,[]>;
 let IsNeverUniform = 1, Constraints = "$vdst = $vdst_in", DisableEncoding="$vdst_in" in {
-def V_WRITELANE_B32 : VOP2_Pseudo<"v_writelane_b32", VOP_WRITELANE,
-  [(set i32:$vdst, (int_amdgcn_writelane i32:$src0, i32:$src1, i32:$vdst_in))]>;
+def V_WRITELANE_B32 : VOP2_Pseudo<"v_writelane_b32", VOP_WRITELANE, []>;
 } // End IsNeverUniform, $vdst = $vdst_in, DisableEncoding $vdst_in
 } // End isConvergent = 1
 
+foreach vt = Reg32Types.types in {
+  def : GCNPat<(vt (AMDGPUreadlane vt:$src0, i32:$src1)),
+        (V_READLANE_B32 VRegOrLdsSrc_32:$src0, SCSrc_b32:$src1)
+  >;
+
+  def : GCNPat<(vt (AMDGPUwritelane vt:$src0, i32:$src1, vt:$src2)),
+        (V_WRITELANE_B32 SCSrc_b32:$src0, SCSrc_b32:$src1, VGPR_32:$src2)
+  >;
+}
+
 let isReMaterializable = 1 in {
 defm V_BFM_B32 : VOP2Inst <"v_bfm_b32", VOP_I32_I32_I32>;
 defm V_BCNT_U32_B32 : VOP2Inst <"v_bcnt_u32_b32", VOP_I32_I32_I32, add_ctpop>;
diff --git a/llvm/test/Analysis/UniformityAnalysis/AMDGPU/intrinsics.ll b/llvm/test/Analysis/UniformityAnalysis/AMDGPU/intrinsics.ll
index 26c85e83b53ad..74d2f53d7b317 100644
--- a/llvm/test/Analysis/UniformityAnalysis/AMDGPU/intrinsics.ll
+++ b/llvm/test/Analysis/UniformityAnalysis/AMDGPU/intrinsics.ll
@@ -56,9 +56,9 @@ define amdgpu_kernel void @mov_dpp8(ptr addrspace(1) %out, i32 %in) #0 {
   ret void
 }
 
-; CHECK: DIVERGENT: %tmp0 = call i32 @llvm.amdgcn.writelane(i32 0, i32 1, i32 2)
+; CHECK: DIVERGENT: %tmp0 = call i32 @llvm.amdgcn.writelane.i32(i32 0, i32 1, i32 2)
 define amdgpu_kernel void @writelane(ptr addrspace(1) %out) #0 {
-  %tmp0 = call i32 @llvm.amdgcn.writelane(i32 0, i32 1, i32 2)
+  %tmp0 = call i32 @llvm.amdgcn.writelane.i32(i32 0, i32 1, i32 2)
   store i32 %tmp0, ptr addrspace(1) %out
   ret void
 }
@@ -237,7 +237,7 @@ declare i32 @llvm.amdgcn.permlanex16.var(i32, i32, i32, i1, i1) #1
 declare i32 @llvm.amdgcn.mov.dpp.i32(i32, i32, i32, i32, i1) #1
 declare i32 @llvm.amdgcn.mov.dpp8.i32(i32, i32) #1
 declare i32 @llvm.amdgcn.update.dpp.i32(i32, i32, i32, i32, i32, i1) #1
-declare i32 @llvm.amdgcn.writelane(i32, i32, i32) #1
+declare i32 @llvm.amdgcn.writelane.i32(i32, i32, i32) #1
 declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16.v8f32.v16f16(<16 x half>, <16 x half> , <8 x float>) #1
 declare <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16.v8f32.v16i16(<16 x i16>, <16 x i16> , <8 x float>) #1
 declare <16 x half> @llvm.amdgcn.wmma.f16.16x16x16.f16.v16f16.v16f16(<16 x half>, <16 x half> , <16 x half>, i1 immarg) #1
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/atomic_optimizations_mul_one.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/atomic_optimizations_mul_one.ll
index 220dc70165e87..bdfafa89cd047 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/atomic_optimizations_mul_one.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/atomic_optimizations_mul_one.ll
@@ -1,5 +1,4 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 3
-; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: opt -S -mtriple=amdgcn-- -passes=amdgpu-atomic-optimizer %s | FileCheck -check-prefix=IR %s
 ; RUN: llc -global-isel -mtriple=amdgcn-- -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
 
@@ -74,7 +73,7 @@ define amdgpu_cs void @atomic_add_and_format(<4 x i32> inreg %arg) {
 ; IR-NEXT:    br label [[TMP11]]
 ; IR:       11:
 ; IR-NEXT:    [[TMP12:%.*]] = phi i32 [ poison, [[DOTENTRY:%.*]] ], [ [[TMP10]], [[TMP9]] ]
-; IR-NEXT:    [[TMP13:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP12]])
+; IR-NEXT:    [[TMP13:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP12]])
 ; IR-NEXT:    [[TMP14:%.*]] = add i32 [[TMP13]], [[TMP5]]
 ; IR-NEXT:    call void @llvm.amdgcn.struct.buffer.store.format.v4i32(<4 x i32> [[ARG]], <4 x i32> [[ARG]], i32 [[TMP14]], i32 0, i32 0, i32 0)
 ; IR-NEXT:    ret void
@@ -172,7 +171,7 @@ define amdgpu_cs void @atomic_sub_and_format(<4 x i32> inreg %arg) {
 ; IR-NEXT:    br label [[TMP11]]
 ; IR:       11:
 ; IR-NEXT:    [[TMP12:%.*]] = phi i32 [ poison, [[DOTENTRY:%.*]] ], [ [[TMP10]], [[TMP9]] ]
-; IR-NEXT:    [[TMP13:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP12]])
+; IR-NEXT:    [[TMP13:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP12]])
 ; IR-NEXT:    [[TMP14:%.*]] = sub i32 [[TMP13]], [[TMP5]]
 ; IR-NEXT:    call void @llvm.amdgcn.struct.buffer.store.format.v4i32(<4 x i32> [[ARG]], <4 x i32> [[ARG]], i32 [[TMP14]], i32 0, i32 0, i32 0)
 ; IR-NEXT:    ret void
@@ -273,7 +272,7 @@ define amdgpu_cs void @atomic_xor_and_format(<4 x i32> inreg %arg) {
 ; IR-NEXT:    br label [[TMP12]]
 ; IR:       12:
 ; IR-NEXT:    [[TMP13:%.*]] = phi i32 [ poison, [[DOTENTRY:%.*]] ], [ [[TMP11]], [[TMP10]] ]
-; IR-NEXT:    [[TMP14:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP13]])
+; IR-NEXT:    [[TMP14:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP13]])
 ; IR-NEXT:    [[TMP15:%.*]] = and i32 [[TMP5]], 1
 ; IR-NEXT:    [[TMP16:%.*]] = xor i32 [[TMP14]], [[TMP15]]
 ; IR-NEXT:    call void @llvm.amdgcn.struct.buffer.store.format.v4i32(<4 x i32> [[ARG]], <4 x i32> [[ARG]], i32 [[TMP16]], i32 0, i32 0, i32 0)
@@ -374,7 +373,7 @@ define amdgpu_cs void @atomic_ptr_add_and_format(ptr addrspace(8) inreg %arg) {
 ; IR-NEXT:    br label [[TMP11]]
 ; IR:       11:
 ; IR-NEXT:    [[TMP12:%.*]] = phi i32 [ poison, [[DOTENTRY:%.*]] ], [ [[TMP10]], [[TMP9]] ]
-; IR-NEXT:    [[TMP13:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP12]])
+; IR-NEXT:    [[TMP13:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP12]])
 ; IR-NEXT:    [[TMP14:%.*]] = add i32 [[TMP13]], [[TMP5]]
 ; IR-NEXT:    [[ARG_INT:%.*]] = ptrtoint ptr addrspace(8) [[ARG]] to i128
 ; IR-NEXT:    [[ARG_VEC:%.*]] = bitcast i128 [[ARG_INT]] to <4 x i32>
@@ -476,7 +475,7 @@ define amdgpu_cs void @atomic_ptr_sub_and_format(ptr addrspace(8) inreg %arg) {
 ; IR-NEXT:    br label [[TMP11]]
 ; IR:       11:
 ; IR-NEXT:    [[TMP12:%.*]] = phi i32 [ poison, [[DOTENTRY:%.*]] ], [ [[TMP10]], [[TMP9]] ]
-; IR-NEXT:    [[TMP13:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP12]])
+; IR-NEXT:    [[TMP13:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP12]])
 ; IR-NEXT:    [[TMP14:%.*]] = sub i32 [[TMP13]], [[TMP5]]
 ; IR-NEXT:    [[ARG_INT:%.*]] = ptrtoint ptr addrspace(8) [[ARG]] to i128
 ; IR-NEXT:    [[ARG_VEC:%.*]] = bitcast i128 [[ARG_INT]] to <4 x i32>
@@ -581,7 +580,7 @@ define amdgpu_cs void @atomic_ptr_xor_and_format(ptr addrspace(8) inreg %arg) {
 ; IR-NEXT:    br label [[TMP12]]
 ; IR:       12:
 ; IR-NEXT:    [[TMP13:%.*]] = phi i32 [ poison, [[DOTENTRY:%.*]] ], [ [[TMP11]], [[TMP10]] ]
-; IR-NEXT:    [[TMP14:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP13]])
+; IR-NEXT:    [[TMP14:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP13]])
 ; IR-NEXT:    [[TMP15:%.*]] = and i32 [[TMP5]], 1
 ; IR-NEXT:    [[TMP16:%.*]] = xor i32 [[TMP14]], [[TMP15]]
 ; IR-NEXT:    [[ARG_INT:%.*]] = ptrtoint ptr addrspace(8) [[ARG]] to i128
diff --git a/llvm/test/CodeGen/AMDGPU/atomic_optimization_split_dt_update.ll b/llvm/test/CodeGen/AMDGPU/atomic_optimization_split_dt_update.ll
index c07cd4e493b9a..019f76aa44a87 100644
--- a/llvm/test/CodeGen/AMDGPU/atomic_optimization_split_dt_update.ll
+++ b/llvm/test/CodeGen/AMDGPU/atomic_optimization_split_dt_update.ll
@@ -48,7 +48,7 @@ define amdgpu_kernel void @ham(ptr addrspace(4) %arg) {
 ; CHECK-NEXT:    [[ACTIVEBITS:%.*]] = phi i64 [ [[TMP6]], [[BB7]] ], [ [[TMP16:%.*]], [[COMPUTELOOP]] ]
 ; CHECK-NEXT:    [[TMP10:%.*]] = call i64 @llvm.cttz.i64(i64 [[ACTIVEBITS]], i1 true)
 ; CHECK-NEXT:    [[TMP11:%.*]] = trunc i64 [[TMP10]] to i32
-; CHECK-NEXT:    [[TMP12:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[PHI]], i32 [[TMP11]])
+; CHECK-NEXT:    [[TMP12:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[PHI]], i32 [[TMP11]])
 ; CHECK-NEXT:    [[TMP13]] = add i32 [[ACCUMULATOR]], [[TMP12]]
 ; CHECK-NEXT:    [[TMP14:%.*]] = shl i64 1, [[TMP10]]
 ; CHECK-NEXT:    [[TMP15:%.*]] = xor i64 [[TMP14]], -1
diff --git a/llvm/test/CodeGen/AMDGPU/global-atomic-scan.ll b/llvm/test/CodeGen/AMDGPU/global-atomic-scan.ll
index 6b47f81bccb71..6c61c837881c4 100644
--- a/llvm/test/CodeGen/AMDGPU/global-atomic-scan.ll
+++ b/llvm/test/CodeGen/AMDGPU/global-atomic-scan.ll
@@ -130,7 +130,7 @@ define amdgpu_kernel void @atomic_add_i32_ret_offset(ptr addrspace(1) %out, ptr
 ; IR-NEXT:    br label [[TMP12]]
 ; IR:       12:
 ; IR-NEXT:    [[TMP13:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[TMP11]], [[TMP10]] ]
-; IR-NEXT:    [[TMP14:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP13]])
+; IR-NEXT:    [[TMP14:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP13]])
 ; IR-NEXT:    [[TMP15:%.*]] = mul i32 [[IN]], [[TMP5]]
 ; IR-NEXT:    [[TMP16:%.*]] = add i32 [[TMP14]], [[TMP15]]
 ; IR-NEXT:    store i32 [[TMP16]], ptr addrspace(1) [[OUT2:%.*]], align 4
@@ -193,7 +193,7 @@ define amdgpu_kernel void @atomic_add_i32_ret_addr64_offset(ptr addrspace(1) %ou
 ; IR-NEXT:    br label [[TMP12]]
 ; IR:       12:
 ; IR-NEXT:    [[TMP13:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[TMP11]], [[TMP10]] ]
-; IR-NEXT:    [[TMP14:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP13]])
+; IR-NEXT:    [[TMP14:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP13]])
 ; IR-NEXT:    [[TMP15:%.*]] = mul i32 [[IN]], [[TMP5]]
 ; IR-NEXT:    [[TMP16:%.*]] = add i32 [[TMP14]], [[TMP15]]
 ; IR-NEXT:    store i32 [[TMP16]], ptr addrspace(1) [[OUT2:%.*]], align 4
@@ -251,7 +251,7 @@ define amdgpu_kernel void @atomic_add_i32_ret(ptr addrspace(1) %out, ptr addrspa
 ; IR-NEXT:    br label [[TMP12]]
 ; IR:       12:
 ; IR-NEXT:    [[TMP13:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[TMP11]], [[TMP10]] ]
-; IR-NEXT:    [[TMP14:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP13]])
+; IR-NEXT:    [[TMP14:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP13]])
 ; IR-NEXT:    [[TMP15:%.*]] = mul i32 [[IN]], [[TMP5]]
 ; IR-NEXT:    [[TMP16:%.*]] = add i32 [[TMP14]], [[TMP15]]
 ; IR-NEXT:    store i32 [[TMP16]], ptr addrspace(1) [[OUT2:%.*]], align 4
@@ -310,7 +310,7 @@ define amdgpu_kernel void @atomic_add_i32_ret_addr64(ptr addrspace(1) %out, ptr
 ; IR-NEXT:    br label [[TMP12]]
 ; IR:       12:
 ; IR-NEXT:    [[TMP13:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[TMP11]], [[TMP10]] ]
-; IR-NEXT:    [[TMP14:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP13]])
+; IR-NEXT:    [[TMP14:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP13]])
 ; IR-NEXT:    [[TMP15:%.*]] = mul i32 [[IN]], [[TMP5]]
 ; IR-NEXT:    [[TMP16:%.*]] = add i32 [[TMP14]], [[TMP15]]
 ; IR-NEXT:    store i32 [[TMP16]], ptr addrspace(1) [[OUT2:%.*]], align 4
@@ -364,7 +364,7 @@ define amdgpu_kernel void @atomic_and_i32_ret_offset(ptr addrspace(1) %out, ptr
 ; IR-NEXT:    br label [[TMP9]]
 ; IR:       9:
 ; IR-NEXT:    [[TMP10:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[TMP8]], [[TMP7]] ]
-; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP10]])
+; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP10]])
 ; IR-NEXT:    [[TMP12:%.*]] = select i1 [[TMP6]], i32 -1, i32 [[IN]]
 ; IR-NEXT:    [[TMP13:%.*]] = and i32 [[TMP11]], [[TMP12]]
 ; IR-NEXT:    store i32 [[TMP13]], ptr addrspace(1) [[OUT2:%.*]], align 4
@@ -421,7 +421,7 @@ define amdgpu_kernel void @atomic_and_i32_ret_addr64_offset(ptr addrspace(1) %ou
 ; IR-NEXT:    br label [[TMP9]]
 ; IR:       9:
 ; IR-NEXT:    [[TMP10:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[TMP8]], [[TMP7]] ]
-; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP10]])
+; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP10]])
 ; IR-NEXT:    [[TMP12:%.*]] = select i1 [[TMP6]], i32 -1, i32 [[IN]]
 ; IR-NEXT:    [[TMP13:%.*]] = and i32 [[TMP11]], [[TMP12]]
 ; IR-NEXT:    store i32 [[TMP13]], ptr addrspace(1) [[OUT2:%.*]], align 4
@@ -473,7 +473,7 @@ define amdgpu_kernel void @atomic_and_i32_ret(ptr addrspace(1) %out, ptr addrspa
 ; IR-NEXT:    br label [[TMP9]]
 ; IR:       9:
 ; IR-NEXT:    [[TMP10:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[TMP8]], [[TMP7]] ]
-; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP10]])
+; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP10]])
 ; IR-NEXT:    [[TMP12:%.*]] = select i1 [[TMP6]], i32 -1, i32 [[IN]]
 ; IR-NEXT:    [[TMP13:%.*]] = and i32 [[TMP11]], [[TMP12]]
 ; IR-NEXT:    store i32 [[TMP13]], ptr addrspace(1) [[OUT2:%.*]], align 4
@@ -526,7 +526,7 @@ define amdgpu_kernel void @atomic_and_i32_ret_addr64(ptr addrspace(1) %out, ptr
 ; IR-NEXT:    br label [[TMP9]]
 ; IR:       9:
 ; IR-NEXT:    [[TMP10:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[TMP8]], [[TMP7]] ]
-; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP10]])
+; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP10]])
 ; IR-NEXT:    [[TMP12:%.*]] = select i1 [[TMP6]], i32 -1, i32 [[IN]]
 ; IR-NEXT:    [[TMP13:%.*]] = and i32 [[TMP11]], [[TMP12]]
 ; IR-NEXT:    store i32 [[TMP13]], ptr addrspace(1) [[OUT2:%.*]], align 4
@@ -586,7 +586,7 @@ define amdgpu_kernel void @atomic_sub_i32_ret_offset(ptr addrspace(1) %out, ptr
 ; IR-NEXT:    br label [[TMP12]]
 ; IR:       12:
 ; IR-NEXT:    [[TMP13:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[TMP11]], [[TMP10]] ]
-; IR-NEXT:    [[TMP14:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP13]])
+; IR-NEXT:    [[TMP14:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP13]])
 ; IR-NEXT:    [[TMP15:%.*]] = mul i32 [[IN]], [[TMP5]]
 ; IR-NEXT:    [[TMP16:%.*]] = sub i32 [[TMP14]], [[TMP15]]
 ; IR-NEXT:    store i32 [[TMP16]], ptr addrspace(1) [[OUT2:%.*]], align 4
@@ -649,7 +649,7 @@ define amdgpu_kernel void @atomic_sub_i32_ret_addr64_offset(ptr addrspace(1) %ou
 ; IR-NEXT:    br label [[TMP12]]
 ; IR:       12:
 ; IR-NEXT:    [[TMP13:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[TMP11]], [[TMP10]] ]
-; IR-NEXT:    [[TMP14:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP13]])
+; IR-NEXT:    [[TMP14:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP13]])
 ; IR-NEXT:    [[TMP15:%.*]] = mul i32 [[IN]], [[TMP5]]
 ; IR-NEXT:    [[TMP16:%.*]] = sub i32 [[TMP14]], [[TMP15]]
 ; IR-NEXT:    store i32 [[TMP16]], ptr addrspace(1) [[OUT2:%.*]], align 4
@@ -707,7 +707,7 @@ define amdgpu_kernel void @atomic_sub_i32_ret(ptr addrspace(1) %out, ptr addrspa
 ; IR-NEXT:    br label [[TMP12]]
 ; IR:       12:
 ; IR-NEXT:    [[TMP13:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[TMP11]], [[TMP10]] ]
-; IR-NEXT:    [[TMP14:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP13]])
+; IR-NEXT:    [[TMP14:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP13]])
 ; IR-NEXT:    [[TMP15:%.*]] = mul i32 [[IN]], [[TMP5]]
 ; IR-NEXT:    [[TMP16:%.*]] = sub i32 [[TMP14]], [[TMP15]]
 ; IR-NEXT:    store i32 [[TMP16]], ptr addrspace(1) [[OUT2:%.*]], align 4
@@ -766,7 +766,7 @@ define amdgpu_kernel void @atomic_sub_i32_ret_addr64(ptr addrspace(1) %out, ptr
 ; IR-NEXT:    br label [[TMP12]]
 ; IR:       12:
 ; IR-NEXT:    [[TMP13:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[TMP11]], [[TMP10]] ]
-; IR-NEXT:    [[TMP14:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP13]])
+; IR-NEXT:    [[TMP14:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP13]])
 ; IR-NEXT:    [[TMP15:%.*]] = mul i32 [[IN]], [[TMP5]]
 ; IR-NEXT:    [[TMP16:%.*]] = sub i32 [[TMP14]], [[TMP15]]
 ; IR-NEXT:    store i32 [[TMP16]], ptr addrspace(1) [[OUT2:%.*]], align 4
@@ -820,7 +820,7 @@ define amdgpu_kernel void @atomic_max_i32_ret_offset(ptr addrspace(1) %out, ptr
 ; IR-NEXT:    br label [[TMP9]]
 ; IR:       9:
 ; IR-NEXT:    [[TMP10:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[TMP8]], [[TMP7]] ]
-; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP10]])
+; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP10]])
 ; IR-NEXT:    [[TMP12:%.*]] = select i1 [[TMP6]], i32 -2147483648, i32 [[IN]]
 ; IR-NEXT:    [[TMP13:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
 ; IR-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP11]], i32 [[TMP12]]
@@ -878,7 +878,7 @@ define amdgpu_kernel void @atomic_max_i32_ret_addr64_offset(ptr addrspace(1) %ou
 ; IR-NEXT:    br label [[TMP9]]
 ; IR:       9:
 ; IR-NEXT:    [[TMP10:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[TMP8]], [[TMP7]] ]
-; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP10]])
+; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP10]])
 ; IR-NEXT:    [[TMP12:%.*]] = select i1 [[TMP6]], i32 -2147483648, i32 [[IN]]
 ; IR-NEXT:    [[TMP13:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
 ; IR-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP11]], i32 [[TMP12]]
@@ -931,7 +931,7 @@ define amdgpu_kernel void @atomic_max_i32_ret(ptr addrspace(1) %out, ptr addrspa
 ; IR-NEXT:    br label [[TMP9]]
 ; IR:       9:
 ; IR-NEXT:    [[TMP10:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[TMP8]], [[TMP7]] ]
-; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP10]])
+; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP10]])
 ; IR-NEXT:    [[TMP12:%.*]] = select i1 [[TMP6]], i32 -2147483648, i32 [[IN]]
 ; IR-NEXT:    [[TMP13:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
 ; IR-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP11]], i32 [[TMP12]]
@@ -985,7 +985,7 @@ define amdgpu_kernel void @atomic_max_i32_ret_addr64(ptr addrspace(1) %out, ptr
 ; IR-NEXT:    br label [[TMP9]]
 ; IR:       9:
 ; IR-NEXT:    [[TMP10:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[TMP8]], [[TMP7]] ]
-; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP10]])
+; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP10]])
 ; IR-NEXT:    [[TMP12:%.*]] = select i1 [[TMP6]], i32 -2147483648, i32 [[IN]]
 ; IR-NEXT:    [[TMP13:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
 ; IR-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP11]], i32 [[TMP12]]
@@ -1040,7 +1040,7 @@ define amdgpu_kernel void @atomic_umax_i32_ret_offset(ptr addrspace(1) %out, ptr
 ; IR-NEXT:    br label [[TMP9]]
 ; IR:       9:
 ; IR-NEXT:    [[TMP10:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[TMP8]], [[TMP7]] ]
-; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP10]])
+; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP10]])
 ; IR-NEXT:    [[TMP12:%.*]] = select i1 [[TMP6]], i32 0, i32 [[IN]]
 ; IR-NEXT:    [[TMP13:%.*]] = icmp ugt i32 [[TMP11]], [[TMP12]]
 ; IR-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP11]], i32 [[TMP12]]
@@ -1098,7 +1098,7 @@ define amdgpu_kernel void @atomic_umax_i32_ret_addr64_offset(ptr addrspace(1) %o
 ; IR-NEXT:    br label [[TMP9]]
 ; IR:       9:
 ; IR-NEXT:    [[TMP10:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[TMP8]], [[TMP7]] ]
-; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP10]])
+; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP10]])
 ; IR-NEXT:    [[TMP12:%.*]] = select i1 [[TMP6]], i32 0, i32 [[IN]]
 ; IR-NEXT:    [[TMP13:%.*]] = icmp ugt i32 [[TMP11]], [[TMP12]]
 ; IR-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP11]], i32 [[TMP12]]
@@ -1151,7 +1151,7 @@ define amdgpu_kernel void @atomic_umax_i32_ret(ptr addrspace(1) %out, ptr addrsp
 ; IR-NEXT:    br label [[TMP9]]
 ; IR:       9:
 ; IR-NEXT:    [[TMP10:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[TMP8]], [[TMP7]] ]
-; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP10]])
+; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP10]])
 ; IR-NEXT:    [[TMP12:%.*]] = select i1 [[TMP6]], i32 0, i32 [[IN]]
 ; IR-NEXT:    [[TMP13:%.*]] = icmp ugt i32 [[TMP11]], [[TMP12]]
 ; IR-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP11]], i32 [[TMP12]]
@@ -1205,7 +1205,7 @@ define amdgpu_kernel void @atomic_umax_i32_ret_addr64(ptr addrspace(1) %out, ptr
 ; IR-NEXT:    br label [[TMP9]]
 ; IR:       9:
 ; IR-NEXT:    [[TMP10:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[TMP8]], [[TMP7]] ]
-; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP10]])
+; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP10]])
 ; IR-NEXT:    [[TMP12:%.*]] = select i1 [[TMP6]], i32 0, i32 [[IN]]
 ; IR-NEXT:    [[TMP13:%.*]] = icmp ugt i32 [[TMP11]], [[TMP12]]
 ; IR-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP11]], i32 [[TMP12]]
@@ -1260,7 +1260,7 @@ define amdgpu_kernel void @atomic_min_i32_ret_offset(ptr addrspace(1) %out, ptr
 ; IR-NEXT:    br label [[TMP9]]
 ; IR:       9:
 ; IR-NEXT:    [[TMP10:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[TMP8]], [[TMP7]] ]
-; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP10]])
+; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP10]])
 ; IR-NEXT:    [[TMP12:%.*]] = select i1 [[TMP6]], i32 2147483647, i32 [[IN]]
 ; IR-NEXT:    [[TMP13:%.*]] = icmp slt i32 [[TMP11]], [[TMP12]]
 ; IR-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP11]], i32 [[TMP12]]
@@ -1318,7 +1318,7 @@ define amdgpu_kernel void @atomic_min_i32_ret_addr64_offset(ptr addrspace(1) %ou
 ; IR-NEXT:    br label [[TMP9]]
 ; IR:       9:
 ; IR-NEXT:    [[TMP10:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[TMP8]], [[TMP7]] ]
-; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP10]])
+; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP10]])
 ; IR-NEXT:    [[TMP12:%.*]] = select i1 [[TMP6]], i32 2147483647, i32 [[IN]]
 ; IR-NEXT:    [[TMP13:%.*]] = icmp slt i32 [[TMP11]], [[TMP12]]
 ; IR-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP11]], i32 [[TMP12]]
@@ -1371,7 +1371,7 @@ define amdgpu_kernel void @atomic_min_i32_ret(ptr addrspace(1) %out, ptr addrspa
 ; IR-NEXT:    br label [[TMP9]]
 ; IR:       9:
 ; IR-NEXT:    [[TMP10:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[TMP8]], [[TMP7]] ]
-; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP10]])
+; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP10]])
 ; IR-NEXT:    [[TMP12:%.*]] = select i1 [[TMP6]], i32 2147483647, i32 [[IN]]
 ; IR-NEXT:    [[TMP13:%.*]] = icmp slt i32 [[TMP11]], [[TMP12]]
 ; IR-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP11]], i32 [[TMP12]]
@@ -1425,7 +1425,7 @@ define amdgpu_kernel void @atomic_min_i32_ret_addr64(ptr addrspace(1) %out, ptr
 ; IR-NEXT:    br label [[TMP9]]
 ; IR:       9:
 ; IR-NEXT:    [[TMP10:%.*]] = phi i32 [ poison, [[ENTRY:%.*]] ], [ [[TMP8]], [[TMP7]] ]
-; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP10]])
+; IR-NEXT:    [[TMP11:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP10]])
 ; IR-NEXT:    [[TMP12:%.*]] = select i1 [[TMP6]], i32 2147483647, i32 [[IN]]
 ; IR-NEXT:    [[TMP13:%.*]] = icmp slt i32 [[TMP11]], [[TMP12]]
 ; IR-NEXT:    [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP11]], i32 [[TMP12]]
diff --git a/llvm/test/CodeGen/AMDGPU/global_atomic_optimizer_fp_rtn.ll b/llvm/test/CodeGen/AMDGPU/global_atomic_optimizer_fp_rtn.ll
index b71728096093c..baaf50377338c 100644
--- a/llvm/test/CodeGen/AMDGPU/global_atomic_optimizer_fp_rtn.ll
+++ b/llvm/test/CodeGen/AMDGPU/global_atomic_optimizer_fp_rtn.ll
@@ -29,7 +29,7 @@ define amdgpu_ps float @global_atomic_fadd_uni_address_uni_value_agent_scope_uns
 ; IR:       16:
 ; IR-NEXT:    [[TMP17:%.*]] = phi float [ poison, [[TMP2]] ], [ [[TMP15]], [[TMP14]] ]
 ; IR-NEXT:    [[TMP18:%.*]] = bitcast float [[TMP17]] to i32
-; IR-NEXT:    [[TMP19:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP18]])
+; IR-NEXT:    [[TMP19:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP18]])
 ; IR-NEXT:    [[TMP20:%.*]] = bitcast i32 [[TMP19]] to float
 ; IR-NEXT:    [[TMP21:%.*]] = uitofp i32 [[TMP8]] to float
 ; IR-NEXT:    [[TMP22:%.*]] = fmul float [[VAL]], [[TMP21]]
@@ -62,7 +62,7 @@ define amdgpu_ps float @global_atomic_fadd_uni_address_div_value_scope_agent_sco
 ; IR-ITERATIVE:       12:
 ; IR-ITERATIVE-NEXT:    [[TMP13:%.*]] = phi float [ poison, [[COMPUTEEND:%.*]] ], [ [[TMP11]], [[TMP10:%.*]] ]
 ; IR-ITERATIVE-NEXT:    [[TMP14:%.*]] = bitcast float [[TMP13]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP15:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP14]])
+; IR-ITERATIVE-NEXT:    [[TMP15:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP14]])
 ; IR-ITERATIVE-NEXT:    [[TMP16:%.*]] = bitcast i32 [[TMP15]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP17:%.*]] = fadd float [[TMP16]], [[TMP28:%.*]]
 ; IR-ITERATIVE-NEXT:    br label [[TMP18]]
@@ -76,11 +76,11 @@ define amdgpu_ps float @global_atomic_fadd_uni_address_div_value_scope_agent_sco
 ; IR-ITERATIVE-NEXT:    [[TMP20:%.*]] = call i64 @llvm.cttz.i64(i64 [[ACTIVEBITS]], i1 true)
 ; IR-ITERATIVE-NEXT:    [[TMP21:%.*]] = trunc i64 [[TMP20]] to i32
 ; IR-ITERATIVE-NEXT:    [[TMP22:%.*]] = bitcast float [[VAL:%.*]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP22]], i32 [[TMP21]])
+; IR-ITERATIVE-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP22]], i32 [[TMP21]])
 ; IR-ITERATIVE-NEXT:    [[TMP24:%.*]] = bitcast i32 [[TMP23]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP25:%.*]] = bitcast float [[ACCUMULATOR]] to i32
 ; IR-ITERATIVE-NEXT:    [[TMP26:%.*]] = bitcast float [[OLDVALUEPHI]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP27:%.*]] = call i32 @llvm.amdgcn.writelane(i32 [[TMP25]], i32 [[TMP21]], i32 [[TMP26]])
+; IR-ITERATIVE-NEXT:    [[TMP27:%.*]] = call i32 @llvm.amdgcn.writelane.i32(i32 [[TMP25]], i32 [[TMP21]], i32 [[TMP26]])
 ; IR-ITERATIVE-NEXT:    [[TMP28]] = bitcast i32 [[TMP27]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP29]] = fadd float [[ACCUMULATOR]], [[TMP24]]
 ; IR-ITERATIVE-NEXT:    [[TMP30:%.*]] = shl i64 1, [[TMP20]]
@@ -120,7 +120,7 @@ define amdgpu_ps float @global_atomic_fadd_uni_address_div_value_scope_agent_sco
 ; IR-DPP-NEXT:    [[TMP24:%.*]] = fadd float [[TMP22]], [[TMP23]]
 ; IR-DPP-NEXT:    [[TMP25:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float -0.000000e+00, float [[TMP24]], i32 312, i32 15, i32 15, i1 false)
 ; IR-DPP-NEXT:    [[TMP26:%.*]] = bitcast float [[TMP24]] to i32
-; IR-DPP-NEXT:    [[TMP27:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP26]], i32 63)
+; IR-DPP-NEXT:    [[TMP27:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP26]], i32 63)
 ; IR-DPP-NEXT:    [[TMP28:%.*]] = bitcast i32 [[TMP27]] to float
 ; IR-DPP-NEXT:    [[TMP29:%.*]] = call float @llvm.amdgcn.strict.wwm.f32(float [[TMP28]])
 ; IR-DPP-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[TMP8]], 0
@@ -131,7 +131,7 @@ define amdgpu_ps float @global_atomic_fadd_uni_address_div_value_scope_agent_sco
 ; IR-DPP:       33:
 ; IR-DPP-NEXT:    [[TMP34:%.*]] = phi float [ poison, [[TMP2]] ], [ [[TMP32]], [[TMP31]] ]
 ; IR-DPP-NEXT:    [[TMP35:%.*]] = bitcast float [[TMP34]] to i32
-; IR-DPP-NEXT:    [[TMP36:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP35]])
+; IR-DPP-NEXT:    [[TMP36:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP35]])
 ; IR-DPP-NEXT:    [[TMP37:%.*]] = bitcast i32 [[TMP36]] to float
 ; IR-DPP-NEXT:    [[TMP38:%.*]] = call float @llvm.amdgcn.strict.wwm.f32(float [[TMP25]])
 ; IR-DPP-NEXT:    [[TMP39:%.*]] = fadd float [[TMP37]], [[TMP38]]
@@ -167,7 +167,7 @@ define amdgpu_ps float @global_atomic_fadd_uni_address_uni_value_one_as_scope_un
 ; IR-ITERATIVE:       16:
 ; IR-ITERATIVE-NEXT:    [[TMP17:%.*]] = phi float [ poison, [[TMP2]] ], [ [[TMP15]], [[TMP14]] ]
 ; IR-ITERATIVE-NEXT:    [[TMP18:%.*]] = bitcast float [[TMP17]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP19:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP18]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP19:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP18]]) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP20:%.*]] = bitcast i32 [[TMP19]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP21:%.*]] = call float @llvm.experimental.constrained.uitofp.f32.i32(i32 [[TMP8]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP22:%.*]] = call float @llvm.experimental.constrained.fmul.f32(float [[VAL]], float [[TMP21]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR7]]
@@ -199,7 +199,7 @@ define amdgpu_ps float @global_atomic_fadd_uni_address_uni_value_one_as_scope_un
 ; IR-DPP:       16:
 ; IR-DPP-NEXT:    [[TMP17:%.*]] = phi float [ poison, [[TMP2]] ], [ [[TMP15]], [[TMP14]] ]
 ; IR-DPP-NEXT:    [[TMP18:%.*]] = bitcast float [[TMP17]] to i32
-; IR-DPP-NEXT:    [[TMP19:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP18]]) #[[ATTR8]]
+; IR-DPP-NEXT:    [[TMP19:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP18]]) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP20:%.*]] = bitcast i32 [[TMP19]] to float
 ; IR-DPP-NEXT:    [[TMP21:%.*]] = call float @llvm.experimental.constrained.uitofp.f32.i32(i32 [[TMP8]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP22:%.*]] = call float @llvm.experimental.constrained.fmul.f32(float [[VAL]], float [[TMP21]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR8]]
@@ -232,7 +232,7 @@ define amdgpu_ps float @global_atomic_fadd_uni_address_div_value_one_as_scope_un
 ; IR-ITERATIVE:       12:
 ; IR-ITERATIVE-NEXT:    [[TMP13:%.*]] = phi float [ poison, [[COMPUTEEND:%.*]] ], [ [[TMP11]], [[TMP10:%.*]] ]
 ; IR-ITERATIVE-NEXT:    [[TMP14:%.*]] = bitcast float [[TMP13]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP15:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP14]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP15:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP14]]) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP16:%.*]] = bitcast i32 [[TMP15]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP17:%.*]] = call float @llvm.experimental.constrained.fadd.f32(float [[TMP16]], float [[TMP28:%.*]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    br label [[TMP18]]
@@ -246,11 +246,11 @@ define amdgpu_ps float @global_atomic_fadd_uni_address_div_value_one_as_scope_un
 ; IR-ITERATIVE-NEXT:    [[TMP20:%.*]] = call i64 @llvm.cttz.i64(i64 [[ACTIVEBITS]], i1 true) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP21:%.*]] = trunc i64 [[TMP20]] to i32
 ; IR-ITERATIVE-NEXT:    [[TMP22:%.*]] = bitcast float [[VAL:%.*]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP22]], i32 [[TMP21]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP22]], i32 [[TMP21]]) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP24:%.*]] = bitcast i32 [[TMP23]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP25:%.*]] = bitcast float [[ACCUMULATOR]] to i32
 ; IR-ITERATIVE-NEXT:    [[TMP26:%.*]] = bitcast float [[OLDVALUEPHI]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP27:%.*]] = call i32 @llvm.amdgcn.writelane(i32 [[TMP25]], i32 [[TMP21]], i32 [[TMP26]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP27:%.*]] = call i32 @llvm.amdgcn.writelane.i32(i32 [[TMP25]], i32 [[TMP21]], i32 [[TMP26]]) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP28]] = bitcast i32 [[TMP27]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP29]] = call float @llvm.experimental.constrained.fadd.f32(float [[ACCUMULATOR]], float [[TMP24]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP30:%.*]] = shl i64 1, [[TMP20]]
@@ -290,7 +290,7 @@ define amdgpu_ps float @global_atomic_fadd_uni_address_div_value_one_as_scope_un
 ; IR-DPP-NEXT:    [[TMP24:%.*]] = call float @llvm.experimental.constrained.fadd.f32(float [[TMP22]], float [[TMP23]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP25:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float -0.000000e+00, float [[TMP24]], i32 312, i32 15, i32 15, i1 false) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP26:%.*]] = bitcast float [[TMP24]] to i32
-; IR-DPP-NEXT:    [[TMP27:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP26]], i32 63) #[[ATTR8]]
+; IR-DPP-NEXT:    [[TMP27:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP26]], i32 63) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP28:%.*]] = bitcast i32 [[TMP27]] to float
 ; IR-DPP-NEXT:    [[TMP29:%.*]] = call float @llvm.amdgcn.strict.wwm.f32(float [[TMP28]]) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[TMP8]], 0
@@ -301,7 +301,7 @@ define amdgpu_ps float @global_atomic_fadd_uni_address_div_value_one_as_scope_un
 ; IR-DPP:       33:
 ; IR-DPP-NEXT:    [[TMP34:%.*]] = phi float [ poison, [[TMP2]] ], [ [[TMP32]], [[TMP31]] ]
 ; IR-DPP-NEXT:    [[TMP35:%.*]] = bitcast float [[TMP34]] to i32
-; IR-DPP-NEXT:    [[TMP36:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP35]]) #[[ATTR8]]
+; IR-DPP-NEXT:    [[TMP36:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP35]]) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP37:%.*]] = bitcast i32 [[TMP36]] to float
 ; IR-DPP-NEXT:    [[TMP38:%.*]] = call float @llvm.amdgcn.strict.wwm.f32(float [[TMP25]]) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP39:%.*]] = call float @llvm.experimental.constrained.fadd.f32(float [[TMP37]], float [[TMP38]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR8]]
@@ -337,7 +337,7 @@ define amdgpu_ps float @global_atomic_fsub_uni_address_uni_value_agent_scope_str
 ; IR-ITERATIVE:       16:
 ; IR-ITERATIVE-NEXT:    [[TMP17:%.*]] = phi float [ poison, [[TMP2]] ], [ [[TMP15]], [[TMP14]] ]
 ; IR-ITERATIVE-NEXT:    [[TMP18:%.*]] = bitcast float [[TMP17]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP19:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP18]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP19:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP18]]) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP20:%.*]] = bitcast i32 [[TMP19]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP21:%.*]] = call float @llvm.experimental.constrained.uitofp.f32.i32(i32 [[TMP8]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP22:%.*]] = call float @llvm.experimental.constrained.fmul.f32(float [[VAL]], float [[TMP21]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR7]]
@@ -369,7 +369,7 @@ define amdgpu_ps float @global_atomic_fsub_uni_address_uni_value_agent_scope_str
 ; IR-DPP:       16:
 ; IR-DPP-NEXT:    [[TMP17:%.*]] = phi float [ poison, [[TMP2]] ], [ [[TMP15]], [[TMP14]] ]
 ; IR-DPP-NEXT:    [[TMP18:%.*]] = bitcast float [[TMP17]] to i32
-; IR-DPP-NEXT:    [[TMP19:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP18]]) #[[ATTR8]]
+; IR-DPP-NEXT:    [[TMP19:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP18]]) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP20:%.*]] = bitcast i32 [[TMP19]] to float
 ; IR-DPP-NEXT:    [[TMP21:%.*]] = call float @llvm.experimental.constrained.uitofp.f32.i32(i32 [[TMP8]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP22:%.*]] = call float @llvm.experimental.constrained.fmul.f32(float [[VAL]], float [[TMP21]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR8]]
@@ -402,7 +402,7 @@ define amdgpu_ps float @global_atomic_fsub_uni_address_div_value_agent_scope_str
 ; IR-ITERATIVE:       12:
 ; IR-ITERATIVE-NEXT:    [[TMP13:%.*]] = phi float [ poison, [[COMPUTEEND:%.*]] ], [ [[TMP11]], [[TMP10:%.*]] ]
 ; IR-ITERATIVE-NEXT:    [[TMP14:%.*]] = bitcast float [[TMP13]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP15:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP14]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP15:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP14]]) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP16:%.*]] = bitcast i32 [[TMP15]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP17:%.*]] = call float @llvm.experimental.constrained.fsub.f32(float [[TMP16]], float [[TMP28:%.*]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    br label [[TMP18]]
@@ -416,11 +416,11 @@ define amdgpu_ps float @global_atomic_fsub_uni_address_div_value_agent_scope_str
 ; IR-ITERATIVE-NEXT:    [[TMP20:%.*]] = call i64 @llvm.cttz.i64(i64 [[ACTIVEBITS]], i1 true) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP21:%.*]] = trunc i64 [[TMP20]] to i32
 ; IR-ITERATIVE-NEXT:    [[TMP22:%.*]] = bitcast float [[VAL:%.*]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP22]], i32 [[TMP21]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP22]], i32 [[TMP21]]) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP24:%.*]] = bitcast i32 [[TMP23]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP25:%.*]] = bitcast float [[ACCUMULATOR]] to i32
 ; IR-ITERATIVE-NEXT:    [[TMP26:%.*]] = bitcast float [[OLDVALUEPHI]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP27:%.*]] = call i32 @llvm.amdgcn.writelane(i32 [[TMP25]], i32 [[TMP21]], i32 [[TMP26]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP27:%.*]] = call i32 @llvm.amdgcn.writelane.i32(i32 [[TMP25]], i32 [[TMP21]], i32 [[TMP26]]) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP28]] = bitcast i32 [[TMP27]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP29]] = call float @llvm.experimental.constrained.fadd.f32(float [[ACCUMULATOR]], float [[TMP24]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP30:%.*]] = shl i64 1, [[TMP20]]
@@ -460,7 +460,7 @@ define amdgpu_ps float @global_atomic_fsub_uni_address_div_value_agent_scope_str
 ; IR-DPP-NEXT:    [[TMP24:%.*]] = call float @llvm.experimental.constrained.fadd.f32(float [[TMP22]], float [[TMP23]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP25:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float -0.000000e+00, float [[TMP24]], i32 312, i32 15, i32 15, i1 false) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP26:%.*]] = bitcast float [[TMP24]] to i32
-; IR-DPP-NEXT:    [[TMP27:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP26]], i32 63) #[[ATTR8]]
+; IR-DPP-NEXT:    [[TMP27:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP26]], i32 63) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP28:%.*]] = bitcast i32 [[TMP27]] to float
 ; IR-DPP-NEXT:    [[TMP29:%.*]] = call float @llvm.amdgcn.strict.wwm.f32(float [[TMP28]]) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[TMP8]], 0
@@ -471,7 +471,7 @@ define amdgpu_ps float @global_atomic_fsub_uni_address_div_value_agent_scope_str
 ; IR-DPP:       33:
 ; IR-DPP-NEXT:    [[TMP34:%.*]] = phi float [ poison, [[TMP2]] ], [ [[TMP32]], [[TMP31]] ]
 ; IR-DPP-NEXT:    [[TMP35:%.*]] = bitcast float [[TMP34]] to i32
-; IR-DPP-NEXT:    [[TMP36:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP35]]) #[[ATTR8]]
+; IR-DPP-NEXT:    [[TMP36:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP35]]) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP37:%.*]] = bitcast i32 [[TMP36]] to float
 ; IR-DPP-NEXT:    [[TMP38:%.*]] = call float @llvm.amdgcn.strict.wwm.f32(float [[TMP25]]) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP39:%.*]] = call float @llvm.experimental.constrained.fsub.f32(float [[TMP37]], float [[TMP38]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR8]]
@@ -503,7 +503,7 @@ define amdgpu_ps float @global_atomic_fmin_uni_address_uni_value_agent_scope_uns
 ; IR:       12:
 ; IR-NEXT:    [[TMP13:%.*]] = phi float [ poison, [[TMP2]] ], [ [[TMP11]], [[TMP10]] ]
 ; IR-NEXT:    [[TMP14:%.*]] = bitcast float [[TMP13]] to i32
-; IR-NEXT:    [[TMP15:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP14]])
+; IR-NEXT:    [[TMP15:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP14]])
 ; IR-NEXT:    [[TMP16:%.*]] = bitcast i32 [[TMP15]] to float
 ; IR-NEXT:    [[TMP17:%.*]] = uitofp i32 [[TMP8]] to float
 ; IR-NEXT:    [[TMP18:%.*]] = select i1 [[TMP9]], float 0x7FF0000000000000, float [[VAL]]
@@ -536,7 +536,7 @@ define amdgpu_ps float @global_atomic_fmin_uni_address_div_value_agent_scope_uns
 ; IR-ITERATIVE:       12:
 ; IR-ITERATIVE-NEXT:    [[TMP13:%.*]] = phi float [ poison, [[COMPUTEEND:%.*]] ], [ [[TMP11]], [[TMP10:%.*]] ]
 ; IR-ITERATIVE-NEXT:    [[TMP14:%.*]] = bitcast float [[TMP13]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP15:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP14]])
+; IR-ITERATIVE-NEXT:    [[TMP15:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP14]])
 ; IR-ITERATIVE-NEXT:    [[TMP16:%.*]] = bitcast i32 [[TMP15]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP17:%.*]] = call float @llvm.minnum.f32(float [[TMP16]], float [[TMP28:%.*]])
 ; IR-ITERATIVE-NEXT:    br label [[TMP18]]
@@ -550,11 +550,11 @@ define amdgpu_ps float @global_atomic_fmin_uni_address_div_value_agent_scope_uns
 ; IR-ITERATIVE-NEXT:    [[TMP20:%.*]] = call i64 @llvm.cttz.i64(i64 [[ACTIVEBITS]], i1 true)
 ; IR-ITERATIVE-NEXT:    [[TMP21:%.*]] = trunc i64 [[TMP20]] to i32
 ; IR-ITERATIVE-NEXT:    [[TMP22:%.*]] = bitcast float [[VAL:%.*]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP22]], i32 [[TMP21]])
+; IR-ITERATIVE-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP22]], i32 [[TMP21]])
 ; IR-ITERATIVE-NEXT:    [[TMP24:%.*]] = bitcast i32 [[TMP23]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP25:%.*]] = bitcast float [[ACCUMULATOR]] to i32
 ; IR-ITERATIVE-NEXT:    [[TMP26:%.*]] = bitcast float [[OLDVALUEPHI]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP27:%.*]] = call i32 @llvm.amdgcn.writelane(i32 [[TMP25]], i32 [[TMP21]], i32 [[TMP26]])
+; IR-ITERATIVE-NEXT:    [[TMP27:%.*]] = call i32 @llvm.amdgcn.writelane.i32(i32 [[TMP25]], i32 [[TMP21]], i32 [[TMP26]])
 ; IR-ITERATIVE-NEXT:    [[TMP28]] = bitcast i32 [[TMP27]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP29]] = call float @llvm.minnum.f32(float [[ACCUMULATOR]], float [[TMP24]])
 ; IR-ITERATIVE-NEXT:    [[TMP30:%.*]] = shl i64 1, [[TMP20]]
@@ -594,7 +594,7 @@ define amdgpu_ps float @global_atomic_fmin_uni_address_div_value_agent_scope_uns
 ; IR-DPP-NEXT:    [[TMP24:%.*]] = call float @llvm.minnum.f32(float [[TMP22]], float [[TMP23]])
 ; IR-DPP-NEXT:    [[TMP25:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float 0x7FF0000000000000, float [[TMP24]], i32 312, i32 15, i32 15, i1 false)
 ; IR-DPP-NEXT:    [[TMP26:%.*]] = bitcast float [[TMP24]] to i32
-; IR-DPP-NEXT:    [[TMP27:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP26]], i32 63)
+; IR-DPP-NEXT:    [[TMP27:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP26]], i32 63)
 ; IR-DPP-NEXT:    [[TMP28:%.*]] = bitcast i32 [[TMP27]] to float
 ; IR-DPP-NEXT:    [[TMP29:%.*]] = call float @llvm.amdgcn.strict.wwm.f32(float [[TMP28]])
 ; IR-DPP-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[TMP8]], 0
@@ -605,7 +605,7 @@ define amdgpu_ps float @global_atomic_fmin_uni_address_div_value_agent_scope_uns
 ; IR-DPP:       33:
 ; IR-DPP-NEXT:    [[TMP34:%.*]] = phi float [ poison, [[TMP2]] ], [ [[TMP32]], [[TMP31]] ]
 ; IR-DPP-NEXT:    [[TMP35:%.*]] = bitcast float [[TMP34]] to i32
-; IR-DPP-NEXT:    [[TMP36:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP35]])
+; IR-DPP-NEXT:    [[TMP36:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP35]])
 ; IR-DPP-NEXT:    [[TMP37:%.*]] = bitcast i32 [[TMP36]] to float
 ; IR-DPP-NEXT:    [[TMP38:%.*]] = call float @llvm.amdgcn.strict.wwm.f32(float [[TMP25]])
 ; IR-DPP-NEXT:    [[TMP39:%.*]] = call float @llvm.minnum.f32(float [[TMP37]], float [[TMP38]])
@@ -637,7 +637,7 @@ define amdgpu_ps float @global_atomic_fmax_uni_address_uni_value_agent_scope_uns
 ; IR-ITERATIVE:       12:
 ; IR-ITERATIVE-NEXT:    [[TMP13:%.*]] = phi float [ poison, [[TMP2]] ], [ [[TMP11]], [[TMP10]] ]
 ; IR-ITERATIVE-NEXT:    [[TMP14:%.*]] = bitcast float [[TMP13]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP15:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP14]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP15:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP14]]) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP16:%.*]] = bitcast i32 [[TMP15]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP17:%.*]] = call float @llvm.experimental.constrained.uitofp.f32.i32(i32 [[TMP8]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP18:%.*]] = select i1 [[TMP9]], float 0xFFF0000000000000, float [[VAL]]
@@ -665,7 +665,7 @@ define amdgpu_ps float @global_atomic_fmax_uni_address_uni_value_agent_scope_uns
 ; IR-DPP:       12:
 ; IR-DPP-NEXT:    [[TMP13:%.*]] = phi float [ poison, [[TMP2]] ], [ [[TMP11]], [[TMP10]] ]
 ; IR-DPP-NEXT:    [[TMP14:%.*]] = bitcast float [[TMP13]] to i32
-; IR-DPP-NEXT:    [[TMP15:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP14]]) #[[ATTR8]]
+; IR-DPP-NEXT:    [[TMP15:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP14]]) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP16:%.*]] = bitcast i32 [[TMP15]] to float
 ; IR-DPP-NEXT:    [[TMP17:%.*]] = call float @llvm.experimental.constrained.uitofp.f32.i32(i32 [[TMP8]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP18:%.*]] = select i1 [[TMP9]], float 0xFFF0000000000000, float [[VAL]]
@@ -698,7 +698,7 @@ define amdgpu_ps float @global_atomic_fmax_uni_address_div_value_agent_scope_uns
 ; IR-ITERATIVE:       12:
 ; IR-ITERATIVE-NEXT:    [[TMP13:%.*]] = phi float [ poison, [[COMPUTEEND:%.*]] ], [ [[TMP11]], [[TMP10:%.*]] ]
 ; IR-ITERATIVE-NEXT:    [[TMP14:%.*]] = bitcast float [[TMP13]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP15:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP14]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP15:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP14]]) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP16:%.*]] = bitcast i32 [[TMP15]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP17:%.*]] = call float @llvm.experimental.constrained.maxnum.f32(float [[TMP16]], float [[TMP28:%.*]], metadata !"fpexcept.strict") #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    br label [[TMP18]]
@@ -712,11 +712,11 @@ define amdgpu_ps float @global_atomic_fmax_uni_address_div_value_agent_scope_uns
 ; IR-ITERATIVE-NEXT:    [[TMP20:%.*]] = call i64 @llvm.cttz.i64(i64 [[ACTIVEBITS]], i1 true) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP21:%.*]] = trunc i64 [[TMP20]] to i32
 ; IR-ITERATIVE-NEXT:    [[TMP22:%.*]] = bitcast float [[VAL:%.*]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP22]], i32 [[TMP21]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP22]], i32 [[TMP21]]) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP24:%.*]] = bitcast i32 [[TMP23]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP25:%.*]] = bitcast float [[ACCUMULATOR]] to i32
 ; IR-ITERATIVE-NEXT:    [[TMP26:%.*]] = bitcast float [[OLDVALUEPHI]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP27:%.*]] = call i32 @llvm.amdgcn.writelane(i32 [[TMP25]], i32 [[TMP21]], i32 [[TMP26]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP27:%.*]] = call i32 @llvm.amdgcn.writelane.i32(i32 [[TMP25]], i32 [[TMP21]], i32 [[TMP26]]) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP28]] = bitcast i32 [[TMP27]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP29]] = call float @llvm.experimental.constrained.maxnum.f32(float [[ACCUMULATOR]], float [[TMP24]], metadata !"fpexcept.strict") #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP30:%.*]] = shl i64 1, [[TMP20]]
@@ -756,7 +756,7 @@ define amdgpu_ps float @global_atomic_fmax_uni_address_div_value_agent_scope_uns
 ; IR-DPP-NEXT:    [[TMP24:%.*]] = call float @llvm.experimental.constrained.maxnum.f32(float [[TMP22]], float [[TMP23]], metadata !"fpexcept.strict") #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP25:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float 0xFFF0000000000000, float [[TMP24]], i32 312, i32 15, i32 15, i1 false) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP26:%.*]] = bitcast float [[TMP24]] to i32
-; IR-DPP-NEXT:    [[TMP27:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP26]], i32 63) #[[ATTR8]]
+; IR-DPP-NEXT:    [[TMP27:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP26]], i32 63) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP28:%.*]] = bitcast i32 [[TMP27]] to float
 ; IR-DPP-NEXT:    [[TMP29:%.*]] = call float @llvm.amdgcn.strict.wwm.f32(float [[TMP28]]) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[TMP8]], 0
@@ -767,7 +767,7 @@ define amdgpu_ps float @global_atomic_fmax_uni_address_div_value_agent_scope_uns
 ; IR-DPP:       33:
 ; IR-DPP-NEXT:    [[TMP34:%.*]] = phi float [ poison, [[TMP2]] ], [ [[TMP32]], [[TMP31]] ]
 ; IR-DPP-NEXT:    [[TMP35:%.*]] = bitcast float [[TMP34]] to i32
-; IR-DPP-NEXT:    [[TMP36:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP35]]) #[[ATTR8]]
+; IR-DPP-NEXT:    [[TMP36:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP35]]) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP37:%.*]] = bitcast i32 [[TMP36]] to float
 ; IR-DPP-NEXT:    [[TMP38:%.*]] = call float @llvm.amdgcn.strict.wwm.f32(float [[TMP25]]) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP39:%.*]] = call float @llvm.experimental.constrained.maxnum.f32(float [[TMP37]], float [[TMP38]], metadata !"fpexcept.strict") #[[ATTR8]]
@@ -803,7 +803,7 @@ define amdgpu_ps float @global_atomic_fadd_uni_address_uni_value_system_scope_st
 ; IR-ITERATIVE:       16:
 ; IR-ITERATIVE-NEXT:    [[TMP17:%.*]] = phi float [ poison, [[TMP2]] ], [ [[TMP15]], [[TMP14]] ]
 ; IR-ITERATIVE-NEXT:    [[TMP18:%.*]] = bitcast float [[TMP17]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP19:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP18]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP19:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP18]]) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP20:%.*]] = bitcast i32 [[TMP19]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP21:%.*]] = call float @llvm.experimental.constrained.uitofp.f32.i32(i32 [[TMP8]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP22:%.*]] = call float @llvm.experimental.constrained.fmul.f32(float [[VAL]], float [[TMP21]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR7]]
@@ -835,7 +835,7 @@ define amdgpu_ps float @global_atomic_fadd_uni_address_uni_value_system_scope_st
 ; IR-DPP:       16:
 ; IR-DPP-NEXT:    [[TMP17:%.*]] = phi float [ poison, [[TMP2]] ], [ [[TMP15]], [[TMP14]] ]
 ; IR-DPP-NEXT:    [[TMP18:%.*]] = bitcast float [[TMP17]] to i32
-; IR-DPP-NEXT:    [[TMP19:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP18]]) #[[ATTR8]]
+; IR-DPP-NEXT:    [[TMP19:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP18]]) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP20:%.*]] = bitcast i32 [[TMP19]] to float
 ; IR-DPP-NEXT:    [[TMP21:%.*]] = call float @llvm.experimental.constrained.uitofp.f32.i32(i32 [[TMP8]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP22:%.*]] = call float @llvm.experimental.constrained.fmul.f32(float [[VAL]], float [[TMP21]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR8]]
@@ -868,7 +868,7 @@ define amdgpu_ps float @global_atomic_fadd_uni_address_div_value_system_scope_st
 ; IR-ITERATIVE:       12:
 ; IR-ITERATIVE-NEXT:    [[TMP13:%.*]] = phi float [ poison, [[COMPUTEEND:%.*]] ], [ [[TMP11]], [[TMP10:%.*]] ]
 ; IR-ITERATIVE-NEXT:    [[TMP14:%.*]] = bitcast float [[TMP13]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP15:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP14]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP15:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP14]]) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP16:%.*]] = bitcast i32 [[TMP15]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP17:%.*]] = call float @llvm.experimental.constrained.fadd.f32(float [[TMP16]], float [[TMP28:%.*]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    br label [[TMP18]]
@@ -882,11 +882,11 @@ define amdgpu_ps float @global_atomic_fadd_uni_address_div_value_system_scope_st
 ; IR-ITERATIVE-NEXT:    [[TMP20:%.*]] = call i64 @llvm.cttz.i64(i64 [[ACTIVEBITS]], i1 true) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP21:%.*]] = trunc i64 [[TMP20]] to i32
 ; IR-ITERATIVE-NEXT:    [[TMP22:%.*]] = bitcast float [[VAL:%.*]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP22]], i32 [[TMP21]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP22]], i32 [[TMP21]]) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP24:%.*]] = bitcast i32 [[TMP23]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP25:%.*]] = bitcast float [[ACCUMULATOR]] to i32
 ; IR-ITERATIVE-NEXT:    [[TMP26:%.*]] = bitcast float [[OLDVALUEPHI]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP27:%.*]] = call i32 @llvm.amdgcn.writelane(i32 [[TMP25]], i32 [[TMP21]], i32 [[TMP26]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP27:%.*]] = call i32 @llvm.amdgcn.writelane.i32(i32 [[TMP25]], i32 [[TMP21]], i32 [[TMP26]]) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP28]] = bitcast i32 [[TMP27]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP29]] = call float @llvm.experimental.constrained.fadd.f32(float [[ACCUMULATOR]], float [[TMP24]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP30:%.*]] = shl i64 1, [[TMP20]]
@@ -926,7 +926,7 @@ define amdgpu_ps float @global_atomic_fadd_uni_address_div_value_system_scope_st
 ; IR-DPP-NEXT:    [[TMP24:%.*]] = call float @llvm.experimental.constrained.fadd.f32(float [[TMP22]], float [[TMP23]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP25:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float -0.000000e+00, float [[TMP24]], i32 312, i32 15, i32 15, i1 false) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP26:%.*]] = bitcast float [[TMP24]] to i32
-; IR-DPP-NEXT:    [[TMP27:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP26]], i32 63) #[[ATTR8]]
+; IR-DPP-NEXT:    [[TMP27:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP26]], i32 63) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP28:%.*]] = bitcast i32 [[TMP27]] to float
 ; IR-DPP-NEXT:    [[TMP29:%.*]] = call float @llvm.amdgcn.strict.wwm.f32(float [[TMP28]]) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP30:%.*]] = icmp eq i32 [[TMP8]], 0
@@ -937,7 +937,7 @@ define amdgpu_ps float @global_atomic_fadd_uni_address_div_value_system_scope_st
 ; IR-DPP:       33:
 ; IR-DPP-NEXT:    [[TMP34:%.*]] = phi float [ poison, [[TMP2]] ], [ [[TMP32]], [[TMP31]] ]
 ; IR-DPP-NEXT:    [[TMP35:%.*]] = bitcast float [[TMP34]] to i32
-; IR-DPP-NEXT:    [[TMP36:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP35]]) #[[ATTR8]]
+; IR-DPP-NEXT:    [[TMP36:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP35]]) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP37:%.*]] = bitcast i32 [[TMP36]] to float
 ; IR-DPP-NEXT:    [[TMP38:%.*]] = call float @llvm.amdgcn.strict.wwm.f32(float [[TMP25]]) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP39:%.*]] = call float @llvm.experimental.constrained.fadd.f32(float [[TMP37]], float [[TMP38]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR8]]
@@ -1084,8 +1084,8 @@ define amdgpu_ps double @global_atomic_fadd_double_uni_address_uni_value_agent_s
 ; IR-NEXT:    [[TMP19:%.*]] = trunc i64 [[TMP18]] to i32
 ; IR-NEXT:    [[TMP20:%.*]] = lshr i64 [[TMP18]], 32
 ; IR-NEXT:    [[TMP21:%.*]] = trunc i64 [[TMP20]] to i32
-; IR-NEXT:    [[TMP22:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP19]])
-; IR-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP21]])
+; IR-NEXT:    [[TMP22:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP19]])
+; IR-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP21]])
 ; IR-NEXT:    [[TMP24:%.*]] = insertelement <2 x i32> poison, i32 [[TMP22]], i32 0
 ; IR-NEXT:    [[TMP25:%.*]] = insertelement <2 x i32> [[TMP24]], i32 [[TMP23]], i32 1
 ; IR-NEXT:    [[TMP26:%.*]] = bitcast <2 x i32> [[TMP25]] to double
@@ -1136,8 +1136,8 @@ define amdgpu_ps double @global_atomic_fadd_double_uni_address_uni_value_one_as_
 ; IR-ITERATIVE-NEXT:    [[TMP19:%.*]] = trunc i64 [[TMP18]] to i32
 ; IR-ITERATIVE-NEXT:    [[TMP20:%.*]] = lshr i64 [[TMP18]], 32
 ; IR-ITERATIVE-NEXT:    [[TMP21:%.*]] = trunc i64 [[TMP20]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP22:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP19]]) #[[ATTR7]]
-; IR-ITERATIVE-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP21]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP22:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP19]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP21]]) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP24:%.*]] = insertelement <2 x i32> poison, i32 [[TMP22]], i32 0
 ; IR-ITERATIVE-NEXT:    [[TMP25:%.*]] = insertelement <2 x i32> [[TMP24]], i32 [[TMP23]], i32 1
 ; IR-ITERATIVE-NEXT:    [[TMP26:%.*]] = bitcast <2 x i32> [[TMP25]] to double
@@ -1174,8 +1174,8 @@ define amdgpu_ps double @global_atomic_fadd_double_uni_address_uni_value_one_as_
 ; IR-DPP-NEXT:    [[TMP19:%.*]] = trunc i64 [[TMP18]] to i32
 ; IR-DPP-NEXT:    [[TMP20:%.*]] = lshr i64 [[TMP18]], 32
 ; IR-DPP-NEXT:    [[TMP21:%.*]] = trunc i64 [[TMP20]] to i32
-; IR-DPP-NEXT:    [[TMP22:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP19]]) #[[ATTR8]]
-; IR-DPP-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP21]]) #[[ATTR8]]
+; IR-DPP-NEXT:    [[TMP22:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP19]]) #[[ATTR8]]
+; IR-DPP-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP21]]) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP24:%.*]] = insertelement <2 x i32> poison, i32 [[TMP22]], i32 0
 ; IR-DPP-NEXT:    [[TMP25:%.*]] = insertelement <2 x i32> [[TMP24]], i32 [[TMP23]], i32 1
 ; IR-DPP-NEXT:    [[TMP26:%.*]] = bitcast <2 x i32> [[TMP25]] to double
@@ -1226,8 +1226,8 @@ define amdgpu_ps double @global_atomic_fsub_double_uni_address_uni_value_agent_s
 ; IR-ITERATIVE-NEXT:    [[TMP19:%.*]] = trunc i64 [[TMP18]] to i32
 ; IR-ITERATIVE-NEXT:    [[TMP20:%.*]] = lshr i64 [[TMP18]], 32
 ; IR-ITERATIVE-NEXT:    [[TMP21:%.*]] = trunc i64 [[TMP20]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP22:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP19]]) #[[ATTR7]]
-; IR-ITERATIVE-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP21]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP22:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP19]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP21]]) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP24:%.*]] = insertelement <2 x i32> poison, i32 [[TMP22]], i32 0
 ; IR-ITERATIVE-NEXT:    [[TMP25:%.*]] = insertelement <2 x i32> [[TMP24]], i32 [[TMP23]], i32 1
 ; IR-ITERATIVE-NEXT:    [[TMP26:%.*]] = bitcast <2 x i32> [[TMP25]] to double
@@ -1264,8 +1264,8 @@ define amdgpu_ps double @global_atomic_fsub_double_uni_address_uni_value_agent_s
 ; IR-DPP-NEXT:    [[TMP19:%.*]] = trunc i64 [[TMP18]] to i32
 ; IR-DPP-NEXT:    [[TMP20:%.*]] = lshr i64 [[TMP18]], 32
 ; IR-DPP-NEXT:    [[TMP21:%.*]] = trunc i64 [[TMP20]] to i32
-; IR-DPP-NEXT:    [[TMP22:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP19]]) #[[ATTR8]]
-; IR-DPP-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP21]]) #[[ATTR8]]
+; IR-DPP-NEXT:    [[TMP22:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP19]]) #[[ATTR8]]
+; IR-DPP-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP21]]) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP24:%.*]] = insertelement <2 x i32> poison, i32 [[TMP22]], i32 0
 ; IR-DPP-NEXT:    [[TMP25:%.*]] = insertelement <2 x i32> [[TMP24]], i32 [[TMP23]], i32 1
 ; IR-DPP-NEXT:    [[TMP26:%.*]] = bitcast <2 x i32> [[TMP25]] to double
@@ -1312,8 +1312,8 @@ define amdgpu_ps double @global_atomic_fmin_double_uni_address_uni_value_agent_s
 ; IR-NEXT:    [[TMP15:%.*]] = trunc i64 [[TMP14]] to i32
 ; IR-NEXT:    [[TMP16:%.*]] = lshr i64 [[TMP14]], 32
 ; IR-NEXT:    [[TMP17:%.*]] = trunc i64 [[TMP16]] to i32
-; IR-NEXT:    [[TMP18:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP15]])
-; IR-NEXT:    [[TMP19:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP17]])
+; IR-NEXT:    [[TMP18:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP15]])
+; IR-NEXT:    [[TMP19:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP17]])
 ; IR-NEXT:    [[TMP20:%.*]] = insertelement <2 x i32> poison, i32 [[TMP18]], i32 0
 ; IR-NEXT:    [[TMP21:%.*]] = insertelement <2 x i32> [[TMP20]], i32 [[TMP19]], i32 1
 ; IR-NEXT:    [[TMP22:%.*]] = bitcast <2 x i32> [[TMP21]] to double
@@ -1360,8 +1360,8 @@ define amdgpu_ps double @global_atomic__fmax_double_uni_address_uni_value_agent_
 ; IR-ITERATIVE-NEXT:    [[TMP15:%.*]] = trunc i64 [[TMP14]] to i32
 ; IR-ITERATIVE-NEXT:    [[TMP16:%.*]] = lshr i64 [[TMP14]], 32
 ; IR-ITERATIVE-NEXT:    [[TMP17:%.*]] = trunc i64 [[TMP16]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP18:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP15]]) #[[ATTR7]]
-; IR-ITERATIVE-NEXT:    [[TMP19:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP17]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP18:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP15]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP19:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP17]]) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP20:%.*]] = insertelement <2 x i32> poison, i32 [[TMP18]], i32 0
 ; IR-ITERATIVE-NEXT:    [[TMP21:%.*]] = insertelement <2 x i32> [[TMP20]], i32 [[TMP19]], i32 1
 ; IR-ITERATIVE-NEXT:    [[TMP22:%.*]] = bitcast <2 x i32> [[TMP21]] to double
@@ -1394,8 +1394,8 @@ define amdgpu_ps double @global_atomic__fmax_double_uni_address_uni_value_agent_
 ; IR-DPP-NEXT:    [[TMP15:%.*]] = trunc i64 [[TMP14]] to i32
 ; IR-DPP-NEXT:    [[TMP16:%.*]] = lshr i64 [[TMP14]], 32
 ; IR-DPP-NEXT:    [[TMP17:%.*]] = trunc i64 [[TMP16]] to i32
-; IR-DPP-NEXT:    [[TMP18:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP15]]) #[[ATTR8]]
-; IR-DPP-NEXT:    [[TMP19:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP17]]) #[[ATTR8]]
+; IR-DPP-NEXT:    [[TMP18:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP15]]) #[[ATTR8]]
+; IR-DPP-NEXT:    [[TMP19:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP17]]) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP20:%.*]] = insertelement <2 x i32> poison, i32 [[TMP18]], i32 0
 ; IR-DPP-NEXT:    [[TMP21:%.*]] = insertelement <2 x i32> [[TMP20]], i32 [[TMP19]], i32 1
 ; IR-DPP-NEXT:    [[TMP22:%.*]] = bitcast <2 x i32> [[TMP21]] to double
@@ -1446,8 +1446,8 @@ define amdgpu_ps double @global_atomic_fadd_double_uni_address_uni_value_system_
 ; IR-ITERATIVE-NEXT:    [[TMP19:%.*]] = trunc i64 [[TMP18]] to i32
 ; IR-ITERATIVE-NEXT:    [[TMP20:%.*]] = lshr i64 [[TMP18]], 32
 ; IR-ITERATIVE-NEXT:    [[TMP21:%.*]] = trunc i64 [[TMP20]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP22:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP19]]) #[[ATTR7]]
-; IR-ITERATIVE-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP21]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP22:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP19]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP21]]) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP24:%.*]] = insertelement <2 x i32> poison, i32 [[TMP22]], i32 0
 ; IR-ITERATIVE-NEXT:    [[TMP25:%.*]] = insertelement <2 x i32> [[TMP24]], i32 [[TMP23]], i32 1
 ; IR-ITERATIVE-NEXT:    [[TMP26:%.*]] = bitcast <2 x i32> [[TMP25]] to double
@@ -1484,8 +1484,8 @@ define amdgpu_ps double @global_atomic_fadd_double_uni_address_uni_value_system_
 ; IR-DPP-NEXT:    [[TMP19:%.*]] = trunc i64 [[TMP18]] to i32
 ; IR-DPP-NEXT:    [[TMP20:%.*]] = lshr i64 [[TMP18]], 32
 ; IR-DPP-NEXT:    [[TMP21:%.*]] = trunc i64 [[TMP20]] to i32
-; IR-DPP-NEXT:    [[TMP22:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP19]]) #[[ATTR8]]
-; IR-DPP-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[TMP21]]) #[[ATTR8]]
+; IR-DPP-NEXT:    [[TMP22:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP19]]) #[[ATTR8]]
+; IR-DPP-NEXT:    [[TMP23:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[TMP21]]) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP24:%.*]] = insertelement <2 x i32> poison, i32 [[TMP22]], i32 0
 ; IR-DPP-NEXT:    [[TMP25:%.*]] = insertelement <2 x i32> [[TMP24]], i32 [[TMP23]], i32 1
 ; IR-DPP-NEXT:    [[TMP26:%.*]] = bitcast <2 x i32> [[TMP25]] to double
diff --git a/llvm/test/CodeGen/AMDGPU/global_atomics_iterative_scan.ll b/llvm/test/CodeGen/AMDGPU/global_atomics_iterative_scan.ll
index f954560d0f5ca..4b4c99b3cd14c 100644
--- a/llvm/test/CodeGen/AMDGPU/global_atomics_iterative_scan.ll
+++ b/llvm/test/CodeGen/AMDGPU/global_atomics_iterative_scan.ll
@@ -83,7 +83,7 @@ define amdgpu_kernel void @divergent_value(ptr addrspace(1) %out, ptr addrspace(
 ; IR-NEXT:    [[ACTIVEBITS:%.*]] = phi i64 [ [[TMP6]], [[ENTRY]] ], [ [[TMP16:%.*]], [[COMPUTELOOP]] ]
 ; IR-NEXT:    [[TMP10:%.*]] = call i64 @llvm.cttz.i64(i64 [[ACTIVEBITS]], i1 true)
 ; IR-NEXT:    [[TMP11:%.*]] = trunc i64 [[TMP10]] to i32
-; IR-NEXT:    [[TMP12:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[VALUE]], i32 [[TMP11]])
+; IR-NEXT:    [[TMP12:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[VALUE]], i32 [[TMP11]])
 ; IR-NEXT:    [[TMP13]] = add i32 [[ACCUMULATOR]], [[TMP12]]
 ; IR-NEXT:    [[TMP14:%.*]] = shl i64 1, [[TMP10]]
 ; IR-NEXT:    [[TMP15:%.*]] = xor i64 [[TMP14]], -1
diff --git a/llvm/test/CodeGen/AMDGPU/global_atomics_iterative_scan_fp.ll b/llvm/test/CodeGen/AMDGPU/global_atomics_iterative_scan_fp.ll
index 86e3d9338e078..38823681d1bb5 100644
--- a/llvm/test/CodeGen/AMDGPU/global_atomics_iterative_scan_fp.ll
+++ b/llvm/test/CodeGen/AMDGPU/global_atomics_iterative_scan_fp.ll
@@ -69,7 +69,7 @@ define amdgpu_kernel void @global_atomic_fadd_div_value(ptr addrspace(1) %ptr) #
 ; IR-ITERATIVE-NEXT:    [[TMP11:%.*]] = call i64 @llvm.cttz.i64(i64 [[ACTIVEBITS]], i1 true)
 ; IR-ITERATIVE-NEXT:    [[TMP12:%.*]] = trunc i64 [[TMP11]] to i32
 ; IR-ITERATIVE-NEXT:    [[TMP13:%.*]] = bitcast float [[DIVVALUE]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP14:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP13]], i32 [[TMP12]])
+; IR-ITERATIVE-NEXT:    [[TMP14:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP13]], i32 [[TMP12]])
 ; IR-ITERATIVE-NEXT:    [[TMP15:%.*]] = bitcast i32 [[TMP14]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP16]] = fadd float [[ACCUMULATOR]], [[TMP15]]
 ; IR-ITERATIVE-NEXT:    [[TMP17:%.*]] = shl i64 1, [[TMP11]]
@@ -107,7 +107,7 @@ define amdgpu_kernel void @global_atomic_fadd_div_value(ptr addrspace(1) %ptr) #
 ; IR-DPP-NEXT:    [[TMP21:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float -0.000000e+00, float [[TMP20]], i32 323, i32 12, i32 15, i1 false)
 ; IR-DPP-NEXT:    [[TMP22:%.*]] = fadd float [[TMP20]], [[TMP21]]
 ; IR-DPP-NEXT:    [[TMP23:%.*]] = bitcast float [[TMP22]] to i32
-; IR-DPP-NEXT:    [[TMP24:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP23]], i32 63)
+; IR-DPP-NEXT:    [[TMP24:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP23]], i32 63)
 ; IR-DPP-NEXT:    [[TMP25:%.*]] = bitcast i32 [[TMP24]] to float
 ; IR-DPP-NEXT:    [[TMP26:%.*]] = call float @llvm.amdgcn.strict.wwm.f32(float [[TMP25]])
 ; IR-DPP-NEXT:    [[TMP27:%.*]] = icmp eq i32 [[TMP6]], 0
@@ -191,7 +191,7 @@ define amdgpu_kernel void @global_atomic_fsub_div_value(ptr addrspace(1) %ptr) #
 ; IR-ITERATIVE-NEXT:    [[TMP11:%.*]] = call i64 @llvm.cttz.i64(i64 [[ACTIVEBITS]], i1 true)
 ; IR-ITERATIVE-NEXT:    [[TMP12:%.*]] = trunc i64 [[TMP11]] to i32
 ; IR-ITERATIVE-NEXT:    [[TMP13:%.*]] = bitcast float [[DIVVALUE]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP14:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP13]], i32 [[TMP12]])
+; IR-ITERATIVE-NEXT:    [[TMP14:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP13]], i32 [[TMP12]])
 ; IR-ITERATIVE-NEXT:    [[TMP15:%.*]] = bitcast i32 [[TMP14]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP16]] = fadd float [[ACCUMULATOR]], [[TMP15]]
 ; IR-ITERATIVE-NEXT:    [[TMP17:%.*]] = shl i64 1, [[TMP11]]
@@ -229,7 +229,7 @@ define amdgpu_kernel void @global_atomic_fsub_div_value(ptr addrspace(1) %ptr) #
 ; IR-DPP-NEXT:    [[TMP21:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float -0.000000e+00, float [[TMP20]], i32 323, i32 12, i32 15, i1 false)
 ; IR-DPP-NEXT:    [[TMP22:%.*]] = fadd float [[TMP20]], [[TMP21]]
 ; IR-DPP-NEXT:    [[TMP23:%.*]] = bitcast float [[TMP22]] to i32
-; IR-DPP-NEXT:    [[TMP24:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP23]], i32 63)
+; IR-DPP-NEXT:    [[TMP24:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP23]], i32 63)
 ; IR-DPP-NEXT:    [[TMP25:%.*]] = bitcast i32 [[TMP24]] to float
 ; IR-DPP-NEXT:    [[TMP26:%.*]] = call float @llvm.amdgcn.strict.wwm.f32(float [[TMP25]])
 ; IR-DPP-NEXT:    [[TMP27:%.*]] = icmp eq i32 [[TMP6]], 0
diff --git a/llvm/test/CodeGen/AMDGPU/global_atomics_optimizer_fp_no_rtn.ll b/llvm/test/CodeGen/AMDGPU/global_atomics_optimizer_fp_no_rtn.ll
index b9234f47df192..83453354320fe 100644
--- a/llvm/test/CodeGen/AMDGPU/global_atomics_optimizer_fp_no_rtn.ll
+++ b/llvm/test/CodeGen/AMDGPU/global_atomics_optimizer_fp_no_rtn.ll
@@ -61,7 +61,7 @@ define amdgpu_ps void @global_atomic_fadd_uni_address_div_value_scope_agent_scop
 ; IR-ITERATIVE-NEXT:    [[TMP14:%.*]] = call i64 @llvm.cttz.i64(i64 [[ACTIVEBITS]], i1 true)
 ; IR-ITERATIVE-NEXT:    [[TMP15:%.*]] = trunc i64 [[TMP14]] to i32
 ; IR-ITERATIVE-NEXT:    [[TMP16:%.*]] = bitcast float [[VAL:%.*]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP17:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP16]], i32 [[TMP15]])
+; IR-ITERATIVE-NEXT:    [[TMP17:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP16]], i32 [[TMP15]])
 ; IR-ITERATIVE-NEXT:    [[TMP18:%.*]] = bitcast i32 [[TMP17]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP19]] = fadd float [[ACCUMULATOR]], [[TMP18]]
 ; IR-ITERATIVE-NEXT:    [[TMP20:%.*]] = shl i64 1, [[TMP14]]
@@ -100,7 +100,7 @@ define amdgpu_ps void @global_atomic_fadd_uni_address_div_value_scope_agent_scop
 ; IR-DPP-NEXT:    [[TMP23:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float -0.000000e+00, float [[TMP22]], i32 323, i32 12, i32 15, i1 false)
 ; IR-DPP-NEXT:    [[TMP24:%.*]] = fadd float [[TMP22]], [[TMP23]]
 ; IR-DPP-NEXT:    [[TMP25:%.*]] = bitcast float [[TMP24]] to i32
-; IR-DPP-NEXT:    [[TMP26:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP25]], i32 63)
+; IR-DPP-NEXT:    [[TMP26:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP25]], i32 63)
 ; IR-DPP-NEXT:    [[TMP27:%.*]] = bitcast i32 [[TMP26]] to float
 ; IR-DPP-NEXT:    [[TMP28:%.*]] = call float @llvm.amdgcn.strict.wwm.f32(float [[TMP27]])
 ; IR-DPP-NEXT:    [[TMP29:%.*]] = icmp eq i32 [[TMP8]], 0
@@ -196,7 +196,7 @@ define amdgpu_ps void @global_atomic_fadd_uni_address_div_value_one_as_scope_uns
 ; IR-ITERATIVE-NEXT:    [[TMP14:%.*]] = call i64 @llvm.cttz.i64(i64 [[ACTIVEBITS]], i1 true) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP15:%.*]] = trunc i64 [[TMP14]] to i32
 ; IR-ITERATIVE-NEXT:    [[TMP16:%.*]] = bitcast float [[VAL:%.*]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP17:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP16]], i32 [[TMP15]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP17:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP16]], i32 [[TMP15]]) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP18:%.*]] = bitcast i32 [[TMP17]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP19]] = call float @llvm.experimental.constrained.fadd.f32(float [[ACCUMULATOR]], float [[TMP18]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP20:%.*]] = shl i64 1, [[TMP14]]
@@ -235,7 +235,7 @@ define amdgpu_ps void @global_atomic_fadd_uni_address_div_value_one_as_scope_uns
 ; IR-DPP-NEXT:    [[TMP23:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float -0.000000e+00, float [[TMP22]], i32 323, i32 12, i32 15, i1 false) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP24:%.*]] = call float @llvm.experimental.constrained.fadd.f32(float [[TMP22]], float [[TMP23]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP25:%.*]] = bitcast float [[TMP24]] to i32
-; IR-DPP-NEXT:    [[TMP26:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP25]], i32 63) #[[ATTR8]]
+; IR-DPP-NEXT:    [[TMP26:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP25]], i32 63) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP27:%.*]] = bitcast i32 [[TMP26]] to float
 ; IR-DPP-NEXT:    [[TMP28:%.*]] = call float @llvm.amdgcn.strict.wwm.f32(float [[TMP27]]) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP29:%.*]] = icmp eq i32 [[TMP8]], 0
@@ -331,7 +331,7 @@ define amdgpu_ps void @global_atomic_fsub_uni_address_div_value_agent_scope_stri
 ; IR-ITERATIVE-NEXT:    [[TMP14:%.*]] = call i64 @llvm.cttz.i64(i64 [[ACTIVEBITS]], i1 true) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP15:%.*]] = trunc i64 [[TMP14]] to i32
 ; IR-ITERATIVE-NEXT:    [[TMP16:%.*]] = bitcast float [[VAL:%.*]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP17:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP16]], i32 [[TMP15]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP17:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP16]], i32 [[TMP15]]) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP18:%.*]] = bitcast i32 [[TMP17]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP19]] = call float @llvm.experimental.constrained.fadd.f32(float [[ACCUMULATOR]], float [[TMP18]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP20:%.*]] = shl i64 1, [[TMP14]]
@@ -370,7 +370,7 @@ define amdgpu_ps void @global_atomic_fsub_uni_address_div_value_agent_scope_stri
 ; IR-DPP-NEXT:    [[TMP23:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float -0.000000e+00, float [[TMP22]], i32 323, i32 12, i32 15, i1 false) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP24:%.*]] = call float @llvm.experimental.constrained.fadd.f32(float [[TMP22]], float [[TMP23]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP25:%.*]] = bitcast float [[TMP24]] to i32
-; IR-DPP-NEXT:    [[TMP26:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP25]], i32 63) #[[ATTR8]]
+; IR-DPP-NEXT:    [[TMP26:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP25]], i32 63) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP27:%.*]] = bitcast i32 [[TMP26]] to float
 ; IR-DPP-NEXT:    [[TMP28:%.*]] = call float @llvm.amdgcn.strict.wwm.f32(float [[TMP27]]) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP29:%.*]] = icmp eq i32 [[TMP8]], 0
@@ -438,7 +438,7 @@ define amdgpu_ps void @global_atomic_fmin_uni_address_div_value_agent_scope_unsa
 ; IR-ITERATIVE-NEXT:    [[TMP14:%.*]] = call i64 @llvm.cttz.i64(i64 [[ACTIVEBITS]], i1 true)
 ; IR-ITERATIVE-NEXT:    [[TMP15:%.*]] = trunc i64 [[TMP14]] to i32
 ; IR-ITERATIVE-NEXT:    [[TMP16:%.*]] = bitcast float [[VAL:%.*]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP17:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP16]], i32 [[TMP15]])
+; IR-ITERATIVE-NEXT:    [[TMP17:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP16]], i32 [[TMP15]])
 ; IR-ITERATIVE-NEXT:    [[TMP18:%.*]] = bitcast i32 [[TMP17]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP19]] = call float @llvm.minnum.f32(float [[ACCUMULATOR]], float [[TMP18]])
 ; IR-ITERATIVE-NEXT:    [[TMP20:%.*]] = shl i64 1, [[TMP14]]
@@ -477,7 +477,7 @@ define amdgpu_ps void @global_atomic_fmin_uni_address_div_value_agent_scope_unsa
 ; IR-DPP-NEXT:    [[TMP23:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float 0x7FF0000000000000, float [[TMP22]], i32 323, i32 12, i32 15, i1 false)
 ; IR-DPP-NEXT:    [[TMP24:%.*]] = call float @llvm.minnum.f32(float [[TMP22]], float [[TMP23]])
 ; IR-DPP-NEXT:    [[TMP25:%.*]] = bitcast float [[TMP24]] to i32
-; IR-DPP-NEXT:    [[TMP26:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP25]], i32 63)
+; IR-DPP-NEXT:    [[TMP26:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP25]], i32 63)
 ; IR-DPP-NEXT:    [[TMP27:%.*]] = bitcast i32 [[TMP26]] to float
 ; IR-DPP-NEXT:    [[TMP28:%.*]] = call float @llvm.amdgcn.strict.wwm.f32(float [[TMP27]])
 ; IR-DPP-NEXT:    [[TMP29:%.*]] = icmp eq i32 [[TMP8]], 0
@@ -565,7 +565,7 @@ define amdgpu_ps void @global_atomic_fmax_uni_address_div_value_agent_scope_unsa
 ; IR-ITERATIVE-NEXT:    [[TMP14:%.*]] = call i64 @llvm.cttz.i64(i64 [[ACTIVEBITS]], i1 true) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP15:%.*]] = trunc i64 [[TMP14]] to i32
 ; IR-ITERATIVE-NEXT:    [[TMP16:%.*]] = bitcast float [[VAL:%.*]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP17:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP16]], i32 [[TMP15]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP17:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP16]], i32 [[TMP15]]) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP18:%.*]] = bitcast i32 [[TMP17]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP19]] = call float @llvm.experimental.constrained.maxnum.f32(float [[ACCUMULATOR]], float [[TMP18]], metadata !"fpexcept.strict") #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP20:%.*]] = shl i64 1, [[TMP14]]
@@ -604,7 +604,7 @@ define amdgpu_ps void @global_atomic_fmax_uni_address_div_value_agent_scope_unsa
 ; IR-DPP-NEXT:    [[TMP23:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float 0xFFF0000000000000, float [[TMP22]], i32 323, i32 12, i32 15, i1 false) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP24:%.*]] = call float @llvm.experimental.constrained.maxnum.f32(float [[TMP22]], float [[TMP23]], metadata !"fpexcept.strict") #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP25:%.*]] = bitcast float [[TMP24]] to i32
-; IR-DPP-NEXT:    [[TMP26:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP25]], i32 63) #[[ATTR8]]
+; IR-DPP-NEXT:    [[TMP26:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP25]], i32 63) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP27:%.*]] = bitcast i32 [[TMP26]] to float
 ; IR-DPP-NEXT:    [[TMP28:%.*]] = call float @llvm.amdgcn.strict.wwm.f32(float [[TMP27]]) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP29:%.*]] = icmp eq i32 [[TMP8]], 0
@@ -700,7 +700,7 @@ define amdgpu_ps void @global_atomic_fadd_uni_address_div_value_system_scope_str
 ; IR-ITERATIVE-NEXT:    [[TMP14:%.*]] = call i64 @llvm.cttz.i64(i64 [[ACTIVEBITS]], i1 true) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP15:%.*]] = trunc i64 [[TMP14]] to i32
 ; IR-ITERATIVE-NEXT:    [[TMP16:%.*]] = bitcast float [[VAL:%.*]] to i32
-; IR-ITERATIVE-NEXT:    [[TMP17:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP16]], i32 [[TMP15]]) #[[ATTR7]]
+; IR-ITERATIVE-NEXT:    [[TMP17:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP16]], i32 [[TMP15]]) #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP18:%.*]] = bitcast i32 [[TMP17]] to float
 ; IR-ITERATIVE-NEXT:    [[TMP19]] = call float @llvm.experimental.constrained.fadd.f32(float [[ACCUMULATOR]], float [[TMP18]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR7]]
 ; IR-ITERATIVE-NEXT:    [[TMP20:%.*]] = shl i64 1, [[TMP14]]
@@ -739,7 +739,7 @@ define amdgpu_ps void @global_atomic_fadd_uni_address_div_value_system_scope_str
 ; IR-DPP-NEXT:    [[TMP23:%.*]] = call float @llvm.amdgcn.update.dpp.f32(float -0.000000e+00, float [[TMP22]], i32 323, i32 12, i32 15, i1 false) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP24:%.*]] = call float @llvm.experimental.constrained.fadd.f32(float [[TMP22]], float [[TMP23]], metadata !"round.dynamic", metadata !"fpexcept.strict") #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP25:%.*]] = bitcast float [[TMP24]] to i32
-; IR-DPP-NEXT:    [[TMP26:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[TMP25]], i32 63) #[[ATTR8]]
+; IR-DPP-NEXT:    [[TMP26:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TMP25]], i32 63) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP27:%.*]] = bitcast i32 [[TMP26]] to float
 ; IR-DPP-NEXT:    [[TMP28:%.*]] = call float @llvm.amdgcn.strict.wwm.f32(float [[TMP27]]) #[[ATTR8]]
 ; IR-DPP-NEXT:    [[TMP29:%.*]] = icmp eq i32 [[TMP8]], 0
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll
index 0284f44f5f14d..732489f22c36f 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll
@@ -1,65 +1,387 @@
-; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=fiji -verify-machineinstrs < %s | FileCheck -enable-var-scope %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=fiji -verify-machineinstrs < %s | FileCheck -check-prefix=CHECK-SDAG -enable-var-scope %s
+; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=fiji -verify-machineinstrs -global-isel < %s | FileCheck -check-prefix=CHECK-GISEL -enable-var-scope %s
 
 declare i32 @llvm.amdgcn.readfirstlane(i32) #0
+declare i64 @llvm.amdgcn.readfirstlane.i64(i64) #0
+declare double @llvm.amdgcn.readfirstlane.f64(double) #0
 
-; CHECK-LABEL: {{^}}test_readfirstlane:
-; CHECK: v_readfirstlane_b32 s{{[0-9]+}}, v2
-define void @test_readfirstlane(ptr addrspace(1) %out, i32 %src) #1 {
-  %readfirstlane = call i32 @llvm.amdgcn.readfirstlane(i32 %src)
+define void @test_readfirstlane_i32(ptr addrspace(1) %out, i32 %src) #1 {
+; CHECK-SDAG-LABEL: test_readfirstlane_i32:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v2, s4
+; CHECK-SDAG-NEXT:    flat_store_dword v[0:1], v2
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; CHECK-GISEL-LABEL: test_readfirstlane_i32:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v2, s4
+; CHECK-GISEL-NEXT:    flat_store_dword v[0:1], v2
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %readfirstlane = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %src)
   store i32 %readfirstlane, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; CHECK-LABEL: {{^}}test_readfirstlane_imm:
-; CHECK: s_mov_b32 [[SGPR_VAL:s[0-9]]], 32
-; CHECK-NOT: [[SGPR_VAL]]
-; CHECK: ; use [[SGPR_VAL]]
-define amdgpu_kernel void @test_readfirstlane_imm(ptr addrspace(1) %out) #1 {
-  %readfirstlane = call i32 @llvm.amdgcn.readfirstlane(i32 32)
+define void @test_readfirstlane_i64(ptr addrspace(1) %out, i64 %src) #1 {
+; CHECK-SDAG-LABEL: test_readfirstlane_i64:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v3
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s5, v2
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v2, s5
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v3, s4
+; CHECK-SDAG-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; CHECK-GISEL-LABEL: test_readfirstlane_i64:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v2, s4
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v3, s5
+; CHECK-GISEL-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %readfirstlane = call i64 @llvm.amdgcn.readfirstlane.i64(i64 %src)
+  store i64 %readfirstlane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define void @test_readfirstlane_f64(ptr addrspace(1) %out, double %src) #1 {
+; CHECK-SDAG-LABEL: test_readfirstlane_f64:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v3
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s5, v2
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v2, s5
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v3, s4
+; CHECK-SDAG-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; CHECK-GISEL-LABEL: test_readfirstlane_f64:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v2, s4
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v3, s5
+; CHECK-GISEL-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %readfirstlane = call double @llvm.amdgcn.readfirstlane.f64(double %src)
+  store double %readfirstlane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define amdgpu_kernel void @test_readfirstlane_imm_i32(ptr addrspace(1) %out) #1 {
+; CHECK-SDAG-LABEL: test_readfirstlane_imm_i32:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_mov_b32 s0, 32
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s0
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readfirstlane_imm_i32:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_mov_b32 s0, 32
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s0
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_endpgm
+  %readfirstlane = call i32 @llvm.amdgcn.readfirstlane.i32(i32 32)
   call void asm sideeffect "; use $0", "s"(i32 %readfirstlane)
   ret void
 }
 
-; CHECK-LABEL: {{^}}test_readfirstlane_imm_fold:
-; CHECK: v_mov_b32_e32 [[VVAL:v[0-9]]], 32
-; CHECK-NOT: [[VVAL]]
-; CHECK: flat_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[VVAL]]
-define amdgpu_kernel void @test_readfirstlane_imm_fold(ptr addrspace(1) %out) #1 {
-  %readfirstlane = call i32 @llvm.amdgcn.readfirstlane(i32 32)
+define amdgpu_kernel void @test_readfirstlane_imm_i64(ptr addrspace(1) %out) #1 {
+; CHECK-SDAG-LABEL: test_readfirstlane_imm_i64:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_mov_b64 s[0:1], 32
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[0:1]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readfirstlane_imm_i64:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_mov_b64 s[0:1], 32
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s[0:1]
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_endpgm
+  %readfirstlane = call i64 @llvm.amdgcn.readfirstlane.i64(i64 32)
+  call void asm sideeffect "; use $0", "s"(i64 %readfirstlane)
+  ret void
+}
+
+define amdgpu_kernel void @test_readfirstlane_imm_f64(ptr addrspace(1) %out) #1 {
+; CHECK-SDAG-LABEL: test_readfirstlane_imm_f64:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_mov_b32 s0, 0
+; CHECK-SDAG-NEXT:    s_mov_b32 s1, 0x40400000
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[0:1]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readfirstlane_imm_f64:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_mov_b32 s0, 0
+; CHECK-GISEL-NEXT:    s_mov_b32 s1, 0x40400000
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s[0:1]
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_endpgm
+  %readfirstlane = call double @llvm.amdgcn.readfirstlane.f64(double 32.0)
+  call void asm sideeffect "; use $0", "s"(double %readfirstlane)
+  ret void
+}
+
+define amdgpu_kernel void @test_readfirstlane_imm_fold_i32(ptr addrspace(1) %out) #1 {
+; CHECK-SDAG-LABEL: test_readfirstlane_imm_fold_i32:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v2, 32
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v1, s1
+; CHECK-SDAG-NEXT:    flat_store_dword v[0:1], v2
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readfirstlane_imm_fold_i32:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v2, 32
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; CHECK-GISEL-NEXT:    flat_store_dword v[0:1], v2
+; CHECK-GISEL-NEXT:    s_endpgm
+  %readfirstlane = call i32 @llvm.amdgcn.readfirstlane.i32(i32 32)
   store i32 %readfirstlane, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; CHECK-LABEL: {{^}}test_readfirstlane_m0:
-; CHECK: s_mov_b32 m0, -1
-; CHECK: v_mov_b32_e32 [[VVAL:v[0-9]]], m0
-; CHECK: flat_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[VVAL]]
+define amdgpu_kernel void @test_readfirstlane_imm_fold_i64(ptr addrspace(1) %out) #1 {
+; CHECK-SDAG-LABEL: test_readfirstlane_imm_fold_i64:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v0, 32
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v1, 0
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-SDAG-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readfirstlane_imm_fold_i64:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-GISEL-NEXT:    s_mov_b64 s[2:3], 32
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-GISEL-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CHECK-GISEL-NEXT:    s_endpgm
+  %readfirstlane = call i64 @llvm.amdgcn.readfirstlane.i64(i64 32)
+  store i64 %readfirstlane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define amdgpu_kernel void @test_readfirstlane_imm_fold_f64(ptr addrspace(1) %out) #1 {
+; CHECK-SDAG-LABEL: test_readfirstlane_imm_fold_f64:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v1, 0x40400000
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-SDAG-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readfirstlane_imm_fold_f64:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-GISEL-NEXT:    s_mov_b32 s2, 0
+; CHECK-GISEL-NEXT:    s_mov_b32 s3, 0x40400000
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-GISEL-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CHECK-GISEL-NEXT:    s_endpgm
+  %readfirstlane = call double @llvm.amdgcn.readfirstlane.f64(double 32.0)
+  store double %readfirstlane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
 define amdgpu_kernel void @test_readfirstlane_m0(ptr addrspace(1) %out) #1 {
+; CHECK-SDAG-LABEL: test_readfirstlane_m0:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    s_mov_b32 m0, -1
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v2, m0
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v1, s1
+; CHECK-SDAG-NEXT:    flat_store_dword v[0:1], v2
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readfirstlane_m0:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    s_mov_b32 m0, -1
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v2, m0
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; CHECK-GISEL-NEXT:    flat_store_dword v[0:1], v2
+; CHECK-GISEL-NEXT:    s_endpgm
   %m0 = call i32 asm "s_mov_b32 m0, -1", "={m0}"()
   %readfirstlane = call i32 @llvm.amdgcn.readfirstlane(i32 %m0)
   store i32 %readfirstlane, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; CHECK-LABEL: {{^}}test_readfirstlane_copy_from_sgpr:
-; CHECK: ;;#ASMSTART
-; CHECK-NEXT: s_mov_b32 [[SGPR:s[0-9]+]]
-; CHECK: ;;#ASMEND
-; CHECK-NOT: [[SGPR]]
-; CHECK-NOT: readfirstlane
-; CHECK: v_mov_b32_e32 [[VCOPY:v[0-9]+]], [[SGPR]]
-; CHECK: flat_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[VCOPY]]
-define amdgpu_kernel void @test_readfirstlane_copy_from_sgpr(ptr addrspace(1) %out) #1 {
+define amdgpu_kernel void @test_readfirstlane_copy_from_sgpr_i32(ptr addrspace(1) %out) #1 {
+; CHECK-SDAG-LABEL: test_readfirstlane_copy_from_sgpr_i32:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    s_mov_b32 s2, 0
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v2, s2
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v1, s1
+; CHECK-SDAG-NEXT:    flat_store_dword v[0:1], v2
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readfirstlane_copy_from_sgpr_i32:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    s_mov_b32 s2, 0
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v2, s2
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; CHECK-GISEL-NEXT:    flat_store_dword v[0:1], v2
+; CHECK-GISEL-NEXT:    s_endpgm
   %sgpr = call i32 asm "s_mov_b32 $0, 0", "=s"()
-  %readfirstlane = call i32 @llvm.amdgcn.readfirstlane(i32 %sgpr)
+  %readfirstlane = call i32 @llvm.amdgcn.readfirstlane.i32(i32 %sgpr)
   store i32 %readfirstlane, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; Make sure this doesn't crash.
-; CHECK-LABEL: {{^}}test_readfirstlane_fi:
-; CHECK: s_mov_b32 [[FIVAL:s[0-9]]], 0
+define amdgpu_kernel void @test_readfirstlane_copy_from_sgpr_i64(ptr addrspace(1) %out) #1 {
+; CHECK-SDAG-LABEL: test_readfirstlane_copy_from_sgpr_i64:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    s_mov_b64 s[2:3], 0
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-SDAG-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readfirstlane_copy_from_sgpr_i64:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    s_mov_b64 s[2:3], 0
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-GISEL-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CHECK-GISEL-NEXT:    s_endpgm
+  %sgpr = call i64 asm "s_mov_b64 $0, 0", "=s"()
+  %readfirstlane = call i64 @llvm.amdgcn.readfirstlane.i64(i64 %sgpr)
+  store i64 %readfirstlane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define amdgpu_kernel void @test_readfirstlane_copy_from_sgpr_f64(ptr addrspace(1) %out) #1 {
+; CHECK-SDAG-LABEL: test_readfirstlane_copy_from_sgpr_f64:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    s_mov_b64 s[2:3], 0
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-SDAG-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readfirstlane_copy_from_sgpr_f64:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    s_mov_b64 s[2:3], 0
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-GISEL-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CHECK-GISEL-NEXT:    s_endpgm
+  %sgpr = call double asm "s_mov_b64 $0, 0", "=s"()
+  %readfirstlane = call double @llvm.amdgcn.readfirstlane.f64(double %sgpr)
+  store double %readfirstlane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
 define amdgpu_kernel void @test_readfirstlane_fi(ptr addrspace(1) %out) #1 {
+; CHECK-SDAG-LABEL: test_readfirstlane_fi:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_add_u32 s0, s0, s9
+; CHECK-SDAG-NEXT:    s_addc_u32 s1, s1, 0
+; CHECK-SDAG-NEXT:    s_mov_b32 s4, 0
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s4
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readfirstlane_fi:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_add_u32 s0, s0, s9
+; CHECK-GISEL-NEXT:    s_addc_u32 s1, s1, 0
+; CHECK-GISEL-NEXT:    s_mov_b32 s4, 0
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s4
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_endpgm
   %alloca = alloca i32, addrspace(5)
   %int = ptrtoint ptr addrspace(5) %alloca to i32
   %readfirstlane = call i32 @llvm.amdgcn.readfirstlane(i32 %int)
@@ -67,5 +389,269 @@ define amdgpu_kernel void @test_readfirstlane_fi(ptr addrspace(1) %out) #1 {
   ret void
 }
 
+define void @test_readfirstlane_half(ptr addrspace(1) %out, half %src) {
+; CHECK-SDAG-LABEL: test_readfirstlane_half:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s4
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; CHECK-GISEL-LABEL: test_readfirstlane_half:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s4
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %x = call half @llvm.amdgcn.readfirstlane.f16(half %src)
+  call void asm sideeffect "; use $0", "s"(half %x)
+  ret void
+}
+
+define void @test_readfirstlane_float(ptr addrspace(1) %out, float %src) {
+; CHECK-SDAG-LABEL: test_readfirstlane_float:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s4
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; CHECK-GISEL-LABEL: test_readfirstlane_float:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s4
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %x = call float @llvm.amdgcn.readfirstlane.f32(float %src)
+  call void asm sideeffect "; use $0", "s"(float %x)
+  ret void
+}
+
+define void @test_readfirstlane_bfloat(ptr addrspace(1) %out, bfloat %src) {
+; CHECK-SDAG-LABEL: test_readfirstlane_bfloat:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s4
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; CHECK-GISEL-LABEL: test_readfirstlane_bfloat:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s4
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %x = call bfloat @llvm.amdgcn.readfirstlane.bf16(bfloat %src)
+  call void asm sideeffect "; use $0", "s"(bfloat %x)
+  ret void
+}
+
+define void @test_readfirstlane_i16(ptr addrspace(1) %out, i16 %src) {
+; CHECK-SDAG-LABEL: test_readfirstlane_i16:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v0, 0xffff
+; CHECK-SDAG-NEXT:    v_and_b32_e32 v0, s4, v0
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use v0
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; CHECK-GISEL-LABEL: test_readfirstlane_i16:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s4
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %x = call i16 @llvm.amdgcn.readfirstlane.i16(i16 %src)
+  call void asm sideeffect "; use $0", "s"(i16 %x)
+  ret void
+}
+
+define void @test_readfirstlane_v2f16(ptr addrspace(1) %out, <2 x half> %src) {
+; CHECK-SDAG-LABEL: test_readfirstlane_v2f16:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s4
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; CHECK-GISEL-LABEL: test_readfirstlane_v2f16:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s4
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %x = call <2 x half> @llvm.amdgcn.readfirstlane.v2f16(<2 x half> %src)
+  call void asm sideeffect "; use $0", "s"(<2 x half> %x)
+  ret void
+}
+
+define void @test_readfirstlane_v2f32(ptr addrspace(1) %out, <2 x float> %src) {
+; CHECK-SDAG-LABEL: test_readfirstlane_v2f32:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s5, v3
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[4:5]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; CHECK-GISEL-LABEL: test_readfirstlane_v2f32:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s[4:5]
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %x = call <2 x float> @llvm.amdgcn.readfirstlane.v2f32(<2 x float> %src)
+  call void asm sideeffect "; use $0", "s"(<2 x float> %x)
+  ret void
+}
+
+define void @test_readfirstlane_v7i32(ptr addrspace(1) %out, <7 x i32> %src) {
+; CHECK-SDAG-LABEL: test_readfirstlane_v7i32:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s10, v8
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s9, v7
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s8, v6
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s7, v5
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s6, v4
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s5, v3
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[4:10]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; CHECK-GISEL-LABEL: test_readfirstlane_v7i32:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s6, v4
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s7, v5
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s8, v6
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s9, v7
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s10, v8
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s[4:10]
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %x = call <7 x i32> @llvm.amdgcn.readfirstlane.v7i32(<7 x i32> %src)
+  call void asm sideeffect "; use $0", "s"(<7 x i32> %x)
+  ret void
+}
+
+define void @test_readfirstlane_p0(ptr addrspace(1) %out, ptr %src) {
+; CHECK-SDAG-LABEL: test_readfirstlane_p0:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s5, v3
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[4:5]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; CHECK-GISEL-LABEL: test_readfirstlane_p0:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s[4:5]
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %x = call ptr @llvm.amdgcn.readfirstlane.p0(ptr %src)
+  call void asm sideeffect "; use $0", "s"(ptr %x)
+  ret void
+}
+
+define void @test_readfirstlane_v3p0(ptr addrspace(1) %out, <3 x ptr> %src) {
+; CHECK-SDAG-LABEL: test_readfirstlane_v3p0:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s9, v7
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s8, v6
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s7, v5
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s6, v4
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s5, v3
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[4:9]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; CHECK-GISEL-LABEL: test_readfirstlane_v3p0:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s6, v4
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s7, v5
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s8, v6
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s9, v7
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s[4:9]
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %x = call <3 x ptr> @llvm.amdgcn.readfirstlane.v3p0(<3 x ptr> %src)
+  call void asm sideeffect "; use $0", "s"(<3 x ptr> %x)
+  ret void
+}
+
+define void @test_readfirstlane_v8i16(ptr addrspace(1) %out, <8 x i16> %src) {
+; CHECK-SDAG-LABEL: test_readfirstlane_v8i16:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s7, v5
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s6, v4
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s5, v3
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[4:7]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; CHECK-GISEL-LABEL: test_readfirstlane_v8i16:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s6, v4
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s7, v5
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s[4:7]
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %x = call <8 x i16> @llvm.amdgcn.readfirstlane.v8i16(<8 x i16> %src)
+  call void asm sideeffect "; use $0", "s"(<8 x i16> %x)
+  ret void
+}
+
 attributes #0 = { nounwind readnone convergent }
 attributes #1 = { nounwind }
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ptr.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ptr.ll
new file mode 100644
index 0000000000000..588f239606f52
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ptr.ll
@@ -0,0 +1,92 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=fiji -verify-machineinstrs < %s | FileCheck -check-prefix=CHECK-SDAG -enable-var-scope %s
+
+define void @test_readfirstlane_p3(ptr addrspace(1) %out, ptr addrspace(3) %src) {
+; CHECK-SDAG-LABEL: test_readfirstlane_p3:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s4
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %x = call ptr addrspace(3) @llvm.amdgcn.readfirstlane.p3(ptr addrspace(3) %src)
+  call void asm sideeffect "; use $0", "s"(ptr addrspace(3) %x)
+  ret void
+}
+
+define void @test_readfirstlane_v3p3(ptr addrspace(1) %out, <3 x ptr addrspace(3)> %src) {
+; CHECK-SDAG-LABEL: test_readfirstlane_v3p3:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s6, v4
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s5, v3
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[4:6]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %x = call <3 x ptr addrspace(3)> @llvm.amdgcn.readfirstlane.v3p3(<3 x ptr addrspace(3)> %src)
+  call void asm sideeffect "; use $0", "s"(<3 x ptr addrspace(3)> %x)
+  ret void
+}
+
+define void @test_readfirstlane_p5(ptr addrspace(1) %out, ptr addrspace(5) %src) {
+; CHECK-SDAG-LABEL: test_readfirstlane_p5:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s4
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %x = call ptr addrspace(5) @llvm.amdgcn.readfirstlane.p5(ptr addrspace(5) %src)
+  call void asm sideeffect "; use $0", "s"(ptr addrspace(5) %x)
+  ret void
+}
+
+define void @test_readfirstlane_v3p5(ptr addrspace(1) %out, <3 x ptr addrspace(5)> %src) {
+; CHECK-SDAG-LABEL: test_readfirstlane_v3p5:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s6, v4
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s5, v3
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[4:6]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %x = call <3 x ptr addrspace(5)> @llvm.amdgcn.readfirstlane.v3p5(<3 x ptr addrspace(5)> %src)
+  call void asm sideeffect "; use $0", "s"(<3 x ptr addrspace(5)> %x)
+  ret void
+}
+
+define void @test_readfirstlane_p6(ptr addrspace(1) %out, ptr addrspace(6) %src) {
+; CHECK-SDAG-LABEL: test_readfirstlane_p6:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s4
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %x = call ptr addrspace(6) @llvm.amdgcn.readfirstlane.p6(ptr addrspace(6) %src)
+  call void asm sideeffect "; use $0", "s"(ptr addrspace(6) %x)
+  ret void
+}
+
+define void @test_readfirstlane_v3p6(ptr addrspace(1) %out, <3 x ptr addrspace(6)> %src) {
+; CHECK-SDAG-LABEL: test_readfirstlane_v3p6:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s6, v4
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s5, v3
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[4:6]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %x = call <3 x ptr addrspace(6)> @llvm.amdgcn.readfirstlane.v3p6(<3 x ptr addrspace(6)> %src)
+  call void asm sideeffect "; use $0", "s"(<3 x ptr addrspace(6)> %x)
+  ret void
+}
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ll
index 51465f6bd10ce..71cd3db81addd 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ll
@@ -1,82 +1,966 @@
-; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=fiji -verify-machineinstrs < %s | FileCheck -enable-var-scope %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=fiji -verify-machineinstrs < %s | FileCheck --check-prefix=CHECK-SDAG -enable-var-scope %s
+; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=fiji -verify-machineinstrs -global-isel < %s | FileCheck --check-prefix=CHECK-GISEL -enable-var-scope %s
 
-declare i32 @llvm.amdgcn.readlane(i32, i32) #0
+declare i32 @llvm.amdgcn.readlane.i32(i32, i32) #0
+declare i64 @llvm.amdgcn.readlane.i64(i64, i32) #0
+declare double @llvm.amdgcn.readlane.f64(double, i32) #0
 
-; CHECK-LABEL: {{^}}test_readlane_sreg_sreg:
-; CHECK-NOT: v_readlane_b32
-define amdgpu_kernel void @test_readlane_sreg_sreg(i32 %src0, i32 %src1) #1 {
-  %readlane = call i32 @llvm.amdgcn.readlane(i32 %src0, i32 %src1)
+define amdgpu_kernel void @test_readlane_sreg_sreg_i32(i32 %src0, i32 %src1) #1 {
+; CHECK-SDAG-LABEL: test_readlane_sreg_sreg_i32:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s0
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readlane_sreg_sreg_i32:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s0
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_endpgm
+  %readlane = call i32 @llvm.amdgcn.readlane.i32(i32 %src0, i32 %src1)
   call void asm sideeffect "; use $0", "s"(i32 %readlane)
   ret void
 }
 
-; CHECK-LABEL: {{^}}test_readlane_vreg_sreg:
-; CHECK: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, s{{[0-9]+}}
-define amdgpu_kernel void @test_readlane_vreg_sreg(i32 %src0, i32 %src1) #1 {
+define amdgpu_kernel void @test_readlane_sreg_sreg_i64(i64 %src0, i32 %src1) #1 {
+; CHECK-SDAG-LABEL: test_readlane_sreg_sreg_i64:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[0:1]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readlane_sreg_sreg_i64:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s[0:1]
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_endpgm
+  %readlane = call i64 @llvm.amdgcn.readlane.i64(i64 %src0, i32 %src1)
+  call void asm sideeffect "; use $0", "s"(i64 %readlane)
+  ret void
+}
+
+define amdgpu_kernel void @test_readlane_sreg_sreg_f64(double %src0, i32 %src1) #1 {
+; CHECK-SDAG-LABEL: test_readlane_sreg_sreg_f64:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[0:1]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readlane_sreg_sreg_f64:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s[0:1]
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_endpgm
+  %readlane = call double @llvm.amdgcn.readlane.f64(double %src0, i32 %src1)
+  call void asm sideeffect "; use $0", "s"(double %readlane)
+  ret void
+}
+
+define amdgpu_kernel void @test_readlane_vreg_sreg_i32(i32 %src0, i32 %src1) #1 {
+; CHECK-SDAG-LABEL: test_readlane_vreg_sreg_i32:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dword s0, s[4:5], 0x4
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; def v0
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readlane_b32 s0, v0, s0
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s0
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readlane_vreg_sreg_i32:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dword s0, s[4:5], 0x4
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; def v0
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readlane_b32 s0, v0, s0
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s0
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_endpgm
   %vgpr = call i32 asm sideeffect "; def $0", "=v"()
-  %readlane = call i32 @llvm.amdgcn.readlane(i32 %vgpr, i32 %src1)
+  %readlane = call i32 @llvm.amdgcn.readlane.i32(i32 %vgpr, i32 %src1)
   call void asm sideeffect "; use $0", "s"(i32 %readlane)
   ret void
 }
 
-; CHECK-LABEL: {{^}}test_readlane_imm_sreg:
-; CHECK-NOT: v_readlane_b32
-define amdgpu_kernel void @test_readlane_imm_sreg(ptr addrspace(1) %out, i32 %src1) #1 {
-  %readlane = call i32 @llvm.amdgcn.readlane(i32 32, i32 %src1)
+define amdgpu_kernel void @test_readlane_vreg_sreg_i64(i64 %src0, i32 %src1) #1 {
+; CHECK-SDAG-LABEL: test_readlane_vreg_sreg_i64:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dword s0, s[4:5], 0x8
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; def v[0:1]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readlane_b32 s1, v1, s0
+; CHECK-SDAG-NEXT:    v_readlane_b32 s0, v0, s0
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[0:1]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readlane_vreg_sreg_i64:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dword s1, s[4:5], 0x8
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; def v[0:1]
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readlane_b32 s0, v0, s1
+; CHECK-GISEL-NEXT:    v_readlane_b32 s1, v1, s1
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s[0:1]
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_endpgm
+  %vgpr = call i64 asm sideeffect "; def $0", "=v"()
+  %readlane = call i64 @llvm.amdgcn.readlane.i64(i64 %vgpr, i32 %src1)
+  call void asm sideeffect "; use $0", "s"(i64 %readlane)
+  ret void
+}
+
+define amdgpu_kernel void @test_readlane_vreg_sreg_f64(double %src0, i32 %src1) #1 {
+; CHECK-SDAG-LABEL: test_readlane_vreg_sreg_f64:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dword s0, s[4:5], 0x8
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; def v[0:1]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readlane_b32 s1, v1, s0
+; CHECK-SDAG-NEXT:    v_readlane_b32 s0, v0, s0
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[0:1]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readlane_vreg_sreg_f64:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dword s1, s[4:5], 0x8
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; def v[0:1]
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readlane_b32 s0, v0, s1
+; CHECK-GISEL-NEXT:    v_readlane_b32 s1, v1, s1
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s[0:1]
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_endpgm
+  %vgpr = call double asm sideeffect "; def $0", "=v"()
+  %readlane = call double @llvm.amdgcn.readlane.f64(double %vgpr, i32 %src1)
+  call void asm sideeffect "; use $0", "s"(double %readlane)
+  ret void
+}
+
+define amdgpu_kernel void @test_readlane_imm_sreg_i32(ptr addrspace(1) %out, i32 %src1) #1 {
+; CHECK-SDAG-LABEL: test_readlane_imm_sreg_i32:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v2, 32
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v1, s1
+; CHECK-SDAG-NEXT:    flat_store_dword v[0:1], v2
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readlane_imm_sreg_i32:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v2, 32
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; CHECK-GISEL-NEXT:    flat_store_dword v[0:1], v2
+; CHECK-GISEL-NEXT:    s_endpgm
+  %readlane = call i32 @llvm.amdgcn.readlane.i32(i32 32, i32 %src1)
   store i32 %readlane, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; CHECK-LABEL: {{^}}test_readlane_vregs:
-; CHECK: v_readfirstlane_b32 [[LANE:s[0-9]+]], v{{[0-9]+}}
-; CHECK: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, [[LANE]]
-define amdgpu_kernel void @test_readlane_vregs(ptr addrspace(1) %out, ptr addrspace(1) %in) #1 {
+define amdgpu_kernel void @test_readlane_imm_sreg_i64(ptr addrspace(1) %out, i32 %src1) #1 {
+; CHECK-SDAG-LABEL: test_readlane_imm_sreg_i64:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v0, 32
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v1, 0
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-SDAG-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readlane_imm_sreg_i64:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-GISEL-NEXT:    s_mov_b64 s[2:3], 32
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-GISEL-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CHECK-GISEL-NEXT:    s_endpgm
+  %readlane = call i64 @llvm.amdgcn.readlane.i64(i64 32, i32 %src1)
+  store i64 %readlane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define amdgpu_kernel void @test_readlane_imm_sreg_f64(ptr addrspace(1) %out, i32 %src1) #1 {
+; CHECK-SDAG-LABEL: test_readlane_imm_sreg_f64:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v1, 0x40400000
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-SDAG-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readlane_imm_sreg_f64:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-GISEL-NEXT:    s_mov_b32 s2, 0
+; CHECK-GISEL-NEXT:    s_mov_b32 s3, 0x40400000
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-GISEL-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CHECK-GISEL-NEXT:    s_endpgm
+  %readlane = call double @llvm.amdgcn.readlane.f64(double 32.0, i32 %src1)
+  store double %readlane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define amdgpu_kernel void @test_readlane_vregs_i32(ptr addrspace(1) %out, ptr addrspace(1) %in) #1 {
+; CHECK-SDAG-LABEL: test_readlane_vregs_i32:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CHECK-SDAG-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-SDAG-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
+; CHECK-SDAG-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CHECK-SDAG-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s0, v1
+; CHECK-SDAG-NEXT:    s_nop 3
+; CHECK-SDAG-NEXT:    v_readlane_b32 s0, v0, s0
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; CHECK-SDAG-NEXT:    flat_store_dword v[2:3], v0
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readlane_vregs_i32:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CHECK-GISEL-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-GISEL-NEXT:    v_add_u32_e32 v0, vcc, v0, v2
+; CHECK-GISEL-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CHECK-GISEL-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s2, v1
+; CHECK-GISEL-NEXT:    s_nop 3
+; CHECK-GISEL-NEXT:    v_readlane_b32 s2, v0, s2
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v2, s2
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; CHECK-GISEL-NEXT:    flat_store_dword v[0:1], v2
+; CHECK-GISEL-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep.in = getelementptr <2 x i32>, ptr addrspace(1) %in, i32 %tid
   %args = load <2 x i32>, ptr addrspace(1) %gep.in
   %value = extractelement <2 x i32> %args, i32 0
   %lane = extractelement <2 x i32> %args, i32 1
-  %readlane = call i32 @llvm.amdgcn.readlane(i32 %value, i32 %lane)
+  %readlane = call i32 @llvm.amdgcn.readlane.i32(i32 %value, i32 %lane)
   store i32 %readlane, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; TODO: m0 should be folded.
-; CHECK-LABEL: {{^}}test_readlane_m0_sreg:
-; CHECK: s_mov_b32 m0, -1
-; CHECK: v_mov_b32_e32 [[VVAL:v[0-9]]], m0
-; CHECK: flat_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[VVAL]]
+define amdgpu_kernel void @test_readlane_vregs_i64(ptr addrspace(1) %out, ptr addrspace(1) %in) #1 {
+; CHECK-SDAG-LABEL: test_readlane_vregs_i64:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CHECK-SDAG-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-SDAG-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
+; CHECK-SDAG-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CHECK-SDAG-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v3, s0
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v4, s1
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s0, v2
+; CHECK-SDAG-NEXT:    s_nop 3
+; CHECK-SDAG-NEXT:    v_readlane_b32 s1, v1, s0
+; CHECK-SDAG-NEXT:    v_readlane_b32 s0, v0, s0
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v1, s1
+; CHECK-SDAG-NEXT:    flat_store_dwordx2 v[3:4], v[0:1]
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readlane_vregs_i64:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CHECK-GISEL-NEXT:    v_lshlrev_b32_e32 v2, 4, v0
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-GISEL-NEXT:    v_add_u32_e32 v0, vcc, v0, v2
+; CHECK-GISEL-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CHECK-GISEL-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s3, v2
+; CHECK-GISEL-NEXT:    s_nop 3
+; CHECK-GISEL-NEXT:    v_readlane_b32 s2, v0, s3
+; CHECK-GISEL-NEXT:    v_readlane_b32 s3, v1, s3
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-GISEL-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CHECK-GISEL-NEXT:    s_endpgm
+  %tid = call i32 @llvm.amdgcn.workitem.id.x()
+  %gep.in = getelementptr <2 x i64>, ptr addrspace(1) %in, i32 %tid
+  %args = load <2 x i64>, ptr addrspace(1) %gep.in
+  %value = extractelement <2 x i64> %args, i32 0
+  %lane = extractelement <2 x i64> %args, i32 1
+  %lane32 = trunc i64 %lane to i32
+  %readlane = call i64 @llvm.amdgcn.readlane.i64(i64 %value, i32 %lane32)
+  store i64 %readlane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define amdgpu_kernel void @test_readlane_vregs_f64(ptr addrspace(1) %out, ptr addrspace(1) %in) #1 {
+; CHECK-SDAG-LABEL: test_readlane_vregs_f64:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CHECK-SDAG-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-SDAG-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
+; CHECK-SDAG-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CHECK-SDAG-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v3, s0
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v4, s1
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s0, v2
+; CHECK-SDAG-NEXT:    s_nop 3
+; CHECK-SDAG-NEXT:    v_readlane_b32 s1, v1, s0
+; CHECK-SDAG-NEXT:    v_readlane_b32 s0, v0, s0
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v1, s1
+; CHECK-SDAG-NEXT:    flat_store_dwordx2 v[3:4], v[0:1]
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readlane_vregs_f64:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; CHECK-GISEL-NEXT:    v_lshlrev_b32_e32 v2, 4, v0
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-GISEL-NEXT:    v_add_u32_e32 v0, vcc, v0, v2
+; CHECK-GISEL-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; CHECK-GISEL-NEXT:    flat_load_dwordx4 v[0:3], v[0:1]
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s3, v2
+; CHECK-GISEL-NEXT:    s_nop 3
+; CHECK-GISEL-NEXT:    v_readlane_b32 s2, v0, s3
+; CHECK-GISEL-NEXT:    v_readlane_b32 s3, v1, s3
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-GISEL-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CHECK-GISEL-NEXT:    s_endpgm
+  %tid = call i32 @llvm.amdgcn.workitem.id.x()
+  %gep.in = getelementptr <2 x double>, ptr addrspace(1) %in, i32 %tid
+  %args = load <2 x double>, ptr addrspace(1) %gep.in
+  %value = extractelement <2 x double> %args, i32 0
+  %lane = extractelement <2 x double> %args, i32 1
+  %lane_cast = bitcast double %lane to i64
+  %lane32 = trunc i64 %lane_cast to i32
+  %readlane = call double @llvm.amdgcn.readlane.f64(double %value, i32 %lane32)
+  store double %readlane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
 define amdgpu_kernel void @test_readlane_m0_sreg(ptr addrspace(1) %out, i32 %src1) #1 {
+; CHECK-SDAG-LABEL: test_readlane_m0_sreg:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    s_mov_b32 m0, -1
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v2, m0
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v1, s1
+; CHECK-SDAG-NEXT:    flat_store_dword v[0:1], v2
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readlane_m0_sreg:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    s_mov_b32 m0, -1
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v2, m0
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; CHECK-GISEL-NEXT:    flat_store_dword v[0:1], v2
+; CHECK-GISEL-NEXT:    s_endpgm
   %m0 = call i32 asm "s_mov_b32 m0, -1", "={m0}"()
   %readlane = call i32 @llvm.amdgcn.readlane(i32 %m0, i32 %src1)
   store i32 %readlane, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; CHECK-LABEL: {{^}}test_readlane_vgpr_imm:
-; CHECK: v_readlane_b32 s{{[0-9]+}}, v{{[0-9]+}}, 32
-define amdgpu_kernel void @test_readlane_vgpr_imm(ptr addrspace(1) %out) #1 {
+define amdgpu_kernel void @test_readlane_vgpr_imm_i32(ptr addrspace(1) %out) #1 {
+; CHECK-SDAG-LABEL: test_readlane_vgpr_imm_i32:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; def v0
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    v_readlane_b32 s2, v0, 32
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v2, s2
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v1, s1
+; CHECK-SDAG-NEXT:    flat_store_dword v[0:1], v2
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readlane_vgpr_imm_i32:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; def v0
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    v_readlane_b32 s2, v0, 32
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v2, s2
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; CHECK-GISEL-NEXT:    flat_store_dword v[0:1], v2
+; CHECK-GISEL-NEXT:    s_endpgm
   %vgpr = call i32 asm sideeffect "; def $0", "=v"()
-  %readlane = call i32 @llvm.amdgcn.readlane(i32 %vgpr, i32 32) #0
+  %readlane = call i32 @llvm.amdgcn.readlane.i32(i32 %vgpr, i32 32) #0
   store i32 %readlane, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; CHECK-LABEL: {{^}}test_readlane_copy_from_sgpr:
-; CHECK: ;;#ASMSTART
-; CHECK-NEXT: s_mov_b32 [[SGPR:s[0-9]+]]
-; CHECK: ;;#ASMEND
-; CHECK-NOT: [[SGPR]]
-; CHECK-NOT: readlane
-; CHECK: v_mov_b32_e32 [[VCOPY:v[0-9]+]], [[SGPR]]
-; CHECK: flat_store_dword v{{\[[0-9]+:[0-9]+\]}}, [[VCOPY]]
-define amdgpu_kernel void @test_readlane_copy_from_sgpr(ptr addrspace(1) %out) #1 {
+define amdgpu_kernel void @test_readlane_vgpr_imm_i64(ptr addrspace(1) %out) #1 {
+; CHECK-SDAG-LABEL: test_readlane_vgpr_imm_i64:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; def v[0:1]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    v_readlane_b32 s2, v1, 32
+; CHECK-SDAG-NEXT:    v_readlane_b32 s3, v0, 32
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v0, s3
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-SDAG-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readlane_vgpr_imm_i64:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; def v[0:1]
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    v_readlane_b32 s2, v0, 32
+; CHECK-GISEL-NEXT:    v_readlane_b32 s3, v1, 32
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-GISEL-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CHECK-GISEL-NEXT:    s_endpgm
+  %vgpr = call i64 asm sideeffect "; def $0", "=v"()
+  %readlane = call i64 @llvm.amdgcn.readlane.i64(i64 %vgpr, i32 32) #0
+  store i64 %readlane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define amdgpu_kernel void @test_readlane_vgpr_imm_f64(ptr addrspace(1) %out) #1 {
+; CHECK-SDAG-LABEL: test_readlane_vgpr_imm_f64:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; def v[0:1]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    v_readlane_b32 s2, v1, 32
+; CHECK-SDAG-NEXT:    v_readlane_b32 s3, v0, 32
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v0, s3
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-SDAG-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readlane_vgpr_imm_f64:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; def v[0:1]
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    v_readlane_b32 s2, v0, 32
+; CHECK-GISEL-NEXT:    v_readlane_b32 s3, v1, 32
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-GISEL-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CHECK-GISEL-NEXT:    s_endpgm
+  %vgpr = call double asm sideeffect "; def $0", "=v"()
+  %readlane = call double @llvm.amdgcn.readlane.f64(double %vgpr, i32 32) #0
+  store double %readlane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define amdgpu_kernel void @test_readlane_copy_from_sgpr_i32(ptr addrspace(1) %out) #1 {
+; CHECK-SDAG-LABEL: test_readlane_copy_from_sgpr_i32:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    s_mov_b32 s2, 0
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v2, s2
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v1, s1
+; CHECK-SDAG-NEXT:    flat_store_dword v[0:1], v2
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readlane_copy_from_sgpr_i32:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    s_mov_b32 s2, 0
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v2, s2
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; CHECK-GISEL-NEXT:    flat_store_dword v[0:1], v2
+; CHECK-GISEL-NEXT:    s_endpgm
   %sgpr = call i32 asm "s_mov_b32 $0, 0", "=s"()
-  %readfirstlane = call i32 @llvm.amdgcn.readlane(i32 %sgpr, i32 7)
+  %readfirstlane = call i32 @llvm.amdgcn.readlane.i32(i32 %sgpr, i32 7)
   store i32 %readfirstlane, ptr addrspace(1) %out, align 4
   ret void
 }
 
+define amdgpu_kernel void @test_readlane_copy_from_sgpr_i64(ptr addrspace(1) %out) #1 {
+; CHECK-SDAG-LABEL: test_readlane_copy_from_sgpr_i64:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    s_mov_b64 s[2:3], 0
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-SDAG-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readlane_copy_from_sgpr_i64:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    s_mov_b64 s[2:3], 0
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-GISEL-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CHECK-GISEL-NEXT:    s_endpgm
+  %sgpr = call i64 asm "s_mov_b64 $0, 0", "=s"()
+  %readfirstlane = call i64 @llvm.amdgcn.readlane.i64(i64 %sgpr, i32 7)
+  store i64 %readfirstlane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define amdgpu_kernel void @test_readlane_copy_from_sgpr_f64(ptr addrspace(1) %out) #1 {
+; CHECK-SDAG-LABEL: test_readlane_copy_from_sgpr_f64:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    s_mov_b64 s[2:3], 0
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-SDAG-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CHECK-SDAG-NEXT:    s_endpgm
+;
+; CHECK-GISEL-LABEL: test_readlane_copy_from_sgpr_f64:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    s_mov_b64 s[2:3], 0
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v3, s1
+; CHECK-GISEL-NEXT:    v_mov_b32_e32 v2, s0
+; CHECK-GISEL-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; CHECK-GISEL-NEXT:    s_endpgm
+  %sgpr = call double asm "s_mov_b64 $0, 0", "=s"()
+  %readfirstlane = call double @llvm.amdgcn.readlane.f64(double %sgpr, i32 7)
+  store double %readfirstlane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define void @test_readlane_half(ptr addrspace(1) %out, half %src, i32 %src1) {
+; CHECK-SDAG-LABEL: test_readlane_half:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v3
+; CHECK-SDAG-NEXT:    s_nop 3
+; CHECK-SDAG-NEXT:    v_readlane_b32 s4, v2, s4
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s4
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; CHECK-GISEL-LABEL: test_readlane_half:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s4, v3
+; CHECK-GISEL-NEXT:    s_nop 3
+; CHECK-GISEL-NEXT:    v_readlane_b32 s4, v2, s4
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s4
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %x = call half @llvm.amdgcn.readlane.f16(half %src, i32 %src1)
+  call void asm sideeffect "; use $0", "s"(half %x)
+  ret void
+}
+
+define void @test_readlane_float(ptr addrspace(1) %out, float %src, i32 %src1) {
+; CHECK-SDAG-LABEL: test_readlane_float:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v3
+; CHECK-SDAG-NEXT:    s_nop 3
+; CHECK-SDAG-NEXT:    v_readlane_b32 s4, v2, s4
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s4
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; CHECK-GISEL-LABEL: test_readlane_float:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s4, v3
+; CHECK-GISEL-NEXT:    s_nop 3
+; CHECK-GISEL-NEXT:    v_readlane_b32 s4, v2, s4
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s4
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %x = call float @llvm.amdgcn.readlane.f32(float %src, i32 %src1)
+  call void asm sideeffect "; use $0", "s"(float %x)
+  ret void
+}
+
+define void @test_readlane_bfloat(ptr addrspace(1) %out, bfloat %src, i32 %src1) {
+; CHECK-SDAG-LABEL: test_readlane_bfloat:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v3
+; CHECK-SDAG-NEXT:    s_nop 3
+; CHECK-SDAG-NEXT:    v_readlane_b32 s4, v2, s4
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s4
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; CHECK-GISEL-LABEL: test_readlane_bfloat:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s4, v3
+; CHECK-GISEL-NEXT:    s_nop 3
+; CHECK-GISEL-NEXT:    v_readlane_b32 s4, v2, s4
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s4
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %x = call bfloat @llvm.amdgcn.readlane.bf16(bfloat %src, i32 %src1)
+  call void asm sideeffect "; use $0", "s"(bfloat %x)
+  ret void
+}
+
+define void @test_readlane_i16(ptr addrspace(1) %out, i16 %src, i32 %src1) {
+; CHECK-SDAG-LABEL: test_readlane_i16:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v3
+; CHECK-SDAG-NEXT:    v_mov_b32_e32 v0, 0xffff
+; CHECK-SDAG-NEXT:    s_nop 2
+; CHECK-SDAG-NEXT:    v_readlane_b32 s4, v2, s4
+; CHECK-SDAG-NEXT:    v_and_b32_e32 v0, s4, v0
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use v0
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; CHECK-GISEL-LABEL: test_readlane_i16:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s4, v3
+; CHECK-GISEL-NEXT:    s_nop 3
+; CHECK-GISEL-NEXT:    v_readlane_b32 s4, v2, s4
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s4
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %x = call i16 @llvm.amdgcn.readlane.i16(i16 %src, i32 %src1)
+  call void asm sideeffect "; use $0", "s"(i16 %x)
+  ret void
+}
+
+define void @test_readlane_v2f16(ptr addrspace(1) %out, <2 x half> %src, i32 %src1) {
+; CHECK-SDAG-LABEL: test_readlane_v2f16:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v3
+; CHECK-SDAG-NEXT:    s_nop 3
+; CHECK-SDAG-NEXT:    v_readlane_b32 s4, v2, s4
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s4
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; CHECK-GISEL-LABEL: test_readlane_v2f16:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s4, v3
+; CHECK-GISEL-NEXT:    s_nop 3
+; CHECK-GISEL-NEXT:    v_readlane_b32 s4, v2, s4
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s4
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %x = call <2 x half> @llvm.amdgcn.readlane.v2f16(<2 x half> %src, i32 %src1)
+  call void asm sideeffect "; use $0", "s"(<2 x half> %x)
+  ret void
+}
+
+define void @test_readlane_v2f32(ptr addrspace(1) %out, <2 x float> %src, i32 %src1) {
+; CHECK-SDAG-LABEL: test_readlane_v2f32:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v4
+; CHECK-SDAG-NEXT:    s_nop 3
+; CHECK-SDAG-NEXT:    v_readlane_b32 s5, v3, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s4, v2, s4
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[4:5]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; CHECK-GISEL-LABEL: test_readlane_v2f32:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s5, v4
+; CHECK-GISEL-NEXT:    s_nop 3
+; CHECK-GISEL-NEXT:    v_readlane_b32 s4, v2, s5
+; CHECK-GISEL-NEXT:    v_readlane_b32 s5, v3, s5
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s[4:5]
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %x = call <2 x float> @llvm.amdgcn.readlane.v2f32(<2 x float> %src, i32 %src1)
+  call void asm sideeffect "; use $0", "s"(<2 x float> %x)
+  ret void
+}
+
+define void @test_readlane_v7i32(ptr addrspace(1) %out, <7 x i32> %src, i32 %src1) {
+; CHECK-SDAG-LABEL: test_readlane_v7i32:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v9
+; CHECK-SDAG-NEXT:    s_nop 3
+; CHECK-SDAG-NEXT:    v_readlane_b32 s10, v8, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s9, v7, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s8, v6, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s7, v5, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s6, v4, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s5, v3, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s4, v2, s4
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[4:10]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; CHECK-GISEL-LABEL: test_readlane_v7i32:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s10, v9
+; CHECK-GISEL-NEXT:    s_nop 3
+; CHECK-GISEL-NEXT:    v_readlane_b32 s4, v2, s10
+; CHECK-GISEL-NEXT:    v_readlane_b32 s5, v3, s10
+; CHECK-GISEL-NEXT:    v_readlane_b32 s6, v4, s10
+; CHECK-GISEL-NEXT:    v_readlane_b32 s7, v5, s10
+; CHECK-GISEL-NEXT:    v_readlane_b32 s8, v6, s10
+; CHECK-GISEL-NEXT:    v_readlane_b32 s9, v7, s10
+; CHECK-GISEL-NEXT:    v_readlane_b32 s10, v8, s10
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s[4:10]
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %x = call <7 x i32> @llvm.amdgcn.readlane.v7i32(<7 x i32> %src, i32 %src1)
+  call void asm sideeffect "; use $0", "s"(<7 x i32> %x)
+  ret void
+}
+
+define void @test_readlane_p0(ptr addrspace(1) %out, ptr %src, i32 %src1) {
+; CHECK-SDAG-LABEL: test_readlane_p0:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v4
+; CHECK-SDAG-NEXT:    s_nop 3
+; CHECK-SDAG-NEXT:    v_readlane_b32 s5, v3, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s4, v2, s4
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[4:5]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; CHECK-GISEL-LABEL: test_readlane_p0:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s5, v4
+; CHECK-GISEL-NEXT:    s_nop 3
+; CHECK-GISEL-NEXT:    v_readlane_b32 s4, v2, s5
+; CHECK-GISEL-NEXT:    v_readlane_b32 s5, v3, s5
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s[4:5]
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %x = call ptr @llvm.amdgcn.readlane.p0(ptr %src, i32 %src1)
+  call void asm sideeffect "; use $0", "s"(ptr %x)
+  ret void
+}
+
+define void @test_readlane_v3p0(ptr addrspace(1) %out, <3 x ptr> %src, i32 %src1) {
+; CHECK-SDAG-LABEL: test_readlane_v3p0:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v8
+; CHECK-SDAG-NEXT:    s_nop 3
+; CHECK-SDAG-NEXT:    v_readlane_b32 s9, v7, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s8, v6, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s7, v5, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s6, v4, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s5, v3, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s4, v2, s4
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[4:9]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; CHECK-GISEL-LABEL: test_readlane_v3p0:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s9, v8
+; CHECK-GISEL-NEXT:    s_nop 3
+; CHECK-GISEL-NEXT:    v_readlane_b32 s4, v2, s9
+; CHECK-GISEL-NEXT:    v_readlane_b32 s5, v3, s9
+; CHECK-GISEL-NEXT:    v_readlane_b32 s6, v4, s9
+; CHECK-GISEL-NEXT:    v_readlane_b32 s7, v5, s9
+; CHECK-GISEL-NEXT:    v_readlane_b32 s8, v6, s9
+; CHECK-GISEL-NEXT:    v_readlane_b32 s9, v7, s9
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s[4:9]
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %x = call <3 x ptr> @llvm.amdgcn.readlane.v3p0(<3 x ptr> %src, i32 %src1)
+  call void asm sideeffect "; use $0", "s"(<3 x ptr> %x)
+  ret void
+}
+
+define void @test_readlane_v8i16(ptr addrspace(1) %out, <8 x i16> %src, i32 %src1) {
+; CHECK-SDAG-LABEL: test_readlane_v8i16:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v6
+; CHECK-SDAG-NEXT:    s_nop 3
+; CHECK-SDAG-NEXT:    v_readlane_b32 s7, v5, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s6, v4, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s5, v3, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s4, v2, s4
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[4:7]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; CHECK-GISEL-LABEL: test_readlane_v8i16:
+; CHECK-GISEL:       ; %bb.0:
+; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s7, v6
+; CHECK-GISEL-NEXT:    s_nop 3
+; CHECK-GISEL-NEXT:    v_readlane_b32 s4, v2, s7
+; CHECK-GISEL-NEXT:    v_readlane_b32 s5, v3, s7
+; CHECK-GISEL-NEXT:    v_readlane_b32 s6, v4, s7
+; CHECK-GISEL-NEXT:    v_readlane_b32 s7, v5, s7
+; CHECK-GISEL-NEXT:    ;;#ASMSTART
+; CHECK-GISEL-NEXT:    ; use s[4:7]
+; CHECK-GISEL-NEXT:    ;;#ASMEND
+; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %x = call <8 x i16> @llvm.amdgcn.readlane.v8i16(<8 x i16> %src, i32 %src1)
+  call void asm sideeffect "; use $0", "s"(<8 x i16> %x)
+  ret void
+}
+
 declare i32 @llvm.amdgcn.workitem.id.x() #2
 
 attributes #0 = { nounwind readnone convergent }
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ptr.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ptr.ll
new file mode 100644
index 0000000000000..1b4ee84c75250
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ptr.ll
@@ -0,0 +1,105 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=fiji -verify-machineinstrs < %s | FileCheck --check-prefix=CHECK-SDAG -enable-var-scope %s
+
+define void @test_readlane_p3(ptr addrspace(1) %out, ptr addrspace(3) %src, i32 %src1) {
+; CHECK-SDAG-LABEL: test_readlane_p3:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v3
+; CHECK-SDAG-NEXT:    s_nop 3
+; CHECK-SDAG-NEXT:    v_readlane_b32 s4, v2, s4
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s4
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %x = call ptr addrspace(3) @llvm.amdgcn.readlane.p3(ptr addrspace(3) %src, i32 %src1)
+  call void asm sideeffect "; use $0", "s"(ptr addrspace(3) %x)
+  ret void
+}
+
+define void @test_readlane_v3p3(ptr addrspace(1) %out, <3 x ptr addrspace(3)> %src, i32 %src1) {
+; CHECK-SDAG-LABEL: test_readlane_v3p3:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v5
+; CHECK-SDAG-NEXT:    s_nop 3
+; CHECK-SDAG-NEXT:    v_readlane_b32 s6, v4, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s5, v3, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s4, v2, s4
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[4:6]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %x = call <3 x ptr addrspace(3)> @llvm.amdgcn.readlane.v3p3(<3 x ptr addrspace(3)> %src, i32 %src1)
+  call void asm sideeffect "; use $0", "s"(<3 x ptr addrspace(3)> %x)
+  ret void
+}
+
+define void @test_readlane_p5(ptr addrspace(1) %out, ptr addrspace(5) %src, i32 %src1) {
+; CHECK-SDAG-LABEL: test_readlane_p5:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v3
+; CHECK-SDAG-NEXT:    s_nop 3
+; CHECK-SDAG-NEXT:    v_readlane_b32 s4, v2, s4
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s4
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %x = call ptr addrspace(5) @llvm.amdgcn.readlane.p5(ptr addrspace(5) %src, i32 %src1)
+  call void asm sideeffect "; use $0", "s"(ptr addrspace(5) %x)
+  ret void
+}
+
+define void @test_readlane_v3p5(ptr addrspace(1) %out, <3 x ptr addrspace(5)> %src, i32 %src1) {
+; CHECK-SDAG-LABEL: test_readlane_v3p5:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v5
+; CHECK-SDAG-NEXT:    s_nop 3
+; CHECK-SDAG-NEXT:    v_readlane_b32 s6, v4, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s5, v3, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s4, v2, s4
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[4:6]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %x = call <3 x ptr addrspace(5)> @llvm.amdgcn.readlane.v3p5(<3 x ptr addrspace(5)> %src, i32 %src1)
+  call void asm sideeffect "; use $0", "s"(<3 x ptr addrspace(5)> %x)
+  ret void
+}
+
+define void @test_readlane_p6(ptr addrspace(1) %out, ptr addrspace(6) %src, i32 %src1) {
+; CHECK-SDAG-LABEL: test_readlane_p6:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v3
+; CHECK-SDAG-NEXT:    s_nop 3
+; CHECK-SDAG-NEXT:    v_readlane_b32 s4, v2, s4
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s4
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %x = call ptr addrspace(6) @llvm.amdgcn.readlane.p6(ptr addrspace(6) %src, i32 %src1)
+  call void asm sideeffect "; use $0", "s"(ptr addrspace(6) %x)
+  ret void
+}
+
+define void @test_readlane_v3p6(ptr addrspace(1) %out, <3 x ptr addrspace(6)> %src, i32 %src1) {
+; CHECK-SDAG-LABEL: test_readlane_v3p6:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v5
+; CHECK-SDAG-NEXT:    s_nop 3
+; CHECK-SDAG-NEXT:    v_readlane_b32 s6, v4, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s5, v3, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s4, v2, s4
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[4:6]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %x = call <3 x ptr addrspace(6)> @llvm.amdgcn.readlane.v3p6(<3 x ptr addrspace(6)> %src, i32 %src1)
+  call void asm sideeffect "; use $0", "s"(<3 x ptr addrspace(6)> %x)
+  ret void
+}
+
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.writelane.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.writelane.ll
index 37951669dbe75..d0a865f565eeb 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.writelane.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.writelane.ll
@@ -1,85 +1,3044 @@
-; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=gfx700 -verify-machineinstrs < %s | FileCheck -check-prefixes=CHECK,CIGFX9 %s
-; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=gfx802 -verify-machineinstrs < %s | FileCheck -check-prefixes=CHECK,CIGFX9 %s
-; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefixes=CHECK,GFX10 %s
-; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=gfx1100 -verify-machineinstrs -amdgpu-enable-vopd=0 < %s | FileCheck -check-prefixes=CHECK,GFX10 %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=gfx802 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX802-SDAG %s
+; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX1010-SDAG %s
+; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=gfx1100 -verify-machineinstrs -amdgpu-enable-vopd=0 < %s | FileCheck -check-prefixes=GFX1100-SDAG %s
+
+; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=gfx802 -verify-machineinstrs -global-isel < %s | FileCheck -check-prefixes=GFX802-GISEL %s
+; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=gfx1010 -verify-machineinstrs -global-isel < %s | FileCheck -check-prefixes=GFX1010-GISEL %s
+; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=gfx1100 -verify-machineinstrs -amdgpu-enable-vopd=0 -global-isel < %s | FileCheck -check-prefixes=GFX1100-GISEL %s
 
 declare i32 @llvm.amdgcn.writelane(i32, i32, i32) #0
+declare i64 @llvm.amdgcn.writelane.i64(i64, i32, i64) #0
+declare double @llvm.amdgcn.writelane.f64(double, i32, double) #0
 
-; CHECK-LABEL: {{^}}test_writelane_sreg:
-; CIGFX9: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, m0
-; GFX10: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, s{{[0-9]+}}
-define amdgpu_kernel void @test_writelane_sreg(ptr addrspace(1) %out, i32 %src0, i32 %src1) #1 {
+define amdgpu_kernel void @test_writelane_sreg_i32(ptr addrspace(1) %out, i32 %src0, i32 %src1) #1 {
+; GFX802-SDAG-LABEL: test_writelane_sreg_i32:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    s_mov_b32 m0, s3
+; GFX802-SDAG-NEXT:    s_load_dword s3, s[0:1], 0x0
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v1, s1
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v2, s3
+; GFX802-SDAG-NEXT:    v_writelane_b32 v2, s2, m0
+; GFX802-SDAG-NEXT:    flat_store_dword v[0:1], v2
+; GFX802-SDAG-NEXT:    s_endpgm
+;
+; GFX1010-SDAG-LABEL: test_writelane_sreg_i32:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    s_load_dword s4, s[0:1], 0x0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v0, s4
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v0, s2, s3
+; GFX1010-SDAG-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX1010-SDAG-NEXT:    s_endpgm
+;
+; GFX1100-SDAG-LABEL: test_writelane_sreg_i32:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    s_load_b32 s4, s[0:1], 0x0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v0, s4
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v0, s2, s3
+; GFX1100-SDAG-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX1100-SDAG-NEXT:    s_nop 0
+; GFX1100-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-SDAG-NEXT:    s_endpgm
+;
+; GFX802-GISEL-LABEL: test_writelane_sreg_i32:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    s_mov_b32 m0, s3
+; GFX802-GISEL-NEXT:    s_load_dword s3, s[0:1], 0x0
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v2, s3
+; GFX802-GISEL-NEXT:    v_writelane_b32 v2, s2, m0
+; GFX802-GISEL-NEXT:    flat_store_dword v[0:1], v2
+; GFX802-GISEL-NEXT:    s_endpgm
+;
+; GFX1010-GISEL-LABEL: test_writelane_sreg_i32:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    s_load_dword s4, s[0:1], 0x0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v0, s4
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v0, s2, s3
+; GFX1010-GISEL-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX1010-GISEL-NEXT:    s_endpgm
+;
+; GFX1100-GISEL-LABEL: test_writelane_sreg_i32:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    s_load_b32 s4, s[0:1], 0x0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v0, s4
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v0, s2, s3
+; GFX1100-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX1100-GISEL-NEXT:    s_nop 0
+; GFX1100-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-GISEL-NEXT:    s_endpgm
   %oldval = load i32, ptr addrspace(1) %out
-  %writelane = call i32 @llvm.amdgcn.writelane(i32 %src0, i32 %src1, i32 %oldval)
+  %writelane = call i32 @llvm.amdgcn.writelane.i32(i32 %src0, i32 %src1, i32 %oldval)
   store i32 %writelane, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; CHECK-LABEL: {{^}}test_writelane_imm_sreg:
-; CHECK: v_writelane_b32 v{{[0-9]+}}, 32, s{{[0-9]+}}
-define amdgpu_kernel void @test_writelane_imm_sreg(ptr addrspace(1) %out, i32 %src1) #1 {
+define amdgpu_kernel void @test_writelane_sreg_i64(ptr addrspace(1) %out, i64 %src0, i32 %src1) #1 {
+; GFX802-SDAG-LABEL: test_writelane_sreg_i64:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-SDAG-NEXT:    s_load_dword s6, s[4:5], 0x10
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
+; GFX802-SDAG-NEXT:    s_mov_b32 m0, s6
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v3, s1
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v1, s5
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v0, s4
+; GFX802-SDAG-NEXT:    v_writelane_b32 v1, s3, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v0, s2, m0
+; GFX802-SDAG-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; GFX802-SDAG-NEXT:    s_endpgm
+;
+; GFX1010-SDAG-LABEL: test_writelane_sreg_i64:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_clause 0x1
+; GFX1010-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-SDAG-NEXT:    s_load_dword s6, s[4:5], 0x10
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v1, s5
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v0, s4
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v1, s3, s6
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v0, s2, s6
+; GFX1010-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX1010-SDAG-NEXT:    s_endpgm
+;
+; GFX1100-SDAG-LABEL: test_writelane_sreg_i64:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_clause 0x1
+; GFX1100-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x0
+; GFX1100-SDAG-NEXT:    s_load_b32 s2, s[0:1], 0x10
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v1, s7, s2
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v0, s6, s2
+; GFX1100-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX1100-SDAG-NEXT:    s_nop 0
+; GFX1100-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-SDAG-NEXT:    s_endpgm
+;
+; GFX802-GISEL-LABEL: test_writelane_sreg_i64:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-GISEL-NEXT:    s_load_dword s6, s[4:5], 0x10
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
+; GFX802-GISEL-NEXT:    s_mov_b32 m0, s6
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v3, s1
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v2, s0
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v0, s4
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v1, s5
+; GFX802-GISEL-NEXT:    v_writelane_b32 v0, s2, m0
+; GFX802-GISEL-NEXT:    v_writelane_b32 v1, s3, m0
+; GFX802-GISEL-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; GFX802-GISEL-NEXT:    s_endpgm
+;
+; GFX1010-GISEL-LABEL: test_writelane_sreg_i64:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_clause 0x1
+; GFX1010-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-GISEL-NEXT:    s_load_dword s6, s[4:5], 0x10
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v0, s4
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v1, s5
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v0, s2, s6
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v1, s3, s6
+; GFX1010-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX1010-GISEL-NEXT:    s_endpgm
+;
+; GFX1100-GISEL-LABEL: test_writelane_sreg_i64:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_clause 0x1
+; GFX1100-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x0
+; GFX1100-GISEL-NEXT:    s_load_b32 s2, s[0:1], 0x10
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v0, s6, s2
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v1, s7, s2
+; GFX1100-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX1100-GISEL-NEXT:    s_nop 0
+; GFX1100-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-GISEL-NEXT:    s_endpgm
+  %oldval = load i64, ptr addrspace(1) %out
+  %writelane = call i64 @llvm.amdgcn.writelane.i64(i64 %src0, i32 %src1, i64 %oldval)
+  store i64 %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define amdgpu_kernel void @test_writelane_sreg_f64(ptr addrspace(1) %out, double %src0, i32 %src1) #1 {
+; GFX802-SDAG-LABEL: test_writelane_sreg_f64:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-SDAG-NEXT:    s_load_dword s6, s[4:5], 0x10
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
+; GFX802-SDAG-NEXT:    s_mov_b32 m0, s6
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v3, s1
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v1, s5
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v0, s4
+; GFX802-SDAG-NEXT:    v_writelane_b32 v1, s3, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v0, s2, m0
+; GFX802-SDAG-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; GFX802-SDAG-NEXT:    s_endpgm
+;
+; GFX1010-SDAG-LABEL: test_writelane_sreg_f64:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_clause 0x1
+; GFX1010-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-SDAG-NEXT:    s_load_dword s6, s[4:5], 0x10
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v1, s5
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v0, s4
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v1, s3, s6
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v0, s2, s6
+; GFX1010-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX1010-SDAG-NEXT:    s_endpgm
+;
+; GFX1100-SDAG-LABEL: test_writelane_sreg_f64:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_clause 0x1
+; GFX1100-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x0
+; GFX1100-SDAG-NEXT:    s_load_b32 s2, s[0:1], 0x10
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v1, s7, s2
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v0, s6, s2
+; GFX1100-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX1100-SDAG-NEXT:    s_nop 0
+; GFX1100-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-SDAG-NEXT:    s_endpgm
+;
+; GFX802-GISEL-LABEL: test_writelane_sreg_f64:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-GISEL-NEXT:    s_load_dword s6, s[4:5], 0x10
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
+; GFX802-GISEL-NEXT:    s_mov_b32 m0, s6
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v3, s1
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v2, s0
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v0, s4
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v1, s5
+; GFX802-GISEL-NEXT:    v_writelane_b32 v0, s2, m0
+; GFX802-GISEL-NEXT:    v_writelane_b32 v1, s3, m0
+; GFX802-GISEL-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; GFX802-GISEL-NEXT:    s_endpgm
+;
+; GFX1010-GISEL-LABEL: test_writelane_sreg_f64:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_clause 0x1
+; GFX1010-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-GISEL-NEXT:    s_load_dword s6, s[4:5], 0x10
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v0, s4
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v1, s5
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v0, s2, s6
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v1, s3, s6
+; GFX1010-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX1010-GISEL-NEXT:    s_endpgm
+;
+; GFX1100-GISEL-LABEL: test_writelane_sreg_f64:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_clause 0x1
+; GFX1100-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x0
+; GFX1100-GISEL-NEXT:    s_load_b32 s2, s[0:1], 0x10
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    s_load_b64 s[0:1], s[4:5], 0x0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v0, s6, s2
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v1, s7, s2
+; GFX1100-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX1100-GISEL-NEXT:    s_nop 0
+; GFX1100-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-GISEL-NEXT:    s_endpgm
+  %oldval = load double, ptr addrspace(1) %out
+  %writelane = call double @llvm.amdgcn.writelane.f64(double %src0, i32 %src1, double %oldval)
+  store double %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define amdgpu_kernel void @test_writelane_imm_sreg_i32(ptr addrspace(1) %out, i32 %src1) #1 {
+; GFX802-SDAG-LABEL: test_writelane_imm_sreg_i32:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX802-SDAG-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    s_load_dword s3, s[0:1], 0x0
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v1, s1
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v2, s3
+; GFX802-SDAG-NEXT:    v_writelane_b32 v2, 32, s2
+; GFX802-SDAG-NEXT:    flat_store_dword v[0:1], v2
+; GFX802-SDAG-NEXT:    s_endpgm
+;
+; GFX1010-SDAG-LABEL: test_writelane_imm_sreg_i32:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_clause 0x1
+; GFX1010-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX1010-SDAG-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    s_load_dword s3, s[0:1], 0x0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v0, s3
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v0, 32, s2
+; GFX1010-SDAG-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX1010-SDAG-NEXT:    s_endpgm
+;
+; GFX1100-SDAG-LABEL: test_writelane_imm_sreg_i32:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_clause 0x1
+; GFX1100-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x0
+; GFX1100-SDAG-NEXT:    s_load_b32 s0, s[0:1], 0x8
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    s_load_b32 s1, s[2:3], 0x0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v0, 32, s0
+; GFX1100-SDAG-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX1100-SDAG-NEXT:    s_nop 0
+; GFX1100-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-SDAG-NEXT:    s_endpgm
+;
+; GFX802-GISEL-LABEL: test_writelane_imm_sreg_i32:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX802-GISEL-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    s_load_dword s3, s[0:1], 0x0
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v2, s3
+; GFX802-GISEL-NEXT:    v_writelane_b32 v2, 32, s2
+; GFX802-GISEL-NEXT:    flat_store_dword v[0:1], v2
+; GFX802-GISEL-NEXT:    s_endpgm
+;
+; GFX1010-GISEL-LABEL: test_writelane_imm_sreg_i32:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_clause 0x1
+; GFX1010-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX1010-GISEL-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    s_load_dword s3, s[0:1], 0x0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v0, s3
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v0, 32, s2
+; GFX1010-GISEL-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX1010-GISEL-NEXT:    s_endpgm
+;
+; GFX1100-GISEL-LABEL: test_writelane_imm_sreg_i32:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_clause 0x1
+; GFX1100-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x0
+; GFX1100-GISEL-NEXT:    s_load_b32 s0, s[0:1], 0x8
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    s_load_b32 s1, s[2:3], 0x0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v0, 32, s0
+; GFX1100-GISEL-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX1100-GISEL-NEXT:    s_nop 0
+; GFX1100-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-GISEL-NEXT:    s_endpgm
   %oldval = load i32, ptr addrspace(1) %out
-  %writelane = call i32 @llvm.amdgcn.writelane(i32 32, i32 %src1, i32 %oldval)
+  %writelane = call i32 @llvm.amdgcn.writelane.i32(i32 32, i32 %src1, i32 %oldval)
   store i32 %writelane, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; CHECK-LABEL: {{^}}test_writelane_vreg_lane:
-; CHECK: v_readfirstlane_b32 [[LANE:s[0-9]+]], v{{[0-9]+}}
-; CHECK: v_writelane_b32 v{{[0-9]+}}, 12, [[LANE]]
-define amdgpu_kernel void @test_writelane_vreg_lane(ptr addrspace(1) %out, ptr addrspace(1) %in) #1 {
+define amdgpu_kernel void @test_writelane_imm_sreg_i64(ptr addrspace(1) %out, i32 %src1) #1 {
+; GFX802-SDAG-LABEL: test_writelane_imm_sreg_i64:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX802-SDAG-NEXT:    s_load_dword s4, s[4:5], 0x8
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v3, s1
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX802-SDAG-NEXT:    v_writelane_b32 v1, 0, s4
+; GFX802-SDAG-NEXT:    v_writelane_b32 v0, 32, s4
+; GFX802-SDAG-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; GFX802-SDAG-NEXT:    s_endpgm
+;
+; GFX1010-SDAG-LABEL: test_writelane_imm_sreg_i64:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_clause 0x1
+; GFX1010-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX1010-SDAG-NEXT:    s_load_dword s6, s[4:5], 0x8
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v1, 0, s6
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v0, 32, s6
+; GFX1010-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX1010-SDAG-NEXT:    s_endpgm
+;
+; GFX1100-SDAG-LABEL: test_writelane_imm_sreg_i64:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_clause 0x1
+; GFX1100-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x0
+; GFX1100-SDAG-NEXT:    s_load_b32 s4, s[0:1], 0x8
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    s_load_b64 s[0:1], s[2:3], 0x0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v1, 0, s4
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v0, 32, s4
+; GFX1100-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[2:3]
+; GFX1100-SDAG-NEXT:    s_nop 0
+; GFX1100-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-SDAG-NEXT:    s_endpgm
+;
+; GFX802-GISEL-LABEL: test_writelane_imm_sreg_i64:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX802-GISEL-NEXT:    s_load_dword s4, s[4:5], 0x8
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v3, s1
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v2, s0
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; GFX802-GISEL-NEXT:    v_writelane_b32 v0, 32, s4
+; GFX802-GISEL-NEXT:    v_writelane_b32 v1, 0, s4
+; GFX802-GISEL-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; GFX802-GISEL-NEXT:    s_endpgm
+;
+; GFX1010-GISEL-LABEL: test_writelane_imm_sreg_i64:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_clause 0x1
+; GFX1010-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX1010-GISEL-NEXT:    s_load_dword s6, s[4:5], 0x8
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v0, 32, s6
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v1, 0, s6
+; GFX1010-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX1010-GISEL-NEXT:    s_endpgm
+;
+; GFX1100-GISEL-LABEL: test_writelane_imm_sreg_i64:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_clause 0x1
+; GFX1100-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x0
+; GFX1100-GISEL-NEXT:    s_load_b32 s4, s[0:1], 0x8
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    s_load_b64 s[0:1], s[2:3], 0x0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v0, 32, s4
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v1, 0, s4
+; GFX1100-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[2:3]
+; GFX1100-GISEL-NEXT:    s_nop 0
+; GFX1100-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-GISEL-NEXT:    s_endpgm
+  %oldval = load i64, ptr addrspace(1) %out
+  %writelane = call i64 @llvm.amdgcn.writelane.i64(i64 32, i32 %src1, i64 %oldval)
+  store i64 %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define amdgpu_kernel void @test_writelane_imm_sreg_f64(ptr addrspace(1) %out, i32 %src1) #1 {
+; GFX802-SDAG-LABEL: test_writelane_imm_sreg_f64:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX802-SDAG-NEXT:    s_load_dword s4, s[4:5], 0x8
+; GFX802-SDAG-NEXT:    s_mov_b32 s5, 0x40400000
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
+; GFX802-SDAG-NEXT:    s_mov_b32 m0, s4
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v3, s1
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX802-SDAG-NEXT:    v_writelane_b32 v1, s5, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v0, 0, s4
+; GFX802-SDAG-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; GFX802-SDAG-NEXT:    s_endpgm
+;
+; GFX1010-SDAG-LABEL: test_writelane_imm_sreg_f64:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_clause 0x1
+; GFX1010-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX1010-SDAG-NEXT:    s_load_dword s6, s[4:5], 0x8
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1010-SDAG-NEXT:    s_mov_b32 s2, 0x40400000
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v1, s2, s6
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v0, 0, s6
+; GFX1010-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX1010-SDAG-NEXT:    s_endpgm
+;
+; GFX1100-SDAG-LABEL: test_writelane_imm_sreg_f64:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_clause 0x1
+; GFX1100-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x0
+; GFX1100-SDAG-NEXT:    s_load_b32 s4, s[0:1], 0x8
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    s_load_b64 s[0:1], s[2:3], 0x0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1100-SDAG-NEXT:    s_mov_b32 s0, 0x40400000
+; GFX1100-SDAG-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v1, s0, s4
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v0, 0, s4
+; GFX1100-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[2:3]
+; GFX1100-SDAG-NEXT:    s_nop 0
+; GFX1100-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-SDAG-NEXT:    s_endpgm
+;
+; GFX802-GISEL-LABEL: test_writelane_imm_sreg_f64:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX802-GISEL-NEXT:    s_load_dword s4, s[4:5], 0x8
+; GFX802-GISEL-NEXT:    s_mov_b32 s5, 0x40400000
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
+; GFX802-GISEL-NEXT:    s_mov_b32 m0, s4
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v3, s1
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v2, s0
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; GFX802-GISEL-NEXT:    v_writelane_b32 v0, 0, s4
+; GFX802-GISEL-NEXT:    v_writelane_b32 v1, s5, m0
+; GFX802-GISEL-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; GFX802-GISEL-NEXT:    s_endpgm
+;
+; GFX1010-GISEL-LABEL: test_writelane_imm_sreg_f64:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_clause 0x1
+; GFX1010-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX1010-GISEL-NEXT:    s_load_dword s6, s[4:5], 0x8
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; GFX1010-GISEL-NEXT:    s_mov_b32 s2, 0x40400000
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v0, 0, s6
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v1, s2, s6
+; GFX1010-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX1010-GISEL-NEXT:    s_endpgm
+;
+; GFX1100-GISEL-LABEL: test_writelane_imm_sreg_f64:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_clause 0x1
+; GFX1100-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x0
+; GFX1100-GISEL-NEXT:    s_load_b32 s4, s[0:1], 0x8
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    s_load_b64 s[0:1], s[2:3], 0x0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1100-GISEL-NEXT:    s_mov_b32 s0, 0x40400000
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v0, 0, s4
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v1, s0, s4
+; GFX1100-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[2:3]
+; GFX1100-GISEL-NEXT:    s_nop 0
+; GFX1100-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-GISEL-NEXT:    s_endpgm
+  %oldval = load double, ptr addrspace(1) %out
+  %writelane = call double @llvm.amdgcn.writelane.f64(double 32.0, i32 %src1, double %oldval)
+  store double %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define amdgpu_kernel void @test_writelane_vreg_lane_i32(ptr addrspace(1) %out, ptr addrspace(1) %in) #1 {
+; GFX802-SDAG-LABEL: test_writelane_vreg_lane_i32:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-SDAG-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX802-SDAG-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
+; GFX802-SDAG-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; GFX802-SDAG-NEXT:    v_add_u32_e32 v0, vcc, 4, v0
+; GFX802-SDAG-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; GFX802-SDAG-NEXT:    flat_load_dword v0, v[0:1]
+; GFX802-SDAG-NEXT:    s_load_dword s2, s[0:1], 0x0
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v2, s2
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; GFX802-SDAG-NEXT:    s_nop 2
+; GFX802-SDAG-NEXT:    v_writelane_b32 v2, 12, s2
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v1, s1
+; GFX802-SDAG-NEXT:    flat_store_dword v[0:1], v2
+; GFX802-SDAG-NEXT:    s_endpgm
+;
+; GFX1010-SDAG-LABEL: test_writelane_vreg_lane_i32:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-SDAG-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    global_load_dword v0, v0, s[2:3] offset:4
+; GFX1010-SDAG-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX1010-SDAG-NEXT:    s_load_dword s2, s[0:1], 0x0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v1, 12, s2
+; GFX1010-SDAG-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX1010-SDAG-NEXT:    s_endpgm
+;
+; GFX1100-SDAG-LABEL: test_writelane_vreg_lane_i32:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
+; GFX1100-SDAG-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    global_load_b32 v0, v0, s[2:3] offset:4
+; GFX1100-SDAG-NEXT:    s_load_b32 s2, s[0:1], 0x0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1100-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v1, 12, s2
+; GFX1100-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1100-SDAG-NEXT:    s_nop 0
+; GFX1100-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-SDAG-NEXT:    s_endpgm
+;
+; GFX802-GISEL-LABEL: test_writelane_vreg_lane_i32:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-GISEL-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; GFX802-GISEL-NEXT:    v_add_u32_e32 v0, vcc, v0, v2
+; GFX802-GISEL-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; GFX802-GISEL-NEXT:    v_add_u32_e32 v0, vcc, 4, v0
+; GFX802-GISEL-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; GFX802-GISEL-NEXT:    flat_load_dword v0, v[0:1]
+; GFX802-GISEL-NEXT:    s_load_dword s2, s[0:1], 0x0
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v2, s2
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; GFX802-GISEL-NEXT:    s_nop 2
+; GFX802-GISEL-NEXT:    v_writelane_b32 v2, 12, s2
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; GFX802-GISEL-NEXT:    flat_store_dword v[0:1], v2
+; GFX802-GISEL-NEXT:    s_endpgm
+;
+; GFX1010-GISEL-LABEL: test_writelane_vreg_lane_i32:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-GISEL-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    global_load_dword v0, v0, s[2:3] offset:4
+; GFX1010-GISEL-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX1010-GISEL-NEXT:    s_load_dword s2, s[0:1], 0x0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v1, 12, s2
+; GFX1010-GISEL-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX1010-GISEL-NEXT:    s_endpgm
+;
+; GFX1100-GISEL-LABEL: test_writelane_vreg_lane_i32:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
+; GFX1100-GISEL-NEXT:    v_lshlrev_b32_e32 v0, 3, v0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    global_load_b32 v0, v0, s[2:3] offset:4
+; GFX1100-GISEL-NEXT:    s_load_b32 s2, s[0:1], 0x0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1100-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v1, 12, s2
+; GFX1100-GISEL-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX1100-GISEL-NEXT:    s_nop 0
+; GFX1100-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-GISEL-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep.in = getelementptr <2 x i32>, ptr addrspace(1) %in, i32 %tid
   %args = load <2 x i32>, ptr addrspace(1) %gep.in
   %oldval = load i32, ptr addrspace(1) %out
   %lane = extractelement <2 x i32> %args, i32 1
-  %writelane = call i32 @llvm.amdgcn.writelane(i32 12, i32 %lane, i32 %oldval)
+  %writelane = call i32 @llvm.amdgcn.writelane.i32(i32 12, i32 %lane, i32 %oldval)
   store i32 %writelane, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; CHECK-LABEL: {{^}}test_writelane_m0_sreg:
-; CHECK: s_mov_b32 m0, -1
-; CIGFX9: s_mov_b32 [[COPY_M0:s[0-9]+]], m0
-; CIGFX9: v_writelane_b32 v{{[0-9]+}}, [[COPY_M0]], m0
-; GFX10: v_writelane_b32 v{{[0-9]+}}, m0, s{{[0-9]+}}
-define amdgpu_kernel void @test_writelane_m0_sreg(ptr addrspace(1) %out, i32 %src1) #1 {
+define amdgpu_kernel void @test_writelane_vreg_lane_i64(ptr addrspace(1) %out, ptr addrspace(1) %in) #1 {
+; GFX802-SDAG-LABEL: test_writelane_vreg_lane_i64:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-SDAG-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX802-SDAG-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
+; GFX802-SDAG-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; GFX802-SDAG-NEXT:    v_add_u32_e32 v0, vcc, 8, v0
+; GFX802-SDAG-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; GFX802-SDAG-NEXT:    flat_load_dword v2, v[0:1]
+; GFX802-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s2, v2
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v3, s1
+; GFX802-SDAG-NEXT:    s_nop 2
+; GFX802-SDAG-NEXT:    v_writelane_b32 v1, 0, s2
+; GFX802-SDAG-NEXT:    v_writelane_b32 v0, 12, s2
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; GFX802-SDAG-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; GFX802-SDAG-NEXT:    s_endpgm
+;
+; GFX1010-SDAG-LABEL: test_writelane_vreg_lane_i64:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-SDAG-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    global_load_dword v0, v0, s[2:3] offset:8
+; GFX1010-SDAG-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX1010-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s3, v0
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v1, 0, s3
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v0, 12, s3
+; GFX1010-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX1010-SDAG-NEXT:    s_endpgm
+;
+; GFX1100-SDAG-LABEL: test_writelane_vreg_lane_i64:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
+; GFX1100-SDAG-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    global_load_b32 v0, v0, s[2:3] offset:8
+; GFX1100-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s3, v0
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1100-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v1, 0, s3
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v0, 12, s3
+; GFX1100-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX1100-SDAG-NEXT:    s_nop 0
+; GFX1100-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-SDAG-NEXT:    s_endpgm
+;
+; GFX802-GISEL-LABEL: test_writelane_vreg_lane_i64:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-GISEL-NEXT:    v_lshlrev_b32_e32 v2, 4, v0
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; GFX802-GISEL-NEXT:    v_add_u32_e32 v0, vcc, v0, v2
+; GFX802-GISEL-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; GFX802-GISEL-NEXT:    v_add_u32_e32 v0, vcc, 8, v0
+; GFX802-GISEL-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; GFX802-GISEL-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; GFX802-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v4, s1
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v3, s0
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v1, s2
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v2, s3
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX802-GISEL-NEXT:    s_nop 3
+; GFX802-GISEL-NEXT:    v_writelane_b32 v1, 12, s2
+; GFX802-GISEL-NEXT:    v_writelane_b32 v2, 0, s2
+; GFX802-GISEL-NEXT:    flat_store_dwordx2 v[3:4], v[1:2]
+; GFX802-GISEL-NEXT:    s_endpgm
+;
+; GFX1010-GISEL-LABEL: test_writelane_vreg_lane_i64:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-GISEL-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    global_load_dwordx2 v[0:1], v0, s[2:3] offset:8
+; GFX1010-GISEL-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX1010-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
+; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v1, 12, s2
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v2, 0, s2
+; GFX1010-GISEL-NEXT:    global_store_dwordx2 v0, v[1:2], s[0:1]
+; GFX1010-GISEL-NEXT:    s_endpgm
+;
+; GFX1100-GISEL-LABEL: test_writelane_vreg_lane_i64:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
+; GFX1100-GISEL-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    global_load_b64 v[0:1], v0, s[2:3] offset:8
+; GFX1100-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x0
+; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1100-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v1, 12, s2
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v2, 0, s2
+; GFX1100-GISEL-NEXT:    global_store_b64 v0, v[1:2], s[0:1]
+; GFX1100-GISEL-NEXT:    s_nop 0
+; GFX1100-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-GISEL-NEXT:    s_endpgm
+  %tid = call i32 @llvm.amdgcn.workitem.id.x()
+  %gep.in = getelementptr <2 x i64>, ptr addrspace(1) %in, i32 %tid
+  %args = load <2 x i64>, ptr addrspace(1) %gep.in
+  %oldval = load i64, ptr addrspace(1) %out
+  %lane = extractelement <2 x i64> %args, i32 1
+  %lane32 = trunc i64 %lane to i32
+  %writelane = call i64 @llvm.amdgcn.writelane.i64(i64 12, i32 %lane32, i64 %oldval)
+  store i64 %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define amdgpu_kernel void @test_writelane_vreg_lane_f64(ptr addrspace(1) %out, ptr addrspace(1) %in) #1 {
+; GFX802-SDAG-LABEL: test_writelane_vreg_lane_f64:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-SDAG-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
+; GFX802-SDAG-NEXT:    s_mov_b32 s4, 0x40280000
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX802-SDAG-NEXT:    v_add_u32_e32 v0, vcc, s2, v0
+; GFX802-SDAG-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; GFX802-SDAG-NEXT:    v_add_u32_e32 v0, vcc, 8, v0
+; GFX802-SDAG-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; GFX802-SDAG-NEXT:    flat_load_dword v2, v[0:1]
+; GFX802-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 m0, v2
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s2, v2
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v3, s1
+; GFX802-SDAG-NEXT:    s_nop 1
+; GFX802-SDAG-NEXT:    v_writelane_b32 v1, s4, m0
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v0, 0, s2
+; GFX802-SDAG-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; GFX802-SDAG-NEXT:    s_endpgm
+;
+; GFX1010-SDAG-LABEL: test_writelane_vreg_lane_f64:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-SDAG-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    global_load_dword v0, v0, s[2:3] offset:8
+; GFX1010-SDAG-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX1010-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s3, v0
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1010-SDAG-NEXT:    s_mov_b32 s2, 0x40280000
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v1, s2, s3
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v0, 0, s3
+; GFX1010-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX1010-SDAG-NEXT:    s_endpgm
+;
+; GFX1100-SDAG-LABEL: test_writelane_vreg_lane_f64:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
+; GFX1100-SDAG-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    global_load_b32 v0, v0, s[2:3] offset:8
+; GFX1100-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s3, v0
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX1100-SDAG-NEXT:    s_mov_b32 s2, 0x40280000
+; GFX1100-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instid1(SALU_CYCLE_1)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v1, s2, s3
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v0, 0, s3
+; GFX1100-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX1100-SDAG-NEXT:    s_nop 0
+; GFX1100-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-SDAG-NEXT:    s_endpgm
+;
+; GFX802-GISEL-LABEL: test_writelane_vreg_lane_f64:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-GISEL-NEXT:    v_lshlrev_b32_e32 v2, 4, v0
+; GFX802-GISEL-NEXT:    s_mov_b32 s4, 0x40280000
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; GFX802-GISEL-NEXT:    v_add_u32_e32 v0, vcc, v0, v2
+; GFX802-GISEL-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; GFX802-GISEL-NEXT:    v_add_u32_e32 v0, vcc, 8, v0
+; GFX802-GISEL-NEXT:    v_addc_u32_e32 v1, vcc, 0, v1, vcc
+; GFX802-GISEL-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
+; GFX802-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v4, s1
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v3, s0
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v1, s2
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v2, s3
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX802-GISEL-NEXT:    s_mov_b32 m0, s2
+; GFX802-GISEL-NEXT:    s_nop 2
+; GFX802-GISEL-NEXT:    v_writelane_b32 v1, 0, s2
+; GFX802-GISEL-NEXT:    v_writelane_b32 v2, s4, m0
+; GFX802-GISEL-NEXT:    flat_store_dwordx2 v[3:4], v[1:2]
+; GFX802-GISEL-NEXT:    s_endpgm
+;
+; GFX1010-GISEL-LABEL: test_writelane_vreg_lane_f64:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-GISEL-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    global_load_dwordx2 v[0:1], v0, s[2:3] offset:8
+; GFX1010-GISEL-NEXT:    s_waitcnt_depctr 0xffe3
+; GFX1010-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
+; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1010-GISEL-NEXT:    s_mov_b32 s3, 0x40280000
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v1, 0, s2
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v2, s3, s2
+; GFX1010-GISEL-NEXT:    global_store_dwordx2 v0, v[1:2], s[0:1]
+; GFX1010-GISEL-NEXT:    s_endpgm
+;
+; GFX1100-GISEL-LABEL: test_writelane_vreg_lane_f64:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
+; GFX1100-GISEL-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    global_load_b64 v[0:1], v0, s[2:3] offset:8
+; GFX1100-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x0
+; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v1, s2
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v2, s3
+; GFX1100-GISEL-NEXT:    s_mov_b32 s3, 0x40280000
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1100-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v1, 0, s2
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v2, s3, s2
+; GFX1100-GISEL-NEXT:    global_store_b64 v0, v[1:2], s[0:1]
+; GFX1100-GISEL-NEXT:    s_nop 0
+; GFX1100-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-GISEL-NEXT:    s_endpgm
+  %tid = call i32 @llvm.amdgcn.workitem.id.x()
+  %gep.in = getelementptr <2 x double>, ptr addrspace(1) %in, i32 %tid
+  %args = load <2 x double>, ptr addrspace(1) %gep.in
+  %oldval = load double, ptr addrspace(1) %out
+  %lane = extractelement <2 x double> %args, i32 1
+  %lane_cast = bitcast double %lane to i64
+  %lane32 = trunc i64 %lane_cast to i32
+  %writelane = call double @llvm.amdgcn.writelane.f64(double 12.0, i32 %lane32, double %oldval)
+  store double %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define amdgpu_kernel void @test_writelane_m0_sreg_i32(ptr addrspace(1) %out, i32 %src1) #1 {
+; GFX802-SDAG-LABEL: test_writelane_m0_sreg_i32:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX802-SDAG-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX802-SDAG-NEXT:    ;;#ASMSTART
+; GFX802-SDAG-NEXT:    s_mov_b32 m0, -1
+; GFX802-SDAG-NEXT:    ;;#ASMEND
+; GFX802-SDAG-NEXT:    s_mov_b32 s4, m0
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    s_load_dword s3, s[0:1], 0x0
+; GFX802-SDAG-NEXT:    s_mov_b32 m0, s2
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v1, s1
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v2, s3
+; GFX802-SDAG-NEXT:    v_writelane_b32 v2, s4, m0
+; GFX802-SDAG-NEXT:    flat_store_dword v[0:1], v2
+; GFX802-SDAG-NEXT:    s_endpgm
+;
+; GFX1010-SDAG-LABEL: test_writelane_m0_sreg_i32:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_clause 0x1
+; GFX1010-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX1010-SDAG-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1010-SDAG-NEXT:    ;;#ASMSTART
+; GFX1010-SDAG-NEXT:    s_mov_b32 m0, -1
+; GFX1010-SDAG-NEXT:    ;;#ASMEND
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    s_load_dword s3, s[0:1], 0x0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v0, s3
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v0, m0, s2
+; GFX1010-SDAG-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX1010-SDAG-NEXT:    s_endpgm
+;
+; GFX1100-SDAG-LABEL: test_writelane_m0_sreg_i32:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_clause 0x1
+; GFX1100-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x0
+; GFX1100-SDAG-NEXT:    s_load_b32 s0, s[0:1], 0x8
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1100-SDAG-NEXT:    ;;#ASMSTART
+; GFX1100-SDAG-NEXT:    s_mov_b32 m0, -1
+; GFX1100-SDAG-NEXT:    ;;#ASMEND
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    s_load_b32 s1, s[2:3], 0x0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v0, m0, s0
+; GFX1100-SDAG-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX1100-SDAG-NEXT:    s_nop 0
+; GFX1100-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-SDAG-NEXT:    s_endpgm
+;
+; GFX802-GISEL-LABEL: test_writelane_m0_sreg_i32:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX802-GISEL-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX802-GISEL-NEXT:    ;;#ASMSTART
+; GFX802-GISEL-NEXT:    s_mov_b32 m0, -1
+; GFX802-GISEL-NEXT:    ;;#ASMEND
+; GFX802-GISEL-NEXT:    s_mov_b32 s4, m0
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    s_load_dword s3, s[0:1], 0x0
+; GFX802-GISEL-NEXT:    s_mov_b32 m0, s2
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v2, s3
+; GFX802-GISEL-NEXT:    v_writelane_b32 v2, s4, m0
+; GFX802-GISEL-NEXT:    flat_store_dword v[0:1], v2
+; GFX802-GISEL-NEXT:    s_endpgm
+;
+; GFX1010-GISEL-LABEL: test_writelane_m0_sreg_i32:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_clause 0x1
+; GFX1010-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX1010-GISEL-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX1010-GISEL-NEXT:    ;;#ASMSTART
+; GFX1010-GISEL-NEXT:    s_mov_b32 m0, -1
+; GFX1010-GISEL-NEXT:    ;;#ASMEND
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    s_load_dword s3, s[0:1], 0x0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v0, s3
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v0, m0, s2
+; GFX1010-GISEL-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX1010-GISEL-NEXT:    s_endpgm
+;
+; GFX1100-GISEL-LABEL: test_writelane_m0_sreg_i32:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_clause 0x1
+; GFX1100-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x0
+; GFX1100-GISEL-NEXT:    s_load_b32 s0, s[0:1], 0x8
+; GFX1100-GISEL-NEXT:    ;;#ASMSTART
+; GFX1100-GISEL-NEXT:    s_mov_b32 m0, -1
+; GFX1100-GISEL-NEXT:    ;;#ASMEND
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    s_load_b32 s1, s[2:3], 0x0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v0, m0, s0
+; GFX1100-GISEL-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX1100-GISEL-NEXT:    s_nop 0
+; GFX1100-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-GISEL-NEXT:    s_endpgm
   %oldval = load i32, ptr addrspace(1) %out
   %m0 = call i32 asm "s_mov_b32 m0, -1", "={m0}"()
-  %writelane = call i32 @llvm.amdgcn.writelane(i32 %m0, i32 %src1, i32 %oldval)
+  %writelane = call i32 @llvm.amdgcn.writelane.i32(i32 %m0, i32 %src1, i32 %oldval)
   store i32 %writelane, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; CHECK-LABEL: {{^}}test_writelane_imm:
-; CHECK: v_writelane_b32 v{{[0-9]+}}, s{{[0-9]+}}, 32
-define amdgpu_kernel void @test_writelane_imm(ptr addrspace(1) %out, i32 %src0) #1 {
+define amdgpu_kernel void @test_writelane_imm_i32(ptr addrspace(1) %out, i32 %src0) #1 {
+; GFX802-SDAG-LABEL: test_writelane_imm_i32:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX802-SDAG-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    s_load_dword s3, s[0:1], 0x0
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v1, s1
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v2, s3
+; GFX802-SDAG-NEXT:    v_writelane_b32 v2, s2, 32
+; GFX802-SDAG-NEXT:    flat_store_dword v[0:1], v2
+; GFX802-SDAG-NEXT:    s_endpgm
+;
+; GFX1010-SDAG-LABEL: test_writelane_imm_i32:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_clause 0x1
+; GFX1010-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX1010-SDAG-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    s_load_dword s3, s[0:1], 0x0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v0, s3
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v0, s2, 32
+; GFX1010-SDAG-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX1010-SDAG-NEXT:    s_endpgm
+;
+; GFX1100-SDAG-LABEL: test_writelane_imm_i32:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_clause 0x1
+; GFX1100-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x0
+; GFX1100-SDAG-NEXT:    s_load_b32 s0, s[0:1], 0x8
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    s_load_b32 s1, s[2:3], 0x0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v0, s0, 32
+; GFX1100-SDAG-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX1100-SDAG-NEXT:    s_nop 0
+; GFX1100-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-SDAG-NEXT:    s_endpgm
+;
+; GFX802-GISEL-LABEL: test_writelane_imm_i32:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX802-GISEL-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    s_load_dword s3, s[0:1], 0x0
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v2, s3
+; GFX802-GISEL-NEXT:    v_writelane_b32 v2, s2, 32
+; GFX802-GISEL-NEXT:    flat_store_dword v[0:1], v2
+; GFX802-GISEL-NEXT:    s_endpgm
+;
+; GFX1010-GISEL-LABEL: test_writelane_imm_i32:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_clause 0x1
+; GFX1010-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX1010-GISEL-NEXT:    s_load_dword s2, s[4:5], 0x8
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    s_load_dword s3, s[0:1], 0x0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v0, s3
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v0, s2, 32
+; GFX1010-GISEL-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX1010-GISEL-NEXT:    s_endpgm
+;
+; GFX1100-GISEL-LABEL: test_writelane_imm_i32:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_clause 0x1
+; GFX1100-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x0
+; GFX1100-GISEL-NEXT:    s_load_b32 s0, s[0:1], 0x8
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    s_load_b32 s1, s[2:3], 0x0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v0, s1
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v0, s0, 32
+; GFX1100-GISEL-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX1100-GISEL-NEXT:    s_nop 0
+; GFX1100-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-GISEL-NEXT:    s_endpgm
   %oldval = load i32, ptr addrspace(1) %out
-  %writelane = call i32 @llvm.amdgcn.writelane(i32 %src0, i32 32, i32 %oldval) #0
+  %writelane = call i32 @llvm.amdgcn.writelane.i32(i32 %src0, i32 32, i32 %oldval) #0
   store i32 %writelane, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; CHECK-LABEL: {{^}}test_writelane_sreg_oldval:
-; CHECK: v_mov_b32_e32 [[OLDVAL:v[0-9]+]], s{{[0-9]+}}
-; CIGFX9: v_writelane_b32 [[OLDVAL]], s{{[0-9]+}}, m0
-; GFX10: v_writelane_b32 [[OLDVAL]], s{{[0-9]+}}, s{{[0-9]+}}
-define amdgpu_kernel void @test_writelane_sreg_oldval(i32 inreg %oldval, ptr addrspace(1) %out, i32 %src0, i32 %src1) #1 {
-  %writelane = call i32 @llvm.amdgcn.writelane(i32 %src0, i32 %src1, i32 %oldval)
+define amdgpu_kernel void @test_writelane_imm_i64(ptr addrspace(1) %out, i64 %src0) #1 {
+; GFX802-SDAG-LABEL: test_writelane_imm_i64:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v3, s1
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v1, s5
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v0, s4
+; GFX802-SDAG-NEXT:    v_writelane_b32 v1, s3, 32
+; GFX802-SDAG-NEXT:    v_writelane_b32 v0, s2, 32
+; GFX802-SDAG-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; GFX802-SDAG-NEXT:    s_endpgm
+;
+; GFX1010-SDAG-LABEL: test_writelane_imm_i64:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v1, s5
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v0, s4
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v1, s3, 32
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v0, s2, 32
+; GFX1010-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX1010-SDAG-NEXT:    s_endpgm
+;
+; GFX1100-SDAG-LABEL: test_writelane_imm_i64:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    s_load_b64 s[4:5], s[0:1], 0x0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v1, s5
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v0, s4
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v1, s3, 32
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v0, s2, 32
+; GFX1100-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX1100-SDAG-NEXT:    s_nop 0
+; GFX1100-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-SDAG-NEXT:    s_endpgm
+;
+; GFX802-GISEL-LABEL: test_writelane_imm_i64:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v3, s1
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v2, s0
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v0, s4
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v1, s5
+; GFX802-GISEL-NEXT:    v_writelane_b32 v0, s2, 32
+; GFX802-GISEL-NEXT:    v_writelane_b32 v1, s3, 32
+; GFX802-GISEL-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; GFX802-GISEL-NEXT:    s_endpgm
+;
+; GFX1010-GISEL-LABEL: test_writelane_imm_i64:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v0, s4
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v1, s5
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v0, s2, 32
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v1, s3, 32
+; GFX1010-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX1010-GISEL-NEXT:    s_endpgm
+;
+; GFX1100-GISEL-LABEL: test_writelane_imm_i64:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    s_load_b64 s[4:5], s[0:1], 0x0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v0, s4
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v1, s5
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v0, s2, 32
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v1, s3, 32
+; GFX1100-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX1100-GISEL-NEXT:    s_nop 0
+; GFX1100-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-GISEL-NEXT:    s_endpgm
+  %oldval = load i64, ptr addrspace(1) %out
+  %writelane = call i64 @llvm.amdgcn.writelane.i64(i64 %src0, i32 32, i64 %oldval) #0
+  store i64 %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define amdgpu_kernel void @test_writelane_imm_f64(ptr addrspace(1) %out, double %src0) #1 {
+; GFX802-SDAG-LABEL: test_writelane_imm_f64:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v3, s1
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v1, s5
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v0, s4
+; GFX802-SDAG-NEXT:    v_writelane_b32 v1, s3, 32
+; GFX802-SDAG-NEXT:    v_writelane_b32 v0, s2, 32
+; GFX802-SDAG-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; GFX802-SDAG-NEXT:    s_endpgm
+;
+; GFX1010-SDAG-LABEL: test_writelane_imm_f64:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v1, s5
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v0, s4
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v1, s3, 32
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v0, s2, 32
+; GFX1010-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX1010-SDAG-NEXT:    s_endpgm
+;
+; GFX1100-SDAG-LABEL: test_writelane_imm_f64:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    s_load_b64 s[4:5], s[0:1], 0x0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v1, s5
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v0, s4
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v1, s3, 32
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v0, s2, 32
+; GFX1100-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX1100-SDAG-NEXT:    s_nop 0
+; GFX1100-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-SDAG-NEXT:    s_endpgm
+;
+; GFX802-GISEL-LABEL: test_writelane_imm_f64:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v3, s1
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v2, s0
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v0, s4
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v1, s5
+; GFX802-GISEL-NEXT:    v_writelane_b32 v0, s2, 32
+; GFX802-GISEL-NEXT:    v_writelane_b32 v1, s3, 32
+; GFX802-GISEL-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; GFX802-GISEL-NEXT:    s_endpgm
+;
+; GFX1010-GISEL-LABEL: test_writelane_imm_f64:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v0, s4
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v1, s5
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v0, s2, 32
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v1, s3, 32
+; GFX1010-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX1010-GISEL-NEXT:    s_endpgm
+;
+; GFX1100-GISEL-LABEL: test_writelane_imm_f64:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    s_load_b64 s[4:5], s[0:1], 0x0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v0, s4
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v1, s5
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v0, s2, 32
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v1, s3, 32
+; GFX1100-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX1100-GISEL-NEXT:    s_nop 0
+; GFX1100-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-GISEL-NEXT:    s_endpgm
+  %oldval = load double, ptr addrspace(1) %out
+  %writelane = call double @llvm.amdgcn.writelane.f64(double %src0, i32 32, double %oldval) #0
+  store double %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define amdgpu_kernel void @test_writelane_sreg_oldval_i32(i32 inreg %oldval, ptr addrspace(1) %out, i32 %src0, i32 %src1) #1 {
+; GFX802-SDAG-LABEL: test_writelane_sreg_oldval_i32:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_load_dword s6, s[4:5], 0x0
+; GFX802-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x8
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v2, s6
+; GFX802-SDAG-NEXT:    s_mov_b32 m0, s3
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v2, s2, m0
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v1, s1
+; GFX802-SDAG-NEXT:    flat_store_dword v[0:1], v2
+; GFX802-SDAG-NEXT:    s_endpgm
+;
+; GFX1010-SDAG-LABEL: test_writelane_sreg_oldval_i32:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_clause 0x1
+; GFX1010-SDAG-NEXT:    s_load_dword s6, s[4:5], 0x0
+; GFX1010-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x8
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v0, s2, s3
+; GFX1010-SDAG-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX1010-SDAG-NEXT:    s_endpgm
+;
+; GFX1100-SDAG-LABEL: test_writelane_sreg_oldval_i32:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_clause 0x1
+; GFX1100-SDAG-NEXT:    s_load_b32 s4, s[0:1], 0x0
+; GFX1100-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x8
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v0, s4
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v0, s2, s3
+; GFX1100-SDAG-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX1100-SDAG-NEXT:    s_nop 0
+; GFX1100-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-SDAG-NEXT:    s_endpgm
+;
+; GFX802-GISEL-LABEL: test_writelane_sreg_oldval_i32:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_load_dword s6, s[4:5], 0x0
+; GFX802-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x8
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v2, s6
+; GFX802-GISEL-NEXT:    s_mov_b32 m0, s3
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; GFX802-GISEL-NEXT:    v_writelane_b32 v2, s2, m0
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; GFX802-GISEL-NEXT:    flat_store_dword v[0:1], v2
+; GFX802-GISEL-NEXT:    s_endpgm
+;
+; GFX1010-GISEL-LABEL: test_writelane_sreg_oldval_i32:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_clause 0x1
+; GFX1010-GISEL-NEXT:    s_load_dword s6, s[4:5], 0x0
+; GFX1010-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x8
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v0, s2, s3
+; GFX1010-GISEL-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX1010-GISEL-NEXT:    s_endpgm
+;
+; GFX1100-GISEL-LABEL: test_writelane_sreg_oldval_i32:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_clause 0x1
+; GFX1100-GISEL-NEXT:    s_load_b32 s4, s[0:1], 0x0
+; GFX1100-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x8
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v0, s4
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v0, s2, s3
+; GFX1100-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX1100-GISEL-NEXT:    s_nop 0
+; GFX1100-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-GISEL-NEXT:    s_endpgm
+  %writelane = call i32 @llvm.amdgcn.writelane.i32(i32 %src0, i32 %src1, i32 %oldval)
   store i32 %writelane, ptr addrspace(1) %out, align 4
   ret void
 }
 
-; CHECK-LABEL: {{^}}test_writelane_imm_oldval:
-; CHECK: v_mov_b32_e32 [[OLDVAL:v[0-9]+]], 42
-; CIGFX9: v_writelane_b32 [[OLDVAL]], s{{[0-9]+}}, m0
-; GFX10: v_writelane_b32 [[OLDVAL]], s{{[0-9]+}}, s{{[0-9]+}}
-define amdgpu_kernel void @test_writelane_imm_oldval(ptr addrspace(1) %out, i32 %src0, i32 %src1) #1 {
-  %writelane = call i32 @llvm.amdgcn.writelane(i32 %src0, i32 %src1, i32 42)
+define amdgpu_kernel void @test_writelane_sreg_oldval_i64(i64 inreg %oldval, ptr addrspace(1) %out, i64 %src0, i32 %src1) #1 {
+; GFX802-SDAG-LABEL: test_writelane_sreg_oldval_i64:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-SDAG-NEXT:    s_load_dword s6, s[4:5], 0x18
+; GFX802-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x10
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v3, s1
+; GFX802-SDAG-NEXT:    s_mov_b32 m0, s6
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX802-SDAG-NEXT:    v_writelane_b32 v3, s5, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v2, s4, m0
+; GFX802-SDAG-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; GFX802-SDAG-NEXT:    s_endpgm
+;
+; GFX1010-SDAG-LABEL: test_writelane_sreg_oldval_i64:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_clause 0x2
+; GFX1010-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX1010-SDAG-NEXT:    s_load_dword s8, s[4:5], 0x18
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v1, s7, s8
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v0, s6, s8
+; GFX1010-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX1010-SDAG-NEXT:    s_endpgm
+;
+; GFX1100-SDAG-LABEL: test_writelane_sreg_oldval_i64:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_clause 0x2
+; GFX1100-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x0
+; GFX1100-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x10
+; GFX1100-SDAG-NEXT:    s_load_b32 s0, s[0:1], 0x18
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v1, s5
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v0, s4
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v1, s3, s0
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v0, s2, s0
+; GFX1100-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[6:7]
+; GFX1100-SDAG-NEXT:    s_nop 0
+; GFX1100-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-SDAG-NEXT:    s_endpgm
+;
+; GFX802-GISEL-LABEL: test_writelane_sreg_oldval_i64:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-GISEL-NEXT:    s_load_dword s6, s[4:5], 0x18
+; GFX802-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x10
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; GFX802-GISEL-NEXT:    s_mov_b32 m0, s6
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v2, s2
+; GFX802-GISEL-NEXT:    v_writelane_b32 v0, s4, m0
+; GFX802-GISEL-NEXT:    v_writelane_b32 v1, s5, m0
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v3, s3
+; GFX802-GISEL-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; GFX802-GISEL-NEXT:    s_endpgm
+;
+; GFX1010-GISEL-LABEL: test_writelane_sreg_oldval_i64:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_clause 0x2
+; GFX1010-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX1010-GISEL-NEXT:    s_load_dword s8, s[4:5], 0x18
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v0, s6, s8
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v1, s7, s8
+; GFX1010-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX1010-GISEL-NEXT:    s_endpgm
+;
+; GFX1100-GISEL-LABEL: test_writelane_sreg_oldval_i64:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_clause 0x2
+; GFX1100-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x0
+; GFX1100-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x10
+; GFX1100-GISEL-NEXT:    s_load_b32 s0, s[0:1], 0x18
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v0, s4
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v1, s5
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v0, s2, s0
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v1, s3, s0
+; GFX1100-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[6:7]
+; GFX1100-GISEL-NEXT:    s_nop 0
+; GFX1100-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-GISEL-NEXT:    s_endpgm
+  %writelane = call i64 @llvm.amdgcn.writelane.i64(i64 %src0, i32 %src1, i64 %oldval)
+  store i64 %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define amdgpu_kernel void @test_writelane_sreg_oldval_f64(double inreg %oldval, ptr addrspace(1) %out, double %src0, i32 %src1) #1 {
+; GFX802-SDAG-LABEL: test_writelane_sreg_oldval_f64:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-SDAG-NEXT:    s_load_dword s6, s[4:5], 0x18
+; GFX802-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x10
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v3, s1
+; GFX802-SDAG-NEXT:    s_mov_b32 m0, s6
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX802-SDAG-NEXT:    v_writelane_b32 v3, s5, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v2, s4, m0
+; GFX802-SDAG-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
+; GFX802-SDAG-NEXT:    s_endpgm
+;
+; GFX1010-SDAG-LABEL: test_writelane_sreg_oldval_f64:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_clause 0x2
+; GFX1010-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX1010-SDAG-NEXT:    s_load_dword s8, s[4:5], 0x18
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v1, s7, s8
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v0, s6, s8
+; GFX1010-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX1010-SDAG-NEXT:    s_endpgm
+;
+; GFX1100-SDAG-LABEL: test_writelane_sreg_oldval_f64:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_clause 0x2
+; GFX1100-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x0
+; GFX1100-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x10
+; GFX1100-SDAG-NEXT:    s_load_b32 s0, s[0:1], 0x18
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v1, s5
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v0, s4
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v1, s3, s0
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v0, s2, s0
+; GFX1100-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[6:7]
+; GFX1100-SDAG-NEXT:    s_nop 0
+; GFX1100-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-SDAG-NEXT:    s_endpgm
+;
+; GFX802-GISEL-LABEL: test_writelane_sreg_oldval_f64:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-GISEL-NEXT:    s_load_dword s6, s[4:5], 0x18
+; GFX802-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x10
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; GFX802-GISEL-NEXT:    s_mov_b32 m0, s6
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v2, s2
+; GFX802-GISEL-NEXT:    v_writelane_b32 v0, s4, m0
+; GFX802-GISEL-NEXT:    v_writelane_b32 v1, s5, m0
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v3, s3
+; GFX802-GISEL-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; GFX802-GISEL-NEXT:    s_endpgm
+;
+; GFX1010-GISEL-LABEL: test_writelane_sreg_oldval_f64:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_clause 0x2
+; GFX1010-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x10
+; GFX1010-GISEL-NEXT:    s_load_dword s8, s[4:5], 0x18
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v0, s6, s8
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v1, s7, s8
+; GFX1010-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[2:3]
+; GFX1010-GISEL-NEXT:    s_endpgm
+;
+; GFX1100-GISEL-LABEL: test_writelane_sreg_oldval_f64:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_clause 0x2
+; GFX1100-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x0
+; GFX1100-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x10
+; GFX1100-GISEL-NEXT:    s_load_b32 s0, s[0:1], 0x18
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v0, s4
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v1, s5
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v0, s2, s0
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v1, s3, s0
+; GFX1100-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[6:7]
+; GFX1100-GISEL-NEXT:    s_nop 0
+; GFX1100-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-GISEL-NEXT:    s_endpgm
+  %writelane = call double @llvm.amdgcn.writelane.f64(double %src0, i32 %src1, double %oldval)
+  store double %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define amdgpu_kernel void @test_writelane_imm_oldval_i32(ptr addrspace(1) %out, i32 %src0, i32 %src1) #1 {
+; GFX802-SDAG-LABEL: test_writelane_imm_oldval_i32:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v2, 42
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    s_mov_b32 m0, s3
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v0, s0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v2, s2, m0
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v1, s1
+; GFX802-SDAG-NEXT:    flat_store_dword v[0:1], v2
+; GFX802-SDAG-NEXT:    s_endpgm
+;
+; GFX1010-SDAG-LABEL: test_writelane_imm_oldval_i32:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v0, 42
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v0, s2, s3
+; GFX1010-SDAG-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX1010-SDAG-NEXT:    s_endpgm
+;
+; GFX1100-SDAG-LABEL: test_writelane_imm_oldval_i32:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v0, 42
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v0, s2, s3
+; GFX1100-SDAG-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX1100-SDAG-NEXT:    s_nop 0
+; GFX1100-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-SDAG-NEXT:    s_endpgm
+;
+; GFX802-GISEL-LABEL: test_writelane_imm_oldval_i32:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v2, 42
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    s_mov_b32 m0, s3
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v0, s0
+; GFX802-GISEL-NEXT:    v_writelane_b32 v2, s2, m0
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v1, s1
+; GFX802-GISEL-NEXT:    flat_store_dword v[0:1], v2
+; GFX802-GISEL-NEXT:    s_endpgm
+;
+; GFX1010-GISEL-LABEL: test_writelane_imm_oldval_i32:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v0, 42
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v0, s2, s3
+; GFX1010-GISEL-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX1010-GISEL-NEXT:    s_endpgm
+;
+; GFX1100-GISEL-LABEL: test_writelane_imm_oldval_i32:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v0, 42
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v0, s2, s3
+; GFX1100-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX1100-GISEL-NEXT:    s_nop 0
+; GFX1100-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-GISEL-NEXT:    s_endpgm
+  %writelane = call i32 @llvm.amdgcn.writelane.i32(i32 %src0, i32 %src1, i32 42)
   store i32 %writelane, ptr addrspace(1) %out, align 4
   ret void
 }
 
+define amdgpu_kernel void @test_writelane_imm_oldval_i64(ptr addrspace(1) %out, i64 %src0, i32 %src1) #1 {
+; GFX802-SDAG-LABEL: test_writelane_imm_oldval_i64:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-SDAG-NEXT:    s_load_dword s4, s[4:5], 0x10
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v1, 0
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v0, 42
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; GFX802-SDAG-NEXT:    s_mov_b32 m0, s4
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v3, s1
+; GFX802-SDAG-NEXT:    v_writelane_b32 v1, s3, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v0, s2, m0
+; GFX802-SDAG-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; GFX802-SDAG-NEXT:    s_endpgm
+;
+; GFX1010-SDAG-LABEL: test_writelane_imm_oldval_i64:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_clause 0x1
+; GFX1010-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-SDAG-NEXT:    s_load_dword s6, s[4:5], 0x10
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v0, 42
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v1, s3, s6
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v0, s2, s6
+; GFX1010-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX1010-SDAG-NEXT:    s_endpgm
+;
+; GFX1100-SDAG-LABEL: test_writelane_imm_oldval_i64:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_clause 0x1
+; GFX1100-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x0
+; GFX1100-SDAG-NEXT:    s_load_b32 s0, s[0:1], 0x10
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v0, 42
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v1, s7, s0
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v0, s6, s0
+; GFX1100-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX1100-SDAG-NEXT:    s_nop 0
+; GFX1100-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-SDAG-NEXT:    s_endpgm
+;
+; GFX802-GISEL-LABEL: test_writelane_imm_oldval_i64:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_load_dword s6, s[4:5], 0x10
+; GFX802-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v0, 42
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    s_mov_b32 m0, s6
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v3, s1
+; GFX802-GISEL-NEXT:    v_writelane_b32 v0, s2, m0
+; GFX802-GISEL-NEXT:    v_writelane_b32 v1, s3, m0
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v2, s0
+; GFX802-GISEL-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; GFX802-GISEL-NEXT:    s_endpgm
+;
+; GFX1010-GISEL-LABEL: test_writelane_imm_oldval_i64:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_clause 0x1
+; GFX1010-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-GISEL-NEXT:    s_load_dword s6, s[4:5], 0x10
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v0, 42
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v0, s2, s6
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v1, s3, s6
+; GFX1010-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX1010-GISEL-NEXT:    s_endpgm
+;
+; GFX1100-GISEL-LABEL: test_writelane_imm_oldval_i64:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_clause 0x1
+; GFX1100-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x0
+; GFX1100-GISEL-NEXT:    s_load_b32 s0, s[0:1], 0x10
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v0, 42
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v0, s6, s0
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v1, s7, s0
+; GFX1100-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX1100-GISEL-NEXT:    s_nop 0
+; GFX1100-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-GISEL-NEXT:    s_endpgm
+  %writelane = call i64 @llvm.amdgcn.writelane.i64(i64 %src0, i32 %src1, i64 42)
+  store i64 %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define amdgpu_kernel void @test_writelane_imm_oldval_f64(ptr addrspace(1) %out, double %src0, i32 %src1) #1 {
+; GFX802-SDAG-LABEL: test_writelane_imm_oldval_f64:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-SDAG-NEXT:    s_load_dword s4, s[4:5], 0x10
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v1, 0x40450000
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX802-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v2, s0
+; GFX802-SDAG-NEXT:    s_mov_b32 m0, s4
+; GFX802-SDAG-NEXT:    v_mov_b32_e32 v3, s1
+; GFX802-SDAG-NEXT:    v_writelane_b32 v1, s3, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v0, s2, m0
+; GFX802-SDAG-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; GFX802-SDAG-NEXT:    s_endpgm
+;
+; GFX1010-SDAG-LABEL: test_writelane_imm_oldval_f64:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_clause 0x1
+; GFX1010-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-SDAG-NEXT:    s_load_dword s6, s[4:5], 0x10
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v1, 0x40450000
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1010-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1010-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v1, s3, s6
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v0, s2, s6
+; GFX1010-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX1010-SDAG-NEXT:    s_endpgm
+;
+; GFX1100-SDAG-LABEL: test_writelane_imm_oldval_f64:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_clause 0x1
+; GFX1100-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x0
+; GFX1100-SDAG-NEXT:    s_load_b32 s0, s[0:1], 0x10
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v1, 0x40450000
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1100-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1100-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v1, s7, s0
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v0, s6, s0
+; GFX1100-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX1100-SDAG-NEXT:    s_nop 0
+; GFX1100-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-SDAG-NEXT:    s_endpgm
+;
+; GFX802-GISEL-LABEL: test_writelane_imm_oldval_f64:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_load_dword s6, s[4:5], 0x10
+; GFX802-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v1, 0x40450000
+; GFX802-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX802-GISEL-NEXT:    s_mov_b32 m0, s6
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v3, s1
+; GFX802-GISEL-NEXT:    v_writelane_b32 v0, s2, m0
+; GFX802-GISEL-NEXT:    v_writelane_b32 v1, s3, m0
+; GFX802-GISEL-NEXT:    v_mov_b32_e32 v2, s0
+; GFX802-GISEL-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
+; GFX802-GISEL-NEXT:    s_endpgm
+;
+; GFX1010-GISEL-LABEL: test_writelane_imm_oldval_f64:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_clause 0x1
+; GFX1010-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
+; GFX1010-GISEL-NEXT:    s_load_dword s6, s[4:5], 0x10
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v1, 0x40450000
+; GFX1010-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1010-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v0, s2, s6
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v1, s3, s6
+; GFX1010-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX1010-GISEL-NEXT:    s_endpgm
+;
+; GFX1100-GISEL-LABEL: test_writelane_imm_oldval_f64:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_clause 0x1
+; GFX1100-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x0
+; GFX1100-GISEL-NEXT:    s_load_b32 s0, s[0:1], 0x10
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v1, 0x40450000
+; GFX1100-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX1100-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v0, s6, s0
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v1, s7, s0
+; GFX1100-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX1100-GISEL-NEXT:    s_nop 0
+; GFX1100-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX1100-GISEL-NEXT:    s_endpgm
+  %writelane = call double @llvm.amdgcn.writelane.f64(double %src0, i32 %src1, double 42.0)
+  store double %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define void @test_writelane_half(ptr addrspace(1) %out, half %src, i32 %src1) {
+; GFX802-SDAG-LABEL: test_writelane_half:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-SDAG-NEXT:    flat_load_ushort v4, v[0:1]
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 m0, v3
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_nop 1
+; GFX802-SDAG-NEXT:    v_writelane_b32 v4, s4, m0
+; GFX802-SDAG-NEXT:    flat_store_short v[0:1], v4
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-SDAG-LABEL: test_writelane_half:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    global_load_ushort v4, v[0:1], off
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v4, s4, s5
+; GFX1010-SDAG-NEXT:    global_store_short v[0:1], v4, off
+; GFX1010-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-SDAG-LABEL: test_writelane_half:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    global_load_u16 v4, v[0:1], off
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s0, v2
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s1, v3
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v4, s0, s1
+; GFX1100-SDAG-NEXT:    global_store_b16 v[0:1], v4, off
+; GFX1100-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX802-GISEL-LABEL: test_writelane_half:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-GISEL-NEXT:    flat_load_ushort v4, v[0:1]
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX802-GISEL-NEXT:    s_mov_b32 m0, s5
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-GISEL-NEXT:    v_writelane_b32 v4, s4, m0
+; GFX802-GISEL-NEXT:    flat_store_short v[0:1], v4
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-GISEL-LABEL: test_writelane_half:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    global_load_ushort v4, v[0:1], off
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v4, s4, s5
+; GFX1010-GISEL-NEXT:    global_store_short v[0:1], v4, off
+; GFX1010-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-GISEL-LABEL: test_writelane_half:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    global_load_u16 v4, v[0:1], off
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s0, v2
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s1, v3
+; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v4, s0, s1
+; GFX1100-GISEL-NEXT:    global_store_b16 v[0:1], v4, off
+; GFX1100-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %oldval = load half, ptr addrspace(1) %out
+  %writelane = call half @llvm.amdgcn.writelane.f16(half %src, i32 %src1, half %oldval)
+  store half %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define void @test_writelane_float(ptr addrspace(1) %out, float %src, i32 %src1) {
+; GFX802-SDAG-LABEL: test_writelane_float:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-SDAG-NEXT:    flat_load_dword v4, v[0:1]
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 m0, v3
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_nop 1
+; GFX802-SDAG-NEXT:    v_writelane_b32 v4, s4, m0
+; GFX802-SDAG-NEXT:    flat_store_dword v[0:1], v4
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-SDAG-LABEL: test_writelane_float:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    global_load_dword v4, v[0:1], off
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v4, s4, s5
+; GFX1010-SDAG-NEXT:    global_store_dword v[0:1], v4, off
+; GFX1010-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-SDAG-LABEL: test_writelane_float:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    global_load_b32 v4, v[0:1], off
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s0, v2
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s1, v3
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v4, s0, s1
+; GFX1100-SDAG-NEXT:    global_store_b32 v[0:1], v4, off
+; GFX1100-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX802-GISEL-LABEL: test_writelane_float:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-GISEL-NEXT:    flat_load_dword v4, v[0:1]
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX802-GISEL-NEXT:    s_mov_b32 m0, s5
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-GISEL-NEXT:    v_writelane_b32 v4, s4, m0
+; GFX802-GISEL-NEXT:    flat_store_dword v[0:1], v4
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-GISEL-LABEL: test_writelane_float:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    global_load_dword v4, v[0:1], off
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v4, s4, s5
+; GFX1010-GISEL-NEXT:    global_store_dword v[0:1], v4, off
+; GFX1010-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-GISEL-LABEL: test_writelane_float:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    global_load_b32 v4, v[0:1], off
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s0, v2
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s1, v3
+; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v4, s0, s1
+; GFX1100-GISEL-NEXT:    global_store_b32 v[0:1], v4, off
+; GFX1100-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %oldval = load float, ptr addrspace(1) %out
+  %writelane = call float @llvm.amdgcn.writelane.f32(float %src, i32 %src1, float %oldval)
+  store float %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define void @test_writelane_bfloat(ptr addrspace(1) %out, bfloat %src, i32 %src1) {
+; GFX802-SDAG-LABEL: test_writelane_bfloat:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-SDAG-NEXT:    flat_load_ushort v4, v[0:1]
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 m0, v3
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_nop 1
+; GFX802-SDAG-NEXT:    v_writelane_b32 v4, s4, m0
+; GFX802-SDAG-NEXT:    flat_store_short v[0:1], v4
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-SDAG-LABEL: test_writelane_bfloat:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    global_load_ushort v4, v[0:1], off
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v4, s4, s5
+; GFX1010-SDAG-NEXT:    global_store_short v[0:1], v4, off
+; GFX1010-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-SDAG-LABEL: test_writelane_bfloat:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    global_load_u16 v4, v[0:1], off
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s0, v2
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s1, v3
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v4, s0, s1
+; GFX1100-SDAG-NEXT:    global_store_b16 v[0:1], v4, off
+; GFX1100-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX802-GISEL-LABEL: test_writelane_bfloat:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-GISEL-NEXT:    flat_load_ushort v4, v[0:1]
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX802-GISEL-NEXT:    s_mov_b32 m0, s5
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-GISEL-NEXT:    v_writelane_b32 v4, s4, m0
+; GFX802-GISEL-NEXT:    flat_store_short v[0:1], v4
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-GISEL-LABEL: test_writelane_bfloat:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    global_load_ushort v4, v[0:1], off
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v4, s4, s5
+; GFX1010-GISEL-NEXT:    global_store_short v[0:1], v4, off
+; GFX1010-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-GISEL-LABEL: test_writelane_bfloat:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    global_load_u16 v4, v[0:1], off
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s0, v2
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s1, v3
+; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v4, s0, s1
+; GFX1100-GISEL-NEXT:    global_store_b16 v[0:1], v4, off
+; GFX1100-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %oldval = load bfloat, ptr addrspace(1) %out
+  %writelane = call bfloat @llvm.amdgcn.writelane.bf16(bfloat %src, i32 %src1, bfloat %oldval)
+  store bfloat %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define void @test_writelane_i16(ptr addrspace(1) %out, i16 %src, i32 %src1) {
+; GFX802-SDAG-LABEL: test_writelane_i16:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-SDAG-NEXT:    flat_load_ushort v4, v[0:1]
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 m0, v3
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_nop 1
+; GFX802-SDAG-NEXT:    v_writelane_b32 v4, s4, m0
+; GFX802-SDAG-NEXT:    flat_store_short v[0:1], v4
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-SDAG-LABEL: test_writelane_i16:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    global_load_ushort v4, v[0:1], off
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v4, s4, s5
+; GFX1010-SDAG-NEXT:    global_store_short v[0:1], v4, off
+; GFX1010-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-SDAG-LABEL: test_writelane_i16:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    global_load_u16 v4, v[0:1], off
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s0, v2
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s1, v3
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v4, s0, s1
+; GFX1100-SDAG-NEXT:    global_store_b16 v[0:1], v4, off
+; GFX1100-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX802-GISEL-LABEL: test_writelane_i16:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-GISEL-NEXT:    flat_load_ushort v4, v[0:1]
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX802-GISEL-NEXT:    s_mov_b32 m0, s5
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-GISEL-NEXT:    v_writelane_b32 v4, s4, m0
+; GFX802-GISEL-NEXT:    flat_store_short v[0:1], v4
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-GISEL-LABEL: test_writelane_i16:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    global_load_ushort v4, v[0:1], off
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v4, s4, s5
+; GFX1010-GISEL-NEXT:    global_store_short v[0:1], v4, off
+; GFX1010-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-GISEL-LABEL: test_writelane_i16:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    global_load_u16 v4, v[0:1], off
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s0, v2
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s1, v3
+; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v4, s0, s1
+; GFX1100-GISEL-NEXT:    global_store_b16 v[0:1], v4, off
+; GFX1100-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %oldval = load i16, ptr addrspace(1) %out
+  %writelane = call i16 @llvm.amdgcn.writelane.i16(i16 %src, i32 %src1, i16 %oldval)
+  store i16 %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define void @test_writelane_v2f16(ptr addrspace(1) %out, <2 x half> %src, i32 %src1) {
+; GFX802-SDAG-LABEL: test_writelane_v2f16:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-SDAG-NEXT:    flat_load_dword v4, v[0:1]
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 m0, v3
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_nop 1
+; GFX802-SDAG-NEXT:    v_writelane_b32 v4, s4, m0
+; GFX802-SDAG-NEXT:    flat_store_dword v[0:1], v4
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-SDAG-LABEL: test_writelane_v2f16:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    global_load_dword v4, v[0:1], off
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v4, s4, s5
+; GFX1010-SDAG-NEXT:    global_store_dword v[0:1], v4, off
+; GFX1010-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-SDAG-LABEL: test_writelane_v2f16:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    global_load_b32 v4, v[0:1], off
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s0, v2
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s1, v3
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v4, s0, s1
+; GFX1100-SDAG-NEXT:    global_store_b32 v[0:1], v4, off
+; GFX1100-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX802-GISEL-LABEL: test_writelane_v2f16:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-GISEL-NEXT:    flat_load_dword v4, v[0:1]
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX802-GISEL-NEXT:    s_mov_b32 m0, s5
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-GISEL-NEXT:    v_writelane_b32 v4, s4, m0
+; GFX802-GISEL-NEXT:    flat_store_dword v[0:1], v4
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-GISEL-LABEL: test_writelane_v2f16:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    global_load_dword v4, v[0:1], off
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v4, s4, s5
+; GFX1010-GISEL-NEXT:    global_store_dword v[0:1], v4, off
+; GFX1010-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-GISEL-LABEL: test_writelane_v2f16:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    global_load_b32 v4, v[0:1], off
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s0, v2
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s1, v3
+; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v4, s0, s1
+; GFX1100-GISEL-NEXT:    global_store_b32 v[0:1], v4, off
+; GFX1100-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %oldval = load <2 x half>, ptr addrspace(1) %out
+  %writelane = call <2 x half> @llvm.amdgcn.writelane.v2f16(<2 x half> %src, i32 %src1, <2 x half> %oldval)
+  store <2 x half> %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define void @test_readlane_v2f32(ptr addrspace(1) %out, <2 x float> %src, i32 %src1) {
+; GFX802-SDAG-LABEL: test_readlane_v2f32:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-SDAG-NEXT:    flat_load_dwordx2 v[5:6], v[0:1]
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 m0, v4
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s4, v3
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s5, v2
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_nop 0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v6, s4, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v5, s5, m0
+; GFX802-SDAG-NEXT:    flat_store_dwordx2 v[0:1], v[5:6]
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-SDAG-LABEL: test_readlane_v2f32:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    global_load_dwordx2 v[5:6], v[0:1], off
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s4, v3
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s5, v4
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s6, v2
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v6, s4, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v5, s6, s5
+; GFX1010-SDAG-NEXT:    global_store_dwordx2 v[0:1], v[5:6], off
+; GFX1010-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-SDAG-LABEL: test_readlane_v2f32:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    global_load_b64 v[5:6], v[0:1], off
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s2, v2
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v6, s0, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v5, s2, s1
+; GFX1100-SDAG-NEXT:    global_store_b64 v[0:1], v[5:6], off
+; GFX1100-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX802-GISEL-LABEL: test_readlane_v2f32:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-GISEL-NEXT:    flat_load_dwordx2 v[5:6], v[0:1]
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s5, v4
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s6, v3
+; GFX802-GISEL-NEXT:    s_mov_b32 m0, s5
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-GISEL-NEXT:    v_writelane_b32 v5, s4, m0
+; GFX802-GISEL-NEXT:    v_writelane_b32 v6, s6, m0
+; GFX802-GISEL-NEXT:    flat_store_dwordx2 v[0:1], v[5:6]
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-GISEL-LABEL: test_readlane_v2f32:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    global_load_dwordx2 v[5:6], v[0:1], off
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s5, v4
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s6, v3
+; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v5, s4, s5
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v6, s6, s5
+; GFX1010-GISEL-NEXT:    global_store_dwordx2 v[0:1], v[5:6], off
+; GFX1010-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-GISEL-LABEL: test_readlane_v2f32:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    global_load_b64 v[5:6], v[0:1], off
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s0, v2
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s2, v3
+; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v5, s0, s1
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v6, s2, s1
+; GFX1100-GISEL-NEXT:    global_store_b64 v[0:1], v[5:6], off
+; GFX1100-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %oldval = load <2 x float>, ptr addrspace(1) %out
+  %writelane = call <2 x float> @llvm.amdgcn.writelane.v2f32(<2 x float> %src, i32 %src1, <2 x float> %oldval)
+  store <2 x float> %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define void @test_writelane_v7i32(ptr addrspace(1) %out, <7 x i32> %src, i32 %src1) {
+; GFX802-SDAG-LABEL: test_writelane_v7i32:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-SDAG-NEXT:    v_add_u32_e32 v17, vcc, 16, v0
+; GFX802-SDAG-NEXT:    flat_load_dwordx4 v[10:13], v[0:1]
+; GFX802-SDAG-NEXT:    v_addc_u32_e32 v18, vcc, 0, v1, vcc
+; GFX802-SDAG-NEXT:    flat_load_dwordx3 v[14:16], v[17:18]
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 m0, v9
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s7, v5
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s8, v4
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s9, v3
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s10, v2
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s4, v8
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s5, v7
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s6, v6
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(1)
+; GFX802-SDAG-NEXT:    v_writelane_b32 v13, s7, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v12, s8, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v11, s9, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v10, s10, m0
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    v_writelane_b32 v16, s4, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v15, s5, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v14, s6, m0
+; GFX802-SDAG-NEXT:    flat_store_dwordx4 v[0:1], v[10:13]
+; GFX802-SDAG-NEXT:    flat_store_dwordx3 v[17:18], v[14:16]
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-SDAG-LABEL: test_writelane_v7i32:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    s_clause 0x1
+; GFX1010-SDAG-NEXT:    global_load_dwordx3 v[14:16], v[0:1], off offset:16
+; GFX1010-SDAG-NEXT:    global_load_dwordx4 v[10:13], v[0:1], off
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s5, v9
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s8, v5
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s9, v4
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s10, v3
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s11, v2
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s4, v8
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s6, v7
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s7, v6
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(1)
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v16, s4, s5
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v13, s8, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v12, s9, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v11, s10, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v10, s11, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v15, s6, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v14, s7, s5
+; GFX1010-SDAG-NEXT:    global_store_dwordx4 v[0:1], v[10:13], off
+; GFX1010-SDAG-NEXT:    global_store_dwordx3 v[0:1], v[14:16], off offset:16
+; GFX1010-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-SDAG-LABEL: test_writelane_v7i32:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    s_clause 0x1
+; GFX1100-SDAG-NEXT:    global_load_b96 v[14:16], v[0:1], off offset:16
+; GFX1100-SDAG-NEXT:    global_load_b128 v[10:13], v[0:1], off
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s1, v9
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s4, v5
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s5, v4
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s6, v3
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s7, v2
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s0, v8
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s2, v7
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s3, v6
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(1)
+; GFX1100-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_3)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v16, s0, s1
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v13, s4, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v12, s5, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v11, s6, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v10, s7, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v15, s2, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v14, s3, s1
+; GFX1100-SDAG-NEXT:    s_clause 0x1
+; GFX1100-SDAG-NEXT:    global_store_b128 v[0:1], v[10:13], off
+; GFX1100-SDAG-NEXT:    global_store_b96 v[0:1], v[14:16], off offset:16
+; GFX1100-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX802-GISEL-LABEL: test_writelane_v7i32:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-GISEL-NEXT:    v_add_u32_e32 v18, vcc, 16, v0
+; GFX802-GISEL-NEXT:    flat_load_dwordx4 v[10:13], v[0:1]
+; GFX802-GISEL-NEXT:    v_addc_u32_e32 v19, vcc, 0, v1, vcc
+; GFX802-GISEL-NEXT:    flat_load_dwordx4 v[14:17], v[18:19]
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s5, v9
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s6, v3
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s7, v4
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s8, v5
+; GFX802-GISEL-NEXT:    s_mov_b32 m0, s5
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s9, v6
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s10, v7
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s11, v8
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(1)
+; GFX802-GISEL-NEXT:    v_writelane_b32 v10, s4, m0
+; GFX802-GISEL-NEXT:    v_writelane_b32 v11, s6, m0
+; GFX802-GISEL-NEXT:    v_writelane_b32 v12, s7, m0
+; GFX802-GISEL-NEXT:    v_writelane_b32 v13, s8, m0
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-GISEL-NEXT:    v_writelane_b32 v14, s9, m0
+; GFX802-GISEL-NEXT:    v_writelane_b32 v15, s10, m0
+; GFX802-GISEL-NEXT:    v_writelane_b32 v16, s11, m0
+; GFX802-GISEL-NEXT:    flat_store_dwordx4 v[0:1], v[10:13]
+; GFX802-GISEL-NEXT:    flat_store_dwordx3 v[18:19], v[14:16]
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-GISEL-LABEL: test_writelane_v7i32:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    s_clause 0x1
+; GFX1010-GISEL-NEXT:    global_load_dwordx4 v[10:13], v[0:1], off
+; GFX1010-GISEL-NEXT:    global_load_dwordx4 v[14:17], v[0:1], off offset:16
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s5, v9
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s6, v3
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s7, v4
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s8, v5
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s9, v6
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s10, v7
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s11, v8
+; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(1)
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v10, s4, s5
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v11, s6, s5
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v12, s7, s5
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v13, s8, s5
+; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v14, s9, s5
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v15, s10, s5
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v16, s11, s5
+; GFX1010-GISEL-NEXT:    global_store_dwordx4 v[0:1], v[10:13], off
+; GFX1010-GISEL-NEXT:    global_store_dwordx3 v[0:1], v[14:16], off offset:16
+; GFX1010-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-GISEL-LABEL: test_writelane_v7i32:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    s_clause 0x1
+; GFX1100-GISEL-NEXT:    global_load_b128 v[10:13], v[0:1], off
+; GFX1100-GISEL-NEXT:    global_load_b128 v[14:17], v[0:1], off offset:16
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s0, v2
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s1, v9
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s2, v3
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s3, v4
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s4, v5
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s5, v6
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s6, v7
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s7, v8
+; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(1)
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v10, s0, s1
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v11, s2, s1
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v12, s3, s1
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v13, s4, s1
+; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v14, s5, s1
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v15, s6, s1
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v16, s7, s1
+; GFX1100-GISEL-NEXT:    s_clause 0x1
+; GFX1100-GISEL-NEXT:    global_store_b128 v[0:1], v[10:13], off
+; GFX1100-GISEL-NEXT:    global_store_b96 v[0:1], v[14:16], off offset:16
+; GFX1100-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %oldval = load <7 x i32>, ptr addrspace(1) %out
+  %writelane = call <7 x i32> @llvm.amdgcn.writelane.v7i32(<7 x i32> %src, i32 %src1, <7 x i32> %oldval)
+  store <7 x i32> %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define void @test_writelane_p0(ptr addrspace(1) %out, ptr %src, i32 %src1) {
+; GFX802-SDAG-LABEL: test_writelane_p0:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-SDAG-NEXT:    flat_load_dwordx2 v[5:6], v[0:1]
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 m0, v4
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s4, v3
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s5, v2
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_nop 0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v6, s4, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v5, s5, m0
+; GFX802-SDAG-NEXT:    flat_store_dwordx2 v[0:1], v[5:6]
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-SDAG-LABEL: test_writelane_p0:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    global_load_dwordx2 v[5:6], v[0:1], off
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s4, v3
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s5, v4
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s6, v2
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v6, s4, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v5, s6, s5
+; GFX1010-SDAG-NEXT:    global_store_dwordx2 v[0:1], v[5:6], off
+; GFX1010-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-SDAG-LABEL: test_writelane_p0:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    global_load_b64 v[5:6], v[0:1], off
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s2, v2
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v6, s0, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v5, s2, s1
+; GFX1100-SDAG-NEXT:    global_store_b64 v[0:1], v[5:6], off
+; GFX1100-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX802-GISEL-LABEL: test_writelane_p0:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-GISEL-NEXT:    flat_load_dwordx2 v[5:6], v[0:1]
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s5, v4
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s6, v3
+; GFX802-GISEL-NEXT:    s_mov_b32 m0, s5
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-GISEL-NEXT:    v_writelane_b32 v5, s4, m0
+; GFX802-GISEL-NEXT:    v_writelane_b32 v6, s6, m0
+; GFX802-GISEL-NEXT:    flat_store_dwordx2 v[0:1], v[5:6]
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-GISEL-LABEL: test_writelane_p0:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    global_load_dwordx2 v[5:6], v[0:1], off
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s5, v4
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s6, v3
+; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v5, s4, s5
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v6, s6, s5
+; GFX1010-GISEL-NEXT:    global_store_dwordx2 v[0:1], v[5:6], off
+; GFX1010-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-GISEL-LABEL: test_writelane_p0:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    global_load_b64 v[5:6], v[0:1], off
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s0, v2
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s2, v3
+; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v5, s0, s1
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v6, s2, s1
+; GFX1100-GISEL-NEXT:    global_store_b64 v[0:1], v[5:6], off
+; GFX1100-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %oldval = load ptr, ptr addrspace(1) %out
+  %writelane = call ptr @llvm.amdgcn.writelane.p0(ptr %src, i32 %src1, ptr %oldval)
+  store ptr %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define void @test_writelane_v3p0(ptr addrspace(1) %out, <4 x ptr> %src, i32 %src1) {
+; GFX802-SDAG-LABEL: test_writelane_v3p0:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-SDAG-NEXT:    v_add_u32_e32 v19, vcc, 16, v0
+; GFX802-SDAG-NEXT:    flat_load_dwordx4 v[11:14], v[0:1]
+; GFX802-SDAG-NEXT:    v_addc_u32_e32 v20, vcc, 0, v1, vcc
+; GFX802-SDAG-NEXT:    flat_load_dwordx4 v[15:18], v[19:20]
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 m0, v10
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s8, v5
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s9, v4
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s10, v3
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s11, v2
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s4, v9
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s5, v8
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s6, v7
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s7, v6
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(1)
+; GFX802-SDAG-NEXT:    v_writelane_b32 v14, s8, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v13, s9, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v12, s10, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v11, s11, m0
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    v_writelane_b32 v18, s4, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v17, s5, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v16, s6, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v15, s7, m0
+; GFX802-SDAG-NEXT:    flat_store_dwordx4 v[0:1], v[11:14]
+; GFX802-SDAG-NEXT:    flat_store_dwordx4 v[19:20], v[15:18]
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-SDAG-LABEL: test_writelane_v3p0:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    s_clause 0x1
+; GFX1010-SDAG-NEXT:    global_load_dwordx4 v[11:14], v[0:1], off offset:16
+; GFX1010-SDAG-NEXT:    global_load_dwordx4 v[15:18], v[0:1], off
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s5, v10
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s9, v5
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s10, v4
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s11, v3
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s12, v2
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s4, v9
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s6, v8
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s7, v7
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s8, v6
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(1)
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v14, s4, s5
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v18, s9, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v17, s10, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v16, s11, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v15, s12, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v13, s6, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v12, s7, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v11, s8, s5
+; GFX1010-SDAG-NEXT:    global_store_dwordx4 v[0:1], v[15:18], off
+; GFX1010-SDAG-NEXT:    global_store_dwordx4 v[0:1], v[11:14], off offset:16
+; GFX1010-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-SDAG-LABEL: test_writelane_v3p0:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    s_clause 0x1
+; GFX1100-SDAG-NEXT:    global_load_b128 v[11:14], v[0:1], off offset:16
+; GFX1100-SDAG-NEXT:    global_load_b128 v[15:18], v[0:1], off
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s1, v10
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s5, v5
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s6, v4
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s7, v3
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s8, v2
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s0, v9
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s2, v8
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s3, v7
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s4, v6
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(1)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v14, s0, s1
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v18, s5, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v17, s6, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v16, s7, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v15, s8, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v13, s2, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v12, s3, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v11, s4, s1
+; GFX1100-SDAG-NEXT:    s_clause 0x1
+; GFX1100-SDAG-NEXT:    global_store_b128 v[0:1], v[15:18], off
+; GFX1100-SDAG-NEXT:    global_store_b128 v[0:1], v[11:14], off offset:16
+; GFX1100-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX802-GISEL-LABEL: test_writelane_v3p0:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-GISEL-NEXT:    v_add_u32_e32 v19, vcc, 16, v0
+; GFX802-GISEL-NEXT:    flat_load_dwordx4 v[11:14], v[0:1]
+; GFX802-GISEL-NEXT:    v_addc_u32_e32 v20, vcc, 0, v1, vcc
+; GFX802-GISEL-NEXT:    flat_load_dwordx4 v[15:18], v[19:20]
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s5, v10
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s6, v3
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s7, v4
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s8, v5
+; GFX802-GISEL-NEXT:    s_mov_b32 m0, s5
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s9, v6
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s10, v7
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s11, v8
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s12, v9
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(1)
+; GFX802-GISEL-NEXT:    v_writelane_b32 v11, s4, m0
+; GFX802-GISEL-NEXT:    v_writelane_b32 v12, s6, m0
+; GFX802-GISEL-NEXT:    v_writelane_b32 v13, s7, m0
+; GFX802-GISEL-NEXT:    v_writelane_b32 v14, s8, m0
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-GISEL-NEXT:    v_writelane_b32 v15, s9, m0
+; GFX802-GISEL-NEXT:    v_writelane_b32 v16, s10, m0
+; GFX802-GISEL-NEXT:    v_writelane_b32 v17, s11, m0
+; GFX802-GISEL-NEXT:    v_writelane_b32 v18, s12, m0
+; GFX802-GISEL-NEXT:    flat_store_dwordx4 v[0:1], v[11:14]
+; GFX802-GISEL-NEXT:    flat_store_dwordx4 v[19:20], v[15:18]
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-GISEL-LABEL: test_writelane_v3p0:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    s_clause 0x1
+; GFX1010-GISEL-NEXT:    global_load_dwordx4 v[11:14], v[0:1], off
+; GFX1010-GISEL-NEXT:    global_load_dwordx4 v[15:18], v[0:1], off offset:16
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s5, v10
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s6, v3
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s7, v4
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s8, v5
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s9, v6
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s10, v7
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s11, v8
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s12, v9
+; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(1)
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v11, s4, s5
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v12, s6, s5
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v13, s7, s5
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v14, s8, s5
+; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v15, s9, s5
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v16, s10, s5
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v17, s11, s5
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v18, s12, s5
+; GFX1010-GISEL-NEXT:    global_store_dwordx4 v[0:1], v[11:14], off
+; GFX1010-GISEL-NEXT:    global_store_dwordx4 v[0:1], v[15:18], off offset:16
+; GFX1010-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-GISEL-LABEL: test_writelane_v3p0:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    s_clause 0x1
+; GFX1100-GISEL-NEXT:    global_load_b128 v[11:14], v[0:1], off
+; GFX1100-GISEL-NEXT:    global_load_b128 v[15:18], v[0:1], off offset:16
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s0, v2
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s1, v10
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s2, v3
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s3, v4
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s4, v5
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s5, v6
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s6, v7
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s7, v8
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s8, v9
+; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(1)
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v11, s0, s1
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v12, s2, s1
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v13, s3, s1
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v14, s4, s1
+; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v15, s5, s1
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v16, s6, s1
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v17, s7, s1
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v18, s8, s1
+; GFX1100-GISEL-NEXT:    s_clause 0x1
+; GFX1100-GISEL-NEXT:    global_store_b128 v[0:1], v[11:14], off
+; GFX1100-GISEL-NEXT:    global_store_b128 v[0:1], v[15:18], off offset:16
+; GFX1100-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %oldval = load <4 x ptr>, ptr addrspace(1) %out
+  %writelane = call <4 x ptr> @llvm.amdgcn.writelane.v3p0(<4 x ptr> %src, i32 %src1, <4 x ptr> %oldval)
+  store <4 x ptr> %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define void @test_writelane_v8i16(ptr addrspace(1) %out, <8 x i16> %src, i32 %src1) {
+; GFX802-SDAG-LABEL: test_writelane_v8i16:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-SDAG-NEXT:    flat_load_dwordx4 v[7:10], v[0:1]
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 m0, v6
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s4, v5
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s5, v4
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s6, v3
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s7, v2
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    v_writelane_b32 v10, s4, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v9, s5, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v8, s6, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v7, s7, m0
+; GFX802-SDAG-NEXT:    flat_store_dwordx4 v[0:1], v[7:10]
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-SDAG-LABEL: test_writelane_v8i16:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    global_load_dwordx4 v[7:10], v[0:1], off
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s4, v5
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s5, v6
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s6, v4
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s7, v3
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s8, v2
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v10, s4, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v9, s6, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v8, s7, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v7, s8, s5
+; GFX1010-SDAG-NEXT:    global_store_dwordx4 v[0:1], v[7:10], off
+; GFX1010-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-SDAG-LABEL: test_writelane_v8i16:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    global_load_b128 v[7:10], v[0:1], off
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s0, v5
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s1, v6
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s2, v4
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s3, v3
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v10, s0, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v9, s2, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v8, s3, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v7, s4, s1
+; GFX1100-SDAG-NEXT:    global_store_b128 v[0:1], v[7:10], off
+; GFX1100-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX802-GISEL-LABEL: test_writelane_v8i16:
+; GFX802-GISEL:       ; %bb.0:
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-GISEL-NEXT:    flat_load_dwordx4 v[7:10], v[0:1]
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s5, v6
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s6, v3
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s7, v4
+; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s8, v5
+; GFX802-GISEL-NEXT:    s_mov_b32 m0, s5
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-GISEL-NEXT:    v_writelane_b32 v7, s4, m0
+; GFX802-GISEL-NEXT:    v_writelane_b32 v8, s6, m0
+; GFX802-GISEL-NEXT:    v_writelane_b32 v9, s7, m0
+; GFX802-GISEL-NEXT:    v_writelane_b32 v10, s8, m0
+; GFX802-GISEL-NEXT:    flat_store_dwordx4 v[0:1], v[7:10]
+; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-GISEL-LABEL: test_writelane_v8i16:
+; GFX1010-GISEL:       ; %bb.0:
+; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-GISEL-NEXT:    global_load_dwordx4 v[7:10], v[0:1], off
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s5, v6
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s6, v3
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s7, v4
+; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s8, v5
+; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v7, s4, s5
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v8, s6, s5
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v9, s7, s5
+; GFX1010-GISEL-NEXT:    v_writelane_b32 v10, s8, s5
+; GFX1010-GISEL-NEXT:    global_store_dwordx4 v[0:1], v[7:10], off
+; GFX1010-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-GISEL-LABEL: test_writelane_v8i16:
+; GFX1100-GISEL:       ; %bb.0:
+; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-GISEL-NEXT:    global_load_b128 v[7:10], v[0:1], off
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s0, v2
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s1, v6
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s2, v3
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s3, v4
+; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s4, v5
+; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v7, s0, s1
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v8, s2, s1
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v9, s3, s1
+; GFX1100-GISEL-NEXT:    v_writelane_b32 v10, s4, s1
+; GFX1100-GISEL-NEXT:    global_store_b128 v[0:1], v[7:10], off
+; GFX1100-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %oldval = load <8 x i16>, ptr addrspace(1) %out
+  %writelane = call <8 x i16> @llvm.amdgcn.writelane.v8i16(<8 x i16> %src, i32 %src1, <8 x i16> %oldval)
+  store <8 x i16> %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
 declare i32 @llvm.amdgcn.workitem.id.x() #2
 
 attributes #0 = { nounwind readnone convergent }
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.writelane.ptr.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.writelane.ptr.ll
new file mode 100644
index 0000000000000..afc394627d356
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.writelane.ptr.ll
@@ -0,0 +1,292 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=gfx802 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX802-SDAG %s
+; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX1010-SDAG %s
+; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=gfx1100 -verify-machineinstrs -amdgpu-enable-vopd=0 < %s | FileCheck -check-prefixes=GFX1100-SDAG %s
+
+define void @test_writelane_p3(ptr addrspace(1) %out, ptr addrspace(3) %src, i32 %src1) {
+; GFX802-SDAG-LABEL: test_writelane_p3:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-SDAG-NEXT:    flat_load_dword v4, v[0:1]
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 m0, v3
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_nop 1
+; GFX802-SDAG-NEXT:    v_writelane_b32 v4, s4, m0
+; GFX802-SDAG-NEXT:    flat_store_dword v[0:1], v4
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-SDAG-LABEL: test_writelane_p3:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    global_load_dword v4, v[0:1], off
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v4, s4, s5
+; GFX1010-SDAG-NEXT:    global_store_dword v[0:1], v4, off
+; GFX1010-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-SDAG-LABEL: test_writelane_p3:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    global_load_b32 v4, v[0:1], off
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s0, v2
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s1, v3
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v4, s0, s1
+; GFX1100-SDAG-NEXT:    global_store_b32 v[0:1], v4, off
+; GFX1100-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %oldval = load ptr addrspace(3), ptr addrspace(1) %out
+  %writelane = call ptr addrspace(3) @llvm.amdgcn.writelane.p3(ptr addrspace(3) %src, i32 %src1, ptr addrspace(3) %oldval)
+  store ptr addrspace(3) %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define void @test_writelane_v3p3(ptr addrspace(1) %out, <3 x ptr addrspace(3)> %src, i32 %src1) {
+; GFX802-SDAG-LABEL: test_writelane_v3p3:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-SDAG-NEXT:    flat_load_dwordx3 v[6:8], v[0:1]
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 m0, v5
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s4, v4
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s6, v2
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    v_writelane_b32 v8, s4, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v7, s5, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v6, s6, m0
+; GFX802-SDAG-NEXT:    flat_store_dwordx3 v[0:1], v[6:8]
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-SDAG-LABEL: test_writelane_v3p3:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    global_load_dwordx3 v[6:8], v[0:1], off
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s4, v4
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s5, v5
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s6, v3
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s7, v2
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v8, s4, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v7, s6, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v6, s7, s5
+; GFX1010-SDAG-NEXT:    global_store_dwordx3 v[0:1], v[6:8], off
+; GFX1010-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-SDAG-LABEL: test_writelane_v3p3:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    global_load_b96 v[6:8], v[0:1], off
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s0, v4
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s1, v5
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s2, v3
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s3, v2
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v8, s0, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v7, s2, s1
+; GFX1100-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_3)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v6, s3, s1
+; GFX1100-SDAG-NEXT:    global_store_b96 v[0:1], v[6:8], off
+; GFX1100-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %oldval = load <3 x ptr addrspace(3)>, ptr addrspace(1) %out
+  %writelane = call <3 x ptr addrspace(3)> @llvm.amdgcn.writelane.v3p3(<3 x ptr addrspace(3)> %src, i32 %src1, <3 x ptr addrspace(3)> %oldval)
+  store <3 x ptr addrspace(3)> %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define void @test_writelane_p5(ptr addrspace(1) %out, ptr addrspace(5) %src, i32 %src1) {
+; GFX802-SDAG-LABEL: test_writelane_p5:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-SDAG-NEXT:    flat_load_dword v4, v[0:1]
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 m0, v3
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_nop 1
+; GFX802-SDAG-NEXT:    v_writelane_b32 v4, s4, m0
+; GFX802-SDAG-NEXT:    flat_store_dword v[0:1], v4
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-SDAG-LABEL: test_writelane_p5:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    global_load_dword v4, v[0:1], off
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v4, s4, s5
+; GFX1010-SDAG-NEXT:    global_store_dword v[0:1], v4, off
+; GFX1010-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-SDAG-LABEL: test_writelane_p5:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    global_load_b32 v4, v[0:1], off
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s0, v2
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s1, v3
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v4, s0, s1
+; GFX1100-SDAG-NEXT:    global_store_b32 v[0:1], v4, off
+; GFX1100-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %oldval = load ptr addrspace(5), ptr addrspace(1) %out
+  %writelane = call ptr addrspace(5) @llvm.amdgcn.writelane.p5(ptr addrspace(5) %src, i32 %src1, ptr addrspace(5) %oldval)
+  store ptr addrspace(5) %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define void @test_writelane_v3p5(ptr addrspace(1) %out, <3 x ptr addrspace(5)> %src, i32 %src1) {
+; GFX802-SDAG-LABEL: test_writelane_v3p5:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-SDAG-NEXT:    flat_load_dwordx3 v[6:8], v[0:1]
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 m0, v5
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s4, v4
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s6, v2
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    v_writelane_b32 v8, s4, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v7, s5, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v6, s6, m0
+; GFX802-SDAG-NEXT:    flat_store_dwordx3 v[0:1], v[6:8]
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-SDAG-LABEL: test_writelane_v3p5:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    global_load_dwordx3 v[6:8], v[0:1], off
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s4, v4
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s5, v5
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s6, v3
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s7, v2
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v8, s4, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v7, s6, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v6, s7, s5
+; GFX1010-SDAG-NEXT:    global_store_dwordx3 v[0:1], v[6:8], off
+; GFX1010-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-SDAG-LABEL: test_writelane_v3p5:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    global_load_b96 v[6:8], v[0:1], off
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s0, v4
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s1, v5
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s2, v3
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s3, v2
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v8, s0, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v7, s2, s1
+; GFX1100-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_3)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v6, s3, s1
+; GFX1100-SDAG-NEXT:    global_store_b96 v[0:1], v[6:8], off
+; GFX1100-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %oldval = load <3 x ptr addrspace(5)>, ptr addrspace(1) %out
+  %writelane = call <3 x ptr addrspace(5)> @llvm.amdgcn.writelane.v3p5(<3 x ptr addrspace(5)> %src, i32 %src1, <3 x ptr addrspace(5)> %oldval)
+  store <3 x ptr addrspace(5)> %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define void @test_writelane_p6(ptr addrspace(1) %out, ptr addrspace(6) %src, i32 %src1) {
+; GFX802-SDAG-LABEL: test_writelane_p6:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-SDAG-NEXT:    flat_load_dword v4, v[0:1]
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 m0, v3
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_nop 1
+; GFX802-SDAG-NEXT:    v_writelane_b32 v4, s4, m0
+; GFX802-SDAG-NEXT:    flat_store_dword v[0:1], v4
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-SDAG-LABEL: test_writelane_p6:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    global_load_dword v4, v[0:1], off
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v4, s4, s5
+; GFX1010-SDAG-NEXT:    global_store_dword v[0:1], v4, off
+; GFX1010-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-SDAG-LABEL: test_writelane_p6:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    global_load_b32 v4, v[0:1], off
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s0, v2
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s1, v3
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v4, s0, s1
+; GFX1100-SDAG-NEXT:    global_store_b32 v[0:1], v4, off
+; GFX1100-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %oldval = load ptr addrspace(6), ptr addrspace(1) %out
+  %writelane = call ptr addrspace(6) @llvm.amdgcn.writelane.p6(ptr addrspace(6) %src, i32 %src1, ptr addrspace(6) %oldval)
+  store ptr addrspace(6) %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define void @test_writelane_v3p6(ptr addrspace(1) %out, <3 x ptr addrspace(6)> %src, i32 %src1) {
+; GFX802-SDAG-LABEL: test_writelane_v3p6:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-SDAG-NEXT:    flat_load_dwordx3 v[6:8], v[0:1]
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 m0, v5
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s4, v4
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s5, v3
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s6, v2
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    v_writelane_b32 v8, s4, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v7, s5, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v6, s6, m0
+; GFX802-SDAG-NEXT:    flat_store_dwordx3 v[0:1], v[6:8]
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-SDAG-LABEL: test_writelane_v3p6:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    global_load_dwordx3 v[6:8], v[0:1], off
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s4, v4
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s5, v5
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s6, v3
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s7, v2
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v8, s4, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v7, s6, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v6, s7, s5
+; GFX1010-SDAG-NEXT:    global_store_dwordx3 v[0:1], v[6:8], off
+; GFX1010-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-SDAG-LABEL: test_writelane_v3p6:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    global_load_b96 v[6:8], v[0:1], off
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s0, v4
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s1, v5
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s2, v3
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s3, v2
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v8, s0, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v7, s2, s1
+; GFX1100-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_3)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v6, s3, s1
+; GFX1100-SDAG-NEXT:    global_store_b96 v[0:1], v[6:8], off
+; GFX1100-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %oldval = load <3 x ptr addrspace(6)>, ptr addrspace(1) %out
+  %writelane = call <3 x ptr addrspace(6)> @llvm.amdgcn.writelane.v3p6(<3 x ptr addrspace(6)> %src, i32 %src1, <3 x ptr addrspace(6)> %oldval)
+  store <3 x ptr addrspace(6)> %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
diff --git a/llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll b/llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll
index 94c32e3cbe99f..483ea8ad57d1b 100644
--- a/llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll
+++ b/llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll
@@ -2714,7 +2714,7 @@ declare i32 @llvm.amdgcn.readfirstlane(i32)
 
 define amdgpu_kernel void @readfirstlane_constant(i32 %arg) {
 ; CHECK-LABEL: @readfirstlane_constant(
-; CHECK-NEXT:    [[VAR:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[ARG:%.*]])
+; CHECK-NEXT:    [[VAR:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[ARG:%.*]])
 ; CHECK-NEXT:    store volatile i32 [[VAR]], ptr undef, align 4
 ; CHECK-NEXT:    store volatile i32 0, ptr undef, align 4
 ; CHECK-NEXT:    store volatile i32 123, ptr undef, align 4
@@ -2737,7 +2737,7 @@ define amdgpu_kernel void @readfirstlane_constant(i32 %arg) {
 
 define i32 @readfirstlane_idempotent(i32 %arg) {
 ; CHECK-LABEL: @readfirstlane_idempotent(
-; CHECK-NEXT:    [[READ0:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[ARG:%.*]])
+; CHECK-NEXT:    [[READ0:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[ARG:%.*]])
 ; CHECK-NEXT:    ret i32 [[READ0]]
 ;
   %read0 = call i32 @llvm.amdgcn.readfirstlane(i32 %arg)
@@ -2748,7 +2748,7 @@ define i32 @readfirstlane_idempotent(i32 %arg) {
 
 define i32 @readfirstlane_readlane(i32 %arg) {
 ; CHECK-LABEL: @readfirstlane_readlane(
-; CHECK-NEXT:    [[READ0:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[ARG:%.*]])
+; CHECK-NEXT:    [[READ0:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[ARG:%.*]])
 ; CHECK-NEXT:    ret i32 [[READ0]]
 ;
   %read0 = call i32 @llvm.amdgcn.readfirstlane(i32 %arg)
@@ -2759,10 +2759,10 @@ define i32 @readfirstlane_readlane(i32 %arg) {
 define i32 @readfirstlane_readfirstlane_different_block(i32 %arg) {
 ; CHECK-LABEL: @readfirstlane_readfirstlane_different_block(
 ; CHECK-NEXT:  bb0:
-; CHECK-NEXT:    [[READ0:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[ARG:%.*]])
+; CHECK-NEXT:    [[READ0:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[ARG:%.*]])
 ; CHECK-NEXT:    br label [[BB1:%.*]]
 ; CHECK:       bb1:
-; CHECK-NEXT:    [[READ1:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[READ0]])
+; CHECK-NEXT:    [[READ1:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[READ0]])
 ; CHECK-NEXT:    ret i32 [[READ1]]
 ;
 bb0:
@@ -2777,10 +2777,10 @@ bb1:
 define i32 @readfirstlane_readlane_different_block(i32 %arg) {
 ; CHECK-LABEL: @readfirstlane_readlane_different_block(
 ; CHECK-NEXT:  bb0:
-; CHECK-NEXT:    [[READ0:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[ARG:%.*]], i32 0)
+; CHECK-NEXT:    [[READ0:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[ARG:%.*]], i32 0)
 ; CHECK-NEXT:    br label [[BB1:%.*]]
 ; CHECK:       bb1:
-; CHECK-NEXT:    [[READ1:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[READ0]])
+; CHECK-NEXT:    [[READ1:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[READ0]])
 ; CHECK-NEXT:    ret i32 [[READ1]]
 ;
 bb0:
@@ -2800,7 +2800,7 @@ declare i32 @llvm.amdgcn.readlane(i32, i32)
 
 define amdgpu_kernel void @readlane_constant(i32 %arg, i32 %lane) {
 ; CHECK-LABEL: @readlane_constant(
-; CHECK-NEXT:    [[VAR:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[ARG:%.*]], i32 7)
+; CHECK-NEXT:    [[VAR:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[ARG:%.*]], i32 7)
 ; CHECK-NEXT:    store volatile i32 [[VAR]], ptr undef, align 4
 ; CHECK-NEXT:    store volatile i32 0, ptr undef, align 4
 ; CHECK-NEXT:    store volatile i32 123, ptr undef, align 4
@@ -2823,7 +2823,7 @@ define amdgpu_kernel void @readlane_constant(i32 %arg, i32 %lane) {
 
 define i32 @readlane_idempotent(i32 %arg, i32 %lane) {
 ; CHECK-LABEL: @readlane_idempotent(
-; CHECK-NEXT:    [[READ0:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[ARG:%.*]], i32 [[LANE:%.*]])
+; CHECK-NEXT:    [[READ0:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[ARG:%.*]], i32 [[LANE:%.*]])
 ; CHECK-NEXT:    ret i32 [[READ0]]
 ;
   %read0 = call i32 @llvm.amdgcn.readlane(i32 %arg, i32 %lane)
@@ -2833,8 +2833,8 @@ define i32 @readlane_idempotent(i32 %arg, i32 %lane) {
 
 define i32 @readlane_idempotent_different_lanes(i32 %arg, i32 %lane0, i32 %lane1) {
 ; CHECK-LABEL: @readlane_idempotent_different_lanes(
-; CHECK-NEXT:    [[READ0:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[ARG:%.*]], i32 [[LANE0:%.*]])
-; CHECK-NEXT:    [[READ1:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[READ0]], i32 [[LANE1:%.*]])
+; CHECK-NEXT:    [[READ0:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[ARG:%.*]], i32 [[LANE0:%.*]])
+; CHECK-NEXT:    [[READ1:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[READ0]], i32 [[LANE1:%.*]])
 ; CHECK-NEXT:    ret i32 [[READ1]]
 ;
   %read0 = call i32 @llvm.amdgcn.readlane(i32 %arg, i32 %lane0)
@@ -2844,7 +2844,7 @@ define i32 @readlane_idempotent_different_lanes(i32 %arg, i32 %lane0, i32 %lane1
 
 define i32 @readlane_readfirstlane(i32 %arg) {
 ; CHECK-LABEL: @readlane_readfirstlane(
-; CHECK-NEXT:    [[READ0:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[ARG:%.*]])
+; CHECK-NEXT:    [[READ0:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[ARG:%.*]])
 ; CHECK-NEXT:    ret i32 [[READ0]]
 ;
   %read0 = call i32 @llvm.amdgcn.readfirstlane(i32 %arg)
@@ -2855,10 +2855,10 @@ define i32 @readlane_readfirstlane(i32 %arg) {
 define i32 @readlane_idempotent_different_block(i32 %arg, i32 %lane) {
 ; CHECK-LABEL: @readlane_idempotent_different_block(
 ; CHECK-NEXT:  bb0:
-; CHECK-NEXT:    [[READ0:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[ARG:%.*]], i32 [[LANE:%.*]])
+; CHECK-NEXT:    [[READ0:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[ARG:%.*]], i32 [[LANE:%.*]])
 ; CHECK-NEXT:    br label [[BB1:%.*]]
 ; CHECK:       bb1:
-; CHECK-NEXT:    [[READ1:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[READ0]], i32 [[LANE]])
+; CHECK-NEXT:    [[READ1:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[READ0]], i32 [[LANE]])
 ; CHECK-NEXT:    ret i32 [[READ1]]
 ;
 bb0:
@@ -2874,10 +2874,10 @@ bb1:
 define i32 @readlane_readfirstlane_different_block(i32 %arg) {
 ; CHECK-LABEL: @readlane_readfirstlane_different_block(
 ; CHECK-NEXT:  bb0:
-; CHECK-NEXT:    [[READ0:%.*]] = call i32 @llvm.amdgcn.readfirstlane(i32 [[ARG:%.*]])
+; CHECK-NEXT:    [[READ0:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[ARG:%.*]])
 ; CHECK-NEXT:    br label [[BB1:%.*]]
 ; CHECK:       bb1:
-; CHECK-NEXT:    [[READ1:%.*]] = call i32 @llvm.amdgcn.readlane(i32 [[READ0]], i32 0)
+; CHECK-NEXT:    [[READ1:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[READ0]], i32 0)
 ; CHECK-NEXT:    ret i32 [[READ1]]
 ;
 bb0:

>From 881e11673545164b0172c42ffd0fbba47ebe38dd Mon Sep 17 00:00:00 2001
From: Vikram <Vikram.Hegde at amd.com>
Date: Sat, 18 May 2024 05:28:50 -0400
Subject: [PATCH 2/9] [AMDGPU] Extend permlane16, permlanex16 and permlane64
 intrinsic lowering for generic types

---
 clang/lib/CodeGen/CGBuiltin.cpp               |   19 +
 llvm/include/llvm/IR/IntrinsicsAMDGPU.td      |   10 +-
 .../Target/AMDGPU/AMDGPUAtomicOptimizer.cpp   |    6 +-
 llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp |    3 +
 llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h   |    3 +
 llvm/lib/Target/AMDGPU/AMDGPUInstrInfo.td     |   25 +
 .../lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp |   74 +-
 llvm/lib/Target/AMDGPU/SIISelLowering.cpp     |   49 +-
 llvm/lib/Target/AMDGPU/VOP1Instructions.td    |   10 +-
 llvm/lib/Target/AMDGPU/VOP3Instructions.td    |   10 +-
 .../UniformityAnalysis/AMDGPU/intrinsics.ll   |   12 +-
 .../CodeGen/AMDGPU/llvm.amdgcn.permlane.ll    | 2694 +++++++++++++++--
 .../InstCombine/AMDGPU/amdgcn-intrinsics.ll   |   28 +-
 llvm/test/Verifier/AMDGPU/intrinsic-immarg.ll |   20 +-
 14 files changed, 2671 insertions(+), 292 deletions(-)

diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index a275ada586d13..9ce2f5b6c103b 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -18479,6 +18479,25 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID,
         CGM.getIntrinsic(Intrinsic::amdgcn_update_dpp, Args[0]->getType());
     return Builder.CreateCall(F, Args);
   }
+  case AMDGPU::BI__builtin_amdgcn_permlane16:
+  case AMDGPU::BI__builtin_amdgcn_permlanex16: {
+    Intrinsic::ID IID;
+    IID = BuiltinID == AMDGPU::BI__builtin_amdgcn_permlane16
+              ? Intrinsic::amdgcn_permlane16
+              : Intrinsic::amdgcn_permlanex16;
+
+    llvm::Value *Src0 = EmitScalarExpr(E->getArg(0));
+    llvm::Value *Src1 = EmitScalarExpr(E->getArg(1));
+    llvm::Value *Src2 = EmitScalarExpr(E->getArg(2));
+    llvm::Value *Src3 = EmitScalarExpr(E->getArg(3));
+    llvm::Value *Src4 = EmitScalarExpr(E->getArg(4));
+    llvm::Value *Src5 = EmitScalarExpr(E->getArg(5));
+
+    llvm::Function *F = CGM.getIntrinsic(IID, Src1->getType());
+    return Builder.CreateCall(F, {Src0, Src1, Src2, Src3, Src4, Src5});
+  }
+  case AMDGPU::BI__builtin_amdgcn_permlane64:
+    return emitUnaryBuiltin(*this, E, Intrinsic::amdgcn_permlane64);
   case AMDGPU::BI__builtin_amdgcn_readlane:
     return emitBinaryBuiltin(*this, E, Intrinsic::amdgcn_readlane);
   case AMDGPU::BI__builtin_amdgcn_readfirstlane:
diff --git a/llvm/include/llvm/IR/IntrinsicsAMDGPU.td b/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
index 457566944069e..63f7f48e82e4a 100644
--- a/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
+++ b/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
@@ -2488,15 +2488,15 @@ def int_amdgcn_global_load_lds : AMDGPUGlobalLoadLDS;
 
 // llvm.amdgcn.permlane16 <old> <src0> <src1> <src2> <fi> <bound_control>
 def int_amdgcn_permlane16 : ClangBuiltin<"__builtin_amdgcn_permlane16">,
-  Intrinsic<[llvm_i32_ty],
-            [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, llvm_i1_ty, llvm_i1_ty],
+  Intrinsic<[llvm_any_ty],
+            [LLVMMatchType<0>, LLVMMatchType<0>, llvm_i32_ty, llvm_i32_ty, llvm_i1_ty, llvm_i1_ty],
             [IntrNoMem, IntrConvergent, IntrWillReturn,
              ImmArg<ArgIndex<4>>, ImmArg<ArgIndex<5>>, IntrNoCallback, IntrNoFree]>;
 
 // llvm.amdgcn.permlanex16 <old> <src0> <src1> <src2> <fi> <bound_control>
 def int_amdgcn_permlanex16 : ClangBuiltin<"__builtin_amdgcn_permlanex16">,
-  Intrinsic<[llvm_i32_ty],
-            [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, llvm_i1_ty, llvm_i1_ty],
+  Intrinsic<[llvm_any_ty],
+            [LLVMMatchType<0>, LLVMMatchType<0>, llvm_i32_ty, llvm_i32_ty, llvm_i1_ty, llvm_i1_ty],
             [IntrNoMem, IntrConvergent, IntrWillReturn,
              ImmArg<ArgIndex<4>>, ImmArg<ArgIndex<5>>, IntrNoCallback, IntrNoFree]>;
 
@@ -2540,7 +2540,7 @@ def int_amdgcn_image_bvh_intersect_ray :
 // llvm.amdgcn.permlane64 <src0>
 def int_amdgcn_permlane64 :
   ClangBuiltin<"__builtin_amdgcn_permlane64">,
-  Intrinsic<[llvm_i32_ty], [llvm_i32_ty],
+  Intrinsic<[llvm_any_ty], [LLVMMatchType<0>],
             [IntrNoMem, IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
 
 def int_amdgcn_ds_add_gs_reg_rtn :
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUAtomicOptimizer.cpp b/llvm/lib/Target/AMDGPU/AMDGPUAtomicOptimizer.cpp
index 9ac0a52f63fd8..5fec5dafe2acd 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUAtomicOptimizer.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUAtomicOptimizer.cpp
@@ -413,7 +413,7 @@ Value *AMDGPUAtomicOptimizerImpl::buildReduction(IRBuilder<> &B,
   assert(ST->hasPermLaneX16());
   V = B.CreateBitCast(V, IntNTy);
   Value *Permlanex16Call = B.CreateIntrinsic(
-      Intrinsic::amdgcn_permlanex16, {},
+      V->getType(), Intrinsic::amdgcn_permlanex16,
       {V, V, B.getInt32(-1), B.getInt32(-1), B.getFalse(), B.getFalse()});
   V = buildNonAtomicBinOp(B, Op, B.CreateBitCast(V, AtomicTy),
                           B.CreateBitCast(Permlanex16Call, AtomicTy));
@@ -425,7 +425,7 @@ Value *AMDGPUAtomicOptimizerImpl::buildReduction(IRBuilder<> &B,
     // Reduce across the upper and lower 32 lanes.
     V = B.CreateBitCast(V, IntNTy);
     Value *Permlane64Call =
-        B.CreateIntrinsic(Intrinsic::amdgcn_permlane64, {}, V);
+        B.CreateIntrinsic(V->getType(), Intrinsic::amdgcn_permlane64, V);
     return buildNonAtomicBinOp(B, Op, B.CreateBitCast(V, AtomicTy),
                                B.CreateBitCast(Permlane64Call, AtomicTy));
   }
@@ -481,7 +481,7 @@ Value *AMDGPUAtomicOptimizerImpl::buildScan(IRBuilder<> &B,
     assert(ST->hasPermLaneX16());
     V = B.CreateBitCast(V, IntNTy);
     Value *PermX = B.CreateIntrinsic(
-        Intrinsic::amdgcn_permlanex16, {},
+        V->getType(), Intrinsic::amdgcn_permlanex16,
         {V, V, B.getInt32(-1), B.getInt32(-1), B.getFalse(), B.getFalse()});
 
     Value *UpdateDPPCall =
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
index 9cd7496ba9f96..12a3ce486a75d 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
@@ -5499,6 +5499,9 @@ const char* AMDGPUTargetLowering::getTargetNodeName(unsigned Opcode) const {
   NODE_NAME_CASE(READLANE)
   NODE_NAME_CASE(READFIRSTLANE)
   NODE_NAME_CASE(WRITELANE)
+  NODE_NAME_CASE(PERMLANE16)
+  NODE_NAME_CASE(PERMLANEX16)
+  NODE_NAME_CASE(PERMLANE64)
   NODE_NAME_CASE(DUMMY_CHAIN)
   case AMDGPUISD::FIRST_MEM_OPCODE_NUMBER: break;
   NODE_NAME_CASE(LOAD_D16_HI)
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h
index 02c3dcf39e554..650ad948184d1 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h
+++ b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h
@@ -561,6 +561,9 @@ enum NodeType : unsigned {
   READLANE,
   READFIRSTLANE,
   WRITELANE,
+  PERMLANE16,
+  PERMLANEX16,
+  PERMLANE64,
 
   DUMMY_CHAIN,
   FIRST_MEM_OPCODE_NUMBER = ISD::FIRST_TARGET_MEMORY_OPCODE,
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUInstrInfo.td b/llvm/lib/Target/AMDGPU/AMDGPUInstrInfo.td
index e4f329b200c86..a82f48950f493 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUInstrInfo.td
+++ b/llvm/lib/Target/AMDGPU/AMDGPUInstrInfo.td
@@ -354,9 +354,21 @@ def AMDGPUDWritelaneOp : SDTypeProfile<1, 3, [
   SDTCisSameAs<0, 1>, SDTCisInt<2>, SDTCisSameAs<0, 3>
 ]>;
 
+def AMDGPUDPermlane16Op : SDTypeProfile<1, 6, [
+  SDTCisSameAs<0, 1>,  // old
+  SDTCisSameAs<0, 2>,  // src0
+  SDTCisInt<3>,        // src1
+  SDTCisInt<4>,        // src2
+  SDTCisInt<5>,        // i1 fi
+  SDTCisInt<6>         // i1 bound_ctrl
+]>;
+
 def AMDGPUreadlane_impl : SDNode<"AMDGPUISD::READLANE", AMDGPUReadlaneOp>;
 def AMDGPUreadfirstlane_impl : SDNode<"AMDGPUISD::READFIRSTLANE", AMDGPUReadfirstlaneOp>;
 def AMDGPUwritelane_impl : SDNode<"AMDGPUISD::WRITELANE", AMDGPUDWritelaneOp>;
+def AMDGPUpermlane16_impl : SDNode<"AMDGPUISD::PERMLANE16", AMDGPUDPermlane16Op>;
+def AMDGPUpermlanex16_impl : SDNode<"AMDGPUISD::PERMLANEX16", AMDGPUDPermlane16Op>;
+def AMDGPUpermlane64_impl : SDNode<"AMDGPUISD::PERMLANE64", AMDGPUReadfirstlaneOp>;
 
 // SI+ export
 def AMDGPUExportOp : SDTypeProfile<0, 8, [
@@ -535,3 +547,16 @@ def AMDGPUwritelane : PatFrags<(ops node:$src0, node:$src1, node:$src2),
   [(int_amdgcn_writelane node:$src0, node:$src1, node:$src2),
    (AMDGPUwritelane_impl node:$src0, node:$src1, node:$src2)]>;
 
+def AMDGPUpermlane16 : PatFrags<(ops node:$src0, node:$src1, node:$src2, node:$src3, node:$src4, node:$src5),
+  [(int_amdgcn_permlane16 node:$src0, node:$src1, node:$src2, node:$src3, node:$src4, node:$src5),
+   (AMDGPUpermlane16_impl node:$src0, node:$src1, node:$src2, node:$src3, node:$src4, node:$src5)]>;
+
+def AMDGPUpermlanex16 : PatFrags<(ops node:$src0, node:$src1, node:$src2, node:$src3, node:$src4, node:$src5),
+  [(int_amdgcn_permlanex16 node:$src0, node:$src1, node:$src2, node:$src3, node:$src4, node:$src5),
+   (AMDGPUpermlanex16_impl node:$src0, node:$src1, node:$src2, node:$src3, node:$src4, node:$src5)]>;
+
+def AMDGPUpermlane64 : PatFrags<(ops node:$src),
+  [(int_amdgcn_permlane64 node:$src),
+   (AMDGPUpermlane64_impl node:$src)]>;
+
+
diff --git a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
index 6ffc8a20f76fa..b28c3521d6336 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
@@ -5397,25 +5397,42 @@ bool AMDGPULegalizerInfo::legalizeLaneOp(LegalizerHelper &Helper,
   Register DstReg = MI.getOperand(0).getReg();
   Register Src0 = MI.getOperand(2).getReg();
 
+  bool IsPermLane16 = IID == Intrinsic::amdgcn_permlane16 ||
+                      IID == Intrinsic::amdgcn_permlanex16;
+
   auto createLaneOp = [&](Register Src0, Register Src1,
                           Register Src2) -> Register {
     auto LaneOp = B.buildIntrinsic(IID, {S32}).addUse(Src0);
     switch (IID) {
     case Intrinsic::amdgcn_readfirstlane:
+    case Intrinsic::amdgcn_permlane64:
       return LaneOp.getReg(0);
     case Intrinsic::amdgcn_readlane:
       return LaneOp.addUse(Src1).getReg(0);
     case Intrinsic::amdgcn_writelane:
       return LaneOp.addUse(Src1).addUse(Src2).getReg(0);
+    case Intrinsic::amdgcn_permlane16:
+    case Intrinsic::amdgcn_permlanex16: {
+      Register Src3 = MI.getOperand(5).getReg();
+      Register Src4 = MI.getOperand(6).getImm();
+      Register Src5 = MI.getOperand(7).getImm();
+      return LaneOp.addUse(Src1)
+          .addUse(Src2)
+          .addUse(Src3)
+          .addImm(Src4)
+          .addImm(Src5)
+          .getReg(0);
+    }
     default:
       llvm_unreachable("unhandled lane op");
     }
   };
 
   Register Src1, Src2;
-  if (IID == Intrinsic::amdgcn_readlane || IID == Intrinsic::amdgcn_writelane) {
+  if (IID == Intrinsic::amdgcn_readlane || IID == Intrinsic::amdgcn_writelane ||
+      IsPermLane16) {
     Src1 = MI.getOperand(3).getReg();
-    if (IID == Intrinsic::amdgcn_writelane) {
+    if (IID == Intrinsic::amdgcn_writelane || IsPermLane16) {
       Src2 = MI.getOperand(4).getReg();
     }
   }
@@ -5433,7 +5450,16 @@ bool AMDGPULegalizerInfo::legalizeLaneOp(LegalizerHelper &Helper,
                             ? Src0
                             : B.buildBitcast(LLT::scalar(Size), Src0).getReg(0);
     Src0 = B.buildAnyExt(S32, Src0Cast).getReg(0);
-    if (Src2.isValid()) {
+
+    if (IsPermLane16) {
+      Register Src1Cast =
+          MRI.getType(Src1).isScalar()
+              ? Src1
+              : B.buildBitcast(LLT::scalar(Size), Src2).getReg(0);
+      Src1 = B.buildAnyExt(LLT::scalar(32), Src1Cast).getReg(0);
+    }
+
+    if (IID == Intrinsic::amdgcn_writelane) {
       Register Src2Cast =
           MRI.getType(Src2).isScalar()
               ? Src2
@@ -5485,46 +5511,45 @@ bool AMDGPULegalizerInfo::legalizeLaneOp(LegalizerHelper &Helper,
       }
       break;
     }
-    case Intrinsic::amdgcn_readfirstlane: {
+    case Intrinsic::amdgcn_readfirstlane:
+    case Intrinsic::amdgcn_permlane64: {
       for (unsigned i = 0; i < NumParts; ++i) {
         Src0 = IsS16Vec ? B.buildBitcast(S32, Src0Parts.getReg(i)).getReg(0)
                         : Src0Parts.getReg(i);
         PartialRes.push_back(
-            (B.buildIntrinsic(Intrinsic::amdgcn_readfirstlane, {S32})
-                 .addUse(Src0)
-                 .getReg(0)));
+            (B.buildIntrinsic(IID, {S32}).addUse(Src0).getReg(0)));
       }
 
       break;
     }
-    case Intrinsic::amdgcn_writelane: {
+    case Intrinsic::amdgcn_writelane:
+    case Intrinsic::amdgcn_permlane16:
+    case Intrinsic::amdgcn_permlanex16: {
       Register Src1 = MI.getOperand(3).getReg();
       Register Src2 = MI.getOperand(4).getReg();
-      MachineInstrBuilder Src2Parts;
+
+      Register SrcX = IsPermLane16 ? Src1 : Src2;
+      MachineInstrBuilder SrcXParts;
 
       if (Ty.isPointer()) {
-        auto PtrToInt = B.buildPtrToInt(S64, Src2);
-        Src2Parts = B.buildUnmerge(S32, PtrToInt);
+        auto PtrToInt = B.buildPtrToInt(S64, SrcX);
+        SrcXParts = B.buildUnmerge(S32, PtrToInt);
       } else if (Ty.isPointerVector()) {
         LLT IntVecTy = Ty.changeElementType(
             LLT::scalar(Ty.getElementType().getSizeInBits()));
-        auto PtrToInt = B.buildPtrToInt(IntVecTy, Src2);
-        Src2Parts = B.buildUnmerge(S32, PtrToInt);
+        auto PtrToInt = B.buildPtrToInt(IntVecTy, SrcX);
+        SrcXParts = B.buildUnmerge(S32, PtrToInt);
       } else
-        Src2Parts =
-            IsS16Vec ? B.buildUnmerge(V2S16, Src2) : B.buildUnmerge(S32, Src2);
+        SrcXParts =
+            IsS16Vec ? B.buildUnmerge(V2S16, SrcX) : B.buildUnmerge(S32, SrcX);
 
       for (unsigned i = 0; i < NumParts; ++i) {
         Src0 = IsS16Vec ? B.buildBitcast(S32, Src0Parts.getReg(i)).getReg(0)
                         : Src0Parts.getReg(i);
-        Src2 = IsS16Vec ? B.buildBitcast(S32, Src2Parts.getReg(i)).getReg(0)
-                        : Src2Parts.getReg(i);
-        PartialRes.push_back(
-            (B.buildIntrinsic(Intrinsic::amdgcn_writelane, {S32})
-                 .addUse(Src0)
-                 .addUse(Src1)
-                 .addUse(Src2))
-                .getReg(0));
+        SrcX = IsS16Vec ? B.buildBitcast(S32, SrcXParts.getReg(i)).getReg(0)
+                        : SrcXParts.getReg(i);
+        PartialRes.push_back(IsPermLane16 ? createLaneOp(Src0, SrcX, Src2)
+                                          : createLaneOp(Src0, Src1, SrcX));
       }
 
       break;
@@ -7519,6 +7544,9 @@ bool AMDGPULegalizerInfo::legalizeIntrinsic(LegalizerHelper &Helper,
   case Intrinsic::amdgcn_readlane:
   case Intrinsic::amdgcn_writelane:
   case Intrinsic::amdgcn_readfirstlane:
+  case Intrinsic::amdgcn_permlane16:
+  case Intrinsic::amdgcn_permlanex16:
+  case Intrinsic::amdgcn_permlane64:
     return legalizeLaneOp(Helper, MI, IntrID);
   default: {
     if (const AMDGPU::ImageDimIntrinsicInfo *ImageDimIntr =
diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index 6c23bdf09974b..5d34ed089f65d 100644
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -6091,22 +6091,39 @@ static SDValue lowerLaneOp(const SITargetLowering &TLI, SDNode *N,
   EVT VT = N->getValueType(0);
   unsigned ValSize = VT.getSizeInBits();
   unsigned IntrinsicID = N->getConstantOperandVal(0);
+  bool IsPermLane16 = IntrinsicID == Intrinsic::amdgcn_permlane16 ||
+                      IntrinsicID == Intrinsic::amdgcn_permlanex16;
+  bool IsPermLane64 = IntrinsicID == Intrinsic::amdgcn_permlane64;
   SDValue Src0 = N->getOperand(1);
   SDLoc SL(N);
   MVT IntVT = MVT::getIntegerVT(ValSize);
 
-  auto createLaneOp = [&DAG, &SL](SDValue Src0, SDValue Src1, SDValue Src2,
-                                  MVT VT) -> SDValue {
-    return (Src2 ? DAG.getNode(AMDGPUISD::WRITELANE, SL, VT, {Src0, Src1, Src2})
-            : Src1 ? DAG.getNode(AMDGPUISD::READLANE, SL, VT, {Src0, Src1})
-                   : DAG.getNode(AMDGPUISD::READFIRSTLANE, SL, VT, {Src0}));
+  auto createLaneOp = [&](SDValue Src0, SDValue Src1, SDValue Src2,
+                          MVT ValueT) -> SDValue {
+    if (IsPermLane16 || IsPermLane64) {
+      if (IsPermLane16) {
+        SDValue Src3 = N->getOperand(4);
+        SDValue Src4 = N->getOperand(5);
+        SDValue Src5 = N->getOperand(6);
+        return DAG.getNode(IntrinsicID == Intrinsic::amdgcn_permlane16
+                               ? AMDGPUISD::PERMLANE16
+                               : AMDGPUISD::PERMLANEX16,
+                           SL, ValueT, {Src0, Src1, Src2, Src3, Src4, Src5});
+      }
+      return DAG.getNode(AMDGPUISD::PERMLANE64, SL, ValueT, {Src0});
+    }
+
+    return (
+        Src2 ? DAG.getNode(AMDGPUISD::WRITELANE, SL, ValueT, {Src0, Src1, Src2})
+        : Src1 ? DAG.getNode(AMDGPUISD::READLANE, SL, ValueT, {Src0, Src1})
+               : DAG.getNode(AMDGPUISD::READFIRSTLANE, SL, ValueT, {Src0}));
   };
 
   SDValue Src1, Src2;
   if (IntrinsicID == Intrinsic::amdgcn_readlane ||
-      IntrinsicID == Intrinsic::amdgcn_writelane) {
+      IntrinsicID == Intrinsic::amdgcn_writelane || IsPermLane16) {
     Src1 = N->getOperand(2);
-    if (IntrinsicID == Intrinsic::amdgcn_writelane)
+    if (IntrinsicID == Intrinsic::amdgcn_writelane || IsPermLane16)
       Src2 = N->getOperand(3);
   }
 
@@ -6118,10 +6135,17 @@ static SDValue lowerLaneOp(const SITargetLowering &TLI, SDNode *N,
   if (ValSize < 32) {
     SDValue InitBitCast = DAG.getBitcast(IntVT, Src0);
     Src0 = DAG.getAnyExtOrTrunc(InitBitCast, SL, MVT::i32);
-    if (Src2.getNode()) {
+
+    if (IsPermLane16) {
+      SDValue Src1Cast = DAG.getBitcast(IntVT, Src1);
+      Src1 = DAG.getAnyExtOrTrunc(Src1Cast, SL, MVT::i32);
+    }
+
+    if (IntrinsicID == Intrinsic::amdgcn_writelane) {
       SDValue Src2Cast = DAG.getBitcast(IntVT, Src2);
       Src2 = DAG.getAnyExtOrTrunc(Src2Cast, SL, MVT::i32);
     }
+
     SDValue LaneOp = createLaneOp(Src0, Src1, Src2, MVT::i32);
     SDValue Trunc = DAG.getAnyExtOrTrunc(LaneOp, SL, IntVT);
     return DAG.getBitcast(VT, Trunc);
@@ -6131,7 +6155,11 @@ static SDValue lowerLaneOp(const SITargetLowering &TLI, SDNode *N,
     MVT VecVT = MVT::getVectorVT(MVT::i32, ValSize / 32);
     Src0 = DAG.getBitcast(VecVT, Src0);
 
-    if (Src2.getNode())
+    if (IsPermLane16) {
+      Src1 = DAG.getBitcast(VecVT, Src1);
+    }
+
+    if (IntrinsicID == Intrinsic::amdgcn_writelane)
       Src2 = DAG.getBitcast(VecVT, Src2);
 
     SDValue LaneOp = createLaneOp(Src0, Src1, Src2, VecVT);
@@ -8612,6 +8640,9 @@ SDValue SITargetLowering::LowerINTRINSIC_WO_CHAIN(SDValue Op,
   case Intrinsic::amdgcn_readlane:
   case Intrinsic::amdgcn_readfirstlane:
   case Intrinsic::amdgcn_writelane:
+  case Intrinsic::amdgcn_permlane16:
+  case Intrinsic::amdgcn_permlanex16:
+  case Intrinsic::amdgcn_permlane64:
     return lowerLaneOp(*this, Op.getNode(), DAG);
   default:
     if (const AMDGPU::ImageDimIntrinsicInfo *ImageDimIntr =
diff --git a/llvm/lib/Target/AMDGPU/VOP1Instructions.td b/llvm/lib/Target/AMDGPU/VOP1Instructions.td
index 9c5d6a7bf6d0b..33d3da3830f57 100644
--- a/llvm/lib/Target/AMDGPU/VOP1Instructions.td
+++ b/llvm/lib/Target/AMDGPU/VOP1Instructions.td
@@ -717,15 +717,19 @@ def V_ACCVGPR_MOV_B32 : VOP1_Pseudo<"v_accvgpr_mov_b32", VOPProfileAccMov, [], 1
 let SubtargetPredicate = isGFX11Plus in {
   // Restrict src0 to be VGPR
   def V_PERMLANE64_B32 : VOP1_Pseudo<"v_permlane64_b32", VOP_MOVRELS,
-                                      getVOP1Pat<int_amdgcn_permlane64,
-                                                 VOP_MOVRELS>.ret,
-                                      /*VOP1Only=*/ 1>;
+                                      [], /*VOP1Only=*/ 1>;
   defm V_MOV_B16_t16    : VOP1Inst<"v_mov_b16_t16", VOPProfile_True16<VOP_I16_I16>>;
   defm V_NOT_B16        : VOP1Inst_t16<"v_not_b16", VOP_I16_I16>;
   defm V_CVT_I32_I16    : VOP1Inst_t16<"v_cvt_i32_i16", VOP_I32_I16>;
   defm V_CVT_U32_U16    : VOP1Inst_t16<"v_cvt_u32_u16", VOP_I32_I16>;
 } // End SubtargetPredicate = isGFX11Plus
 
+foreach vt = Reg32Types.types in {
+  def : GCNPat<(AMDGPUpermlane64 (vt VRegSrc_32:$src0)),
+        (vt (V_PERMLANE64_B32 (vt VRegSrc_32:$src0)))
+  >;
+}
+
 //===----------------------------------------------------------------------===//
 // Target-specific instruction encodings.
 //===----------------------------------------------------------------------===//
diff --git a/llvm/lib/Target/AMDGPU/VOP3Instructions.td b/llvm/lib/Target/AMDGPU/VOP3Instructions.td
index 616bc7684753e..eee2085eab341 100644
--- a/llvm/lib/Target/AMDGPU/VOP3Instructions.td
+++ b/llvm/lib/Target/AMDGPU/VOP3Instructions.td
@@ -838,8 +838,8 @@ def gi_opsel_i1timm : GICustomOperandRenderer<"renderOpSelTImm">,
   GISDNodeXFormEquiv<opsel_i1timm>;
 
 class PermlanePat<SDPatternOperator permlane,
-  Instruction inst> : GCNPat<
-  (permlane i32:$vdst_in, i32:$src0, i32:$src1, i32:$src2,
+  Instruction inst, ValueType vt> : GCNPat<
+  (permlane vt:$vdst_in, vt:$src0, i32:$src1, i32:$src2,
             timm:$fi, timm:$bc),
   (inst (opsel_i1timm $fi), VGPR_32:$src0, (opsel_i1timm $bc),
         SCSrc_b32:$src1, 0, SCSrc_b32:$src2, VGPR_32:$vdst_in)
@@ -864,8 +864,10 @@ let SubtargetPredicate = isGFX10Plus in {
     defm V_PERMLANEX16_B32 : VOP3Inst<"v_permlanex16_b32", VOP3_PERMLANE_Profile>;
   } // End $vdst = $vdst_in, DisableEncoding $vdst_in
 
-  def : PermlanePat<int_amdgcn_permlane16, V_PERMLANE16_B32_e64>;
-  def : PermlanePat<int_amdgcn_permlanex16, V_PERMLANEX16_B32_e64>;
+  foreach vt = Reg32Types.types in {
+    def : PermlanePat<AMDGPUpermlane16, V_PERMLANE16_B32_e64, vt>;
+    def : PermlanePat<AMDGPUpermlanex16, V_PERMLANEX16_B32_e64, vt>;
+  }
 
   defm V_ADD_NC_U16 : VOP3Inst <"v_add_nc_u16", VOP3_Profile<VOP_I16_I16_I16, VOP3_OPSEL>, add>;
   defm V_SUB_NC_U16 : VOP3Inst <"v_sub_nc_u16", VOP3_Profile<VOP_I16_I16_I16, VOP3_OPSEL>, sub>;
diff --git a/llvm/test/Analysis/UniformityAnalysis/AMDGPU/intrinsics.ll b/llvm/test/Analysis/UniformityAnalysis/AMDGPU/intrinsics.ll
index 74d2f53d7b317..680c998a4b39f 100644
--- a/llvm/test/Analysis/UniformityAnalysis/AMDGPU/intrinsics.ll
+++ b/llvm/test/Analysis/UniformityAnalysis/AMDGPU/intrinsics.ll
@@ -7,16 +7,16 @@ define amdgpu_kernel void @ds_swizzle(ptr addrspace(1) %out, i32 %src) #0 {
   ret void
 }
 
-; CHECK: DIVERGENT: %v = call i32 @llvm.amdgcn.permlane16(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false) #0
+; CHECK: DIVERGENT: %v = call i32 @llvm.amdgcn.permlane16.i32(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false) #0
 define amdgpu_kernel void @v_permlane16_b32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) #0 {
-  %v = call i32 @llvm.amdgcn.permlane16(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false) #0
+  %v = call i32 @llvm.amdgcn.permlane16.i32(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false) #0
   store i32 %v, ptr addrspace(1) %out
   ret void
 }
 
-; CHECK: DIVERGENT: %v = call i32 @llvm.amdgcn.permlanex16(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false) #0
+; CHECK: DIVERGENT: %v = call i32 @llvm.amdgcn.permlanex16.i32(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false) #0
 define amdgpu_kernel void @v_permlanex16_b32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) #0 {
-  %v = call i32 @llvm.amdgcn.permlanex16(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false) #0
+  %v = call i32 @llvm.amdgcn.permlanex16.i32(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false) #0
   store i32 %v, ptr addrspace(1) %out
   ret void
 }
@@ -230,8 +230,8 @@ bb:
 }
 
 declare i32 @llvm.amdgcn.ds.swizzle(i32, i32) #1
-declare i32 @llvm.amdgcn.permlane16(i32, i32, i32, i32, i1, i1) #1
-declare i32 @llvm.amdgcn.permlanex16(i32, i32, i32, i32, i1, i1) #1
+declare i32 @llvm.amdgcn.permlane16.i32(i32, i32, i32, i32, i1, i1) #1
+declare i32 @llvm.amdgcn.permlanex16.i32(i32, i32, i32, i32, i1, i1) #1
 declare i32 @llvm.amdgcn.permlane16.var(i32, i32, i32, i1, i1) #1
 declare i32 @llvm.amdgcn.permlanex16.var(i32, i32, i32, i1, i1) #1
 declare i32 @llvm.amdgcn.mov.dpp.i32(i32, i32, i32, i32, i1) #1
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane.ll
index 265d64f47bb23..b8dab361346c3 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane.ll
@@ -6,13 +6,16 @@
 ; RUN: llc -global-isel=0 -amdgpu-load-store-vectorizer=0 -mtriple=amdgcn -mcpu=gfx1200 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX12,GFX12-SDAG %s
 ; RUN: llc -global-isel=1 -amdgpu-load-store-vectorizer=0 -mtriple=amdgcn -mcpu=gfx1200 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX12,GFX12-GISEL %s
 
-declare i32 @llvm.amdgcn.permlane16(i32, i32, i32, i32, i1, i1)
+declare i32 @llvm.amdgcn.permlane16.i32(i32, i32, i32, i32, i1, i1)
+declare float @llvm.amdgcn.permlane16.f32(float, float, i32, i32, i1, i1)
+declare i64 @llvm.amdgcn.permlane16.i64(i64, i64, i32, i32, i1, i1)
+declare double @llvm.amdgcn.permlane16.f64(double, double, i32, i32, i1, i1)
 declare i32 @llvm.amdgcn.permlanex16(i32, i32, i32, i32, i1, i1)
 declare i32 @llvm.amdgcn.workitem.id.x()
 declare i32 @llvm.amdgcn.workitem.id.y()
 
-define amdgpu_kernel void @v_permlane16_b32_vss(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-LABEL: v_permlane16_b32_vss:
+define amdgpu_kernel void @v_permlane16_b32_vss_i32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlane16_b32_vss_i32:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_clause 0x1
 ; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
@@ -24,7 +27,7 @@ define amdgpu_kernel void @v_permlane16_b32_vss(ptr addrspace(1) %out, i32 %src0
 ; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
 ; GFX10-NEXT:    s_endpgm
 ;
-; GFX11-LABEL: v_permlane16_b32_vss:
+; GFX11-LABEL: v_permlane16_b32_vss_i32:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_clause 0x1
 ; GFX11-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
@@ -38,7 +41,7 @@ define amdgpu_kernel void @v_permlane16_b32_vss(ptr addrspace(1) %out, i32 %src0
 ; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 ;
-; GFX12-LABEL: v_permlane16_b32_vss:
+; GFX12-LABEL: v_permlane16_b32_vss_i32:
 ; GFX12:       ; %bb.0:
 ; GFX12-NEXT:    s_clause 0x1
 ; GFX12-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
@@ -51,13 +54,252 @@ define amdgpu_kernel void @v_permlane16_b32_vss(ptr addrspace(1) %out, i32 %src0
 ; GFX12-NEXT:    s_nop 0
 ; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-NEXT:    s_endpgm
-  %v = call i32 @llvm.amdgcn.permlane16(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  %v = call i32 @llvm.amdgcn.permlane16.i32(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false)
   store i32 %v, ptr addrspace(1) %out
   ret void
 }
 
-define amdgpu_kernel void @v_permlane16_b32_vii(ptr addrspace(1) %out, i32 %src0) {
-; GFX10-LABEL: v_permlane16_b32_vii:
+define amdgpu_kernel void @v_permlane16_b32_vss_f32(ptr addrspace(1) %out, float %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlane16_b32_vss_f32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-NEXT:    v_permlane16_b32 v0, v0, s7, s2
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlane16_b32_vss_f32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_permlane16_b32 v0, v0, s7, s0
+; GFX11-NEXT:    global_store_b32 v1, v0, s[4:5]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlane16_b32_vss_f32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_permlane16_b32 v0, v0, s7, s0
+; GFX12-NEXT:    global_store_b32 v1, v0, s[4:5]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %v = call float @llvm.amdgcn.permlane16.f32(float %src0, float %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vss_i64(ptr addrspace(1) %out, i64 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_vss_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_vss_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_vss_i64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s0, s1
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_vss_i64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s0, s1
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v1, s0, s1
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_vss_i64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s0, s1
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_vss_i64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s0, s1
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v1, s0, s1
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %v = call i64 @llvm.amdgcn.permlane16.i64(i64 %src0, i64 %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vss_f64(ptr addrspace(1) %out, double %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_vss_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_vss_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_vss_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s0, s1
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_vss_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s0, s1
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v1, s0, s1
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_vss_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s0, s1
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_vss_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s0, s1
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v1, s0, s1
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %v = call double @llvm.amdgcn.permlane16.f64(double %src0, double %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vii_i32(ptr addrspace(1) %out, i32 %src0) {
+; GFX10-LABEL: v_permlane16_b32_vii_i32:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_clause 0x1
 ; GFX10-NEXT:    s_load_dword s4, s[0:1], 0x2c
@@ -69,7 +311,7 @@ define amdgpu_kernel void @v_permlane16_b32_vii(ptr addrspace(1) %out, i32 %src0
 ; GFX10-NEXT:    global_store_dword v1, v0, s[2:3]
 ; GFX10-NEXT:    s_endpgm
 ;
-; GFX11-LABEL: v_permlane16_b32_vii:
+; GFX11-LABEL: v_permlane16_b32_vii_i32:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_clause 0x1
 ; GFX11-NEXT:    s_load_b32 s2, s[0:1], 0x2c
@@ -83,7 +325,7 @@ define amdgpu_kernel void @v_permlane16_b32_vii(ptr addrspace(1) %out, i32 %src0
 ; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 ;
-; GFX12-LABEL: v_permlane16_b32_vii:
+; GFX12-LABEL: v_permlane16_b32_vii_i32:
 ; GFX12:       ; %bb.0:
 ; GFX12-NEXT:    s_load_b96 s[0:2], s[0:1], 0x24
 ; GFX12-NEXT:    s_wait_kmcnt 0x0
@@ -94,326 +336,2057 @@ define amdgpu_kernel void @v_permlane16_b32_vii(ptr addrspace(1) %out, i32 %src0
 ; GFX12-NEXT:    s_nop 0
 ; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-NEXT:    s_endpgm
-  %v = call i32 @llvm.amdgcn.permlane16(i32 %src0, i32 %src0, i32 1, i32 2, i1 false, i1 false)
+  %v = call i32 @llvm.amdgcn.permlane16.i32(i32 %src0, i32 %src0, i32 1, i32 2, i1 false, i1 false)
   store i32 %v, ptr addrspace(1) %out
   ret void
 }
 
-; FIXME-GFX10PLUS: It is allowed to have both immediates as literals
-define amdgpu_kernel void @v_permlane16_b32_vll(ptr addrspace(1) %out, i32 %src0) {
-; GFX10-LABEL: v_permlane16_b32_vll:
+define amdgpu_kernel void @v_permlane16_b32_vii_f32(ptr addrspace(1) %out, float %src0) {
+; GFX10-LABEL: v_permlane16_b32_vii_f32:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_clause 0x1
 ; GFX10-NEXT:    s_load_dword s4, s[0:1], 0x2c
 ; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10-NEXT:    s_movk_i32 s0, 0x1234
 ; GFX10-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    v_mov_b32_e32 v0, s4
-; GFX10-NEXT:    v_permlane16_b32 v0, v0, s0, 0xc1d1
+; GFX10-NEXT:    v_permlane16_b32 v0, v0, 1, 2
 ; GFX10-NEXT:    global_store_dword v1, v0, s[2:3]
 ; GFX10-NEXT:    s_endpgm
 ;
-; GFX11-LABEL: v_permlane16_b32_vll:
+; GFX11-LABEL: v_permlane16_b32_vii_f32:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_clause 0x1
 ; GFX11-NEXT:    s_load_b32 s2, s[0:1], 0x2c
 ; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
-; GFX11-NEXT:    s_movk_i32 s2, 0x1234
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX11-NEXT:    v_permlane16_b32 v0, v0, s2, 0xc1d1
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_permlane16_b32 v0, v0, 1, 2
 ; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX11-NEXT:    s_nop 0
 ; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 ;
-; GFX12-LABEL: v_permlane16_b32_vll:
+; GFX12-LABEL: v_permlane16_b32_vii_f32:
 ; GFX12:       ; %bb.0:
 ; GFX12-NEXT:    s_load_b96 s[0:2], s[0:1], 0x24
 ; GFX12-NEXT:    s_wait_kmcnt 0x0
 ; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
-; GFX12-NEXT:    s_movk_i32 s2, 0x1234
-; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX12-NEXT:    v_permlane16_b32 v0, v0, s2, 0xc1d1
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_permlane16_b32 v0, v0, 1, 2
 ; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX12-NEXT:    s_nop 0
 ; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-NEXT:    s_endpgm
-  %v = call i32 @llvm.amdgcn.permlane16(i32 %src0, i32 %src0, i32 4660, i32 49617, i1 false, i1 false)
-  store i32 %v, ptr addrspace(1) %out
+  %v = call float @llvm.amdgcn.permlane16.f32(float %src0, float %src0, i32 1, i32 2, i1 false, i1 false)
+  store float %v, ptr addrspace(1) %out
   ret void
 }
 
-define amdgpu_kernel void @v_permlane16_b32_vvv(ptr addrspace(1) %out, i32 %src0) {
-; GFX10-LABEL: v_permlane16_b32_vvv:
-; GFX10:       ; %bb.0:
-; GFX10-NEXT:    s_clause 0x1
-; GFX10-NEXT:    s_load_dword s4, s[0:1], 0x2c
-; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10-NEXT:    s_mov_b32 null, 0
-; GFX10-NEXT:    v_readfirstlane_b32 s0, v0
-; GFX10-NEXT:    v_readfirstlane_b32 s1, v1
-; GFX10-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    v_mov_b32_e32 v0, s4
-; GFX10-NEXT:    v_permlane16_b32 v0, v0, s0, s1
-; GFX10-NEXT:    global_store_dword v1, v0, s[2:3]
-; GFX10-NEXT:    s_endpgm
+define amdgpu_kernel void @v_permlane16_b32_vii_i64(ptr addrspace(1) %out, i64 %src0) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_vii_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, 1, 2
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, 1, 2
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_vii_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, 1, 2
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v1, 1, 2
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX10-GISEL-NEXT:    s_endpgm
 ;
-; GFX11-SDAG-LABEL: v_permlane16_b32_vvv:
+; GFX11-SDAG-LABEL: v_permlane16_b32_vii_i64:
 ; GFX11-SDAG:       ; %bb.0:
-; GFX11-SDAG-NEXT:    s_clause 0x1
-; GFX11-SDAG-NEXT:    s_load_b32 s2, s[0:1], 0x2c
-; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX11-SDAG-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
-; GFX11-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
-; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX11-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
 ; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-SDAG-NEXT:    v_mov_b32_e32 v1, s2
-; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, 0
-; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s3, s2
-; GFX11-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s3
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, 1, 2
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, 1, 2
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
 ; GFX11-SDAG-NEXT:    s_nop 0
 ; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-SDAG-NEXT:    s_endpgm
 ;
-; GFX11-GISEL-LABEL: v_permlane16_b32_vvv:
+; GFX11-GISEL-LABEL: v_permlane16_b32_vii_i64:
 ; GFX11-GISEL:       ; %bb.0:
-; GFX11-GISEL-NEXT:    s_clause 0x1
-; GFX11-GISEL-NEXT:    s_load_b32 s2, s[0:1], 0x2c
-; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX11-GISEL-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
-; GFX11-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
-; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_2)
-; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX11-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-GISEL-NEXT:    v_mov_b32_e32 v0, s2
-; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
-; GFX11-GISEL-NEXT:    v_mov_b32_e32 v1, 0
-; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s3, s4
-; GFX11-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, 1, 2
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v1, 1, 2
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
 ; GFX11-GISEL-NEXT:    s_nop 0
 ; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-GISEL-NEXT:    s_endpgm
 ;
-; GFX12-SDAG-LABEL: v_permlane16_b32_vvv:
+; GFX12-SDAG-LABEL: v_permlane16_b32_vii_i64:
 ; GFX12-SDAG:       ; %bb.0:
-; GFX12-SDAG-NEXT:    s_load_b96 s[0:2], s[0:1], 0x24
-; GFX12-SDAG-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
-; GFX12-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
-; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX12-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
 ; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
-; GFX12-SDAG-NEXT:    v_mov_b32_e32 v1, s2
-; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, 0
-; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s3, s2
-; GFX12-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s3
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, 1, 2
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, 1, 2
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
 ; GFX12-SDAG-NEXT:    s_nop 0
 ; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-SDAG-NEXT:    s_endpgm
 ;
-; GFX12-GISEL-LABEL: v_permlane16_b32_vvv:
+; GFX12-GISEL-LABEL: v_permlane16_b32_vii_i64:
 ; GFX12-GISEL:       ; %bb.0:
-; GFX12-GISEL-NEXT:    s_load_b96 s[0:2], s[0:1], 0x24
-; GFX12-GISEL-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
-; GFX12-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
-; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_2)
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX12-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX12-GISEL-NEXT:    v_mov_b32_e32 v0, s2
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
-; GFX12-GISEL-NEXT:    v_mov_b32_e32 v1, 0
-; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s3, s4
-; GFX12-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, 1, 2
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v1, 1, 2
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
 ; GFX12-GISEL-NEXT:    s_nop 0
 ; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-GISEL-NEXT:    s_endpgm
-  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
-  %tidy = call i32 @llvm.amdgcn.workitem.id.y()
-  %v = call i32 @llvm.amdgcn.permlane16(i32 %src0, i32 %src0, i32 %tidx, i32 %tidy, i1 false, i1 false)
-  store i32 %v, ptr addrspace(1) %out
+  %v = call i64 @llvm.amdgcn.permlane16.i64(i64 %src0, i64 %src0, i32 1, i32 2, i1 false, i1 false)
+  store i64 %v, ptr addrspace(1) %out
   ret void
 }
 
-define amdgpu_kernel void @v_permlane16_b32_vvs(ptr addrspace(1) %out, i32 %src0, i32 %src2) {
-; GFX10-SDAG-LABEL: v_permlane16_b32_vvs:
+define amdgpu_kernel void @v_permlane16_b32_vii_f64(ptr addrspace(1) %out, double %src0) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_vii_f64:
 ; GFX10-SDAG:       ; %bb.0:
 ; GFX10-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s2
-; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, 0
-; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3
-; GFX10-SDAG-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, 1, 2
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, 1, 2
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
 ; GFX10-SDAG-NEXT:    s_endpgm
 ;
-; GFX10-GISEL-LABEL: v_permlane16_b32_vvs:
+; GFX10-GISEL-LABEL: v_permlane16_b32_vii_f64:
 ; GFX10-GISEL:       ; %bb.0:
 ; GFX10-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
-; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
-; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s2
-; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s4, s3
-; GFX10-GISEL-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, 1, 2
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v1, 1, 2
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
 ; GFX10-GISEL-NEXT:    s_endpgm
 ;
-; GFX11-SDAG-LABEL: v_permlane16_b32_vvs:
+; GFX11-SDAG-LABEL: v_permlane16_b32_vii_f64:
 ; GFX11-SDAG:       ; %bb.0:
 ; GFX11-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
 ; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-SDAG-NEXT:    v_mov_b32_e32 v1, s2
-; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, 0
-; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3
-; GFX11-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s3
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, 1, 2
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, 1, 2
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_vii_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, 1, 2
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v1, 1, 2
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_vii_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s3
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, 1, 2
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, 1, 2
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_vii_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, 1, 2
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v1, 1, 2
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %v = call double @llvm.amdgcn.permlane16.f64(double %src0, double %src0, i32 1, i32 2, i1 false, i1 false)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+; FIXME-GFX10PLUS: It is allowed to have both immediates as literals
+define amdgpu_kernel void @v_permlane16_b32_vll_i32(ptr addrspace(1) %out, i32 %src0) {
+; GFX10-LABEL: v_permlane16_b32_vll_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dword s4, s[0:1], 0x2c
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX10-NEXT:    s_movk_i32 s0, 0x1234
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v0, s4
+; GFX10-NEXT:    v_permlane16_b32 v0, v0, s0, 0xc1d1
+; GFX10-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlane16_b32_vll_i32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX11-NEXT:    s_movk_i32 s2, 0x1234
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
+; GFX11-NEXT:    v_permlane16_b32 v0, v0, s2, 0xc1d1
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlane16_b32_vll_i32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_load_b96 s[0:2], s[0:1], 0x24
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX12-NEXT:    s_movk_i32 s2, 0x1234
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
+; GFX12-NEXT:    v_permlane16_b32 v0, v0, s2, 0xc1d1
+; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %v = call i32 @llvm.amdgcn.permlane16.i32(i32 %src0, i32 %src0, i32 4660, i32 49617, i1 false, i1 false)
+  store i32 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vll_i64(ptr addrspace(1) %out, i64 %src0) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_vll_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-SDAG-NEXT:    s_movk_i32 s2, 0x1234
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, 0xc1d1
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, 0xc1d1
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_vll_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-GISEL-NEXT:    s_movk_i32 s2, 0x1234
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, 0xc1d1
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, 0xc1d1
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_vll_i64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s3
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX11-SDAG-NEXT:    s_movk_i32 s2, 0x1234
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instid1(SALU_CYCLE_1)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, 0xc1d1
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, 0xc1d1
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_vll_i64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX11-GISEL-NEXT:    s_movk_i32 s2, 0x1234
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, 0xc1d1
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, 0xc1d1
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_vll_i64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s3
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX12-SDAG-NEXT:    s_movk_i32 s2, 0x1234
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instid1(SALU_CYCLE_1)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, 0xc1d1
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, 0xc1d1
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_vll_i64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX12-GISEL-NEXT:    s_movk_i32 s2, 0x1234
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, 0xc1d1
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, 0xc1d1
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %v = call i64 @llvm.amdgcn.permlane16.i64(i64 %src0, i64 %src0, i32 4660, i32 49617, i1 false, i1 false)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vll_f32(ptr addrspace(1) %out,float %src0) {
+; GFX10-LABEL: v_permlane16_b32_vll_f32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dword s4, s[0:1], 0x2c
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX10-NEXT:    s_movk_i32 s0, 0x1234
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v0, s4
+; GFX10-NEXT:    v_permlane16_b32 v0, v0, s0, 0xc1d1
+; GFX10-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlane16_b32_vll_f32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX11-NEXT:    s_movk_i32 s2, 0x1234
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
+; GFX11-NEXT:    v_permlane16_b32 v0, v0, s2, 0xc1d1
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlane16_b32_vll_f32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_load_b96 s[0:2], s[0:1], 0x24
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX12-NEXT:    s_movk_i32 s2, 0x1234
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
+; GFX12-NEXT:    v_permlane16_b32 v0, v0, s2, 0xc1d1
+; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %v = call float @llvm.amdgcn.permlane16.f32(float %src0, float %src0, i32 4660, i32 49617, i1 false, i1 false)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vll_f64(ptr addrspace(1) %out, double %src0) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_vll_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-SDAG-NEXT:    s_movk_i32 s2, 0x1234
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, 0xc1d1
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, 0xc1d1
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_vll_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-GISEL-NEXT:    s_movk_i32 s2, 0x1234
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, 0xc1d1
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, 0xc1d1
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_vll_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s3
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX11-SDAG-NEXT:    s_movk_i32 s2, 0x1234
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instid1(SALU_CYCLE_1)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, 0xc1d1
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, 0xc1d1
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_vll_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX11-GISEL-NEXT:    s_movk_i32 s2, 0x1234
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, 0xc1d1
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, 0xc1d1
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_vll_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s3
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX12-SDAG-NEXT:    s_movk_i32 s2, 0x1234
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instid1(SALU_CYCLE_1)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, 0xc1d1
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, 0xc1d1
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_vll_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX12-GISEL-NEXT:    s_movk_i32 s2, 0x1234
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, 0xc1d1
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, 0xc1d1
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %v = call double @llvm.amdgcn.permlane16.f64(double %src0, double %src0, i32 4660, i32 49617, i1 false, i1 false)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vvv_i32(ptr addrspace(1) %out, i32 %src0) {
+; GFX10-LABEL: v_permlane16_b32_vvv_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dword s4, s[0:1], 0x2c
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX10-NEXT:    s_mov_b32 null, 0
+; GFX10-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10-NEXT:    v_readfirstlane_b32 s1, v1
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v0, s4
+; GFX10-NEXT:    v_permlane16_b32 v0, v0, s0, s1
+; GFX10-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_vvv_i32:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b32 s2, s[0:1], 0x2c
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
+; GFX11-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s3, s2
+; GFX11-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_vvv_i32:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b32 s2, s[0:1], 0x2c
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
+; GFX11-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s3, s4
+; GFX11-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_vvv_i32:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_load_b96 s[0:2], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
+; GFX12-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s3, s2
+; GFX12-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_vvv_i32:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_load_b96 s[0:2], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
+; GFX12-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s3, s4
+; GFX12-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidy = call i32 @llvm.amdgcn.workitem.id.y()
+  %v = call i32 @llvm.amdgcn.permlane16.i32(i32 %src0, i32 %src0, i32 %tidx, i32 %tidy, i1 false, i1 false)
+  store i32 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vvv_i64(ptr addrspace(1) %out, i64 %src0) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_vvv_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v1
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, s4, s5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_vvv_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s5, v1
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s4, s5
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v1, s4, s5
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlane16_b32_vvv_i64:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
+; GFX11-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_4) | instid1(VALU_DEP_2)
+; GFX11-NEXT:    v_readfirstlane_b32 s5, v0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_mov_b32_e32 v0, s2
+; GFX11-NEXT:    v_readfirstlane_b32 s4, v1
+; GFX11-NEXT:    v_mov_b32_e32 v1, s3
+; GFX11-NEXT:    v_permlane16_b32 v0, v0, s4, s5
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-NEXT:    v_permlane16_b32 v1, v1, s4, s5
+; GFX11-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlane16_b32_vvv_i64:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
+; GFX12-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_4) | instid1(VALU_DEP_2)
+; GFX12-NEXT:    v_readfirstlane_b32 s5, v0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_mov_b32_e32 v0, s2
+; GFX12-NEXT:    v_readfirstlane_b32 s4, v1
+; GFX12-NEXT:    v_mov_b32_e32 v1, s3
+; GFX12-NEXT:    v_permlane16_b32 v0, v0, s4, s5
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-NEXT:    v_permlane16_b32 v1, v1, s4, s5
+; GFX12-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidy = call i32 @llvm.amdgcn.workitem.id.y()
+  %v = call i64 @llvm.amdgcn.permlane16.i64(i64 %src0, i64 %src0, i32 %tidx, i32 %tidy, i1 false, i1 false)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vvv_f32(ptr addrspace(1) %out, float %src0) {
+; GFX10-LABEL: v_permlane16_b32_vvv_f32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dword s4, s[0:1], 0x2c
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX10-NEXT:    s_mov_b32 null, 0
+; GFX10-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10-NEXT:    v_readfirstlane_b32 s1, v1
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v0, s4
+; GFX10-NEXT:    v_permlane16_b32 v0, v0, s0, s1
+; GFX10-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_vvv_f32:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b32 s2, s[0:1], 0x2c
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
+; GFX11-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s3, s2
+; GFX11-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_vvv_f32:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b32 s2, s[0:1], 0x2c
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
+; GFX11-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s3, s4
+; GFX11-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_vvv_f32:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_load_b96 s[0:2], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
+; GFX12-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s3, s2
+; GFX12-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_vvv_f32:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_load_b96 s[0:2], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
+; GFX12-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s3, s4
+; GFX12-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidy = call i32 @llvm.amdgcn.workitem.id.y()
+  %v = call float @llvm.amdgcn.permlane16.f32(float %src0, float %src0, i32 %tidx, i32 %tidy, i1 false, i1 false)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vvv_f64(ptr addrspace(1) %out, double %src0) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_vvv_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v1
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, s4, s5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_vvv_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s5, v1
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s4, s5
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v1, s4, s5
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlane16_b32_vvv_f64:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
+; GFX11-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_4) | instid1(VALU_DEP_2)
+; GFX11-NEXT:    v_readfirstlane_b32 s5, v0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_mov_b32_e32 v0, s2
+; GFX11-NEXT:    v_readfirstlane_b32 s4, v1
+; GFX11-NEXT:    v_mov_b32_e32 v1, s3
+; GFX11-NEXT:    v_permlane16_b32 v0, v0, s4, s5
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-NEXT:    v_permlane16_b32 v1, v1, s4, s5
+; GFX11-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlane16_b32_vvv_f64:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
+; GFX12-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_4) | instid1(VALU_DEP_2)
+; GFX12-NEXT:    v_readfirstlane_b32 s5, v0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_mov_b32_e32 v0, s2
+; GFX12-NEXT:    v_readfirstlane_b32 s4, v1
+; GFX12-NEXT:    v_mov_b32_e32 v1, s3
+; GFX12-NEXT:    v_permlane16_b32 v0, v0, s4, s5
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-NEXT:    v_permlane16_b32 v1, v1, s4, s5
+; GFX12-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidy = call i32 @llvm.amdgcn.workitem.id.y()
+  %v = call double @llvm.amdgcn.permlane16.f64(double %src0, double %src0, i32 %tidx, i32 %tidy, i1 false, i1 false)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vvs_i32(ptr addrspace(1) %out, i32 %src0, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_vvs_i32:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX10-SDAG-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_vvs_i32:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s4, s3
+; GFX10-GISEL-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_vvs_i32:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX11-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_vvs_i32:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s4, s3
+; GFX11-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_vvs_i32:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX12-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_vvs_i32:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s4, s3
+; GFX12-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %v = call i32 @llvm.amdgcn.permlane16.i32(i32 %src0, i32 %src0, i32 %tidx, i32 %src2, i1 false, i1 false)
+  store i32 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vvs_i64(ptr addrspace(1) %out, i64 %src0, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_vvs_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-SDAG-NEXT:    s_mov_b32 null, 0
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, s0, s2
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s0, s2
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_vvs_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-GISEL-NEXT:    s_mov_b32 null, 0
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s0, s2
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v1, s0, s2
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_vvs_i64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s1, s0
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, s1, s0
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_vvs_i64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s1, s0
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v1, s1, s0
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_vvs_i64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s1, s0
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, s1, s0
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_vvs_i64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s1, s0
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v1, s1, s0
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %v = call i64 @llvm.amdgcn.permlane16.i64(i64 %src0, i64 %src0, i32 %tidx, i32 %src2, i1 false, i1 false)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vvs_f32(ptr addrspace(1) %out, float %src0, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_vvs_f32:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX10-SDAG-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_vvs_f32:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s4, s3
+; GFX10-GISEL-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_vvs_f32:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX11-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_vvs_f32:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s4, s3
+; GFX11-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_vvs_f32:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX12-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_vvs_f32:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s4, s3
+; GFX12-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %v = call float @llvm.amdgcn.permlane16.f32(float %src0, float %src0, i32 %tidx, i32 %src2, i1 false, i1 false)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vvs_f64(ptr addrspace(1) %out, double %src0, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_vvs_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-SDAG-NEXT:    s_mov_b32 null, 0
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, s0, s2
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s0, s2
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_vvs_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-GISEL-NEXT:    s_mov_b32 null, 0
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s0, s2
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v1, s0, s2
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_vvs_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s1, s0
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, s1, s0
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_vvs_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s1, s0
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v1, s1, s0
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_vvs_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s1, s0
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, s1, s0
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_vvs_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s1, s0
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v1, s1, s0
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %v = call double @llvm.amdgcn.permlane16.f64(double %src0, double %src0, i32 %tidx, i32 %src2, i1 false, i1 false)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vsv_i32(ptr addrspace(1) %out, i32 %src0, i32 %src1) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_vsv_i32:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s3, s2
+; GFX10-SDAG-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_vsv_i32:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s4, v1
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s3, s4
+; GFX10-GISEL-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_vsv_i32:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s3, s2
+; GFX11-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_vsv_i32:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s3, s4
+; GFX11-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_vsv_i32:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s3, s2
+; GFX12-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_vsv_i32:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s3, s4
+; GFX12-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidy = call i32 @llvm.amdgcn.workitem.id.y()
+  %v = call i32 @llvm.amdgcn.permlane16.i32(i32 %src0, i32 %src0, i32 %src1, i32 %tidy, i1 false, i1 false)
+  store i32 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vsv_i64(ptr addrspace(1) %out, i64 %src0, i32 %src1) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_vsv_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-SDAG-NEXT:    s_mov_b32 null, 0
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s0
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s0
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_vsv_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-GISEL-NEXT:    s_mov_b32 null, 0
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s0
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s0
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_vsv_i64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v1, s7 :: v_dual_mov_b32 v0, s6
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s0, s1
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_vsv_i64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s0, s1
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v1, s0, s1
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_vsv_i64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v1, s7 :: v_dual_mov_b32 v0, s6
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s0, s1
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_vsv_i64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s0, s1
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v1, s0, s1
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidy = call i32 @llvm.amdgcn.workitem.id.y()
+  %v = call i64 @llvm.amdgcn.permlane16(i64 %src0, i64 %src0, i32 %src1, i32 %tidy, i1 false, i1 false)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vsv_f32(ptr addrspace(1) %out, float %src0, i32 %src1) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_vsv_f32:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s3, s2
+; GFX10-SDAG-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_vsv_f32:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s4, v1
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s3, s4
+; GFX10-GISEL-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_vsv_f32:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s3, s2
+; GFX11-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_vsv_f32:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s3, s4
+; GFX11-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_vsv_f32:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s3, s2
+; GFX12-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_vsv_f32:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s3, s4
+; GFX12-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidy = call i32 @llvm.amdgcn.workitem.id.y()
+  %v = call float @llvm.amdgcn.permlane16.f32(float %src0, float %src0, i32 %src1, i32 %tidy, i1 false, i1 false)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vsv_f64(ptr addrspace(1) %out, double %src0, i32 %src1) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_vsv_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-SDAG-NEXT:    s_mov_b32 null, 0
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s0
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s0
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_vsv_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-GISEL-NEXT:    s_mov_b32 null, 0
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s0
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s0
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_vsv_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v1, s7 :: v_dual_mov_b32 v0, s6
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s0, s1
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_vsv_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s0, s1
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v1, s0, s1
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_vsv_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v1, s7 :: v_dual_mov_b32 v0, s6
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s0, s1
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_vsv_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s0, s1
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v1, s0, s1
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidy = call i32 @llvm.amdgcn.workitem.id.y()
+  %v = call double @llvm.amdgcn.permlane16.f64(double %src0, double %src0, i32 %src1, i32 %tidy, i1 false, i1 false)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vss_fi_i32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlane16_b32_vss_fi_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-NEXT:    v_permlane16_b32 v0, v0, s7, s2 op_sel:[1,0]
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlane16_b32_vss_fi_i32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_permlane16_b32 v0, v0, s7, s0 op_sel:[1,0]
+; GFX11-NEXT:    global_store_b32 v1, v0, s[4:5]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlane16_b32_vss_fi_i32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_permlane16_b32 v0, v0, s7, s0 op_sel:[1,0]
+; GFX12-NEXT:    global_store_b32 v1, v0, s[4:5]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %v = call i32 @llvm.amdgcn.permlane16.i32(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 true, i1 false)
+  store i32 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vss_fi_i64(ptr addrspace(1) %out, i64 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_vss_fi_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[1,0]
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_vss_fi_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[1,0]
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_vss_fi_i64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s0, s1 op_sel:[1,0]
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, s0, s1 op_sel:[1,0]
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_vss_fi_i64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s0, s1 op_sel:[1,0]
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v1, s0, s1 op_sel:[1,0]
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_vss_fi_i64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s0, s1 op_sel:[1,0]
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, s0, s1 op_sel:[1,0]
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_vss_fi_i64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s0, s1 op_sel:[1,0]
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v1, s0, s1 op_sel:[1,0]
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %v = call i64 @llvm.amdgcn.permlane16.i64(i64 %src0, i64 %src0, i32 %src1, i32 %src2, i1 true, i1 false)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vss_fi_f32(ptr addrspace(1) %out, float %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlane16_b32_vss_fi_f32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-NEXT:    v_permlane16_b32 v0, v0, s7, s2 op_sel:[1,0]
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlane16_b32_vss_fi_f32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_permlane16_b32 v0, v0, s7, s0 op_sel:[1,0]
+; GFX11-NEXT:    global_store_b32 v1, v0, s[4:5]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlane16_b32_vss_fi_f32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_permlane16_b32 v0, v0, s7, s0 op_sel:[1,0]
+; GFX12-NEXT:    global_store_b32 v1, v0, s[4:5]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %v = call float @llvm.amdgcn.permlane16.f32(float %src0, float %src0, i32 %src1, i32 %src2, i1 true, i1 false)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vss_fi_f64(ptr addrspace(1) %out, double %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_vss_fi_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[1,0]
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_vss_fi_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[1,0]
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_vss_fi_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s0, s1 op_sel:[1,0]
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, s0, s1 op_sel:[1,0]
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
 ; GFX11-SDAG-NEXT:    s_nop 0
 ; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-SDAG-NEXT:    s_endpgm
 ;
-; GFX11-GISEL-LABEL: v_permlane16_b32_vvs:
+; GFX11-GISEL-LABEL: v_permlane16_b32_vss_fi_f64:
 ; GFX11-GISEL:       ; %bb.0:
-; GFX11-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
-; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
-; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s4, s3
-; GFX11-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s0, s1 op_sel:[1,0]
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v1, s0, s1 op_sel:[1,0]
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
 ; GFX11-GISEL-NEXT:    s_nop 0
 ; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-GISEL-NEXT:    s_endpgm
 ;
-; GFX12-SDAG-LABEL: v_permlane16_b32_vvs:
+; GFX12-SDAG-LABEL: v_permlane16_b32_vss_fi_f64:
 ; GFX12-SDAG:       ; %bb.0:
-; GFX12-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
 ; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
-; GFX12-SDAG-NEXT:    v_mov_b32_e32 v1, s2
-; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, 0
-; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3
-; GFX12-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s0, s1 op_sel:[1,0]
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, s0, s1 op_sel:[1,0]
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
 ; GFX12-SDAG-NEXT:    s_nop 0
 ; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-SDAG-NEXT:    s_endpgm
 ;
-; GFX12-GISEL-LABEL: v_permlane16_b32_vvs:
+; GFX12-GISEL-LABEL: v_permlane16_b32_vss_fi_f64:
 ; GFX12-GISEL:       ; %bb.0:
-; GFX12-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
-; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s4, s3
-; GFX12-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s0, s1 op_sel:[1,0]
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v1, s0, s1 op_sel:[1,0]
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
 ; GFX12-GISEL-NEXT:    s_nop 0
 ; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-GISEL-NEXT:    s_endpgm
-  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
-  %v = call i32 @llvm.amdgcn.permlane16(i32 %src0, i32 %src0, i32 %tidx, i32 %src2, i1 false, i1 false)
+  %v = call double @llvm.amdgcn.permlane16.f64(double %src0, double %src0, i32 %src1, i32 %src2, i1 true, i1 false)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vss_bc_i32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlane16_b32_vss_bc_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-NEXT:    v_permlane16_b32 v0, v0, s7, s2 op_sel:[0,1]
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlane16_b32_vss_bc_i32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_permlane16_b32 v0, v0, s7, s0 op_sel:[0,1]
+; GFX11-NEXT:    global_store_b32 v1, v0, s[4:5]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlane16_b32_vss_bc_i32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_permlane16_b32 v0, v0, s7, s0 op_sel:[0,1]
+; GFX12-NEXT:    global_store_b32 v1, v0, s[4:5]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %v = call i32 @llvm.amdgcn.permlane16.i32(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 true)
   store i32 %v, ptr addrspace(1) %out
   ret void
 }
 
-define amdgpu_kernel void @v_permlane16_b32_vsv(ptr addrspace(1) %out, i32 %src0, i32 %src1) {
-; GFX10-SDAG-LABEL: v_permlane16_b32_vsv:
+define amdgpu_kernel void @v_permlane16_b32_vss_bc_i64(ptr addrspace(1) %out, i64 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_vss_bc_i64:
 ; GFX10-SDAG:       ; %bb.0:
-; GFX10-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s2
-; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s2, v1
-; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s3, s2
-; GFX10-SDAG-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[0,1]
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
 ; GFX10-SDAG-NEXT:    s_endpgm
 ;
-; GFX10-GISEL-LABEL: v_permlane16_b32_vsv:
+; GFX10-GISEL-LABEL: v_permlane16_b32_vss_bc_i64:
 ; GFX10-GISEL:       ; %bb.0:
-; GFX10-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
-; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s4, v1
-; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s2
-; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s3, s4
-; GFX10-GISEL-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[0,1]
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
 ; GFX10-GISEL-NEXT:    s_endpgm
 ;
-; GFX11-SDAG-LABEL: v_permlane16_b32_vsv:
+; GFX11-SDAG-LABEL: v_permlane16_b32_vss_bc_i64:
 ; GFX11-SDAG:       ; %bb.0:
-; GFX11-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
-; GFX11-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
 ; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-SDAG-NEXT:    v_mov_b32_e32 v1, s2
-; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, 0
-; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s3, s2
-; GFX11-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s0, s1 op_sel:[0,1]
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, s0, s1 op_sel:[0,1]
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
 ; GFX11-SDAG-NEXT:    s_nop 0
 ; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-SDAG-NEXT:    s_endpgm
 ;
-; GFX11-GISEL-LABEL: v_permlane16_b32_vsv:
+; GFX11-GISEL-LABEL: v_permlane16_b32_vss_bc_i64:
 ; GFX11-GISEL:       ; %bb.0:
-; GFX11-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
-; GFX11-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
-; GFX11-GISEL-NEXT:    v_mov_b32_e32 v1, 0
-; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
-; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-GISEL-NEXT:    v_mov_b32_e32 v0, s2
-; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s3, s4
-; GFX11-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s0, s1 op_sel:[0,1]
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v1, s0, s1 op_sel:[0,1]
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
 ; GFX11-GISEL-NEXT:    s_nop 0
 ; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-GISEL-NEXT:    s_endpgm
 ;
-; GFX12-SDAG-LABEL: v_permlane16_b32_vsv:
+; GFX12-SDAG-LABEL: v_permlane16_b32_vss_bc_i64:
 ; GFX12-SDAG:       ; %bb.0:
-; GFX12-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
-; GFX12-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
 ; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
-; GFX12-SDAG-NEXT:    v_mov_b32_e32 v1, s2
-; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, 0
-; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s3, s2
-; GFX12-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s0, s1 op_sel:[0,1]
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, s0, s1 op_sel:[0,1]
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
 ; GFX12-SDAG-NEXT:    s_nop 0
 ; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-SDAG-NEXT:    s_endpgm
 ;
-; GFX12-GISEL-LABEL: v_permlane16_b32_vsv:
+; GFX12-GISEL-LABEL: v_permlane16_b32_vss_bc_i64:
 ; GFX12-GISEL:       ; %bb.0:
-; GFX12-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
-; GFX12-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
-; GFX12-GISEL-NEXT:    v_mov_b32_e32 v1, 0
-; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX12-GISEL-NEXT:    v_mov_b32_e32 v0, s2
-; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s3, s4
-; GFX12-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s0, s1 op_sel:[0,1]
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v1, s0, s1 op_sel:[0,1]
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
 ; GFX12-GISEL-NEXT:    s_nop 0
 ; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-GISEL-NEXT:    s_endpgm
-  %tidy = call i32 @llvm.amdgcn.workitem.id.y()
-  %v = call i32 @llvm.amdgcn.permlane16(i32 %src0, i32 %src0, i32 %src1, i32 %tidy, i1 false, i1 false)
-  store i32 %v, ptr addrspace(1) %out
+  %v = call i64 @llvm.amdgcn.permlane16.i64(i64 %src0, i64 %src0, i32 %src1, i32 %src2, i1 false, i1 true)
+  store i64 %v, ptr addrspace(1) %out
   ret void
 }
 
-define amdgpu_kernel void @v_permlane16_b32_vss_fi(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-LABEL: v_permlane16_b32_vss_fi:
+define amdgpu_kernel void @v_permlane16_b32_vss_bc_f32(ptr addrspace(1) %out, float %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlane16_b32_vss_bc_f32:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_clause 0x1
 ; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
@@ -421,11 +2394,11 @@ define amdgpu_kernel void @v_permlane16_b32_vss_fi(ptr addrspace(1) %out, i32 %s
 ; GFX10-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    v_mov_b32_e32 v0, s6
-; GFX10-NEXT:    v_permlane16_b32 v0, v0, s7, s2 op_sel:[1,0]
+; GFX10-NEXT:    v_permlane16_b32 v0, v0, s7, s2 op_sel:[0,1]
 ; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
 ; GFX10-NEXT:    s_endpgm
 ;
-; GFX11-LABEL: v_permlane16_b32_vss_fi:
+; GFX11-LABEL: v_permlane16_b32_vss_bc_f32:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_clause 0x1
 ; GFX11-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
@@ -433,13 +2406,13 @@ define amdgpu_kernel void @v_permlane16_b32_vss_fi(ptr addrspace(1) %out, i32 %s
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11-NEXT:    v_permlane16_b32 v0, v0, s7, s0 op_sel:[1,0]
+; GFX11-NEXT:    v_permlane16_b32 v0, v0, s7, s0 op_sel:[0,1]
 ; GFX11-NEXT:    global_store_b32 v1, v0, s[4:5]
 ; GFX11-NEXT:    s_nop 0
 ; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 ;
-; GFX12-LABEL: v_permlane16_b32_vss_fi:
+; GFX12-LABEL: v_permlane16_b32_vss_bc_f32:
 ; GFX12:       ; %bb.0:
 ; GFX12-NEXT:    s_clause 0x1
 ; GFX12-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
@@ -447,18 +2420,115 @@ define amdgpu_kernel void @v_permlane16_b32_vss_fi(ptr addrspace(1) %out, i32 %s
 ; GFX12-NEXT:    s_wait_kmcnt 0x0
 ; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
 ; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX12-NEXT:    v_permlane16_b32 v0, v0, s7, s0 op_sel:[1,0]
+; GFX12-NEXT:    v_permlane16_b32 v0, v0, s7, s0 op_sel:[0,1]
 ; GFX12-NEXT:    global_store_b32 v1, v0, s[4:5]
 ; GFX12-NEXT:    s_nop 0
 ; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-NEXT:    s_endpgm
-  %v = call i32 @llvm.amdgcn.permlane16(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 true, i1 false)
-  store i32 %v, ptr addrspace(1) %out
+  %v = call float @llvm.amdgcn.permlane16.f32(float %src0, float %src0, i32 %src1, i32 %src2, i1 false, i1 true)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vss_bc_f64(ptr addrspace(1) %out, double %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_vss_bc_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[0,1]
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_vss_bc_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[0,1]
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_vss_bc_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s0, s1 op_sel:[0,1]
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, s0, s1 op_sel:[0,1]
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_vss_bc_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s0, s1 op_sel:[0,1]
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v1, s0, s1 op_sel:[0,1]
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_vss_bc_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s0, s1 op_sel:[0,1]
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, s0, s1 op_sel:[0,1]
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_vss_bc_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s0, s1 op_sel:[0,1]
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v1, s0, s1 op_sel:[0,1]
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %v = call double @llvm.amdgcn.permlane16.f64(double %src0, double %src0, i32 %src1, i32 %src2, i1 false, i1 true)
+  store double %v, ptr addrspace(1) %out
   ret void
 }
 
-define amdgpu_kernel void @v_permlane16_b32_vss_bc(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-LABEL: v_permlane16_b32_vss_bc:
+define amdgpu_kernel void @v_permlane16_b32_vss_fi_bc_i32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlane16_b32_vss_fi_bc_i32:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_clause 0x1
 ; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
@@ -466,11 +2536,11 @@ define amdgpu_kernel void @v_permlane16_b32_vss_bc(ptr addrspace(1) %out, i32 %s
 ; GFX10-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    v_mov_b32_e32 v0, s6
-; GFX10-NEXT:    v_permlane16_b32 v0, v0, s7, s2 op_sel:[0,1]
+; GFX10-NEXT:    v_permlane16_b32 v0, v0, s7, s2 op_sel:[1,1]
 ; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
 ; GFX10-NEXT:    s_endpgm
 ;
-; GFX11-LABEL: v_permlane16_b32_vss_bc:
+; GFX11-LABEL: v_permlane16_b32_vss_fi_bc_i32:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_clause 0x1
 ; GFX11-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
@@ -478,13 +2548,13 @@ define amdgpu_kernel void @v_permlane16_b32_vss_bc(ptr addrspace(1) %out, i32 %s
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11-NEXT:    v_permlane16_b32 v0, v0, s7, s0 op_sel:[0,1]
+; GFX11-NEXT:    v_permlane16_b32 v0, v0, s7, s0 op_sel:[1,1]
 ; GFX11-NEXT:    global_store_b32 v1, v0, s[4:5]
 ; GFX11-NEXT:    s_nop 0
 ; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 ;
-; GFX12-LABEL: v_permlane16_b32_vss_bc:
+; GFX12-LABEL: v_permlane16_b32_vss_fi_bc_i32:
 ; GFX12:       ; %bb.0:
 ; GFX12-NEXT:    s_clause 0x1
 ; GFX12-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
@@ -492,18 +2562,115 @@ define amdgpu_kernel void @v_permlane16_b32_vss_bc(ptr addrspace(1) %out, i32 %s
 ; GFX12-NEXT:    s_wait_kmcnt 0x0
 ; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
 ; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX12-NEXT:    v_permlane16_b32 v0, v0, s7, s0 op_sel:[0,1]
+; GFX12-NEXT:    v_permlane16_b32 v0, v0, s7, s0 op_sel:[1,1]
 ; GFX12-NEXT:    global_store_b32 v1, v0, s[4:5]
 ; GFX12-NEXT:    s_nop 0
 ; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-NEXT:    s_endpgm
-  %v = call i32 @llvm.amdgcn.permlane16(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 true)
+  %v = call i32 @llvm.amdgcn.permlane16.i32(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 true, i1 true)
   store i32 %v, ptr addrspace(1) %out
   ret void
 }
 
-define amdgpu_kernel void @v_permlane16_b32_vss_fi_bc(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-LABEL: v_permlane16_b32_vss_fi_bc:
+define amdgpu_kernel void @v_permlane16_b32_vss_fi_bc_i64(ptr addrspace(1) %out, i64 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_vss_fi_bc_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[1,1]
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_vss_fi_bc_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[1,1]
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_vss_fi_bc_i64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s0, s1 op_sel:[1,1]
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, s0, s1 op_sel:[1,1]
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_vss_fi_bc_i64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s0, s1 op_sel:[1,1]
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v1, s0, s1 op_sel:[1,1]
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_vss_fi_bc_i64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s0, s1 op_sel:[1,1]
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, s0, s1 op_sel:[1,1]
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_vss_fi_bc_i64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s0, s1 op_sel:[1,1]
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v1, s0, s1 op_sel:[1,1]
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %v = call i64 @llvm.amdgcn.permlane16.i64(i64 %src0, i64 %src0, i32 %src1, i32 %src2, i1 true, i1 true)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vss_fi_bc_f32(ptr addrspace(1) %out, float %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlane16_b32_vss_fi_bc_f32:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_clause 0x1
 ; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
@@ -515,7 +2682,7 @@ define amdgpu_kernel void @v_permlane16_b32_vss_fi_bc(ptr addrspace(1) %out, i32
 ; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
 ; GFX10-NEXT:    s_endpgm
 ;
-; GFX11-LABEL: v_permlane16_b32_vss_fi_bc:
+; GFX11-LABEL: v_permlane16_b32_vss_fi_bc_f32:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_clause 0x1
 ; GFX11-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
@@ -529,7 +2696,7 @@ define amdgpu_kernel void @v_permlane16_b32_vss_fi_bc(ptr addrspace(1) %out, i32
 ; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 ;
-; GFX12-LABEL: v_permlane16_b32_vss_fi_bc:
+; GFX12-LABEL: v_permlane16_b32_vss_fi_bc_f32:
 ; GFX12:       ; %bb.0:
 ; GFX12-NEXT:    s_clause 0x1
 ; GFX12-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
@@ -542,8 +2709,105 @@ define amdgpu_kernel void @v_permlane16_b32_vss_fi_bc(ptr addrspace(1) %out, i32
 ; GFX12-NEXT:    s_nop 0
 ; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-NEXT:    s_endpgm
-  %v = call i32 @llvm.amdgcn.permlane16(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 true, i1 true)
-  store i32 %v, ptr addrspace(1) %out
+  %v = call float @llvm.amdgcn.permlane16.f32(float %src0, float %src0, i32 %src1, i32 %src2, i1 true, i1 true)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_vss_fi_bc_f64(ptr addrspace(1) %out, double %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_vss_fi_bc_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[1,1]
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_vss_fi_bc_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[1,1]
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_vss_fi_bc_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s0, s1 op_sel:[1,1]
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, s0, s1 op_sel:[1,1]
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_vss_fi_bc_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s0, s1 op_sel:[1,1]
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v1, s0, s1 op_sel:[1,1]
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_vss_fi_bc_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s0, s1 op_sel:[1,1]
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, s0, s1 op_sel:[1,1]
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_vss_fi_bc_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s0, s1 op_sel:[1,1]
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v1, s0, s1 op_sel:[1,1]
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %v = call double @llvm.amdgcn.permlane16.f64(double %src0, double %src0, i32 %src1, i32 %src2, i1 true, i1 true)
+  store double %v, ptr addrspace(1) %out
   ret void
 }
 
diff --git a/llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll b/llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll
index 483ea8ad57d1b..925e88d041715 100644
--- a/llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll
+++ b/llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll
@@ -2933,37 +2933,37 @@ define amdgpu_kernel void @update_dpp_undef_old(ptr addrspace(1) %out, i32 %in1)
 ; llvm.amdgcn.permlane16
 ; --------------------------------------------------------------------
 
-declare i32 @llvm.amdgcn.permlane16(i32, i32, i32, i32, i1 immarg, i1 immarg)
+declare i32 @llvm.amdgcn.permlane16.i32(i32, i32, i32, i32, i1 immarg, i1 immarg)
 
 define amdgpu_kernel void @permlane16(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
 ; CHECK-LABEL: @permlane16(
-; CHECK-NEXT:    [[RES:%.*]] = call i32 @llvm.amdgcn.permlane16(i32 12345, i32 [[SRC0:%.*]], i32 [[SRC1:%.*]], i32 [[SRC2:%.*]], i1 false, i1 false)
+; CHECK-NEXT:    [[RES:%.*]] = call i32 @llvm.amdgcn.permlane16.i32(i32 12345, i32 [[SRC0:%.*]], i32 [[SRC1:%.*]], i32 [[SRC2:%.*]], i1 false, i1 false)
 ; CHECK-NEXT:    store i32 [[RES]], ptr addrspace(1) [[OUT:%.*]], align 4
 ; CHECK-NEXT:    ret void
 ;
-  %res = call i32 @llvm.amdgcn.permlane16(i32 12345, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  %res = call i32 @llvm.amdgcn.permlane16.i32(i32 12345, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false)
   store i32 %res, ptr addrspace(1) %out
   ret void
 }
 
 define amdgpu_kernel void @permlane16_bound_ctrl(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
 ; CHECK-LABEL: @permlane16_bound_ctrl(
-; CHECK-NEXT:    [[RES:%.*]] = call i32 @llvm.amdgcn.permlane16(i32 undef, i32 [[SRC0:%.*]], i32 [[SRC1:%.*]], i32 [[SRC2:%.*]], i1 false, i1 true)
+; CHECK-NEXT:    [[RES:%.*]] = call i32 @llvm.amdgcn.permlane16.i32(i32 undef, i32 [[SRC0:%.*]], i32 [[SRC1:%.*]], i32 [[SRC2:%.*]], i1 false, i1 true)
 ; CHECK-NEXT:    store i32 [[RES]], ptr addrspace(1) [[OUT:%.*]], align 4
 ; CHECK-NEXT:    ret void
 ;
-  %res = call i32 @llvm.amdgcn.permlane16(i32 12345, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 true)
+  %res = call i32 @llvm.amdgcn.permlane16.i32(i32 12345, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 true)
   store i32 %res, ptr addrspace(1) %out
   ret void
 }
 
 define amdgpu_kernel void @permlane16_fetch_invalid_bound_ctrl(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
 ; CHECK-LABEL: @permlane16_fetch_invalid_bound_ctrl(
-; CHECK-NEXT:    [[RES:%.*]] = call i32 @llvm.amdgcn.permlane16(i32 undef, i32 [[SRC0:%.*]], i32 [[SRC1:%.*]], i32 [[SRC2:%.*]], i1 true, i1 true)
+; CHECK-NEXT:    [[RES:%.*]] = call i32 @llvm.amdgcn.permlane16.i32(i32 undef, i32 [[SRC0:%.*]], i32 [[SRC1:%.*]], i32 [[SRC2:%.*]], i1 true, i1 true)
 ; CHECK-NEXT:    store i32 [[RES]], ptr addrspace(1) [[OUT:%.*]], align 4
 ; CHECK-NEXT:    ret void
 ;
-  %res = call i32 @llvm.amdgcn.permlane16(i32 12345, i32 %src0, i32 %src1, i32 %src2, i1 true, i1 true)
+  %res = call i32 @llvm.amdgcn.permlane16.i32(i32 12345, i32 %src0, i32 %src1, i32 %src2, i1 true, i1 true)
   store i32 %res, ptr addrspace(1) %out
   ret void
 }
@@ -2972,37 +2972,37 @@ define amdgpu_kernel void @permlane16_fetch_invalid_bound_ctrl(ptr addrspace(1)
 ; llvm.amdgcn.permlanex16
 ; --------------------------------------------------------------------
 
-declare i32 @llvm.amdgcn.permlanex16(i32, i32, i32, i32, i1 immarg, i1 immarg)
+declare i32 @llvm.amdgcn.permlanex16.i32(i32, i32, i32, i32, i1 immarg, i1 immarg)
 
 define amdgpu_kernel void @permlanex16(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
 ; CHECK-LABEL: @permlanex16(
-; CHECK-NEXT:    [[RES:%.*]] = call i32 @llvm.amdgcn.permlanex16(i32 12345, i32 [[SRC0:%.*]], i32 [[SRC1:%.*]], i32 [[SRC2:%.*]], i1 false, i1 false)
+; CHECK-NEXT:    [[RES:%.*]] = call i32 @llvm.amdgcn.permlanex16.i32(i32 12345, i32 [[SRC0:%.*]], i32 [[SRC1:%.*]], i32 [[SRC2:%.*]], i1 false, i1 false)
 ; CHECK-NEXT:    store i32 [[RES]], ptr addrspace(1) [[OUT:%.*]], align 4
 ; CHECK-NEXT:    ret void
 ;
-  %res = call i32 @llvm.amdgcn.permlanex16(i32 12345, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  %res = call i32 @llvm.amdgcn.permlanex16.i32(i32 12345, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false)
   store i32 %res, ptr addrspace(1) %out
   ret void
 }
 
 define amdgpu_kernel void @permlanex16_bound_ctrl(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
 ; CHECK-LABEL: @permlanex16_bound_ctrl(
-; CHECK-NEXT:    [[RES:%.*]] = call i32 @llvm.amdgcn.permlanex16(i32 undef, i32 [[SRC0:%.*]], i32 [[SRC1:%.*]], i32 [[SRC2:%.*]], i1 false, i1 true)
+; CHECK-NEXT:    [[RES:%.*]] = call i32 @llvm.amdgcn.permlanex16.i32(i32 undef, i32 [[SRC0:%.*]], i32 [[SRC1:%.*]], i32 [[SRC2:%.*]], i1 false, i1 true)
 ; CHECK-NEXT:    store i32 [[RES]], ptr addrspace(1) [[OUT:%.*]], align 4
 ; CHECK-NEXT:    ret void
 ;
-  %res = call i32 @llvm.amdgcn.permlanex16(i32 12345, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 true)
+  %res = call i32 @llvm.amdgcn.permlanex16.i32(i32 12345, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 true)
   store i32 %res, ptr addrspace(1) %out
   ret void
 }
 
 define amdgpu_kernel void @permlanex16_fetch_invalid_bound_ctrl(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
 ; CHECK-LABEL: @permlanex16_fetch_invalid_bound_ctrl(
-; CHECK-NEXT:    [[RES:%.*]] = call i32 @llvm.amdgcn.permlanex16(i32 undef, i32 [[SRC0:%.*]], i32 [[SRC1:%.*]], i32 [[SRC2:%.*]], i1 true, i1 true)
+; CHECK-NEXT:    [[RES:%.*]] = call i32 @llvm.amdgcn.permlanex16.i32(i32 undef, i32 [[SRC0:%.*]], i32 [[SRC1:%.*]], i32 [[SRC2:%.*]], i1 true, i1 true)
 ; CHECK-NEXT:    store i32 [[RES]], ptr addrspace(1) [[OUT:%.*]], align 4
 ; CHECK-NEXT:    ret void
 ;
-  %res = call i32 @llvm.amdgcn.permlanex16(i32 12345, i32 %src0, i32 %src1, i32 %src2, i1 true, i1 true)
+  %res = call i32 @llvm.amdgcn.permlanex16.i32(i32 12345, i32 %src0, i32 %src1, i32 %src2, i1 true, i1 true)
   store i32 %res, ptr addrspace(1) %out
   ret void
 }
diff --git a/llvm/test/Verifier/AMDGPU/intrinsic-immarg.ll b/llvm/test/Verifier/AMDGPU/intrinsic-immarg.ll
index bb370a6d1dfeb..5a061fc286935 100644
--- a/llvm/test/Verifier/AMDGPU/intrinsic-immarg.ll
+++ b/llvm/test/Verifier/AMDGPU/intrinsic-immarg.ll
@@ -513,31 +513,31 @@ define i32 @test_udot4(i32 %arg0, i32 %arg1, i32 %arg2, i1 %arg3) {
   ret i32 %val
 }
 
-declare i32 @llvm.amdgcn.permlane16(i32, i32, i32, i32, i1, i1)
+declare i32 @llvm.amdgcn.permlane16.i32(i32, i32, i32, i32, i1, i1)
 define i32 @test_permlane16(ptr addrspace(1) %out, i32 %arg0, i32 %arg1, i32 %arg2, i1 %arg3, i1 %arg4) {
   ; CHECK: immarg operand has non-immediate parameter
   ; CHECK-NEXT: i1 %arg3
-  ; CHECK-NEXT: %v1 = call i32 @llvm.amdgcn.permlane16(i32 %arg0, i32 %arg0, i32 %arg1, i32 %arg2, i1 %arg3, i1 false)
-  %v1 = call i32 @llvm.amdgcn.permlane16(i32 %arg0, i32 %arg0, i32 %arg1, i32 %arg2, i1 %arg3, i1 false)
+  ; CHECK-NEXT: %v1 = call i32 @llvm.amdgcn.permlane16.i32(i32 %arg0, i32 %arg0, i32 %arg1, i32 %arg2, i1 %arg3, i1 false)
+  %v1 = call i32 @llvm.amdgcn.permlane16.i32(i32 %arg0, i32 %arg0, i32 %arg1, i32 %arg2, i1 %arg3, i1 false)
 
   ; CHECK: immarg operand has non-immediate parameter
   ; CHECK-NEXT: i1 %arg4
-  ; CHECK-NEXT: call i32 @llvm.amdgcn.permlane16(i32 %v2, i32 %arg0, i32 %arg1, i32 %arg2, i1 false, i1 %arg4)
-  %v2 = call i32 @llvm.amdgcn.permlane16(i32 %v2, i32 %arg0, i32 %arg1, i32 %arg2, i1 false, i1 %arg4)
+  ; CHECK-NEXT: call i32 @llvm.amdgcn.permlane16.i32(i32 %v2, i32 %arg0, i32 %arg1, i32 %arg2, i1 false, i1 %arg4)
+  %v2 = call i32 @llvm.amdgcn.permlane16.i32(i32 %v2, i32 %arg0, i32 %arg1, i32 %arg2, i1 false, i1 %arg4)
   ret i32 %v2
 }
 
-declare i32 @llvm.amdgcn.permlanex16(i32, i32, i32, i32, i1, i1)
+declare i32 @llvm.amdgcn.permlanex16.i32(i32, i32, i32, i32, i1, i1)
 define i32 @test_permlanex16(ptr addrspace(1) %out, i32 %arg0, i32 %arg1, i32 %arg2, i1 %arg3, i1 %arg4) {
   ; CHECK: immarg operand has non-immediate parameter
   ; CHECK-NEXT: i1 %arg3
-  ; CHECK-NEXT: %v1 = call i32 @llvm.amdgcn.permlanex16(i32 %arg0, i32 %arg0, i32 %arg1, i32 %arg2, i1 %arg3, i1 false)
-  %v1 = call i32 @llvm.amdgcn.permlanex16(i32 %arg0, i32 %arg0, i32 %arg1, i32 %arg2, i1 %arg3, i1 false)
+  ; CHECK-NEXT: %v1 = call i32 @llvm.amdgcn.permlanex16.i32(i32 %arg0, i32 %arg0, i32 %arg1, i32 %arg2, i1 %arg3, i1 false)
+  %v1 = call i32 @llvm.amdgcn.permlanex16.i32(i32 %arg0, i32 %arg0, i32 %arg1, i32 %arg2, i1 %arg3, i1 false)
 
   ; CHECK: immarg operand has non-immediate parameter
   ; CHECK-NEXT: i1 %arg4
-  ; CHECK-NEXT: call i32 @llvm.amdgcn.permlanex16(i32 %v2, i32 %arg0, i32 %arg1, i32 %arg2, i1 false, i1 %arg4)
-  %v2 = call i32 @llvm.amdgcn.permlanex16(i32 %v2, i32 %arg0, i32 %arg1, i32 %arg2, i1 false, i1 %arg4)
+  ; CHECK-NEXT: call i32 @llvm.amdgcn.permlanex16.i32(i32 %v2, i32 %arg0, i32 %arg1, i32 %arg2, i1 false, i1 %arg4)
+  %v2 = call i32 @llvm.amdgcn.permlanex16.i32(i32 %v2, i32 %arg0, i32 %arg1, i32 %arg2, i1 false, i1 %arg4)
   ret i32 %v2
 }
 

>From 827d209bd87738da673da40c8e7d9ce870fe9140 Mon Sep 17 00:00:00 2001
From: Vikram <Vikram.Hegde at amd.com>
Date: Mon, 20 May 2024 07:36:01 -0400
Subject: [PATCH 3/9] fix builtin handling

---
 clang/lib/CodeGen/CGBuiltin.cpp                   | 2 +-
 clang/test/CodeGenOpenCL/builtins-amdgcn-gfx10.cl | 4 ++--
 clang/test/CodeGenOpenCL/builtins-amdgcn-gfx11.cl | 2 +-
 llvm/include/llvm/IR/IntrinsicsAMDGPU.td          | 5 ++---
 4 files changed, 6 insertions(+), 7 deletions(-)

diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index 9ce2f5b6c103b..4184074860bcc 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -18493,7 +18493,7 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID,
     llvm::Value *Src4 = EmitScalarExpr(E->getArg(4));
     llvm::Value *Src5 = EmitScalarExpr(E->getArg(5));
 
-    llvm::Function *F = CGM.getIntrinsic(IID, Src1->getType());
+    llvm::Function *F = CGM.getIntrinsic(IID, Src0->getType());
     return Builder.CreateCall(F, {Src0, Src1, Src2, Src3, Src4, Src5});
   }
   case AMDGPU::BI__builtin_amdgcn_permlane64:
diff --git a/clang/test/CodeGenOpenCL/builtins-amdgcn-gfx10.cl b/clang/test/CodeGenOpenCL/builtins-amdgcn-gfx10.cl
index 3c40370e7f107..bc4933edeb32e 100644
--- a/clang/test/CodeGenOpenCL/builtins-amdgcn-gfx10.cl
+++ b/clang/test/CodeGenOpenCL/builtins-amdgcn-gfx10.cl
@@ -7,13 +7,13 @@ typedef unsigned int uint;
 typedef unsigned long ulong;
 
 // CHECK-LABEL: @test_permlane16(
-// CHECK: call i32 @llvm.amdgcn.permlane16(i32 %a, i32 %b, i32 %c, i32 %d, i1 false, i1 false)
+// CHECK: call i32 @llvm.amdgcn.permlane16.i32(i32 %a, i32 %b, i32 %c, i32 %d, i1 false, i1 false)
 void test_permlane16(global uint* out, uint a, uint b, uint c, uint d) {
   *out = __builtin_amdgcn_permlane16(a, b, c, d, 0, 0);
 }
 
 // CHECK-LABEL: @test_permlanex16(
-// CHECK: call i32 @llvm.amdgcn.permlanex16(i32 %a, i32 %b, i32 %c, i32 %d, i1 false, i1 false)
+// CHECK: call i32 @llvm.amdgcn.permlanex16.i32(i32 %a, i32 %b, i32 %c, i32 %d, i1 false, i1 false)
 void test_permlanex16(global uint* out, uint a, uint b, uint c, uint d) {
   *out = __builtin_amdgcn_permlanex16(a, b, c, d, 0, 0);
 }
diff --git a/clang/test/CodeGenOpenCL/builtins-amdgcn-gfx11.cl b/clang/test/CodeGenOpenCL/builtins-amdgcn-gfx11.cl
index d17ff81e5d43c..2f45ef326eef2 100644
--- a/clang/test/CodeGenOpenCL/builtins-amdgcn-gfx11.cl
+++ b/clang/test/CodeGenOpenCL/builtins-amdgcn-gfx11.cl
@@ -35,7 +35,7 @@ void test_ds_bvh_stack_rtn(global uint2* out, uint addr, uint data, uint4 data1)
 }
 
 // CHECK-LABEL: @test_permlane64(
-// CHECK: call i32 @llvm.amdgcn.permlane64(i32 %a)
+// CHECK: call i32 @llvm.amdgcn.permlane64.i32(i32 %a)
 void test_permlane64(global uint* out, uint a) {
   *out = __builtin_amdgcn_permlane64(a);
 }
diff --git a/llvm/include/llvm/IR/IntrinsicsAMDGPU.td b/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
index 63f7f48e82e4a..ba2582588db94 100644
--- a/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
+++ b/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
@@ -2487,14 +2487,14 @@ def int_amdgcn_global_load_lds : AMDGPUGlobalLoadLDS;
 //===----------------------------------------------------------------------===//
 
 // llvm.amdgcn.permlane16 <old> <src0> <src1> <src2> <fi> <bound_control>
-def int_amdgcn_permlane16 : ClangBuiltin<"__builtin_amdgcn_permlane16">,
+def int_amdgcn_permlane16 :
   Intrinsic<[llvm_any_ty],
             [LLVMMatchType<0>, LLVMMatchType<0>, llvm_i32_ty, llvm_i32_ty, llvm_i1_ty, llvm_i1_ty],
             [IntrNoMem, IntrConvergent, IntrWillReturn,
              ImmArg<ArgIndex<4>>, ImmArg<ArgIndex<5>>, IntrNoCallback, IntrNoFree]>;
 
 // llvm.amdgcn.permlanex16 <old> <src0> <src1> <src2> <fi> <bound_control>
-def int_amdgcn_permlanex16 : ClangBuiltin<"__builtin_amdgcn_permlanex16">,
+def int_amdgcn_permlanex16 :
   Intrinsic<[llvm_any_ty],
             [LLVMMatchType<0>, LLVMMatchType<0>, llvm_i32_ty, llvm_i32_ty, llvm_i1_ty, llvm_i1_ty],
             [IntrNoMem, IntrConvergent, IntrWillReturn,
@@ -2539,7 +2539,6 @@ def int_amdgcn_image_bvh_intersect_ray :
 
 // llvm.amdgcn.permlane64 <src0>
 def int_amdgcn_permlane64 :
-  ClangBuiltin<"__builtin_amdgcn_permlane64">,
   Intrinsic<[llvm_any_ty], [LLVMMatchType<0>],
             [IntrNoMem, IntrConvergent, IntrWillReturn, IntrNoCallback, IntrNoFree]>;
 

>From 6047848af702cc5d781442ee8fb339a9ab567e65 Mon Sep 17 00:00:00 2001
From: Vikram <Vikram.Hegde at amd.com>
Date: Mon, 27 May 2024 06:47:09 +0000
Subject: [PATCH 4/9] Review comments

---
 clang/lib/CodeGen/CGBuiltin.cpp |  9 ++++-----
 llvm/docs/AMDGPUUsage.rst       | 19 +++++++++++++++++++
 2 files changed, 23 insertions(+), 5 deletions(-)

diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index 4184074860bcc..d9ab0051f0d23 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -18481,11 +18481,6 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID,
   }
   case AMDGPU::BI__builtin_amdgcn_permlane16:
   case AMDGPU::BI__builtin_amdgcn_permlanex16: {
-    Intrinsic::ID IID;
-    IID = BuiltinID == AMDGPU::BI__builtin_amdgcn_permlane16
-              ? Intrinsic::amdgcn_permlane16
-              : Intrinsic::amdgcn_permlanex16;
-
     llvm::Value *Src0 = EmitScalarExpr(E->getArg(0));
     llvm::Value *Src1 = EmitScalarExpr(E->getArg(1));
     llvm::Value *Src2 = EmitScalarExpr(E->getArg(2));
@@ -18493,6 +18488,10 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID,
     llvm::Value *Src4 = EmitScalarExpr(E->getArg(4));
     llvm::Value *Src5 = EmitScalarExpr(E->getArg(5));
 
+    Intrinsic::ID IID = BuiltinID == AMDGPU::BI__builtin_amdgcn_permlane16
+                            ? Intrinsic::amdgcn_permlane16
+                            : Intrinsic::amdgcn_permlanex16;
+
     llvm::Function *F = CGM.getIntrinsic(IID, Src0->getType());
     return Builder.CreateCall(F, {Src0, Src1, Src2, Src3, Src4, Src5});
   }
diff --git a/llvm/docs/AMDGPUUsage.rst b/llvm/docs/AMDGPUUsage.rst
index 51969be85648f..97e2aecd614ed 100644
--- a/llvm/docs/AMDGPUUsage.rst
+++ b/llvm/docs/AMDGPUUsage.rst
@@ -1190,6 +1190,25 @@ The AMDGPU backend implements the following LLVM IR intrinsics.
                                                    reduction will be performed using default iterative strategy.
                                                    Intrinsic is currently only implemented for i32.
 
+  llvm.amdgcn.permlane16                           Provides direct access to v_permlane16_b32. Performs arbitrary gather-style
+                                                   operation within a row (16 contiguous lanes) of the second input operand.
+                                                   The third and fourth inputs must be scalar values. these are combined into
+                                                   a single 64-bit value representing lane selects used to swizzle within each
+                                                   row. Currently implemented for i16, i32, float, half, bf16, v2i16, v2f16 and
+                                                   types whose sizes are multiples of 32-bit.
+
+  llvm.amdgcn.permlanex16                          Provides direct access to v_permlanex16_b32. Performs arbitrary gather-style
+                                                   operation across two rows of the second input operand (each row is 16 contiguous
+                                                   lanes). The third and fourth inputs must be scalar values. these are combined
+                                                   into a single 64-bit value representing lane selects used to swizzle within each
+                                                   row. Currently implemented for i16, i32, float, half, bf16, v2i16, v2f16 and types
+                                                   whose sizes are multiples of 32-bit.
+
+  llvm.amdgcn.permlane64                           Provides direct access to v_permlane64_b32. Performs a specific permutation across
+                                                   lanes of the input operand where the high half and low half of a wave64 are swapped.
+                                                   Performs no operation in wave32 mode. Currently implemented for i16, i32, float,
+                                                   half, bf16, v2i16, v2f16 and types whose sizes are multiples of 32-bit.
+
   llvm.amdgcn.udot2                                Provides direct access to v_dot2_u32_u16 across targets which
                                                    support such instructions. This performs unsigned dot product
                                                    with two v2i16 operands, summed with the third i32 operand. The

>From 8a36f078ea65a267713cfe477f62500fa1964e21 Mon Sep 17 00:00:00 2001
From: Vikram <Vikram.Hegde at amd.com>
Date: Wed, 29 May 2024 11:54:16 +0000
Subject: [PATCH 5/9] updated test cases, added new pointer/vector tests

---
 .../CodeGen/AMDGPU/llvm.amdgcn.permlane.ll    | 7558 +++++++++++++++--
 .../AMDGPU/llvm.amdgcn.permlane.ptr.ll        |  694 ++
 .../CodeGen/AMDGPU/llvm.amdgcn.permlane64.ll  |  401 +-
 .../AMDGPU/llvm.amdgcn.permlane64.ptr.ll      |  180 +
 4 files changed, 8081 insertions(+), 752 deletions(-)
 create mode 100644 llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane.ptr.ll
 create mode 100644 llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ptr.ll

diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane.ll
index b8dab361346c3..1ae22c3eec185 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane.ll
@@ -6,10 +6,7 @@
 ; RUN: llc -global-isel=0 -amdgpu-load-store-vectorizer=0 -mtriple=amdgcn -mcpu=gfx1200 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX12,GFX12-SDAG %s
 ; RUN: llc -global-isel=1 -amdgpu-load-store-vectorizer=0 -mtriple=amdgcn -mcpu=gfx1200 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX12,GFX12-GISEL %s
 
-declare i32 @llvm.amdgcn.permlane16.i32(i32, i32, i32, i32, i1, i1)
-declare float @llvm.amdgcn.permlane16.f32(float, float, i32, i32, i1, i1)
-declare i64 @llvm.amdgcn.permlane16.i64(i64, i64, i32, i32, i1, i1)
-declare double @llvm.amdgcn.permlane16.f64(double, double, i32, i32, i1, i1)
+declare i32 @llvm.amdgcn.permlane16(i32, i32, i32, i32, i1, i1)
 declare i32 @llvm.amdgcn.permlanex16(i32, i32, i32, i32, i1, i1)
 declare i32 @llvm.amdgcn.workitem.id.x()
 declare i32 @llvm.amdgcn.workitem.id.y()
@@ -1752,7 +1749,7 @@ define amdgpu_kernel void @v_permlane16_b32_vsv_i64(ptr addrspace(1) %out, i64 %
 ; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-GISEL-NEXT:    s_endpgm
   %tidy = call i32 @llvm.amdgcn.workitem.id.y()
-  %v = call i64 @llvm.amdgcn.permlane16(i64 %src0, i64 %src0, i32 %src1, i32 %tidy, i1 false, i1 false)
+  %v = call i64 @llvm.amdgcn.permlane16.i64(i64 %src0, i64 %src0, i32 %src1, i32 %tidy, i1 false, i1 false)
   store i64 %v, ptr addrspace(1) %out
   ret void
 }
@@ -2811,8 +2808,8 @@ define amdgpu_kernel void @v_permlane16_b32_vss_fi_bc_f64(ptr addrspace(1) %out,
   ret void
 }
 
-define amdgpu_kernel void @v_permlanex16_b32_vss(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-LABEL: v_permlanex16_b32_vss:
+define amdgpu_kernel void @v_permlanex16_b32_vss_i32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlanex16_b32_vss_i32:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_clause 0x1
 ; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
@@ -2824,7 +2821,7 @@ define amdgpu_kernel void @v_permlanex16_b32_vss(ptr addrspace(1) %out, i32 %src
 ; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
 ; GFX10-NEXT:    s_endpgm
 ;
-; GFX11-LABEL: v_permlanex16_b32_vss:
+; GFX11-LABEL: v_permlanex16_b32_vss_i32:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_clause 0x1
 ; GFX11-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
@@ -2838,7 +2835,7 @@ define amdgpu_kernel void @v_permlanex16_b32_vss(ptr addrspace(1) %out, i32 %src
 ; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 ;
-; GFX12-LABEL: v_permlanex16_b32_vss:
+; GFX12-LABEL: v_permlanex16_b32_vss_i32:
 ; GFX12:       ; %bb.0:
 ; GFX12-NEXT:    s_clause 0x1
 ; GFX12-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
@@ -2851,1112 +2848,7189 @@ define amdgpu_kernel void @v_permlanex16_b32_vss(ptr addrspace(1) %out, i32 %src
 ; GFX12-NEXT:    s_nop 0
 ; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-NEXT:    s_endpgm
-  %v = call i32 @llvm.amdgcn.permlanex16(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  %v = call i32 @llvm.amdgcn.permlanex16.i32(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 false)
   store i32 %v, ptr addrspace(1) %out
   ret void
 }
 
-define amdgpu_kernel void @v_permlanex16_b32_vii(ptr addrspace(1) %out, i32 %src0) {
-; GFX10-LABEL: v_permlanex16_b32_vii:
+define amdgpu_kernel void @v_permlanex16_b32_vss_f32(ptr addrspace(1) %out, float %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlanex16_b32_vss_f32:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_clause 0x1
-; GFX10-NEXT:    s_load_dword s4, s[0:1], 0x2c
-; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dword s2, s[0:1], 0x34
 ; GFX10-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    v_mov_b32_e32 v0, s4
-; GFX10-NEXT:    v_permlanex16_b32 v0, v0, 1, 2
-; GFX10-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s7, s2
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
 ; GFX10-NEXT:    s_endpgm
 ;
-; GFX11-LABEL: v_permlanex16_b32_vii:
+; GFX11-LABEL: v_permlanex16_b32_vss_f32:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_clause 0x1
-; GFX11-NEXT:    s_load_b32 s2, s[0:1], 0x2c
-; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-NEXT:    s_load_b32 s0, s[0:1], 0x34
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11-NEXT:    v_permlanex16_b32 v0, v0, 1, 2
-; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s7, s0
+; GFX11-NEXT:    global_store_b32 v1, v0, s[4:5]
 ; GFX11-NEXT:    s_nop 0
 ; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 ;
-; GFX12-LABEL: v_permlanex16_b32_vii:
+; GFX12-LABEL: v_permlanex16_b32_vss_f32:
 ; GFX12:       ; %bb.0:
-; GFX12-NEXT:    s_load_b96 s[0:2], s[0:1], 0x24
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-NEXT:    s_load_b32 s0, s[0:1], 0x34
 ; GFX12-NEXT:    s_wait_kmcnt 0x0
-; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
 ; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX12-NEXT:    v_permlanex16_b32 v0, v0, 1, 2
-; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s7, s0
+; GFX12-NEXT:    global_store_b32 v1, v0, s[4:5]
 ; GFX12-NEXT:    s_nop 0
 ; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-NEXT:    s_endpgm
-  %v = call i32 @llvm.amdgcn.permlanex16(i32 %src0, i32 %src0, i32 1, i32 2, i1 false, i1 false)
-  store i32 %v, ptr addrspace(1) %out
+  %v = call float @llvm.amdgcn.permlanex16.f32(float %src0, float %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store float %v, ptr addrspace(1) %out
   ret void
 }
 
-; FIXME-GFX10PLUS: It is allowed to have both immediates as literals
-define amdgpu_kernel void @v_permlanex16_b32_vll(ptr addrspace(1) %out, i32 %src0) {
-; GFX10-LABEL: v_permlanex16_b32_vll:
-; GFX10:       ; %bb.0:
-; GFX10-NEXT:    s_clause 0x1
-; GFX10-NEXT:    s_load_dword s4, s[0:1], 0x2c
-; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10-NEXT:    s_movk_i32 s0, 0x1234
-; GFX10-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    v_mov_b32_e32 v0, s4
-; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s0, 0xc1d1
-; GFX10-NEXT:    global_store_dword v1, v0, s[2:3]
-; GFX10-NEXT:    s_endpgm
-;
-; GFX11-LABEL: v_permlanex16_b32_vll:
-; GFX11:       ; %bb.0:
-; GFX11-NEXT:    s_clause 0x1
-; GFX11-NEXT:    s_load_b32 s2, s[0:1], 0x2c
-; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
-; GFX11-NEXT:    s_movk_i32 s2, 0x1234
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s2, 0xc1d1
-; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
-; GFX11-NEXT:    s_nop 0
-; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
-; GFX11-NEXT:    s_endpgm
+define amdgpu_kernel void @v_permlanex16_b32_vss_i64(ptr addrspace(1) %out, i64 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_vss_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
 ;
-; GFX12-LABEL: v_permlanex16_b32_vll:
-; GFX12:       ; %bb.0:
-; GFX12-NEXT:    s_load_b96 s[0:2], s[0:1], 0x24
-; GFX12-NEXT:    s_wait_kmcnt 0x0
-; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
-; GFX12-NEXT:    s_movk_i32 s2, 0x1234
-; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
-; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s2, 0xc1d1
-; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
-; GFX12-NEXT:    s_nop 0
-; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
-; GFX12-NEXT:    s_endpgm
-  %v = call i32 @llvm.amdgcn.permlanex16(i32 %src0, i32 %src0, i32 4660, i32 49617, i1 false, i1 false)
-  store i32 %v, ptr addrspace(1) %out
-  ret void
-}
-
-define amdgpu_kernel void @v_permlanex16_b32_vvv(ptr addrspace(1) %out, i32 %src0) {
-; GFX10-LABEL: v_permlanex16_b32_vvv:
-; GFX10:       ; %bb.0:
-; GFX10-NEXT:    s_clause 0x1
-; GFX10-NEXT:    s_load_dword s4, s[0:1], 0x2c
-; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
-; GFX10-NEXT:    s_mov_b32 null, 0
-; GFX10-NEXT:    v_readfirstlane_b32 s0, v0
-; GFX10-NEXT:    v_readfirstlane_b32 s1, v1
-; GFX10-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    v_mov_b32_e32 v0, s4
-; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s0, s1
-; GFX10-NEXT:    global_store_dword v1, v0, s[2:3]
-; GFX10-NEXT:    s_endpgm
+; GFX10-GISEL-LABEL: v_permlanex16_b32_vss_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
 ;
-; GFX11-SDAG-LABEL: v_permlanex16_b32_vvv:
+; GFX11-SDAG-LABEL: v_permlanex16_b32_vss_i64:
 ; GFX11-SDAG:       ; %bb.0:
 ; GFX11-SDAG-NEXT:    s_clause 0x1
-; GFX11-SDAG-NEXT:    s_load_b32 s2, s[0:1], 0x2c
-; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX11-SDAG-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
-; GFX11-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
-; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
 ; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-SDAG-NEXT:    v_mov_b32_e32 v1, s2
-; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, 0
-; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s3, s2
-; GFX11-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s0, s1
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
 ; GFX11-SDAG-NEXT:    s_nop 0
 ; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-SDAG-NEXT:    s_endpgm
 ;
-; GFX11-GISEL-LABEL: v_permlanex16_b32_vvv:
+; GFX11-GISEL-LABEL: v_permlanex16_b32_vss_i64:
 ; GFX11-GISEL:       ; %bb.0:
 ; GFX11-GISEL-NEXT:    s_clause 0x1
-; GFX11-GISEL-NEXT:    s_load_b32 s2, s[0:1], 0x2c
-; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX11-GISEL-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
-; GFX11-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
-; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_2)
-; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX11-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-GISEL-NEXT:    v_mov_b32_e32 v0, s2
-; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
-; GFX11-GISEL-NEXT:    v_mov_b32_e32 v1, 0
-; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s3, s4
-; GFX11-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s0, s1
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s0, s1
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
 ; GFX11-GISEL-NEXT:    s_nop 0
 ; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-GISEL-NEXT:    s_endpgm
 ;
-; GFX12-SDAG-LABEL: v_permlanex16_b32_vvv:
+; GFX12-SDAG-LABEL: v_permlanex16_b32_vss_i64:
 ; GFX12-SDAG:       ; %bb.0:
-; GFX12-SDAG-NEXT:    s_load_b96 s[0:2], s[0:1], 0x24
-; GFX12-SDAG-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
-; GFX12-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
-; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
-; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
 ; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
-; GFX12-SDAG-NEXT:    v_mov_b32_e32 v1, s2
-; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, 0
-; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s3, s2
-; GFX12-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s0, s1
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
 ; GFX12-SDAG-NEXT:    s_nop 0
 ; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-SDAG-NEXT:    s_endpgm
 ;
-; GFX12-GISEL-LABEL: v_permlanex16_b32_vvv:
+; GFX12-GISEL-LABEL: v_permlanex16_b32_vss_i64:
 ; GFX12-GISEL:       ; %bb.0:
-; GFX12-GISEL-NEXT:    s_load_b96 s[0:2], s[0:1], 0x24
-; GFX12-GISEL-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
-; GFX12-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
-; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_2)
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX12-GISEL-NEXT:    v_mov_b32_e32 v0, s2
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
-; GFX12-GISEL-NEXT:    v_mov_b32_e32 v1, 0
-; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s3, s4
-; GFX12-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s0, s1
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s0, s1
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
 ; GFX12-GISEL-NEXT:    s_nop 0
 ; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-GISEL-NEXT:    s_endpgm
-  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
-  %tidy = call i32 @llvm.amdgcn.workitem.id.y()
-  %v = call i32 @llvm.amdgcn.permlanex16(i32 %src0, i32 %src0, i32 %tidx, i32 %tidy, i1 false, i1 false)
-  store i32 %v, ptr addrspace(1) %out
+  %v = call i64 @llvm.amdgcn.permlanex16.i64(i64 %src0, i64 %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store i64 %v, ptr addrspace(1) %out
   ret void
 }
 
-define amdgpu_kernel void @v_permlanex16_b32_vvs(ptr addrspace(1) %out, i32 %src0, i32 %src2) {
-; GFX10-SDAG-LABEL: v_permlanex16_b32_vvs:
+define amdgpu_kernel void @v_permlanex16_b32_vss_f64(ptr addrspace(1) %out, double %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_vss_f64:
 ; GFX10-SDAG:       ; %bb.0:
-; GFX10-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s2
-; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s6
 ; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
-; GFX10-SDAG-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
 ; GFX10-SDAG-NEXT:    s_endpgm
 ;
-; GFX10-GISEL-LABEL: v_permlanex16_b32_vvs:
+; GFX10-GISEL-LABEL: v_permlanex16_b32_vss_f64:
 ; GFX10-GISEL:       ; %bb.0:
-; GFX10-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
-; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
-; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s2
-; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s4, s3
-; GFX10-GISEL-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
 ; GFX10-GISEL-NEXT:    s_endpgm
 ;
-; GFX11-SDAG-LABEL: v_permlanex16_b32_vvs:
+; GFX11-SDAG-LABEL: v_permlanex16_b32_vss_f64:
 ; GFX11-SDAG:       ; %bb.0:
-; GFX11-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
 ; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-SDAG-NEXT:    v_mov_b32_e32 v1, s2
-; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, 0
-; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
-; GFX11-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s0, s1
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
 ; GFX11-SDAG-NEXT:    s_nop 0
 ; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-SDAG-NEXT:    s_endpgm
 ;
-; GFX11-GISEL-LABEL: v_permlanex16_b32_vvs:
+; GFX11-GISEL-LABEL: v_permlanex16_b32_vss_f64:
 ; GFX11-GISEL:       ; %bb.0:
-; GFX11-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
-; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-GISEL-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
-; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s4, s3
-; GFX11-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s0, s1
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s0, s1
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
 ; GFX11-GISEL-NEXT:    s_nop 0
 ; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-GISEL-NEXT:    s_endpgm
 ;
-; GFX12-SDAG-LABEL: v_permlanex16_b32_vvs:
+; GFX12-SDAG-LABEL: v_permlanex16_b32_vss_f64:
 ; GFX12-SDAG:       ; %bb.0:
-; GFX12-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
 ; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
-; GFX12-SDAG-NEXT:    v_mov_b32_e32 v1, s2
-; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, 0
-; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
-; GFX12-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s0, s1
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
 ; GFX12-SDAG-NEXT:    s_nop 0
 ; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-SDAG-NEXT:    s_endpgm
 ;
-; GFX12-GISEL-LABEL: v_permlanex16_b32_vvs:
+; GFX12-GISEL-LABEL: v_permlanex16_b32_vss_f64:
 ; GFX12-GISEL:       ; %bb.0:
-; GFX12-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX12-GISEL-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
-; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s4, s3
-; GFX12-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s0, s1
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s0, s1
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
 ; GFX12-GISEL-NEXT:    s_nop 0
 ; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-GISEL-NEXT:    s_endpgm
-  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
-  %v = call i32 @llvm.amdgcn.permlanex16(i32 %src0, i32 %src0, i32 %tidx, i32 %src2, i1 false, i1 false)
+  %v = call double @llvm.amdgcn.permlanex16.f64(double %src0, double %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vii_i32(ptr addrspace(1) %out, i32 %src0) {
+; GFX10-LABEL: v_permlanex16_b32_vii_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dword s4, s[0:1], 0x2c
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v0, s4
+; GFX10-NEXT:    v_permlanex16_b32 v0, v0, 1, 2
+; GFX10-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlanex16_b32_vii_i32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_permlanex16_b32 v0, v0, 1, 2
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlanex16_b32_vii_i32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_load_b96 s[0:2], s[0:1], 0x24
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_permlanex16_b32 v0, v0, 1, 2
+; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %v = call i32 @llvm.amdgcn.permlanex16.i32(i32 %src0, i32 %src0, i32 1, i32 2, i1 false, i1 false)
   store i32 %v, ptr addrspace(1) %out
   ret void
 }
 
-define amdgpu_kernel void @v_permlanex16_b32_vsv(ptr addrspace(1) %out, i32 %src0, i32 %src1) {
-; GFX10-SDAG-LABEL: v_permlanex16_b32_vsv:
+define amdgpu_kernel void @v_permlanex16_b32_vii_f32(ptr addrspace(1) %out, float %src0) {
+; GFX10-LABEL: v_permlanex16_b32_vii_f32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dword s4, s[0:1], 0x2c
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v0, s4
+; GFX10-NEXT:    v_permlanex16_b32 v0, v0, 1, 2
+; GFX10-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlanex16_b32_vii_f32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_permlanex16_b32 v0, v0, 1, 2
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlanex16_b32_vii_f32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_load_b96 s[0:2], s[0:1], 0x24
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_permlanex16_b32 v0, v0, 1, 2
+; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %v = call float @llvm.amdgcn.permlanex16.f32(float %src0, float %src0, i32 1, i32 2, i1 false, i1 false)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vii_i64(ptr addrspace(1) %out, i64 %src0) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_vii_i64:
 ; GFX10-SDAG:       ; %bb.0:
 ; GFX10-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s2
-; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s2, v1
-; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s3, s2
-; GFX10-SDAG-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, 1, 2
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, 1, 2
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
 ; GFX10-SDAG-NEXT:    s_endpgm
 ;
-; GFX10-GISEL-LABEL: v_permlanex16_b32_vsv:
+; GFX10-GISEL-LABEL: v_permlanex16_b32_vii_i64:
 ; GFX10-GISEL:       ; %bb.0:
 ; GFX10-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
-; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s4, v1
-; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s2
-; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s3, s4
-; GFX10-GISEL-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, 1, 2
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v1, 1, 2
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
 ; GFX10-GISEL-NEXT:    s_endpgm
 ;
-; GFX11-SDAG-LABEL: v_permlanex16_b32_vsv:
+; GFX11-SDAG-LABEL: v_permlanex16_b32_vii_i64:
 ; GFX11-SDAG:       ; %bb.0:
 ; GFX11-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
-; GFX11-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
 ; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-SDAG-NEXT:    v_mov_b32_e32 v1, s2
-; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, 0
-; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s3, s2
-; GFX11-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s3
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, 1, 2
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, 1, 2
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
 ; GFX11-SDAG-NEXT:    s_nop 0
 ; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-SDAG-NEXT:    s_endpgm
 ;
-; GFX11-GISEL-LABEL: v_permlanex16_b32_vsv:
+; GFX11-GISEL-LABEL: v_permlanex16_b32_vii_i64:
 ; GFX11-GISEL:       ; %bb.0:
 ; GFX11-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
-; GFX11-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
-; GFX11-GISEL-NEXT:    v_mov_b32_e32 v1, 0
-; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
-; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-GISEL-NEXT:    v_mov_b32_e32 v0, s2
-; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s3, s4
-; GFX11-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, 1, 2
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v1, 1, 2
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
 ; GFX11-GISEL-NEXT:    s_nop 0
 ; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-GISEL-NEXT:    s_endpgm
 ;
-; GFX12-SDAG-LABEL: v_permlanex16_b32_vsv:
+; GFX12-SDAG-LABEL: v_permlanex16_b32_vii_i64:
 ; GFX12-SDAG:       ; %bb.0:
 ; GFX12-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
-; GFX12-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
 ; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
-; GFX12-SDAG-NEXT:    v_mov_b32_e32 v1, s2
-; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, 0
-; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s3, s2
-; GFX12-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s3
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, 1, 2
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, 1, 2
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
 ; GFX12-SDAG-NEXT:    s_nop 0
 ; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-SDAG-NEXT:    s_endpgm
 ;
-; GFX12-GISEL-LABEL: v_permlanex16_b32_vsv:
+; GFX12-GISEL-LABEL: v_permlanex16_b32_vii_i64:
 ; GFX12-GISEL:       ; %bb.0:
 ; GFX12-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
-; GFX12-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
-; GFX12-GISEL-NEXT:    v_mov_b32_e32 v1, 0
-; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
-; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX12-GISEL-NEXT:    v_mov_b32_e32 v0, s2
-; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s3, s4
-; GFX12-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, 1, 2
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v1, 1, 2
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
 ; GFX12-GISEL-NEXT:    s_nop 0
 ; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-GISEL-NEXT:    s_endpgm
-  %tidy = call i32 @llvm.amdgcn.workitem.id.y()
-  %v = call i32 @llvm.amdgcn.permlanex16(i32 %src0, i32 %src0, i32 %src1, i32 %tidy, i1 false, i1 false)
-  store i32 %v, ptr addrspace(1) %out
+  %v = call i64 @llvm.amdgcn.permlanex16.i64(i64 %src0, i64 %src0, i32 1, i32 2, i1 false, i1 false)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vii_f64(ptr addrspace(1) %out, double %src0) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_vii_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, 1, 2
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, 1, 2
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_vii_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, 1, 2
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v1, 1, 2
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_vii_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s3
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, 1, 2
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, 1, 2
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_vii_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, 1, 2
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v1, 1, 2
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_vii_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s3
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, 1, 2
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, 1, 2
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_vii_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, 1, 2
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v1, 1, 2
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %v = call double @llvm.amdgcn.permlanex16.f64(double %src0, double %src0, i32 1, i32 2, i1 false, i1 false)
+  store double %v, ptr addrspace(1) %out
   ret void
 }
 
-define amdgpu_kernel void @v_permlanex16_b32_vss_fi(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-LABEL: v_permlanex16_b32_vss_fi:
+; FIXME-GFX10PLUS: It is allowed to have both immediates as literals
+define amdgpu_kernel void @v_permlanex16_b32_vll_i32(ptr addrspace(1) %out, i32 %src0) {
+; GFX10-LABEL: v_permlanex16_b32_vll_i32:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_clause 0x1
-; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
-; GFX10-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-NEXT:    s_load_dword s4, s[0:1], 0x2c
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX10-NEXT:    s_movk_i32 s0, 0x1234
 ; GFX10-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    v_mov_b32_e32 v0, s6
-; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s7, s2 op_sel:[1,0]
-; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    v_mov_b32_e32 v0, s4
+; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s0, 0xc1d1
+; GFX10-NEXT:    global_store_dword v1, v0, s[2:3]
 ; GFX10-NEXT:    s_endpgm
 ;
-; GFX11-LABEL: v_permlanex16_b32_vss_fi:
+; GFX11-LABEL: v_permlanex16_b32_vll_i32:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_clause 0x1
-; GFX11-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
-; GFX11-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-NEXT:    s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s7, s0 op_sel:[1,0]
-; GFX11-NEXT:    global_store_b32 v1, v0, s[4:5]
+; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX11-NEXT:    s_movk_i32 s2, 0x1234
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
+; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s2, 0xc1d1
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX11-NEXT:    s_nop 0
 ; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 ;
-; GFX12-LABEL: v_permlanex16_b32_vss_fi:
+; GFX12-LABEL: v_permlanex16_b32_vll_i32:
 ; GFX12:       ; %bb.0:
-; GFX12-NEXT:    s_clause 0x1
-; GFX12-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
-; GFX12-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-NEXT:    s_load_b96 s[0:2], s[0:1], 0x24
 ; GFX12-NEXT:    s_wait_kmcnt 0x0
-; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
-; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s7, s0 op_sel:[1,0]
-; GFX12-NEXT:    global_store_b32 v1, v0, s[4:5]
+; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX12-NEXT:    s_movk_i32 s2, 0x1234
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
+; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s2, 0xc1d1
+; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX12-NEXT:    s_nop 0
 ; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-NEXT:    s_endpgm
-  %v = call i32 @llvm.amdgcn.permlanex16(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 true, i1 false)
+  %v = call i32 @llvm.amdgcn.permlanex16.i32(i32 %src0, i32 %src0, i32 4660, i32 49617, i1 false, i1 false)
   store i32 %v, ptr addrspace(1) %out
   ret void
 }
 
-define amdgpu_kernel void @v_permlanex16_b32_vss_bc(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-LABEL: v_permlanex16_b32_vss_bc:
+define amdgpu_kernel void @v_permlanex16_b32_vll_f32(ptr addrspace(1) %out, float %src0) {
+; GFX10-LABEL: v_permlanex16_b32_vll_f32:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_clause 0x1
-; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
-; GFX10-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-NEXT:    s_load_dword s4, s[0:1], 0x2c
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX10-NEXT:    s_movk_i32 s0, 0x1234
 ; GFX10-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    v_mov_b32_e32 v0, s6
-; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s7, s2 op_sel:[0,1]
-; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    v_mov_b32_e32 v0, s4
+; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s0, 0xc1d1
+; GFX10-NEXT:    global_store_dword v1, v0, s[2:3]
 ; GFX10-NEXT:    s_endpgm
 ;
-; GFX11-LABEL: v_permlanex16_b32_vss_bc:
+; GFX11-LABEL: v_permlanex16_b32_vll_f32:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_clause 0x1
-; GFX11-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
-; GFX11-NEXT:    s_load_b32 s0, s[0:1], 0x34
-; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s7, s0 op_sel:[0,1]
-; GFX11-NEXT:    global_store_b32 v1, v0, s[4:5]
-; GFX11-NEXT:    s_nop 0
-; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
-; GFX11-NEXT:    s_endpgm
-;
-; GFX12-LABEL: v_permlanex16_b32_vss_bc:
-; GFX12:       ; %bb.0:
-; GFX12-NEXT:    s_clause 0x1
-; GFX12-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
-; GFX12-NEXT:    s_load_b32 s0, s[0:1], 0x34
-; GFX12-NEXT:    s_wait_kmcnt 0x0
-; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
-; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s7, s0 op_sel:[0,1]
-; GFX12-NEXT:    global_store_b32 v1, v0, s[4:5]
-; GFX12-NEXT:    s_nop 0
-; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
-; GFX12-NEXT:    s_endpgm
-  %v = call i32 @llvm.amdgcn.permlanex16(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 true)
-  store i32 %v, ptr addrspace(1) %out
-  ret void
-}
-
-define amdgpu_kernel void @v_permlanex16_b32_vss_fi_bc(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-LABEL: v_permlanex16_b32_vss_fi_bc:
-; GFX10:       ; %bb.0:
-; GFX10-NEXT:    s_clause 0x1
-; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
-; GFX10-NEXT:    s_load_dword s2, s[0:1], 0x34
-; GFX10-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    v_mov_b32_e32 v0, s6
-; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s7, s2 op_sel:[1,1]
-; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
-; GFX10-NEXT:    s_endpgm
-;
-; GFX11-LABEL: v_permlanex16_b32_vss_fi_bc:
-; GFX11:       ; %bb.0:
-; GFX11-NEXT:    s_clause 0x1
-; GFX11-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
-; GFX11-NEXT:    s_load_b32 s0, s[0:1], 0x34
-; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s7, s0 op_sel:[1,1]
-; GFX11-NEXT:    global_store_b32 v1, v0, s[4:5]
-; GFX11-NEXT:    s_nop 0
-; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
-; GFX11-NEXT:    s_endpgm
-;
-; GFX12-LABEL: v_permlanex16_b32_vss_fi_bc:
-; GFX12:       ; %bb.0:
-; GFX12-NEXT:    s_clause 0x1
-; GFX12-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
-; GFX12-NEXT:    s_load_b32 s0, s[0:1], 0x34
-; GFX12-NEXT:    s_wait_kmcnt 0x0
-; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
-; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s7, s0 op_sel:[1,1]
-; GFX12-NEXT:    global_store_b32 v1, v0, s[4:5]
-; GFX12-NEXT:    s_nop 0
-; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
-; GFX12-NEXT:    s_endpgm
-  %v = call i32 @llvm.amdgcn.permlanex16(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 true, i1 true)
-  store i32 %v, ptr addrspace(1) %out
-  ret void
-}
-
-define amdgpu_kernel void @v_permlane16_b32_tid_tid(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-LABEL: v_permlane16_b32_tid_tid:
-; GFX10:       ; %bb.0:
-; GFX10-NEXT:    s_clause 0x1
-; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
-; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
-; GFX10-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    v_permlane16_b32 v0, v0, s2, s3
-; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
-; GFX10-NEXT:    s_endpgm
-;
-; GFX11-LABEL: v_permlane16_b32_tid_tid:
-; GFX11:       ; %bb.0:
-; GFX11-NEXT:    s_clause 0x1
-; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
-; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX11-NEXT:    v_mov_b32_e32 v1, 0
-; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-NEXT:    v_permlane16_b32 v0, v0, s2, s3
-; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
-; GFX11-NEXT:    s_nop 0
-; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
-; GFX11-NEXT:    s_endpgm
-;
-; GFX12-LABEL: v_permlane16_b32_tid_tid:
-; GFX12:       ; %bb.0:
-; GFX12-NEXT:    s_clause 0x1
-; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
-; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX12-NEXT:    v_mov_b32_e32 v1, 0
-; GFX12-NEXT:    s_wait_kmcnt 0x0
-; GFX12-NEXT:    v_permlane16_b32 v0, v0, s2, s3
-; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
-; GFX12-NEXT:    s_nop 0
-; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
-; GFX12-NEXT:    s_endpgm
-  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
-  %v = call i32 @llvm.amdgcn.permlane16(i32 %tidx, i32 %tidx, i32 %src1, i32 %src2, i1 false, i1 false)
-  store i32 %v, ptr addrspace(1) %out
-  ret void
-}
-
-define amdgpu_kernel void @v_permlane16_b32_undef_tid(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-LABEL: v_permlane16_b32_undef_tid:
-; GFX10:       ; %bb.0:
-; GFX10-NEXT:    s_clause 0x1
-; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
-; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
-; GFX10-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    v_permlane16_b32 v0, v0, s2, s3
-; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
-; GFX10-NEXT:    s_endpgm
-;
-; GFX11-LABEL: v_permlane16_b32_undef_tid:
-; GFX11:       ; %bb.0:
-; GFX11-NEXT:    s_clause 0x1
-; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-NEXT:    s_load_b32 s2, s[0:1], 0x2c
 ; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX11-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX11-NEXT:    s_movk_i32 s2, 0x1234
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
+; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s2, 0xc1d1
 ; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX11-NEXT:    s_nop 0
 ; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 ;
-; GFX12-LABEL: v_permlane16_b32_undef_tid:
+; GFX12-LABEL: v_permlanex16_b32_vll_f32:
 ; GFX12:       ; %bb.0:
-; GFX12-NEXT:    s_clause 0x1
-; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
-; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX12-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-NEXT:    s_load_b96 s[0:2], s[0:1], 0x24
 ; GFX12-NEXT:    s_wait_kmcnt 0x0
-; GFX12-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX12-NEXT:    s_movk_i32 s2, 0x1234
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
+; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s2, 0xc1d1
 ; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX12-NEXT:    s_nop 0
 ; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-NEXT:    s_endpgm
-  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
-  %undef = freeze i32 poison
-  %v = call i32 @llvm.amdgcn.permlane16(i32 %undef, i32 %tidx, i32 %src1, i32 %src2, i1 false, i1 false)
-  store i32 %v, ptr addrspace(1) %out
+  %v = call float @llvm.amdgcn.permlanex16.f32(float %src0, float %src0, i32 4660, i32 49617, i1 false, i1 false)
+  store float %v, ptr addrspace(1) %out
   ret void
 }
 
-define amdgpu_kernel void @v_permlane16_b32_i_tid(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-SDAG-LABEL: v_permlane16_b32_i_tid:
+define amdgpu_kernel void @v_permlanex16_b32_vll_i64(ptr addrspace(1) %out, i64 %src0) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_vll_i64:
 ; GFX10-SDAG:       ; %bb.0:
-; GFX10-SDAG-NEXT:    s_clause 0x1
-; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
-; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
-; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, 0x3039
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
 ; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v0, s2, s3
-; GFX10-SDAG-NEXT:    global_store_dword v2, v1, s[4:5]
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-SDAG-NEXT:    s_movk_i32 s2, 0x1234
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, 0xc1d1
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, 0xc1d1
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
 ; GFX10-SDAG-NEXT:    s_endpgm
 ;
-; GFX10-GISEL-LABEL: v_permlane16_b32_i_tid:
+; GFX10-GISEL-LABEL: v_permlanex16_b32_vll_i64:
 ; GFX10-GISEL:       ; %bb.0:
-; GFX10-GISEL-NEXT:    s_clause 0x1
-; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
-; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
-; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, 0x3039
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v0, s2, s3
-; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, 0
-; GFX10-GISEL-NEXT:    global_store_dword v0, v1, s[4:5]
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-GISEL-NEXT:    s_movk_i32 s2, 0x1234
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, 0xc1d1
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, 0xc1d1
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
 ; GFX10-GISEL-NEXT:    s_endpgm
 ;
-; GFX11-SDAG-LABEL: v_permlane16_b32_i_tid:
+; GFX11-SDAG-LABEL: v_permlanex16_b32_vll_i64:
 ; GFX11-SDAG:       ; %bb.0:
-; GFX11-SDAG-NEXT:    s_clause 0x1
-; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
-; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX11-SDAG-NEXT:    v_dual_mov_b32 v1, 0x3039 :: v_dual_mov_b32 v2, 0
+; GFX11-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
 ; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v0, s2, s3
-; GFX11-SDAG-NEXT:    global_store_b32 v2, v1, s[0:1]
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s3
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX11-SDAG-NEXT:    s_movk_i32 s2, 0x1234
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instid1(SALU_CYCLE_1)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, 0xc1d1
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, 0xc1d1
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
 ; GFX11-SDAG-NEXT:    s_nop 0
 ; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-SDAG-NEXT:    s_endpgm
 ;
-; GFX11-GISEL-LABEL: v_permlane16_b32_i_tid:
+; GFX11-GISEL-LABEL: v_permlanex16_b32_vll_i64:
 ; GFX11-GISEL:       ; %bb.0:
-; GFX11-GISEL-NEXT:    s_clause 0x1
-; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
-; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX11-GISEL-NEXT:    v_mov_b32_e32 v1, 0x3039
+; GFX11-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v0, s2, s3
-; GFX11-GISEL-NEXT:    v_mov_b32_e32 v0, 0
-; GFX11-GISEL-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX11-GISEL-NEXT:    s_movk_i32 s2, 0x1234
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, 0xc1d1
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, 0xc1d1
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
 ; GFX11-GISEL-NEXT:    s_nop 0
 ; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-GISEL-NEXT:    s_endpgm
 ;
-; GFX12-SDAG-LABEL: v_permlane16_b32_i_tid:
+; GFX12-SDAG-LABEL: v_permlanex16_b32_vll_i64:
 ; GFX12-SDAG:       ; %bb.0:
-; GFX12-SDAG-NEXT:    s_clause 0x1
-; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
-; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX12-SDAG-NEXT:    v_dual_mov_b32 v1, 0x3039 :: v_dual_mov_b32 v2, 0
+; GFX12-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
 ; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
-; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v0, s2, s3
-; GFX12-SDAG-NEXT:    global_store_b32 v2, v1, s[0:1]
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s3
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX12-SDAG-NEXT:    s_movk_i32 s2, 0x1234
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instid1(SALU_CYCLE_1)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, 0xc1d1
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, 0xc1d1
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
 ; GFX12-SDAG-NEXT:    s_nop 0
 ; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-SDAG-NEXT:    s_endpgm
 ;
-; GFX12-GISEL-LABEL: v_permlane16_b32_i_tid:
+; GFX12-GISEL-LABEL: v_permlanex16_b32_vll_i64:
 ; GFX12-GISEL:       ; %bb.0:
-; GFX12-GISEL-NEXT:    s_clause 0x1
-; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
-; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX12-GISEL-NEXT:    v_mov_b32_e32 v1, 0x3039
+; GFX12-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
-; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v0, s2, s3
-; GFX12-GISEL-NEXT:    v_mov_b32_e32 v0, 0
-; GFX12-GISEL-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX12-GISEL-NEXT:    s_movk_i32 s2, 0x1234
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, 0xc1d1
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, 0xc1d1
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
 ; GFX12-GISEL-NEXT:    s_nop 0
 ; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-GISEL-NEXT:    s_endpgm
-  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
-  %v = call i32 @llvm.amdgcn.permlane16(i32 12345, i32 %tidx, i32 %src1, i32 %src2, i1 false, i1 false)
-  store i32 %v, ptr addrspace(1) %out
+  %v = call i64 @llvm.amdgcn.permlanex16.i64(i64 %src0, i64 %src0, i32 4660, i32 49617, i1 false, i1 false)
+  store i64 %v, ptr addrspace(1) %out
   ret void
 }
 
-define amdgpu_kernel void @v_permlane16_b32_i_tid_fi(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-LABEL: v_permlane16_b32_i_tid_fi:
-; GFX10:       ; %bb.0:
-; GFX10-NEXT:    s_clause 0x1
-; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
-; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
-; GFX10-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
-; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
-; GFX10-NEXT:    s_endpgm
-;
-; GFX11-LABEL: v_permlane16_b32_i_tid_fi:
-; GFX11:       ; %bb.0:
-; GFX11-NEXT:    s_clause 0x1
-; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
-; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX11-NEXT:    v_mov_b32_e32 v1, 0
-; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
-; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
-; GFX11-NEXT:    s_nop 0
-; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
-; GFX11-NEXT:    s_endpgm
+define amdgpu_kernel void @v_permlanex16_b32_vll_f64(ptr addrspace(1) %out, double %src0) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_vll_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-SDAG-NEXT:    s_movk_i32 s2, 0x1234
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, 0xc1d1
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, 0xc1d1
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX10-SDAG-NEXT:    s_endpgm
 ;
-; GFX12-LABEL: v_permlane16_b32_i_tid_fi:
-; GFX12:       ; %bb.0:
-; GFX12-NEXT:    s_clause 0x1
+; GFX10-GISEL-LABEL: v_permlanex16_b32_vll_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-GISEL-NEXT:    s_movk_i32 s2, 0x1234
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, 0xc1d1
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, 0xc1d1
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_vll_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s3
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX11-SDAG-NEXT:    s_movk_i32 s2, 0x1234
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instid1(SALU_CYCLE_1)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, 0xc1d1
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, 0xc1d1
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_vll_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX11-GISEL-NEXT:    s_movk_i32 s2, 0x1234
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, 0xc1d1
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, 0xc1d1
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_vll_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s3
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX12-SDAG-NEXT:    s_movk_i32 s2, 0x1234
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instid1(SALU_CYCLE_1)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, 0xc1d1
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, 0xc1d1
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_vll_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX12-GISEL-NEXT:    s_movk_i32 s2, 0x1234
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, 0xc1d1
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, 0xc1d1
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %v = call double @llvm.amdgcn.permlanex16.f64(double %src0, double %src0, i32 4660, i32 49617, i1 false, i1 false)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vvv_i32(ptr addrspace(1) %out, i32 %src0) {
+; GFX10-LABEL: v_permlanex16_b32_vvv_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dword s4, s[0:1], 0x2c
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX10-NEXT:    s_mov_b32 null, 0
+; GFX10-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10-NEXT:    v_readfirstlane_b32 s1, v1
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v0, s4
+; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s0, s1
+; GFX10-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_vvv_i32:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b32 s2, s[0:1], 0x2c
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
+; GFX11-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s3, s2
+; GFX11-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_vvv_i32:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b32 s2, s[0:1], 0x2c
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
+; GFX11-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s3, s4
+; GFX11-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_vvv_i32:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_load_b96 s[0:2], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
+; GFX12-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s3, s2
+; GFX12-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_vvv_i32:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_load_b96 s[0:2], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
+; GFX12-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s3, s4
+; GFX12-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidy = call i32 @llvm.amdgcn.workitem.id.y()
+  %v = call i32 @llvm.amdgcn.permlanex16.i32(i32 %src0, i32 %src0, i32 %tidx, i32 %tidy, i1 false, i1 false)
+  store i32 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vvv_f32(ptr addrspace(1) %out, float %src0) {
+; GFX10-LABEL: v_permlanex16_b32_vvv_f32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dword s4, s[0:1], 0x2c
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX10-NEXT:    s_mov_b32 null, 0
+; GFX10-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10-NEXT:    v_readfirstlane_b32 s1, v1
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v0, s4
+; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s0, s1
+; GFX10-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_vvv_f32:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b32 s2, s[0:1], 0x2c
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
+; GFX11-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s3, s2
+; GFX11-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_vvv_f32:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b32 s2, s[0:1], 0x2c
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
+; GFX11-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s3, s4
+; GFX11-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_vvv_f32:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_load_b96 s[0:2], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
+; GFX12-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_3)
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s3, s2
+; GFX12-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_vvv_f32:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_load_b96 s[0:2], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
+; GFX12-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_4) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s3, v1
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s3, s4
+; GFX12-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidy = call i32 @llvm.amdgcn.workitem.id.y()
+  %v = call float @llvm.amdgcn.permlanex16.f32(float %src0, float %src0, i32 %tidx, i32 %tidy, i1 false, i1 false)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vvv_i64(ptr addrspace(1) %out, i64 %src0) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_vvv_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v1
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s4, s5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_vvv_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s5, v1
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s4, s5
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s4, s5
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlanex16_b32_vvv_i64:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
+; GFX11-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_4) | instid1(VALU_DEP_2)
+; GFX11-NEXT:    v_readfirstlane_b32 s5, v0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_mov_b32_e32 v0, s2
+; GFX11-NEXT:    v_readfirstlane_b32 s4, v1
+; GFX11-NEXT:    v_mov_b32_e32 v1, s3
+; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s4, s5
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-NEXT:    v_permlanex16_b32 v1, v1, s4, s5
+; GFX11-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlanex16_b32_vvv_i64:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
+; GFX12-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_4) | instid1(VALU_DEP_2)
+; GFX12-NEXT:    v_readfirstlane_b32 s5, v0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_mov_b32_e32 v0, s2
+; GFX12-NEXT:    v_readfirstlane_b32 s4, v1
+; GFX12-NEXT:    v_mov_b32_e32 v1, s3
+; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s4, s5
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-NEXT:    v_permlanex16_b32 v1, v1, s4, s5
+; GFX12-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidy = call i32 @llvm.amdgcn.workitem.id.y()
+  %v = call i64 @llvm.amdgcn.permlanex16.i64(i64 %src0, i64 %src0, i32 %tidx, i32 %tidy, i1 false, i1 false)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vvv_f64(ptr addrspace(1) %out, double %src0) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_vvv_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v1
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s4, s5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_vvv_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s5, v1
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s3
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s4, s5
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s4, s5
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlanex16_b32_vvv_f64:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
+; GFX11-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_4) | instid1(VALU_DEP_2)
+; GFX11-NEXT:    v_readfirstlane_b32 s5, v0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_mov_b32_e32 v0, s2
+; GFX11-NEXT:    v_readfirstlane_b32 s4, v1
+; GFX11-NEXT:    v_mov_b32_e32 v1, s3
+; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s4, s5
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-NEXT:    v_permlanex16_b32 v1, v1, s4, s5
+; GFX11-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlanex16_b32_vvv_f64:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-NEXT:    v_and_b32_e32 v1, 0x3ff, v0
+; GFX12-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_4) | instid1(VALU_DEP_2)
+; GFX12-NEXT:    v_readfirstlane_b32 s5, v0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_mov_b32_e32 v0, s2
+; GFX12-NEXT:    v_readfirstlane_b32 s4, v1
+; GFX12-NEXT:    v_mov_b32_e32 v1, s3
+; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s4, s5
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-NEXT:    v_permlanex16_b32 v1, v1, s4, s5
+; GFX12-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidy = call i32 @llvm.amdgcn.workitem.id.y()
+  %v = call double @llvm.amdgcn.permlanex16.f64(double %src0, double %src0, i32 %tidx, i32 %tidy, i1 false, i1 false)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vvs_i32(ptr addrspace(1) %out, i32 %src0, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_vvs_i32:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX10-SDAG-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_vvs_i32:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s4, s3
+; GFX10-GISEL-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_vvs_i32:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX11-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_vvs_i32:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s4, s3
+; GFX11-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_vvs_i32:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX12-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_vvs_i32:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s4, s3
+; GFX12-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %v = call i32 @llvm.amdgcn.permlanex16.i32(i32 %src0, i32 %src0, i32 %tidx, i32 %src2, i1 false, i1 false)
+  store i32 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vvs_f32(ptr addrspace(1) %out, float %src0, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_vvs_f32:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX10-SDAG-NEXT:    global_store_dword v0, v1, s[0:1]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_vvs_f32:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s4, s3
+; GFX10-GISEL-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_vvs_f32:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX11-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_vvs_f32:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s4, s3
+; GFX11-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_vvs_f32:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX12-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_vvs_f32:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s4, s3
+; GFX12-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %v = call float @llvm.amdgcn.permlanex16.f32(float %src0, float %src0, i32 %tidx, i32 %src2, i1 false, i1 false)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vvs_i64(ptr addrspace(1) %out, i64 %src0, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_vvs_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-SDAG-NEXT:    s_mov_b32 null, 0
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s0, s2
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s0, s2
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_vvs_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-GISEL-NEXT:    s_mov_b32 null, 0
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s0, s2
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s0, s2
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_vvs_i64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s1, s0
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s1, s0
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_vvs_i64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s1, s0
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s1, s0
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_vvs_i64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s1, s0
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s1, s0
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_vvs_i64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s1, s0
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s1, s0
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %v = call i64 @llvm.amdgcn.permlanex16.i64(i64 %src0, i64 %src0, i32 %tidx, i32 %src2, i1 false, i1 false)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vvs_f64(ptr addrspace(1) %out, double %src0, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_vvs_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-SDAG-NEXT:    s_mov_b32 null, 0
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s0, s2
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s0, s2
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_vvs_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-GISEL-NEXT:    s_mov_b32 null, 0
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s0, v0
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s0, s2
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s0, s2
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_vvs_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s1, s0
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s1, s0
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_vvs_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s1, s0
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s1, s0
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_vvs_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s1, s0
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s1, s0
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_vvs_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s1, s0
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s1, s0
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %v = call double @llvm.amdgcn.permlanex16.f64(double %src0, double %src0, i32 %tidx, i32 %src2, i1 false, i1 false)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vsv_i32(ptr addrspace(1) %out, i32 %src0, i32 %src1) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_vsv_i32:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s3, s2
+; GFX10-SDAG-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_vsv_i32:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s4, v1
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s3, s4
+; GFX10-GISEL-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_vsv_i32:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s3, s2
+; GFX11-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_vsv_i32:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s3, s4
+; GFX11-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_vsv_i32:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s3, s2
+; GFX12-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_vsv_i32:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s3, s4
+; GFX12-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidy = call i32 @llvm.amdgcn.workitem.id.y()
+  %v = call i32 @llvm.amdgcn.permlanex16.i32(i32 %src0, i32 %src0, i32 %src1, i32 %tidy, i1 false, i1 false)
+  store i32 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vsv_f32(ptr addrspace(1) %out, float %src0, i32 %src1) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_vsv_f32:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s2, v1
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s3, s2
+; GFX10-SDAG-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_vsv_f32:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s4, v1
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s3, s4
+; GFX10-GISEL-NEXT:    global_store_dword v1, v0, s[0:1]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_vsv_f32:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s3, s2
+; GFX11-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_vsv_f32:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s3, s4
+; GFX11-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_vsv_f32:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v1, s2
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s3, s2
+; GFX12-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_vsv_f32:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s4, v0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v0, s2
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s3, s4
+; GFX12-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidy = call i32 @llvm.amdgcn.workitem.id.y()
+  %v = call float @llvm.amdgcn.permlanex16.f32(float %src0, float %src0, i32 %src1, i32 %tidy, i1 false, i1 false)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vsv_i64(ptr addrspace(1) %out, i64 %src0, i32 %src1) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_vsv_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-SDAG-NEXT:    s_mov_b32 null, 0
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s0
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s0
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_vsv_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-GISEL-NEXT:    s_mov_b32 null, 0
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s0
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s0
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_vsv_i64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v1, s7 :: v_dual_mov_b32 v0, s6
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s0, s1
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_vsv_i64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s0, s1
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s0, s1
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_vsv_i64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v1, s7 :: v_dual_mov_b32 v0, s6
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s0, s1
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_vsv_i64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s0, s1
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s0, s1
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidy = call i32 @llvm.amdgcn.workitem.id.y()
+  %v = call i64 @llvm.amdgcn.permlanex16.i64(i64 %src0, i64 %src0, i32 %src1, i32 %tidy, i1 false, i1 false)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vsv_f64(ptr addrspace(1) %out, double %src0, i32 %src1) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_vsv_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-SDAG-NEXT:    s_mov_b32 null, 0
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s0
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s0
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_vsv_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-GISEL-NEXT:    s_mov_b32 null, 0
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s0, v1
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s0
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s0
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_vsv_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v1, s7 :: v_dual_mov_b32 v0, s6
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s0, s1
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_vsv_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s0, s1
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s0, s1
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_vsv_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-SDAG-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v1, s7 :: v_dual_mov_b32 v0, s6
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s0, s1
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_vsv_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-GISEL-NEXT:    v_bfe_u32 v0, v0, 10, 10
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s1, v0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s0, s1
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s0, s1
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidy = call i32 @llvm.amdgcn.workitem.id.y()
+  %v = call double @llvm.amdgcn.permlanex16.f64(double %src0, double %src0, i32 %src1, i32 %tidy, i1 false, i1 false)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vss_fi_i32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlanex16_b32_vss_fi_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s7, s2 op_sel:[1,0]
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlanex16_b32_vss_fi_i32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s7, s0 op_sel:[1,0]
+; GFX11-NEXT:    global_store_b32 v1, v0, s[4:5]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlanex16_b32_vss_fi_i32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s7, s0 op_sel:[1,0]
+; GFX12-NEXT:    global_store_b32 v1, v0, s[4:5]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %v = call i32 @llvm.amdgcn.permlanex16.i32(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 true, i1 false)
+  store i32 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vss_fi_f32(ptr addrspace(1) %out, float %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlanex16_b32_vss_fi_f32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s7, s2 op_sel:[1,0]
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlanex16_b32_vss_fi_f32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s7, s0 op_sel:[1,0]
+; GFX11-NEXT:    global_store_b32 v1, v0, s[4:5]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlanex16_b32_vss_fi_f32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s7, s0 op_sel:[1,0]
+; GFX12-NEXT:    global_store_b32 v1, v0, s[4:5]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %v = call float @llvm.amdgcn.permlanex16.f32(float %src0, float %src0, i32 %src1, i32 %src2, i1 true, i1 false)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vss_fi_i64(ptr addrspace(1) %out, i64 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_vss_fi_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[1,0]
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_vss_fi_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[1,0]
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_vss_fi_i64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s0, s1 op_sel:[1,0]
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s0, s1 op_sel:[1,0]
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_vss_fi_i64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s0, s1 op_sel:[1,0]
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s0, s1 op_sel:[1,0]
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_vss_fi_i64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s0, s1 op_sel:[1,0]
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s0, s1 op_sel:[1,0]
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_vss_fi_i64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s0, s1 op_sel:[1,0]
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s0, s1 op_sel:[1,0]
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %v = call i64 @llvm.amdgcn.permlanex16.i64(i64 %src0, i64 %src0, i32 %src1, i32 %src2, i1 true, i1 false)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vss_fi_f64(ptr addrspace(1) %out, double %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_vss_fi_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[1,0]
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_vss_fi_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[1,0]
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_vss_fi_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s0, s1 op_sel:[1,0]
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s0, s1 op_sel:[1,0]
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_vss_fi_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s0, s1 op_sel:[1,0]
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s0, s1 op_sel:[1,0]
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_vss_fi_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s0, s1 op_sel:[1,0]
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s0, s1 op_sel:[1,0]
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_vss_fi_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s0, s1 op_sel:[1,0]
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s0, s1 op_sel:[1,0]
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %v = call double @llvm.amdgcn.permlanex16.f64(double %src0, double %src0, i32 %src1, i32 %src2, i1 true, i1 false)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vss_bc_i32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlanex16_b32_vss_bc_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s7, s2 op_sel:[0,1]
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlanex16_b32_vss_bc_i32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s7, s0 op_sel:[0,1]
+; GFX11-NEXT:    global_store_b32 v1, v0, s[4:5]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlanex16_b32_vss_bc_i32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s7, s0 op_sel:[0,1]
+; GFX12-NEXT:    global_store_b32 v1, v0, s[4:5]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %v = call i32 @llvm.amdgcn.permlanex16.i32(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 true)
+  store i32 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vss_bc_f32(ptr addrspace(1) %out, float %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlanex16_b32_vss_bc_f32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s7, s2 op_sel:[0,1]
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlanex16_b32_vss_bc_f32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s7, s0 op_sel:[0,1]
+; GFX11-NEXT:    global_store_b32 v1, v0, s[4:5]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlanex16_b32_vss_bc_f32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s7, s0 op_sel:[0,1]
+; GFX12-NEXT:    global_store_b32 v1, v0, s[4:5]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %v = call float @llvm.amdgcn.permlanex16.f32(float %src0, float %src0, i32 %src1, i32 %src2, i1 false, i1 true)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vss_bc_i64(ptr addrspace(1) %out, i64 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_vss_bc_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[0,1]
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_vss_bc_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[0,1]
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_vss_bc_i64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s0, s1 op_sel:[0,1]
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s0, s1 op_sel:[0,1]
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_vss_bc_i64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s0, s1 op_sel:[0,1]
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s0, s1 op_sel:[0,1]
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_vss_bc_i64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s0, s1 op_sel:[0,1]
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s0, s1 op_sel:[0,1]
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_vss_bc_i64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s0, s1 op_sel:[0,1]
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s0, s1 op_sel:[0,1]
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %v = call i64 @llvm.amdgcn.permlanex16.i64(i64 %src0, i64 %src0, i32 %src1, i32 %src2, i1 false, i1 true)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vss_bc_f64(ptr addrspace(1) %out, double %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_vss_bc_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[0,1]
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_vss_bc_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[0,1]
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_vss_bc_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s0, s1 op_sel:[0,1]
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s0, s1 op_sel:[0,1]
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_vss_bc_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s0, s1 op_sel:[0,1]
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s0, s1 op_sel:[0,1]
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_vss_bc_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s0, s1 op_sel:[0,1]
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s0, s1 op_sel:[0,1]
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_vss_bc_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s0, s1 op_sel:[0,1]
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s0, s1 op_sel:[0,1]
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %v = call double @llvm.amdgcn.permlanex16.f64(double %src0, double %src0, i32 %src1, i32 %src2, i1 false, i1 true)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vss_fi_bc_i32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlanex16_b32_vss_fi_bc_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s7, s2 op_sel:[1,1]
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlanex16_b32_vss_fi_bc_i32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s7, s0 op_sel:[1,1]
+; GFX11-NEXT:    global_store_b32 v1, v0, s[4:5]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlanex16_b32_vss_fi_bc_i32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s7, s0 op_sel:[1,1]
+; GFX12-NEXT:    global_store_b32 v1, v0, s[4:5]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %v = call i32 @llvm.amdgcn.permlanex16.i32(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 true, i1 true)
+  store i32 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vss_fi_bc_f32(ptr addrspace(1) %out, float %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlanex16_b32_vss_fi_bc_f32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-NEXT:    s_load_dword s2, s[0:1], 0x34
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s7, s2 op_sel:[1,1]
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlanex16_b32_vss_fi_bc_f32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s7, s0 op_sel:[1,1]
+; GFX11-NEXT:    global_store_b32 v1, v0, s[4:5]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlanex16_b32_vss_fi_bc_f32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-NEXT:    s_load_b32 s0, s[0:1], 0x34
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s6
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s7, s0 op_sel:[1,1]
+; GFX12-NEXT:    global_store_b32 v1, v0, s[4:5]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %v = call float @llvm.amdgcn.permlanex16.f32(float %src0, float %src0, i32 %src1, i32 %src2, i1 true, i1 true)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vss_fi_bc_i64(ptr addrspace(1) %out, i64 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_vss_fi_bc_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[1,1]
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_vss_fi_bc_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[1,1]
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_vss_fi_bc_i64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s0, s1 op_sel:[1,1]
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s0, s1 op_sel:[1,1]
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_vss_fi_bc_i64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s0, s1 op_sel:[1,1]
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s0, s1 op_sel:[1,1]
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_vss_fi_bc_i64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s0, s1 op_sel:[1,1]
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s0, s1 op_sel:[1,1]
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_vss_fi_bc_i64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s0, s1 op_sel:[1,1]
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s0, s1 op_sel:[1,1]
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %v = call i64 @llvm.amdgcn.permlanex16.i64(i64 %src0, i64 %src0, i32 %src1, i32 %src2, i1 true, i1 true)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_vss_fi_bc_f64(ptr addrspace(1) %out, double %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_vss_fi_bc_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[1,1]
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_vss_fi_bc_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x34
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, s6
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, s7
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[1,1]
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_vss_fi_bc_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s0, s1 op_sel:[1,1]
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s0, s1 op_sel:[1,1]
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_vss_fi_bc_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s0, s1 op_sel:[1,1]
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s0, s1 op_sel:[1,1]
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_vss_fi_bc_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v0, s6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s0, s1 op_sel:[1,1]
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s0, s1 op_sel:[1,1]
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_vss_fi_bc_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b128 s[4:7], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x34
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s7
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s0, s1 op_sel:[1,1]
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s0, s1 op_sel:[1,1]
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[4:5]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %v = call double @llvm.amdgcn.permlanex16.f64(double %src0, double %src0, i32 %src1, i32 %src2, i1 true, i1 true)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_tid_tid_i32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlane16_b32_tid_tid_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlane16_b32_tid_tid_i32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlane16_b32_tid_tid_i32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %v = call i32 @llvm.amdgcn.permlane16.i32(i32 %tidx, i32 %tidx, i32 %src1, i32 %src2, i1 false, i1 false)
+  store i32 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_tid_tid_f32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlane16_b32_tid_tid_f32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlane16_b32_tid_tid_f32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlane16_b32_tid_tid_f32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_f32 = bitcast i32 %tidx to float
+  %v = call float @llvm.amdgcn.permlane16.f32(float %tidx_f32, float %tidx_f32, i32 %src1, i32 %src2, i1 false, i1 false)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_tid_tid_i64(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_tid_tid_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_tid_tid_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_tid_tid_i64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_tid_tid_i64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_tid_tid_i64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_tid_tid_i64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_i64 = zext i32 %tidx to i64
+  %v = call i64 @llvm.amdgcn.permlane16.i64(i64 %tidx_i64, i64 %tidx_i64, i32 %src1, i32 %src2, i1 false, i1 false)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_tid_tid_f64(ptr addrspace(1) %out, float %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_tid_tid_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_tid_tid_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_tid_tid_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_tid_tid_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_tid_tid_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_tid_tid_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_f32 = bitcast i32 %tidx to float
+  %tidx_f64 = fpext float %tidx_f32 to double
+  %v = call double @llvm.amdgcn.permlane16.f64(double %tidx_f64, double %tidx_f64, i32 %src1, i32 %src2, i1 false, i1 false)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_undef_tid_i32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlane16_b32_undef_tid_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlane16_b32_undef_tid_i32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlane16_b32_undef_tid_i32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %undef = freeze i32 poison
+  %v = call i32 @llvm.amdgcn.permlane16.i32(i32 %undef, i32 %tidx, i32 %src1, i32 %src2, i1 false, i1 false)
+  store i32 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_undef_tid_f32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlane16_b32_undef_tid_f32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlane16_b32_undef_tid_f32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlane16_b32_undef_tid_f32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_f32 = bitcast i32 %tidx to float
+  %undef = freeze float poison
+  %v = call float @llvm.amdgcn.permlane16.f32(float %undef, float %tidx_f32, i32 %src1, i32 %src2, i1 false, i1 false)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_undef_tid_i64(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_undef_tid_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v2, s2, s3
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_undef_tid_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v2, s2, s3
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_undef_tid_i64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v2, s2, s3
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_undef_tid_i64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v2, s2, s3
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_undef_tid_i64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v2, s2, s3
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_undef_tid_i64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v2, s2, s3
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_i64 = zext i32 %tidx to i64
+  %undef = freeze i64 poison
+  %v = call i64 @llvm.amdgcn.permlane16.i64(i64 %undef, i64 %tidx_i64, i32 %src1, i32 %src2, i1 false, i1 false)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_undef_tid_f64(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_undef_tid_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_undef_tid_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_undef_tid_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_undef_tid_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_undef_tid_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_undef_tid_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_f32 = bitcast i32 %tidx to float
+  %tidx_f64 = fpext float %tidx_f32 to double
+  %undef = freeze double poison
+  %v = call double @llvm.amdgcn.permlane16.f64(double %undef, double %tidx_f64, i32 %src1, i32 %src2, i1 false, i1 false)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_i_tid_i32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_i_tid_i32:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, 0x3039
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v0, s2, s3
+; GFX10-SDAG-NEXT:    global_store_dword v2, v1, s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_i_tid_i32:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, 0x3039
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v0, s2, s3
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-GISEL-NEXT:    global_store_dword v0, v1, s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_i_tid_i32:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v1, 0x3039 :: v_dual_mov_b32 v2, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v0, s2, s3
+; GFX11-SDAG-NEXT:    global_store_b32 v2, v1, s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_i_tid_i32:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v1, 0x3039
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v0, s2, s3
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-GISEL-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_i_tid_i32:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v1, 0x3039 :: v_dual_mov_b32 v2, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v0, s2, s3
+; GFX12-SDAG-NEXT:    global_store_b32 v2, v1, s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_i_tid_i32:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v1, 0x3039
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v0, s2, s3
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-GISEL-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %v = call i32 @llvm.amdgcn.permlane16.i32(i32 12345, i32 %tidx, i32 %src1, i32 %src2, i1 false, i1 false)
+  store i32 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_i_tid_f32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_i_tid_f32:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, 0x449a5000
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v0, s2, s3
+; GFX10-SDAG-NEXT:    global_store_dword v2, v1, s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_i_tid_f32:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, 0x449a5000
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v0, s2, s3
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-GISEL-NEXT:    global_store_dword v0, v1, s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_i_tid_f32:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v1, 0x449a5000 :: v_dual_mov_b32 v2, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v0, s2, s3
+; GFX11-SDAG-NEXT:    global_store_b32 v2, v1, s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_i_tid_f32:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v1, 0x449a5000
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v0, s2, s3
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-GISEL-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_i_tid_f32:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v1, 0x449a5000 :: v_dual_mov_b32 v2, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v0, s2, s3
+; GFX12-SDAG-NEXT:    global_store_b32 v2, v1, s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_i_tid_f32:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v1, 0x449a5000
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v0, s2, s3
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-GISEL-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_f32 = bitcast i32 %tidx to float
+  %v = call float @llvm.amdgcn.permlane16.f32(float 1234.5, float %tidx_f32, i32 %src1, i32 %src2, i1 false, i1 false)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_i_tid_i64(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_i_tid_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, 0x3039
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v3, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v2, v2, s2, s3
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v0, s2, s3
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v3, v[1:2], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_i_tid_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, 0x3039
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v3, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v0, s2, s3
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v2, v2, s2, s3
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v3, v[1:2], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_i_tid_i64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, 0x3039
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v3, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v2, v2, s2, s3
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v0, s2, s3
+; GFX11-SDAG-NEXT:    global_store_b64 v3, v[1:2], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_i_tid_i64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v1, 0x3039 :: v_dual_mov_b32 v2, 0
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v3, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v0, s2, s3
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v2, v2, s2, s3
+; GFX11-GISEL-NEXT:    global_store_b64 v3, v[1:2], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_i_tid_i64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, 0x3039
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v3, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v2, v2, s2, s3
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v0, s2, s3
+; GFX12-SDAG-NEXT:    global_store_b64 v3, v[1:2], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_i_tid_i64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v1, 0x3039 :: v_dual_mov_b32 v2, 0
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v3, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v0, s2, s3
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v2, v2, s2, s3
+; GFX12-GISEL-NEXT:    global_store_b64 v3, v[1:2], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_i64 = zext i32 %tidx to i64
+  %v = call i64 @llvm.amdgcn.permlane16.i64(i64 12345, i64 %tidx_i64, i32 %src1, i32 %src2, i1 false, i1 false)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_i_tid_f64(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_i_tid_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v3, 0x40934a00
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v4, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v3, v1, s2, s3
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v2, v0, s2, s3
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v4, v[2:3], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_i_tid_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v3, 0x40934a00
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v4, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v2, v0, s2, s3
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v3, v1, s2, s3
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v4, v[2:3], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_i_tid_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v3, 0x40934a00 :: v_dual_mov_b32 v2, 0
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v4, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v3, v1, s2, s3
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v2, v0, s2, s3
+; GFX11-SDAG-NEXT:    global_store_b64 v4, v[2:3], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_i_tid_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v3, 0x40934a00
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v4, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v2, v0, s2, s3
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v3, v1, s2, s3
+; GFX11-GISEL-NEXT:    global_store_b64 v4, v[2:3], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_i_tid_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v3, 0x40934a00 :: v_dual_mov_b32 v2, 0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v4, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v3, v1, s2, s3
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v2, v0, s2, s3
+; GFX12-SDAG-NEXT:    global_store_b64 v4, v[2:3], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_i_tid_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v3, 0x40934a00
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v4, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v2, v0, s2, s3
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v3, v1, s2, s3
+; GFX12-GISEL-NEXT:    global_store_b64 v4, v[2:3], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_f32 = bitcast i32 %tidx to float
+  %tidx_f64 = fpext float %tidx_f32 to double
+  %v = call double @llvm.amdgcn.permlane16.f64(double 1234.5, double %tidx_f64, i32 %src1, i32 %src2, i1 false, i1 false)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_i_tid_fi_i32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlane16_b32_i_tid_fi_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlane16_b32_i_tid_fi_i32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlane16_b32_i_tid_fi_i32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %undef = freeze i32 poison
+  %v = call i32 @llvm.amdgcn.permlane16.i32(i32 %undef, i32 %tidx, i32 %src1, i32 %src2, i1 true, i1 false)
+  store i32 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_i_tid_fi_f32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlane16_b32_i_tid_fi_f32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlane16_b32_i_tid_fi_f32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlane16_b32_i_tid_fi_f32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_f32 = bitcast i32 %tidx to float
+  %undef = freeze float poison
+  %v = call float @llvm.amdgcn.permlane16.f32(float %undef, float %tidx_f32, i32 %src1, i32 %src2, i1 true, i1 false)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_i_tid_fi_i64(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_i_tid_fi_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v2, s2, s3 op_sel:[1,0]
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_i_tid_fi_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v2, s2, s3 op_sel:[1,0]
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_i_tid_fi_i64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v2, s2, s3 op_sel:[1,0]
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_i_tid_fi_i64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v2, s2, s3 op_sel:[1,0]
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_i_tid_fi_i64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v2, s2, s3 op_sel:[1,0]
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_i_tid_fi_i64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v2, s2, s3 op_sel:[1,0]
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_i64 = zext i32 %tidx to i64
+  %undef = freeze i64 poison
+  %v = call i64 @llvm.amdgcn.permlane16.i64(i64 %undef, i64 %tidx_i64, i32 %src1, i32 %src2, i1 true, i1 false)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_i_tid_fi_f64(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_i_tid_fi_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[1,0]
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_i_tid_fi_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[1,0]
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_i_tid_fi_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[1,0]
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_i_tid_fi_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[1,0]
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_i_tid_fi_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[1,0]
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_i_tid_fi_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[1,0]
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_f32 = bitcast i32 %tidx to float
+  %tidx_f64 = fpext float %tidx_f32 to double
+  %undef = freeze double poison
+  %v = call double @llvm.amdgcn.permlane16.f64(double %undef, double %tidx_f64, i32 %src1, i32 %src2, i1 true, i1 false)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_i_tid_bc_i32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlane16_b32_i_tid_bc_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlane16_b32_i_tid_bc_i32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlane16_b32_i_tid_bc_i32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %undef = freeze i32 poison
+  %v = call i32 @llvm.amdgcn.permlane16.i32(i32 %undef, i32 %tidx, i32 %src1, i32 %src2, i1 false, i1 true)
+  store i32 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_i_tid_bc_f32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlane16_b32_i_tid_bc_f32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlane16_b32_i_tid_bc_f32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlane16_b32_i_tid_bc_f32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_f32 = bitcast i32 %tidx to float
+  %undef = freeze float poison
+  %v = call float @llvm.amdgcn.permlane16.f32(float %undef, float %tidx_f32, i32 %src1, i32 %src2, i1 false, i1 true)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_i_tid_bc_i64(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_i_tid_bc_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v2, s2, s3 op_sel:[0,1]
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_i_tid_bc_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v2, s2, s3 op_sel:[0,1]
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_i_tid_bc_i64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v2, s2, s3 op_sel:[0,1]
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_i_tid_bc_i64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v2, s2, s3 op_sel:[0,1]
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_i_tid_bc_i64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v2, s2, s3 op_sel:[0,1]
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_i_tid_bc_i64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v2, s2, s3 op_sel:[0,1]
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_i64 = zext i32 %tidx to i64
+  %undef = freeze i64 poison
+  %v = call i64 @llvm.amdgcn.permlane16.i64(i64 %undef, i64 %tidx_i64, i32 %src1, i32 %src2, i1 false, i1 true)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_i_tid_bc_f64(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_i_tid_bc_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[0,1]
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_i_tid_bc_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[0,1]
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_i_tid_bc_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[0,1]
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_i_tid_bc_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[0,1]
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_i_tid_bc_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[0,1]
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_i_tid_bc_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[0,1]
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_f32 = bitcast i32 %tidx to float
+  %tidx_f64 = fpext float %tidx_f32 to double
+  %undef = freeze double poison
+  %v = call double @llvm.amdgcn.permlane16.f64(double %undef, double %tidx_f64, i32 %src1, i32 %src2, i1 false, i1 true)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_i_tid_fi_bc_i32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlane16_b32_i_tid_fi_bc_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlane16_b32_i_tid_fi_bc_i32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlane16_b32_i_tid_fi_bc_i32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %undef = freeze i32 poison
+  %v = call i32 @llvm.amdgcn.permlane16.i32(i32 %undef, i32 %tidx, i32 %src1, i32 %src2, i1 true, i1 true)
+  store i32 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_i_tid_fi_bc_f32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlane16_b32_i_tid_fi_bc_f32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlane16_b32_i_tid_fi_bc_f32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlane16_b32_i_tid_fi_bc_f32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_f32 = bitcast i32 %tidx to float
+  %undef = freeze float poison
+  %v = call float @llvm.amdgcn.permlane16.f32(float %undef, float %tidx_f32, i32 %src1, i32 %src2, i1 true, i1 true)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_i_tid_fi_bc_i64(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_i_tid_fi_bc_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v2, s2, s3 op_sel:[1,1]
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_i_tid_fi_bc_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v2, s2, s3 op_sel:[1,1]
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_i_tid_fi_bc_i64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v2, s2, s3 op_sel:[1,1]
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_i_tid_fi_bc_i64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v2, s2, s3 op_sel:[1,1]
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_i_tid_fi_bc_i64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v2, s2, s3 op_sel:[1,1]
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_i_tid_fi_bc_i64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v2, s2, s3 op_sel:[1,1]
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_i64 = zext i32 %tidx to i64
+  %undef = freeze i64 poison
+  %v = call i64 @llvm.amdgcn.permlane16.i64(i64 %undef, i64 %tidx_i64, i32 %src1, i32 %src2, i1 true, i1 true)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlane16_b32_i_tid_fi_bc_f64(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_b32_i_tid_fi_bc_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[1,1]
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlane16_b32_i_tid_fi_bc_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[1,1]
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlane16_b32_i_tid_fi_bc_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[1,1]
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlane16_b32_i_tid_fi_bc_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[1,1]
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlane16_b32_i_tid_fi_bc_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[1,1]
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlane16_b32_i_tid_fi_bc_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v1, v1, s2, s3 op_sel:[1,1]
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_f32 = bitcast i32 %tidx to float
+  %tidx_f64 = fpext float %tidx_f32 to double
+  %undef = freeze double poison
+  %v = call double @llvm.amdgcn.permlane16.f64(double %undef, double %tidx_f64, i32 %src1, i32 %src2, i1 true, i1 true)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_tid_tid_i32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlanex16_b32_tid_tid_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlanex16_b32_tid_tid_i32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlanex16_b32_tid_tid_i32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %v = call i32 @llvm.amdgcn.permlanex16.i32(i32 %tidx, i32 %tidx, i32 %src1, i32 %src2, i1 false, i1 false)
+  store i32 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_tid_tid_f32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlanex16_b32_tid_tid_f32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlanex16_b32_tid_tid_f32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlanex16_b32_tid_tid_f32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_f32 = bitcast i32 %tidx to float
+  %v = call float @llvm.amdgcn.permlanex16.f32(float %tidx_f32, float %tidx_f32, i32 %src1, i32 %src2, i1 false, i1 false)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_tid_tid_i64(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_tid_tid_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_tid_tid_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_tid_tid_i64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_tid_tid_i64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_tid_tid_i64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_tid_tid_i64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_i64 = zext i32 %tidx to i64
+  %v = call i64 @llvm.amdgcn.permlanex16.i64(i64 %tidx_i64, i64 %tidx_i64, i32 %src1, i32 %src2, i1 false, i1 false)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_tid_tid_f64(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_tid_tid_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_tid_tid_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_tid_tid_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_tid_tid_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_tid_tid_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_tid_tid_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_f32 = bitcast i32 %tidx to float
+  %tidx_f64 = fpext float %tidx_f32 to double
+  %v = call double @llvm.amdgcn.permlanex16.f64(double %tidx_f64, double %tidx_f64, i32 %src1, i32 %src2, i1 false, i1 false)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_undef_tid_i32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlanex16_b32_undef_tid_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlanex16_b32_undef_tid_i32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlanex16_b32_undef_tid_i32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %undef = freeze i32 poison
+  %v = call i32 @llvm.amdgcn.permlanex16.i32(i32 %undef, i32 %tidx, i32 %src1, i32 %src2, i1 false, i1 false)
+  store i32 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_undef_tid_f32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlanex16_b32_undef_tid_f32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlanex16_b32_undef_tid_f32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlanex16_b32_undef_tid_f32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_f32 = bitcast i32 %tidx to float
+  %undef = freeze float poison
+  %v = call float @llvm.amdgcn.permlanex16.f32(float %undef, float %tidx_f32, i32 %src1, i32 %src2, i1 false, i1 false)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_undef_tid_i64(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_undef_tid_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v2, s2, s3
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_undef_tid_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v2, s2, s3
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_undef_tid_i64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v2, s2, s3
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_undef_tid_i64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v2, s2, s3
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_undef_tid_i64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v2, s2, s3
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_undef_tid_i64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v2, s2, s3
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_i64 = zext i32 %tidx to i64
+  %undef = freeze i64 poison
+  %v = call i64 @llvm.amdgcn.permlanex16.i64(i64 %undef, i64 %tidx_i64, i32 %src1, i32 %src2, i1 false, i1 false)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_undef_tid_f64(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_undef_tid_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_undef_tid_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_undef_tid_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_undef_tid_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_undef_tid_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_undef_tid_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_f32 = bitcast i32 %tidx to float
+  %tidx_f64 = fpext float %tidx_f32 to double
+  %undef = freeze double poison
+  %v = call double @llvm.amdgcn.permlanex16.f64(double %undef, double %tidx_f64, i32 %src1, i32 %src2, i1 false, i1 false)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_i_tid_i32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_i_tid_i32:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, 0x3039
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v0, s2, s3
+; GFX10-SDAG-NEXT:    global_store_dword v2, v1, s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_i_tid_i32:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, 0x3039
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v0, s2, s3
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-GISEL-NEXT:    global_store_dword v0, v1, s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_i_tid_i32:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v1, 0x3039 :: v_dual_mov_b32 v2, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v0, s2, s3
+; GFX11-SDAG-NEXT:    global_store_b32 v2, v1, s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_i_tid_i32:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v1, 0x3039
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v0, s2, s3
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-GISEL-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_i_tid_i32:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v1, 0x3039 :: v_dual_mov_b32 v2, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v0, s2, s3
+; GFX12-SDAG-NEXT:    global_store_b32 v2, v1, s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_i_tid_i32:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v1, 0x3039
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v0, s2, s3
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-GISEL-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %v = call i32 @llvm.amdgcn.permlanex16.i32(i32 12345, i32 %tidx, i32 %src1, i32 %src2, i1 false, i1 false)
+  store i32 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_i_tid_f32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_i_tid_f32:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, 0x449a5000
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v0, s2, s3
+; GFX10-SDAG-NEXT:    global_store_dword v2, v1, s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_i_tid_f32:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, 0x449a5000
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v0, s2, s3
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX10-GISEL-NEXT:    global_store_dword v0, v1, s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_i_tid_f32:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v1, 0x449a5000 :: v_dual_mov_b32 v2, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v0, s2, s3
+; GFX11-SDAG-NEXT:    global_store_b32 v2, v1, s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_i_tid_f32:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v1, 0x449a5000
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v0, s2, s3
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX11-GISEL-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_i_tid_f32:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v1, 0x449a5000 :: v_dual_mov_b32 v2, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v0, s2, s3
+; GFX12-SDAG-NEXT:    global_store_b32 v2, v1, s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_i_tid_f32:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v1, 0x449a5000
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v0, s2, s3
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX12-GISEL-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_f32 = bitcast i32 %tidx to float
+  %v = call float @llvm.amdgcn.permlanex16.f32(float 1234.5, float %tidx_f32, i32 %src1, i32 %src2, i1 false, i1 false)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_i_tid_i64(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_i_tid_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, 0x3039
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v3, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s2, s3
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v0, s2, s3
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v3, v[1:2], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_i_tid_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, 0x3039
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v3, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v0, s2, s3
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v2, v2, s2, s3
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v3, v[1:2], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_i_tid_i64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, 0x3039
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v3, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s2, s3
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v0, s2, s3
+; GFX11-SDAG-NEXT:    global_store_b64 v3, v[1:2], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_i_tid_i64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v1, 0x3039 :: v_dual_mov_b32 v2, 0
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v3, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v0, s2, s3
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v2, v2, s2, s3
+; GFX11-GISEL-NEXT:    global_store_b64 v3, v[1:2], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_i_tid_i64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, 0x3039
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v3, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s2, s3
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v0, s2, s3
+; GFX12-SDAG-NEXT:    global_store_b64 v3, v[1:2], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_i_tid_i64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v1, 0x3039 :: v_dual_mov_b32 v2, 0
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v3, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v0, s2, s3
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v2, v2, s2, s3
+; GFX12-GISEL-NEXT:    global_store_b64 v3, v[1:2], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_i64 = zext i32 %tidx to i64
+  %v = call i64 @llvm.amdgcn.permlanex16.i64(i64 12345, i64 %tidx_i64, i32 %src1, i32 %src2, i1 false, i1 false)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_i_tid_f64(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_i_tid_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v3, 0x40934a00
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v4, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v3, v1, s2, s3
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v2, v0, s2, s3
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v4, v[2:3], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_i_tid_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v3, 0x40934a00
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v4, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v2, v0, s2, s3
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v3, v1, s2, s3
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v4, v[2:3], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_i_tid_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v3, 0x40934a00 :: v_dual_mov_b32 v2, 0
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v4, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v3, v1, s2, s3
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v2, v0, s2, s3
+; GFX11-SDAG-NEXT:    global_store_b64 v4, v[2:3], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_i_tid_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v3, 0x40934a00
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v4, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v2, v0, s2, s3
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v3, v1, s2, s3
+; GFX11-GISEL-NEXT:    global_store_b64 v4, v[2:3], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_i_tid_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_dual_mov_b32 v3, 0x40934a00 :: v_dual_mov_b32 v2, 0
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v4, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v3, v1, s2, s3
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v2, v0, s2, s3
+; GFX12-SDAG-NEXT:    global_store_b64 v4, v[2:3], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_i_tid_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v3, 0x40934a00
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v4, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v2, v0, s2, s3
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v3, v1, s2, s3
+; GFX12-GISEL-NEXT:    global_store_b64 v4, v[2:3], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_f32 = bitcast i32 %tidx to float
+  %tidx_f64 = fpext float %tidx_f32 to double
+  %v = call double @llvm.amdgcn.permlanex16.f64(double 1234.5, double %tidx_f64, i32 %src1, i32 %src2, i1 false, i1 false)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_i_tid_fi_i32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlanex16_b32_i_tid_fi_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlanex16_b32_i_tid_fi_i32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlanex16_b32_i_tid_fi_i32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %undef = freeze i32 poison
+  %v = call i32 @llvm.amdgcn.permlanex16.i32(i32 %undef, i32 %tidx, i32 %src1, i32 %src2, i1 true, i1 false)
+  store i32 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_i_tid_fi_f32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlanex16_b32_i_tid_fi_f32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlanex16_b32_i_tid_fi_f32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlanex16_b32_i_tid_fi_f32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
+; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX12-NEXT:    s_nop 0
+; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_f32 = bitcast i32 %tidx to float
+  %undef = freeze float poison
+  %v = call float @llvm.amdgcn.permlanex16.f32(float %undef, float %tidx_f32, i32 %src1, i32 %src2, i1 true, i1 false)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_i_tid_fi_i64(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_i_tid_fi_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v2, s2, s3 op_sel:[1,0]
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_i_tid_fi_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v2, s2, s3 op_sel:[1,0]
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_i_tid_fi_i64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v2, s2, s3 op_sel:[1,0]
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_i_tid_fi_i64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v2, s2, s3 op_sel:[1,0]
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_i_tid_fi_i64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v2, s2, s3 op_sel:[1,0]
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_i_tid_fi_i64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v2, s2, s3 op_sel:[1,0]
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_i64 = zext i32 %tidx to i64
+  %undef = freeze i64 poison
+  %v = call i64 @llvm.amdgcn.permlanex16.i64(i64 %undef, i64 %tidx_i64, i32 %src1, i32 %src2, i1 true, i1 false)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_i_tid_fi_f64(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_i_tid_fi_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[1,0]
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_i_tid_fi_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[1,0]
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_i_tid_fi_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[1,0]
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_i_tid_fi_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[1,0]
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_i_tid_fi_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[1,0]
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_i_tid_fi_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[1,0]
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_f32 = bitcast i32 %tidx to float
+  %tidx_f64 = fpext float %tidx_f32 to double
+  %undef = freeze double poison
+  %v = call double @llvm.amdgcn.permlanex16.f64(double %undef, double %tidx_f64, i32 %src1, i32 %src2, i1 true, i1 false)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_i_tid_bc_i32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlanex16_b32_i_tid_bc_i32:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_clause 0x1
+; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-NEXT:    v_mov_b32_e32 v1, 0
+; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlanex16_b32_i_tid_bc_i32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+;
+; GFX12-LABEL: v_permlanex16_b32_i_tid_bc_i32:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_clause 0x1
 ; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
 ; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX12-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX12-NEXT:    s_wait_kmcnt 0x0
-; GFX12-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
+; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[0,1]
 ; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX12-NEXT:    s_nop 0
 ; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-NEXT:    s_endpgm
   %tidx = call i32 @llvm.amdgcn.workitem.id.x()
   %undef = freeze i32 poison
-  %v = call i32 @llvm.amdgcn.permlane16(i32 %undef, i32 %tidx, i32 %src1, i32 %src2, i1 true, i1 false)
+  %v = call i32 @llvm.amdgcn.permlanex16.i32(i32 %undef, i32 %tidx, i32 %src1, i32 %src2, i1 false, i1 true)
   store i32 %v, ptr addrspace(1) %out
   ret void
 }
 
-define amdgpu_kernel void @v_permlane16_b32_i_tid_bc(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-LABEL: v_permlane16_b32_i_tid_bc:
+define amdgpu_kernel void @v_permlanex16_b32_i_tid_bc_f32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlanex16_b32_i_tid_bc_f32:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_clause 0x1
 ; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
 ; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
 ; GFX10-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[0,1]
 ; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
 ; GFX10-NEXT:    s_endpgm
 ;
-; GFX11-LABEL: v_permlane16_b32_i_tid_bc:
+; GFX11-LABEL: v_permlanex16_b32_i_tid_bc_f32:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_clause 0x1
 ; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
 ; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[0,1]
 ; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX11-NEXT:    s_nop 0
 ; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 ;
-; GFX12-LABEL: v_permlane16_b32_i_tid_bc:
+; GFX12-LABEL: v_permlanex16_b32_i_tid_bc_f32:
 ; GFX12:       ; %bb.0:
 ; GFX12-NEXT:    s_clause 0x1
 ; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
 ; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX12-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX12-NEXT:    s_wait_kmcnt 0x0
-; GFX12-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[0,1]
 ; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX12-NEXT:    s_nop 0
 ; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-NEXT:    s_endpgm
   %tidx = call i32 @llvm.amdgcn.workitem.id.x()
-  %undef = freeze i32 poison
-  %v = call i32 @llvm.amdgcn.permlane16(i32 %undef, i32 %tidx, i32 %src1, i32 %src2, i1 false, i1 true)
-  store i32 %v, ptr addrspace(1) %out
+  %tidx_f32 = bitcast i32 %tidx to float
+  %undef = freeze float poison
+  %v = call float @llvm.amdgcn.permlanex16.f32(float %undef, float %tidx_f32, i32 %src1, i32 %src2, i1 false, i1 true)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_i_tid_bc_i64(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_i_tid_bc_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v2, s2, s3 op_sel:[0,1]
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_i_tid_bc_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v2, s2, s3 op_sel:[0,1]
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_i_tid_bc_i64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v2, s2, s3 op_sel:[0,1]
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_i_tid_bc_i64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v2, s2, s3 op_sel:[0,1]
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_i_tid_bc_i64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v2, s2, s3 op_sel:[0,1]
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_i_tid_bc_i64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v2, s2, s3 op_sel:[0,1]
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_i64 = zext i32 %tidx to i64
+  %undef = freeze i64 poison
+  %v = call i64 @llvm.amdgcn.permlanex16.i64(i64 %undef, i64 %tidx_i64, i32 %src1, i32 %src2, i1 false, i1 true)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_i_tid_bc_f64(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_i_tid_bc_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[0,1]
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_i_tid_bc_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[0,1]
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_i_tid_bc_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[0,1]
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_i_tid_bc_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[0,1]
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_i_tid_bc_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[0,1]
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_i_tid_bc_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[0,1]
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[0,1]
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_f32 = bitcast i32 %tidx to float
+  %tidx_f64 = fpext float %tidx_f32 to double
+  %undef = freeze double poison
+  %v = call double @llvm.amdgcn.permlanex16.f64(double %undef, double %tidx_f64, i32 %src1, i32 %src2, i1 false, i1 true)
+  store double %v, ptr addrspace(1) %out
   ret void
 }
 
-define amdgpu_kernel void @v_permlane16_b32_i_tid_fi_bc(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-LABEL: v_permlane16_b32_i_tid_fi_bc:
+define amdgpu_kernel void @v_permlanex16_b32_i_tid_fi_bc_i32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlanex16_b32_i_tid_fi_bc_i32:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_clause 0x1
 ; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
 ; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
 ; GFX10-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,1]
 ; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
 ; GFX10-NEXT:    s_endpgm
 ;
-; GFX11-LABEL: v_permlane16_b32_i_tid_fi_bc:
+; GFX11-LABEL: v_permlanex16_b32_i_tid_fi_bc_i32:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_clause 0x1
 ; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
 ; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,1]
 ; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX11-NEXT:    s_nop 0
 ; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 ;
-; GFX12-LABEL: v_permlane16_b32_i_tid_fi_bc:
+; GFX12-LABEL: v_permlanex16_b32_i_tid_fi_bc_i32:
 ; GFX12:       ; %bb.0:
 ; GFX12-NEXT:    s_clause 0x1
 ; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
 ; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX12-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX12-NEXT:    s_wait_kmcnt 0x0
-; GFX12-NEXT:    v_permlane16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,1]
 ; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX12-NEXT:    s_nop 0
 ; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-NEXT:    s_endpgm
   %tidx = call i32 @llvm.amdgcn.workitem.id.x()
   %undef = freeze i32 poison
-  %v = call i32 @llvm.amdgcn.permlane16(i32 %undef, i32 %tidx, i32 %src1, i32 %src2, i1 true, i1 true)
+  %v = call i32 @llvm.amdgcn.permlanex16.i32(i32 %undef, i32 %tidx, i32 %src1, i32 %src2, i1 true, i1 true)
   store i32 %v, ptr addrspace(1) %out
   ret void
 }
 
-define amdgpu_kernel void @v_permlanex16_b32_tid_tid(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-LABEL: v_permlanex16_b32_tid_tid:
+define amdgpu_kernel void @v_permlanex16_b32_i_tid_fi_bc_f32(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlanex16_b32_i_tid_fi_bc_f32:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_clause 0x1
 ; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
 ; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
 ; GFX10-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,1]
 ; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
 ; GFX10-NEXT:    s_endpgm
 ;
-; GFX11-LABEL: v_permlanex16_b32_tid_tid:
+; GFX11-LABEL: v_permlanex16_b32_i_tid_fi_bc_f32:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_clause 0x1
 ; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
 ; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,1]
 ; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX11-NEXT:    s_nop 0
 ; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 ;
-; GFX12-LABEL: v_permlanex16_b32_tid_tid:
+; GFX12-LABEL: v_permlanex16_b32_i_tid_fi_bc_f32:
 ; GFX12:       ; %bb.0:
 ; GFX12-NEXT:    s_clause 0x1
 ; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
 ; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX12-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX12-NEXT:    s_wait_kmcnt 0x0
-; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
+; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,1]
 ; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
 ; GFX12-NEXT:    s_nop 0
 ; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX12-NEXT:    s_endpgm
   %tidx = call i32 @llvm.amdgcn.workitem.id.x()
-  %v = call i32 @llvm.amdgcn.permlanex16(i32 %tidx, i32 %tidx, i32 %src1, i32 %src2, i1 false, i1 false)
-  store i32 %v, ptr addrspace(1) %out
+  %tidx_f32 = bitcast i32 %tidx to float
+  %undef = freeze float poison
+  %v = call float @llvm.amdgcn.permlanex16.f32(float %undef, float %tidx_f32, i32 %src1, i32 %src2, i1 true, i1 true)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_i_tid_fi_bc_i64(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_i_tid_fi_bc_i64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v2, s2, s3 op_sel:[1,1]
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_i_tid_fi_bc_i64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v2, s2, s3 op_sel:[1,1]
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_i_tid_fi_bc_i64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v2, s2, s3 op_sel:[1,1]
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_i_tid_fi_bc_i64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v2, s2, s3 op_sel:[1,1]
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_i_tid_fi_bc_i64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v2, s2, s3 op_sel:[1,1]
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_i_tid_fi_bc_i64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v2, s2, s3 op_sel:[1,1]
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_i64 = zext i32 %tidx to i64
+  %undef = freeze i64 poison
+  %v = call i64 @llvm.amdgcn.permlanex16.i64(i64 %undef, i64 %tidx_i64, i32 %src1, i32 %src2, i1 true, i1 true)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @v_permlanex16_b32_i_tid_fi_bc_f64(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_b32_i_tid_fi_bc_f64:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX10-SDAG-NEXT:    s_clause 0x1
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[1,1]
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-SDAG-NEXT:    s_endpgm
+;
+; GFX10-GISEL-LABEL: v_permlanex16_b32_i_tid_fi_bc_f64:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX10-GISEL-NEXT:    s_clause 0x1
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
+; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
+; GFX10-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[1,1]
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v2, v[0:1], s[4:5]
+; GFX10-GISEL-NEXT:    s_endpgm
+;
+; GFX11-SDAG-LABEL: v_permlanex16_b32_i_tid_fi_bc_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[1,1]
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: v_permlanex16_b32_i_tid_fi_bc_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[1,1]
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+;
+; GFX12-SDAG-LABEL: v_permlanex16_b32_i_tid_fi_bc_f64:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[1,1]
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX12-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-SDAG-NEXT:    s_nop 0
+; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-SDAG-NEXT:    s_endpgm
+;
+; GFX12-GISEL-LABEL: v_permlanex16_b32_i_tid_fi_bc_f64:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
+; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX12-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,1]
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v1, s2, s3 op_sel:[1,1]
+; GFX12-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX12-GISEL-NEXT:    s_nop 0
+; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX12-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_f32 = bitcast i32 %tidx to float
+  %tidx_f64 = fpext float %tidx_f32 to double
+  %undef = freeze double poison
+  %v = call double @llvm.amdgcn.permlanex16.f64(double %undef, double %tidx_f64, i32 %src1, i32 %src2, i1 true, i1 true)
+  store double %v, ptr addrspace(1) %out
   ret void
 }
 
-define amdgpu_kernel void @v_permlanex16_b32_undef_tid(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-LABEL: v_permlanex16_b32_undef_tid:
+define void @v_permlane16_half(ptr addrspace(1) %out, half %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlane16_half:
 ; GFX10:       ; %bb.0:
-; GFX10-NEXT:    s_clause 0x1
-; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
-; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
-; GFX10-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
-; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
-; GFX10-NEXT:    s_endpgm
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    v_readfirstlane_b32 s4, v3
+; GFX10-NEXT:    v_readfirstlane_b32 s5, v4
+; GFX10-NEXT:    v_permlane16_b32 v2, v2, s4, s5
+; GFX10-NEXT:    global_store_short v[0:1], v2, off
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: v_permlane16_half:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX11-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX11-NEXT:    global_store_b16 v[0:1], v2, off
+; GFX11-NEXT:    s_setpc_b64 s[30:31]
 ;
-; GFX11-LABEL: v_permlanex16_b32_undef_tid:
+; GFX12-LABEL: v_permlane16_half:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-NEXT:    s_wait_expcnt 0x0
+; GFX12-NEXT:    s_wait_samplecnt 0x0
+; GFX12-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX12-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX12-NEXT:    global_store_b16 v[0:1], v2, off
+; GFX12-NEXT:    s_setpc_b64 s[30:31]
+  %v = call half @llvm.amdgcn.permlane16.f16(half %src0, half %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store half %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @v_permlanex16_half(ptr addrspace(1) %out, half %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlanex16_half:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    v_readfirstlane_b32 s4, v3
+; GFX10-NEXT:    v_readfirstlane_b32 s5, v4
+; GFX10-NEXT:    v_permlanex16_b32 v2, v2, s4, s5
+; GFX10-NEXT:    global_store_short v[0:1], v2, off
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: v_permlanex16_half:
 ; GFX11:       ; %bb.0:
-; GFX11-NEXT:    s_clause 0x1
-; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
-; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX11-NEXT:    v_mov_b32_e32 v1, 0
-; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
-; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
-; GFX11-NEXT:    s_nop 0
-; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
-; GFX11-NEXT:    s_endpgm
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX11-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX11-NEXT:    global_store_b16 v[0:1], v2, off
+; GFX11-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-LABEL: v_permlanex16_half:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-NEXT:    s_wait_expcnt 0x0
+; GFX12-NEXT:    s_wait_samplecnt 0x0
+; GFX12-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX12-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX12-NEXT:    global_store_b16 v[0:1], v2, off
+; GFX12-NEXT:    s_setpc_b64 s[30:31]
+  %v = call half @llvm.amdgcn.permlanex16.f16(half %src0, half %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store half %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @v_permlane16_bfloat(ptr addrspace(1) %out, bfloat %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlane16_bfloat:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    v_readfirstlane_b32 s4, v3
+; GFX10-NEXT:    v_readfirstlane_b32 s5, v4
+; GFX10-NEXT:    v_permlane16_b32 v2, v2, s4, s5
+; GFX10-NEXT:    global_store_short v[0:1], v2, off
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: v_permlane16_bfloat:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX11-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX11-NEXT:    global_store_b16 v[0:1], v2, off
+; GFX11-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-LABEL: v_permlane16_bfloat:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-NEXT:    s_wait_expcnt 0x0
+; GFX12-NEXT:    s_wait_samplecnt 0x0
+; GFX12-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX12-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX12-NEXT:    global_store_b16 v[0:1], v2, off
+; GFX12-NEXT:    s_setpc_b64 s[30:31]
+  %v = call bfloat @llvm.amdgcn.permlane16.f16(bfloat %src0, bfloat %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store bfloat %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @v_permlanex16_bfloat(ptr addrspace(1) %out, bfloat %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlanex16_bfloat:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    v_readfirstlane_b32 s4, v3
+; GFX10-NEXT:    v_readfirstlane_b32 s5, v4
+; GFX10-NEXT:    v_permlanex16_b32 v2, v2, s4, s5
+; GFX10-NEXT:    global_store_short v[0:1], v2, off
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: v_permlanex16_bfloat:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX11-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX11-NEXT:    global_store_b16 v[0:1], v2, off
+; GFX11-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-LABEL: v_permlanex16_bfloat:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-NEXT:    s_wait_expcnt 0x0
+; GFX12-NEXT:    s_wait_samplecnt 0x0
+; GFX12-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX12-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX12-NEXT:    global_store_b16 v[0:1], v2, off
+; GFX12-NEXT:    s_setpc_b64 s[30:31]
+  %v = call bfloat @llvm.amdgcn.permlanex16.f16(bfloat %src0, bfloat %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store bfloat %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @v_permlane16_i16(ptr addrspace(1) %out, i16 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlane16_i16:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    v_readfirstlane_b32 s4, v3
+; GFX10-NEXT:    v_readfirstlane_b32 s5, v4
+; GFX10-NEXT:    v_permlane16_b32 v2, v2, s4, s5
+; GFX10-NEXT:    global_store_short v[0:1], v2, off
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: v_permlane16_i16:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX11-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX11-NEXT:    global_store_b16 v[0:1], v2, off
+; GFX11-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-LABEL: v_permlane16_i16:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-NEXT:    s_wait_expcnt 0x0
+; GFX12-NEXT:    s_wait_samplecnt 0x0
+; GFX12-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX12-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX12-NEXT:    global_store_b16 v[0:1], v2, off
+; GFX12-NEXT:    s_setpc_b64 s[30:31]
+  %v = call i16 @llvm.amdgcn.permlane16.i16(i16 %src0, i16 %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store i16 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @v_permlanex16_i16(ptr addrspace(1) %out, i16 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlanex16_i16:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    v_readfirstlane_b32 s4, v3
+; GFX10-NEXT:    v_readfirstlane_b32 s5, v4
+; GFX10-NEXT:    v_permlanex16_b32 v2, v2, s4, s5
+; GFX10-NEXT:    global_store_short v[0:1], v2, off
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: v_permlanex16_i16:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX11-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX11-NEXT:    global_store_b16 v[0:1], v2, off
+; GFX11-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-LABEL: v_permlanex16_i16:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-NEXT:    s_wait_expcnt 0x0
+; GFX12-NEXT:    s_wait_samplecnt 0x0
+; GFX12-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX12-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX12-NEXT:    global_store_b16 v[0:1], v2, off
+; GFX12-NEXT:    s_setpc_b64 s[30:31]
+  %v = call i16 @llvm.amdgcn.permlanex16.i16(i16 %src0, i16 %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store i16 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @v_permlane16_v2f16(ptr addrspace(1) %out, <2 x half> %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlane16_v2f16:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    v_readfirstlane_b32 s4, v3
+; GFX10-NEXT:    v_readfirstlane_b32 s5, v4
+; GFX10-NEXT:    v_permlane16_b32 v2, v2, s4, s5
+; GFX10-NEXT:    global_store_dword v[0:1], v2, off
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: v_permlane16_v2f16:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX11-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX11-NEXT:    global_store_b32 v[0:1], v2, off
+; GFX11-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-LABEL: v_permlane16_v2f16:
+; GFX12:       ; %bb.0:
+; GFX12-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-NEXT:    s_wait_expcnt 0x0
+; GFX12-NEXT:    s_wait_samplecnt 0x0
+; GFX12-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-NEXT:    s_wait_kmcnt 0x0
+; GFX12-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX12-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX12-NEXT:    global_store_b32 v[0:1], v2, off
+; GFX12-NEXT:    s_setpc_b64 s[30:31]
+  %v = call <2 x half> @llvm.amdgcn.permlane16.v2f16(<2 x half> %src0, <2 x half> %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store <2 x half> %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @v_permlanex16_v2f16(ptr addrspace(1) %out, <2 x half> %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlanex16_v2f16:
+; GFX10:       ; %bb.0:
+; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT:    v_readfirstlane_b32 s4, v3
+; GFX10-NEXT:    v_readfirstlane_b32 s5, v4
+; GFX10-NEXT:    v_permlanex16_b32 v2, v2, s4, s5
+; GFX10-NEXT:    global_store_dword v[0:1], v2, off
+; GFX10-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-LABEL: v_permlanex16_v2f16:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX11-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX11-NEXT:    global_store_b32 v[0:1], v2, off
+; GFX11-NEXT:    s_setpc_b64 s[30:31]
 ;
-; GFX12-LABEL: v_permlanex16_b32_undef_tid:
+; GFX12-LABEL: v_permlanex16_v2f16:
 ; GFX12:       ; %bb.0:
-; GFX12-NEXT:    s_clause 0x1
-; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
-; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX12-NEXT:    v_mov_b32_e32 v1, 0
+; GFX12-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-NEXT:    s_wait_expcnt 0x0
+; GFX12-NEXT:    s_wait_samplecnt 0x0
+; GFX12-NEXT:    s_wait_bvhcnt 0x0
 ; GFX12-NEXT:    s_wait_kmcnt 0x0
-; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s2, s3
-; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
-; GFX12-NEXT:    s_nop 0
-; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
-; GFX12-NEXT:    s_endpgm
-  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
-  %undef = freeze i32 poison
-  %v = call i32 @llvm.amdgcn.permlanex16(i32 %undef, i32 %tidx, i32 %src1, i32 %src2, i1 false, i1 false)
-  store i32 %v, ptr addrspace(1) %out
+; GFX12-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX12-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX12-NEXT:    global_store_b32 v[0:1], v2, off
+; GFX12-NEXT:    s_setpc_b64 s[30:31]
+  %v = call <2 x half> @llvm.amdgcn.permlanex16.v2f16(<2 x half> %src0, <2 x half> %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store <2 x half> %v, ptr addrspace(1) %out
   ret void
 }
 
-define amdgpu_kernel void @v_permlanex16_b32_i_tid(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-SDAG-LABEL: v_permlanex16_b32_i_tid:
+define void @v_permlane16_v2f32(ptr addrspace(1) %out, <2 x float> %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_v2f32:
 ; GFX10-SDAG:       ; %bb.0:
-; GFX10-SDAG-NEXT:    s_clause 0x1
-; GFX10-SDAG-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
-; GFX10-SDAG-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
-; GFX10-SDAG-NEXT:    v_mov_b32_e32 v1, 0x3039
-; GFX10-SDAG-NEXT:    v_mov_b32_e32 v2, 0
-; GFX10-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-SDAG-NEXT:    v_permlanex16_b32 v1, v0, s2, s3
-; GFX10-SDAG-NEXT:    global_store_dword v2, v1, s[4:5]
-; GFX10-SDAG-NEXT:    s_endpgm
-;
-; GFX10-GISEL-LABEL: v_permlanex16_b32_i_tid:
+; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v4
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v3, v3, s4, s5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v2, v2, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v[0:1], v[2:3], off
+; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-GISEL-LABEL: v_permlane16_v2f32:
 ; GFX10-GISEL:       ; %bb.0:
-; GFX10-GISEL-NEXT:    s_clause 0x1
-; GFX10-GISEL-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
-; GFX10-GISEL-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
-; GFX10-GISEL-NEXT:    v_mov_b32_e32 v1, 0x3039
-; GFX10-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-GISEL-NEXT:    v_permlanex16_b32 v1, v0, s2, s3
-; GFX10-GISEL-NEXT:    v_mov_b32_e32 v0, 0
-; GFX10-GISEL-NEXT:    global_store_dword v0, v1, s[4:5]
-; GFX10-GISEL-NEXT:    s_endpgm
-;
-; GFX11-SDAG-LABEL: v_permlanex16_b32_i_tid:
+; GFX10-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s4, v4
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s5, v5
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v2, v2, s4, s5
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v3, v3, s4, s5
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v[0:1], v[2:3], off
+; GFX10-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-LABEL: v_permlane16_v2f32:
 ; GFX11-SDAG:       ; %bb.0:
-; GFX11-SDAG-NEXT:    s_clause 0x1
-; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
-; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX11-SDAG-NEXT:    v_dual_mov_b32 v1, 0x3039 :: v_dual_mov_b32 v2, 0
-; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s0, v4
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v5
 ; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11-SDAG-NEXT:    v_permlanex16_b32 v1, v0, s2, s3
-; GFX11-SDAG-NEXT:    global_store_b32 v2, v1, s[0:1]
-; GFX11-SDAG-NEXT:    s_nop 0
-; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
-; GFX11-SDAG-NEXT:    s_endpgm
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v3, v3, s0, s1
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b64 v[0:1], v[2:3], off
+; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
 ;
-; GFX11-GISEL-LABEL: v_permlanex16_b32_i_tid:
+; GFX11-GISEL-LABEL: v_permlane16_v2f32:
 ; GFX11-GISEL:       ; %bb.0:
-; GFX11-GISEL-NEXT:    s_clause 0x1
-; GFX11-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
-; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX11-GISEL-NEXT:    v_mov_b32_e32 v1, 0x3039
-; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s0, v4
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s1, v5
 ; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX11-GISEL-NEXT:    v_permlanex16_b32 v1, v0, s2, s3
-; GFX11-GISEL-NEXT:    v_mov_b32_e32 v0, 0
-; GFX11-GISEL-NEXT:    global_store_b32 v0, v1, s[0:1]
-; GFX11-GISEL-NEXT:    s_nop 0
-; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
-; GFX11-GISEL-NEXT:    s_endpgm
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v3, v3, s0, s1
+; GFX11-GISEL-NEXT:    global_store_b64 v[0:1], v[2:3], off
+; GFX11-GISEL-NEXT:    s_setpc_b64 s[30:31]
 ;
-; GFX12-SDAG-LABEL: v_permlanex16_b32_i_tid:
+; GFX12-SDAG-LABEL: v_permlane16_v2f32:
 ; GFX12-SDAG:       ; %bb.0:
-; GFX12-SDAG-NEXT:    s_clause 0x1
-; GFX12-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
-; GFX12-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX12-SDAG-NEXT:    v_dual_mov_b32 v1, 0x3039 :: v_dual_mov_b32 v2, 0
+; GFX12-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_expcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_samplecnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_bvhcnt 0x0
 ; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s0, v4
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v5
 ; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX12-SDAG-NEXT:    v_permlanex16_b32 v1, v0, s2, s3
-; GFX12-SDAG-NEXT:    global_store_b32 v2, v1, s[0:1]
-; GFX12-SDAG-NEXT:    s_nop 0
-; GFX12-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
-; GFX12-SDAG-NEXT:    s_endpgm
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v3, v3, s0, s1
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b64 v[0:1], v[2:3], off
+; GFX12-SDAG-NEXT:    s_setpc_b64 s[30:31]
 ;
-; GFX12-GISEL-LABEL: v_permlanex16_b32_i_tid:
+; GFX12-GISEL-LABEL: v_permlane16_v2f32:
 ; GFX12-GISEL:       ; %bb.0:
-; GFX12-GISEL-NEXT:    s_clause 0x1
-; GFX12-GISEL-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
-; GFX12-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX12-GISEL-NEXT:    v_mov_b32_e32 v1, 0x3039
+; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-GISEL-NEXT:    s_wait_expcnt 0x0
+; GFX12-GISEL-NEXT:    s_wait_samplecnt 0x0
+; GFX12-GISEL-NEXT:    s_wait_bvhcnt 0x0
 ; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s0, v4
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s1, v5
 ; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX12-GISEL-NEXT:    v_permlanex16_b32 v1, v0, s2, s3
-; GFX12-GISEL-NEXT:    v_mov_b32_e32 v0, 0
-; GFX12-GISEL-NEXT:    global_store_b32 v0, v1, s[0:1]
-; GFX12-GISEL-NEXT:    s_nop 0
-; GFX12-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
-; GFX12-GISEL-NEXT:    s_endpgm
-  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
-  %v = call i32 @llvm.amdgcn.permlanex16(i32 12345, i32 %tidx, i32 %src1, i32 %src2, i1 false, i1 false)
-  store i32 %v, ptr addrspace(1) %out
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v3, v3, s0, s1
+; GFX12-GISEL-NEXT:    global_store_b64 v[0:1], v[2:3], off
+; GFX12-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %v = call <2 x float> @llvm.amdgcn.permlane16.v2f32(<2 x float> %src0, <2 x float> %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store <2 x float> %v, ptr addrspace(1) %out
   ret void
 }
 
-define amdgpu_kernel void @v_permlanex16_b32_i_tid_fi(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-LABEL: v_permlanex16_b32_i_tid_fi:
-; GFX10:       ; %bb.0:
-; GFX10-NEXT:    s_clause 0x1
-; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
-; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
-; GFX10-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,0]
-; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
-; GFX10-NEXT:    s_endpgm
+define void @v_permlanex16_v2f32(ptr addrspace(1) %out, <2 x float> %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_v2f32:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v4
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v3, v3, s4, s5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v[0:1], v[2:3], off
+; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-GISEL-LABEL: v_permlanex16_v2f32:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s4, v4
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s5, v5
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v2, v2, s4, s5
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v3, v3, s4, s5
+; GFX10-GISEL-NEXT:    global_store_dwordx2 v[0:1], v[2:3], off
+; GFX10-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-LABEL: v_permlanex16_v2f32:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s0, v4
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v5
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v3, v3, s0, s1
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b64 v[0:1], v[2:3], off
+; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
 ;
-; GFX11-LABEL: v_permlanex16_b32_i_tid_fi:
-; GFX11:       ; %bb.0:
-; GFX11-NEXT:    s_clause 0x1
-; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
-; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX11-NEXT:    v_mov_b32_e32 v1, 0
-; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,0]
-; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
-; GFX11-NEXT:    s_nop 0
-; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
-; GFX11-NEXT:    s_endpgm
+; GFX11-GISEL-LABEL: v_permlanex16_v2f32:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s0, v4
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s1, v5
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v3, v3, s0, s1
+; GFX11-GISEL-NEXT:    global_store_b64 v[0:1], v[2:3], off
+; GFX11-GISEL-NEXT:    s_setpc_b64 s[30:31]
 ;
-; GFX12-LABEL: v_permlanex16_b32_i_tid_fi:
-; GFX12:       ; %bb.0:
-; GFX12-NEXT:    s_clause 0x1
-; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
-; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX12-NEXT:    v_mov_b32_e32 v1, 0
-; GFX12-NEXT:    s_wait_kmcnt 0x0
-; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,0]
-; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
-; GFX12-NEXT:    s_nop 0
-; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
-; GFX12-NEXT:    s_endpgm
-  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
-  %undef = freeze i32 poison
-  %v = call i32 @llvm.amdgcn.permlanex16(i32 %undef, i32 %tidx, i32 %src1, i32 %src2, i1 true, i1 false)
-  store i32 %v, ptr addrspace(1) %out
+; GFX12-SDAG-LABEL: v_permlanex16_v2f32:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_expcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_samplecnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s0, v4
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v5
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v3, v3, s0, s1
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b64 v[0:1], v[2:3], off
+; GFX12-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-GISEL-LABEL: v_permlanex16_v2f32:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-GISEL-NEXT:    s_wait_expcnt 0x0
+; GFX12-GISEL-NEXT:    s_wait_samplecnt 0x0
+; GFX12-GISEL-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s0, v4
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s1, v5
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v3, v3, s0, s1
+; GFX12-GISEL-NEXT:    global_store_b64 v[0:1], v[2:3], off
+; GFX12-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %v = call <2 x float> @llvm.amdgcn.permlanex16.v2f32(<2 x float> %src0, <2 x float> %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store <2 x float> %v, ptr addrspace(1) %out
   ret void
 }
 
-define amdgpu_kernel void @v_permlanex16_b32_i_tid_bc(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-LABEL: v_permlanex16_b32_i_tid_bc:
-; GFX10:       ; %bb.0:
-; GFX10-NEXT:    s_clause 0x1
-; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
-; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
-; GFX10-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[0,1]
-; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
-; GFX10-NEXT:    s_endpgm
+define void @v_permlane16_v7i32(ptr addrspace(1) %out, <7 x i32> %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_v7i32:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v9
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v10
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v8, v8, s4, s5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v7, v7, s4, s5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v6, v6, s4, s5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v5, v5, s4, s5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v4, v4, s4, s5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v3, v3, s4, s5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v2, v2, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dwordx3 v[0:1], v[6:8], off offset:16
+; GFX10-SDAG-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off
+; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-GISEL-LABEL: v_permlane16_v7i32:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s4, v9
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s5, v10
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v2, v2, s4, s5
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v3, v3, s4, s5
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v4, v4, s4, s5
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v5, v5, s4, s5
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v6, v6, s4, s5
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v7, v7, s4, s5
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v8, v8, s4, s5
+; GFX10-GISEL-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off
+; GFX10-GISEL-NEXT:    global_store_dwordx3 v[0:1], v[6:8], off offset:16
+; GFX10-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-LABEL: v_permlane16_v7i32:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s0, v9
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v10
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v8, v8, s0, s1
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v7, v7, s0, s1
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v6, v6, s0, s1
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v5, v5, s0, s1
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v4, v4, s0, s1
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v3, v3, s0, s1
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    global_store_b96 v[0:1], v[6:8], off offset:16
+; GFX11-SDAG-NEXT:    global_store_b128 v[0:1], v[2:5], off
+; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
 ;
-; GFX11-LABEL: v_permlanex16_b32_i_tid_bc:
-; GFX11:       ; %bb.0:
-; GFX11-NEXT:    s_clause 0x1
-; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
-; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX11-NEXT:    v_mov_b32_e32 v1, 0
-; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[0,1]
-; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
-; GFX11-NEXT:    s_nop 0
-; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
-; GFX11-NEXT:    s_endpgm
+; GFX11-GISEL-LABEL: v_permlane16_v7i32:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s0, v9
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s1, v10
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v3, v3, s0, s1
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v4, v4, s0, s1
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v5, v5, s0, s1
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v6, v6, s0, s1
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v7, v7, s0, s1
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v8, v8, s0, s1
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    global_store_b128 v[0:1], v[2:5], off
+; GFX11-GISEL-NEXT:    global_store_b96 v[0:1], v[6:8], off offset:16
+; GFX11-GISEL-NEXT:    s_setpc_b64 s[30:31]
 ;
-; GFX12-LABEL: v_permlanex16_b32_i_tid_bc:
-; GFX12:       ; %bb.0:
-; GFX12-NEXT:    s_clause 0x1
-; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
-; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX12-NEXT:    v_mov_b32_e32 v1, 0
-; GFX12-NEXT:    s_wait_kmcnt 0x0
-; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[0,1]
-; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
-; GFX12-NEXT:    s_nop 0
-; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
-; GFX12-NEXT:    s_endpgm
-  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
-  %undef = freeze i32 poison
-  %v = call i32 @llvm.amdgcn.permlanex16(i32 %undef, i32 %tidx, i32 %src1, i32 %src2, i1 false, i1 true)
-  store i32 %v, ptr addrspace(1) %out
+; GFX12-SDAG-LABEL: v_permlane16_v7i32:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_expcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_samplecnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s0, v9
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v10
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v8, v8, s0, s1
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v7, v7, s0, s1
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v6, v6, s0, s1
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v5, v5, s0, s1
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v4, v4, s0, s1
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v3, v3, s0, s1
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    global_store_b96 v[0:1], v[6:8], off offset:16
+; GFX12-SDAG-NEXT:    global_store_b128 v[0:1], v[2:5], off
+; GFX12-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-GISEL-LABEL: v_permlane16_v7i32:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-GISEL-NEXT:    s_wait_expcnt 0x0
+; GFX12-GISEL-NEXT:    s_wait_samplecnt 0x0
+; GFX12-GISEL-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s0, v9
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s1, v10
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v3, v3, s0, s1
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v4, v4, s0, s1
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v5, v5, s0, s1
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v6, v6, s0, s1
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v7, v7, s0, s1
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v8, v8, s0, s1
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    global_store_b128 v[0:1], v[2:5], off
+; GFX12-GISEL-NEXT:    global_store_b96 v[0:1], v[6:8], off offset:16
+; GFX12-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %v = call <7 x i32> @llvm.amdgcn.permlane16.v7i32(<7 x i32> %src0, <7 x i32> %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store <7 x i32> %v, ptr addrspace(1) %out
   ret void
 }
 
-define amdgpu_kernel void @v_permlanex16_b32_i_tid_fi_bc(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-LABEL: v_permlanex16_b32_i_tid_fi_bc:
-; GFX10:       ; %bb.0:
-; GFX10-NEXT:    s_clause 0x1
-; GFX10-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x30
-; GFX10-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x24
-; GFX10-NEXT:    v_mov_b32_e32 v1, 0
-; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,1]
-; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
-; GFX10-NEXT:    s_endpgm
+define void @v_permlanex16_v7i32(ptr addrspace(1) %out, <7 x i32> %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_v7i32:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v9
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v10
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v8, v8, s4, s5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v7, v7, s4, s5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v6, v6, s4, s5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v5, v5, s4, s5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v4, v4, s4, s5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v3, v3, s4, s5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dwordx3 v[0:1], v[6:8], off offset:16
+; GFX10-SDAG-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off
+; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-GISEL-LABEL: v_permlanex16_v7i32:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s4, v9
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s5, v10
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v2, v2, s4, s5
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v3, v3, s4, s5
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v4, v4, s4, s5
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v5, v5, s4, s5
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v6, v6, s4, s5
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v7, v7, s4, s5
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v8, v8, s4, s5
+; GFX10-GISEL-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off
+; GFX10-GISEL-NEXT:    global_store_dwordx3 v[0:1], v[6:8], off offset:16
+; GFX10-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-LABEL: v_permlanex16_v7i32:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s0, v9
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v10
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v8, v8, s0, s1
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v7, v7, s0, s1
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v6, v6, s0, s1
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v5, v5, s0, s1
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v4, v4, s0, s1
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v3, v3, s0, s1
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    global_store_b96 v[0:1], v[6:8], off offset:16
+; GFX11-SDAG-NEXT:    global_store_b128 v[0:1], v[2:5], off
+; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
 ;
-; GFX11-LABEL: v_permlanex16_b32_i_tid_fi_bc:
-; GFX11:       ; %bb.0:
-; GFX11-NEXT:    s_clause 0x1
-; GFX11-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
-; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX11-NEXT:    v_mov_b32_e32 v1, 0
-; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX11-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,1]
-; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
-; GFX11-NEXT:    s_nop 0
-; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
-; GFX11-NEXT:    s_endpgm
+; GFX11-GISEL-LABEL: v_permlanex16_v7i32:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s0, v9
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s1, v10
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v3, v3, s0, s1
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v4, v4, s0, s1
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v5, v5, s0, s1
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v6, v6, s0, s1
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v7, v7, s0, s1
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v8, v8, s0, s1
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    global_store_b128 v[0:1], v[2:5], off
+; GFX11-GISEL-NEXT:    global_store_b96 v[0:1], v[6:8], off offset:16
+; GFX11-GISEL-NEXT:    s_setpc_b64 s[30:31]
 ;
-; GFX12-LABEL: v_permlanex16_b32_i_tid_fi_bc:
-; GFX12:       ; %bb.0:
-; GFX12-NEXT:    s_clause 0x1
-; GFX12-NEXT:    s_load_b64 s[2:3], s[0:1], 0x30
-; GFX12-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX12-NEXT:    v_mov_b32_e32 v1, 0
-; GFX12-NEXT:    s_wait_kmcnt 0x0
-; GFX12-NEXT:    v_permlanex16_b32 v0, v0, s2, s3 op_sel:[1,1]
-; GFX12-NEXT:    global_store_b32 v1, v0, s[0:1]
-; GFX12-NEXT:    s_nop 0
-; GFX12-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
-; GFX12-NEXT:    s_endpgm
-  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
-  %undef = freeze i32 poison
-  %v = call i32 @llvm.amdgcn.permlanex16(i32 %undef, i32 %tidx, i32 %src1, i32 %src2, i1 true, i1 true)
-  store i32 %v, ptr addrspace(1) %out
+; GFX12-SDAG-LABEL: v_permlanex16_v7i32:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_expcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_samplecnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s0, v9
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v10
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v8, v8, s0, s1
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v7, v7, s0, s1
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v6, v6, s0, s1
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v5, v5, s0, s1
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v4, v4, s0, s1
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v3, v3, s0, s1
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    global_store_b96 v[0:1], v[6:8], off offset:16
+; GFX12-SDAG-NEXT:    global_store_b128 v[0:1], v[2:5], off
+; GFX12-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-GISEL-LABEL: v_permlanex16_v7i32:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-GISEL-NEXT:    s_wait_expcnt 0x0
+; GFX12-GISEL-NEXT:    s_wait_samplecnt 0x0
+; GFX12-GISEL-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s0, v9
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s1, v10
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v3, v3, s0, s1
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v4, v4, s0, s1
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v5, v5, s0, s1
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v6, v6, s0, s1
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v7, v7, s0, s1
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v8, v8, s0, s1
+; GFX12-GISEL-NEXT:    s_clause 0x1
+; GFX12-GISEL-NEXT:    global_store_b128 v[0:1], v[2:5], off
+; GFX12-GISEL-NEXT:    global_store_b96 v[0:1], v[6:8], off offset:16
+; GFX12-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %v = call <7 x i32> @llvm.amdgcn.permlanex16.v7i32(<7 x i32> %src0, <7 x i32> %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store <7 x i32> %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @v_permlane16_v8i16(ptr addrspace(1) %out, <8 x i16> %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_v8i16:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v6
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v7
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v5, v5, s4, s5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v4, v4, s4, s5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v3, v3, s4, s5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v2, v2, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off
+; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-GISEL-LABEL: v_permlane16_v8i16:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s4, v6
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s5, v7
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v2, v2, s4, s5
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v3, v3, s4, s5
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v4, v4, s4, s5
+; GFX10-GISEL-NEXT:    v_permlane16_b32 v5, v5, s4, s5
+; GFX10-GISEL-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off
+; GFX10-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-LABEL: v_permlane16_v8i16:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s0, v6
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v7
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v5, v5, s0, s1
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v4, v4, s0, s1
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v3, v3, s0, s1
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b128 v[0:1], v[2:5], off
+; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-GISEL-LABEL: v_permlane16_v8i16:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s0, v6
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s1, v7
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v3, v3, s0, s1
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v4, v4, s0, s1
+; GFX11-GISEL-NEXT:    v_permlane16_b32 v5, v5, s0, s1
+; GFX11-GISEL-NEXT:    global_store_b128 v[0:1], v[2:5], off
+; GFX11-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-SDAG-LABEL: v_permlane16_v8i16:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_expcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_samplecnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s0, v6
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v7
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v5, v5, s0, s1
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v4, v4, s0, s1
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v3, v3, s0, s1
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b128 v[0:1], v[2:5], off
+; GFX12-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-GISEL-LABEL: v_permlane16_v8i16:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-GISEL-NEXT:    s_wait_expcnt 0x0
+; GFX12-GISEL-NEXT:    s_wait_samplecnt 0x0
+; GFX12-GISEL-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s0, v6
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s1, v7
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v3, v3, s0, s1
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v4, v4, s0, s1
+; GFX12-GISEL-NEXT:    v_permlane16_b32 v5, v5, s0, s1
+; GFX12-GISEL-NEXT:    global_store_b128 v[0:1], v[2:5], off
+; GFX12-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %v = call <8 x i16> @llvm.amdgcn.permlane16.v8i16(<8 x i16> %src0, <8 x i16> %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store <8 x i16> %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @v_permlanex16_v8i16(ptr addrspace(1) %out, <8 x i16> %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_v8i16:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v6
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v7
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v5, v5, s4, s5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v4, v4, s4, s5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v3, v3, s4, s5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off
+; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX10-GISEL-LABEL: v_permlanex16_v8i16:
+; GFX10-GISEL:       ; %bb.0:
+; GFX10-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s4, v6
+; GFX10-GISEL-NEXT:    v_readfirstlane_b32 s5, v7
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v2, v2, s4, s5
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v3, v3, s4, s5
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v4, v4, s4, s5
+; GFX10-GISEL-NEXT:    v_permlanex16_b32 v5, v5, s4, s5
+; GFX10-GISEL-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off
+; GFX10-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-LABEL: v_permlanex16_v8i16:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s0, v6
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v7
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v5, v5, s0, s1
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v4, v4, s0, s1
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v3, v3, s0, s1
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b128 v[0:1], v[2:5], off
+; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-GISEL-LABEL: v_permlanex16_v8i16:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s0, v6
+; GFX11-GISEL-NEXT:    v_readfirstlane_b32 s1, v7
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v3, v3, s0, s1
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v4, v4, s0, s1
+; GFX11-GISEL-NEXT:    v_permlanex16_b32 v5, v5, s0, s1
+; GFX11-GISEL-NEXT:    global_store_b128 v[0:1], v[2:5], off
+; GFX11-GISEL-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-SDAG-LABEL: v_permlanex16_v8i16:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_expcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_samplecnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s0, v6
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v7
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v5, v5, s0, s1
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v4, v4, s0, s1
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v3, v3, s0, s1
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b128 v[0:1], v[2:5], off
+; GFX12-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-GISEL-LABEL: v_permlanex16_v8i16:
+; GFX12-GISEL:       ; %bb.0:
+; GFX12-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-GISEL-NEXT:    s_wait_expcnt 0x0
+; GFX12-GISEL-NEXT:    s_wait_samplecnt 0x0
+; GFX12-GISEL-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s0, v6
+; GFX12-GISEL-NEXT:    v_readfirstlane_b32 s1, v7
+; GFX12-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v3, v3, s0, s1
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v4, v4, s0, s1
+; GFX12-GISEL-NEXT:    v_permlanex16_b32 v5, v5, s0, s1
+; GFX12-GISEL-NEXT:    global_store_b128 v[0:1], v[2:5], off
+; GFX12-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %v = call <8 x i16> @llvm.amdgcn.permlanex16.v8i16(<8 x i16> %src0, <8 x i16> %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store <8 x i16> %v, ptr addrspace(1) %out
   ret void
 }
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane.ptr.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane.ptr.ll
new file mode 100644
index 0000000000000..bb42834221681
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane.ptr.ll
@@ -0,0 +1,694 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc -global-isel=0 -amdgpu-load-store-vectorizer=0 -mtriple=amdgcn -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX10-SDAG %s
+; RUN: llc -global-isel=0 -amdgpu-load-store-vectorizer=0 -mtriple=amdgcn -mcpu=gfx1100 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX11-SDAG %s
+; RUN: llc -global-isel=0 -amdgpu-load-store-vectorizer=0 -mtriple=amdgcn -mcpu=gfx1200 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX12-SDAG %s
+
+define void @v_permlane16_p0(ptr addrspace(1) %out, ptr %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_p0:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v4
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v3, v3, s4, s5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v2, v2, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v[0:1], v[2:3], off
+; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-LABEL: v_permlane16_p0:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s0, v4
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v5
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v3, v3, s0, s1
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b64 v[0:1], v[2:3], off
+; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-SDAG-LABEL: v_permlane16_p0:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_expcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_samplecnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s0, v4
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v5
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v3, v3, s0, s1
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b64 v[0:1], v[2:3], off
+; GFX12-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %v = call ptr @llvm.amdgcn.permlane16.p0(ptr %src0, ptr %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store ptr %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @v_permlanex16_p0(ptr addrspace(1) %out, ptr %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_p0:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v4
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v3, v3, s4, s5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v[0:1], v[2:3], off
+; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-LABEL: v_permlanex16_p0:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s0, v4
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v5
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v3, v3, s0, s1
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b64 v[0:1], v[2:3], off
+; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-SDAG-LABEL: v_permlanex16_p0:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_expcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_samplecnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s0, v4
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v5
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v3, v3, s0, s1
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b64 v[0:1], v[2:3], off
+; GFX12-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %v = call ptr @llvm.amdgcn.permlanex16.p0(ptr %src0, ptr %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store ptr %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @v_permlane16_v3p0(ptr addrspace(1) %out, <3 x ptr> %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_v3p0:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v8
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v9
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v7, v7, s4, s5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v6, v6, s4, s5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v5, v5, s4, s5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v4, v4, s4, s5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v3, v3, s4, s5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v2, v2, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v[0:1], v[6:7], off offset:16
+; GFX10-SDAG-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off
+; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-LABEL: v_permlane16_v3p0:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s0, v8
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v9
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v7, v7, s0, s1
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v6, v6, s0, s1
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v5, v5, s0, s1
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v4, v4, s0, s1
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v3, v3, s0, s1
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    global_store_b64 v[0:1], v[6:7], off offset:16
+; GFX11-SDAG-NEXT:    global_store_b128 v[0:1], v[2:5], off
+; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-SDAG-LABEL: v_permlane16_v3p0:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_expcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_samplecnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s0, v8
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v9
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v7, v7, s0, s1
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v6, v6, s0, s1
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v5, v5, s0, s1
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v4, v4, s0, s1
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v3, v3, s0, s1
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    global_store_b64 v[0:1], v[6:7], off offset:16
+; GFX12-SDAG-NEXT:    global_store_b128 v[0:1], v[2:5], off
+; GFX12-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %v = call <3 x ptr> @llvm.amdgcn.permlane16.v3p0(<3 x ptr> %src0, <3 x ptr> %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store <3 x ptr> %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @v_permlanex16_v3p0(ptr addrspace(1) %out, <3 x ptr> %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_v3p0:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v8
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v9
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v7, v7, s4, s5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v6, v6, s4, s5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v5, v5, s4, s5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v4, v4, s4, s5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v3, v3, s4, s5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dwordx2 v[0:1], v[6:7], off offset:16
+; GFX10-SDAG-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off
+; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-LABEL: v_permlanex16_v3p0:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s0, v8
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v9
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v7, v7, s0, s1
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v6, v6, s0, s1
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v5, v5, s0, s1
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v4, v4, s0, s1
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v3, v3, s0, s1
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    global_store_b64 v[0:1], v[6:7], off offset:16
+; GFX11-SDAG-NEXT:    global_store_b128 v[0:1], v[2:5], off
+; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-SDAG-LABEL: v_permlanex16_v3p0:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_expcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_samplecnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s0, v8
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v9
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v7, v7, s0, s1
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v6, v6, s0, s1
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v5, v5, s0, s1
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v4, v4, s0, s1
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v3, v3, s0, s1
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX12-SDAG-NEXT:    s_clause 0x1
+; GFX12-SDAG-NEXT:    global_store_b64 v[0:1], v[6:7], off offset:16
+; GFX12-SDAG-NEXT:    global_store_b128 v[0:1], v[2:5], off
+; GFX12-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %v = call <3 x ptr> @llvm.amdgcn.permlanex16.v3p0(<3 x ptr> %src0, <3 x ptr> %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store <3 x ptr> %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @v_permlane16_p3(ptr addrspace(1) %out, ptr addrspace(3) %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_p3:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v3
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v4
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v2, v2, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dword v[0:1], v2, off
+; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-LABEL: v_permlane16_p3:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b32 v[0:1], v2, off
+; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-SDAG-LABEL: v_permlane16_p3:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_expcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_samplecnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b32 v[0:1], v2, off
+; GFX12-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %v = call ptr addrspace(3) @llvm.amdgcn.permlane16.p3(ptr addrspace(3) %src0, ptr addrspace(3) %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store ptr addrspace(3) %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @v_permlanex16_p3(ptr addrspace(1) %out, ptr addrspace(3) %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_p3:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v3
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v4
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dword v[0:1], v2, off
+; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-LABEL: v_permlanex16_p3:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b32 v[0:1], v2, off
+; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-SDAG-LABEL: v_permlanex16_p3:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_expcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_samplecnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b32 v[0:1], v2, off
+; GFX12-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %v = call ptr addrspace(3) @llvm.amdgcn.permlanex16.p3(ptr addrspace(3) %src0, ptr addrspace(3) %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store ptr addrspace(3) %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @v_permlane16_v3p3(ptr addrspace(1) %out, <3 x ptr addrspace(3)> %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_v3p3:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v5
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v6
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v4, v4, s4, s5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v3, v3, s4, s5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v2, v2, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dwordx3 v[0:1], v[2:4], off
+; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-LABEL: v_permlane16_v3p3:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s0, v5
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v4, v4, s0, s1
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v3, v3, s0, s1
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b96 v[0:1], v[2:4], off
+; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-SDAG-LABEL: v_permlane16_v3p3:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_expcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_samplecnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s0, v5
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v4, v4, s0, s1
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v3, v3, s0, s1
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b96 v[0:1], v[2:4], off
+; GFX12-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %v = call <3 x ptr addrspace(3)> @llvm.amdgcn.permlane16.v3p3(<3 x ptr addrspace(3)> %src0, <3 x ptr addrspace(3)> %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store <3 x ptr addrspace(3)> %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @v_permlanex16_v3p3(ptr addrspace(1) %out, <3 x ptr addrspace(3)> %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_v3p3:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v5
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v6
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v4, v4, s4, s5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v3, v3, s4, s5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dwordx3 v[0:1], v[2:4], off
+; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-LABEL: v_permlanex16_v3p3:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s0, v5
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v4, v4, s0, s1
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v3, v3, s0, s1
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b96 v[0:1], v[2:4], off
+; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-SDAG-LABEL: v_permlanex16_v3p3:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_expcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_samplecnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s0, v5
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v4, v4, s0, s1
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v3, v3, s0, s1
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b96 v[0:1], v[2:4], off
+; GFX12-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %v = call <3 x ptr addrspace(3)> @llvm.amdgcn.permlanex16.v3p3(<3 x ptr addrspace(3)> %src0, <3 x ptr addrspace(3)> %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store <3 x ptr addrspace(3)> %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @v_permlane16_p5(ptr addrspace(1) %out, ptr addrspace(5) %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_p5:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v3
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v4
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v2, v2, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dword v[0:1], v2, off
+; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-LABEL: v_permlane16_p5:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b32 v[0:1], v2, off
+; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-SDAG-LABEL: v_permlane16_p5:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_expcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_samplecnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b32 v[0:1], v2, off
+; GFX12-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %v = call ptr addrspace(5) @llvm.amdgcn.permlane16.p5(ptr addrspace(5) %src0, ptr addrspace(5) %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store ptr addrspace(5) %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @v_permlanex16_p5(ptr addrspace(1) %out, ptr addrspace(5) %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_p5:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v3
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v4
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dword v[0:1], v2, off
+; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-LABEL: v_permlanex16_p5:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b32 v[0:1], v2, off
+; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-SDAG-LABEL: v_permlanex16_p5:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_expcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_samplecnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b32 v[0:1], v2, off
+; GFX12-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %v = call ptr addrspace(5) @llvm.amdgcn.permlanex16.p5(ptr addrspace(5) %src0, ptr addrspace(5) %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store ptr addrspace(5) %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @v_permlane16_v3p5(ptr addrspace(1) %out, <3 x ptr addrspace(5)> %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_v3p5:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v5
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v6
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v4, v4, s4, s5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v3, v3, s4, s5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v2, v2, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dwordx3 v[0:1], v[2:4], off
+; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-LABEL: v_permlane16_v3p5:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s0, v5
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v4, v4, s0, s1
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v3, v3, s0, s1
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b96 v[0:1], v[2:4], off
+; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-SDAG-LABEL: v_permlane16_v3p5:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_expcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_samplecnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s0, v5
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v4, v4, s0, s1
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v3, v3, s0, s1
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b96 v[0:1], v[2:4], off
+; GFX12-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %v = call <3 x ptr addrspace(5)> @llvm.amdgcn.permlane16.v3p5(<3 x ptr addrspace(5)> %src0, <3 x ptr addrspace(5)> %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store <3 x ptr addrspace(5)> %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @v_permlanex16_v3p5(ptr addrspace(1) %out, <3 x ptr addrspace(5)> %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_v3p5:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v5
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v6
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v4, v4, s4, s5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v3, v3, s4, s5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dwordx3 v[0:1], v[2:4], off
+; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-LABEL: v_permlanex16_v3p5:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s0, v5
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v4, v4, s0, s1
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v3, v3, s0, s1
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b96 v[0:1], v[2:4], off
+; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-SDAG-LABEL: v_permlanex16_v3p5:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_expcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_samplecnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s0, v5
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v4, v4, s0, s1
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v3, v3, s0, s1
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b96 v[0:1], v[2:4], off
+; GFX12-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %v = call <3 x ptr addrspace(5)> @llvm.amdgcn.permlanex16.v3p5(<3 x ptr addrspace(5)> %src0, <3 x ptr addrspace(5)> %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store <3 x ptr addrspace(5)> %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @v_permlane16_p6(ptr addrspace(1) %out, ptr addrspace(6) %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_p6:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v3
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v4
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v2, v2, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dword v[0:1], v2, off
+; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-LABEL: v_permlane16_p6:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b32 v[0:1], v2, off
+; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-SDAG-LABEL: v_permlane16_p6:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_expcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_samplecnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b32 v[0:1], v2, off
+; GFX12-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %v = call ptr addrspace(6) @llvm.amdgcn.permlane16.p6(ptr addrspace(6) %src0, ptr addrspace(6) %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store ptr addrspace(6) %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @v_permlanex16_p6(ptr addrspace(1) %out, ptr addrspace(6) %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_p6:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v3
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v4
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dword v[0:1], v2, off
+; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-LABEL: v_permlanex16_p6:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b32 v[0:1], v2, off
+; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-SDAG-LABEL: v_permlanex16_p6:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_expcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_samplecnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b32 v[0:1], v2, off
+; GFX12-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %v = call ptr addrspace(6) @llvm.amdgcn.permlanex16.p6(ptr addrspace(6) %src0, ptr addrspace(6) %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store ptr addrspace(6) %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @v_permlane16_v3p6(ptr addrspace(1) %out, <3 x ptr addrspace(6)> %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlane16_v3p6:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v5
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v6
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v4, v4, s4, s5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v3, v3, s4, s5
+; GFX10-SDAG-NEXT:    v_permlane16_b32 v2, v2, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dwordx3 v[0:1], v[2:4], off
+; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-LABEL: v_permlane16_v3p6:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s0, v5
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v4, v4, s0, s1
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v3, v3, s0, s1
+; GFX11-SDAG-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b96 v[0:1], v[2:4], off
+; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-SDAG-LABEL: v_permlane16_v3p6:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_expcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_samplecnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s0, v5
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v4, v4, s0, s1
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v3, v3, s0, s1
+; GFX12-SDAG-NEXT:    v_permlane16_b32 v2, v2, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b96 v[0:1], v[2:4], off
+; GFX12-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %v = call <3 x ptr addrspace(6)> @llvm.amdgcn.permlane16.v3p6(<3 x ptr addrspace(6)> %src0, <3 x ptr addrspace(6)> %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store <3 x ptr addrspace(6)> %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @v_permlanex16_v3p6(ptr addrspace(1) %out, <3 x ptr addrspace(6)> %src0, i32 %src1, i32 %src2) {
+; GFX10-SDAG-LABEL: v_permlanex16_v3p6:
+; GFX10-SDAG:       ; %bb.0:
+; GFX10-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s4, v5
+; GFX10-SDAG-NEXT:    v_readfirstlane_b32 s5, v6
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v4, v4, s4, s5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v3, v3, s4, s5
+; GFX10-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s4, s5
+; GFX10-SDAG-NEXT:    global_store_dwordx3 v[0:1], v[2:4], off
+; GFX10-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-SDAG-LABEL: v_permlanex16_v3p6:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s0, v5
+; GFX11-SDAG-NEXT:    v_readfirstlane_b32 s1, v6
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v4, v4, s0, s1
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v3, v3, s0, s1
+; GFX11-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX11-SDAG-NEXT:    global_store_b96 v[0:1], v[2:4], off
+; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX12-SDAG-LABEL: v_permlanex16_v3p6:
+; GFX12-SDAG:       ; %bb.0:
+; GFX12-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_expcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_samplecnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_bvhcnt 0x0
+; GFX12-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s0, v5
+; GFX12-SDAG-NEXT:    v_readfirstlane_b32 s1, v6
+; GFX12-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v4, v4, s0, s1
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v3, v3, s0, s1
+; GFX12-SDAG-NEXT:    v_permlanex16_b32 v2, v2, s0, s1
+; GFX12-SDAG-NEXT:    global_store_b96 v[0:1], v[2:4], off
+; GFX12-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %v = call <3 x ptr addrspace(6)> @llvm.amdgcn.permlanex16.v3p6(<3 x ptr addrspace(6)> %src0, <3 x ptr addrspace(6)> %src0, i32 %src1, i32 %src2, i1 false, i1 false)
+  store <3 x ptr addrspace(6)> %v, ptr addrspace(1) %out
+  ret void
+}
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ll
index b81cb97725648..f653baa7365c7 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ll
@@ -5,8 +5,8 @@
 declare i32 @llvm.amdgcn.permlane64(i32)
 declare i32 @llvm.amdgcn.workitem.id.x()
 
-define amdgpu_kernel void @test_s(ptr addrspace(1) %out, i32 %src0) {
-; GFX11-LABEL: test_s:
+define amdgpu_kernel void @test_s_i32(ptr addrspace(1) %out, i32 %src0) {
+; GFX11-LABEL: test_s_i32:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_clause 0x1
 ; GFX11-NEXT:    s_load_b32 s2, s[0:1], 0x2c
@@ -19,13 +19,98 @@ define amdgpu_kernel void @test_s(ptr addrspace(1) %out, i32 %src0) {
 ; GFX11-NEXT:    s_nop 0
 ; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
-  %v = call i32 @llvm.amdgcn.permlane64(i32 %src0)
+  %v = call i32 @llvm.amdgcn.permlane64.i32(i32 %src0)
   store i32 %v, ptr addrspace(1) %out
   ret void
 }
 
-define amdgpu_kernel void @test_i(ptr addrspace(1) %out) {
-; GFX11-LABEL: test_i:
+define amdgpu_kernel void @test_s_f32(ptr addrspace(1) %out, float %src0) {
+; GFX11-LABEL: test_s_f32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_clause 0x1
+; GFX11-NEXT:    s_load_b32 s2, s[0:1], 0x2c
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_permlane64_b32 v0, v0
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+  %v = call float @llvm.amdgcn.permlane64.f32(float %src0)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @test_s_i64(ptr addrspace(1) %out, i64 %src0) {
+; GFX11-SDAG-LABEL: test_s_i64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v3, 0 :: v_dual_mov_b32 v0, s3
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, s2
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v1, v0
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v0, v2
+; GFX11-SDAG-NEXT:    global_store_b64 v3, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: test_s_i64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlane64_b32 v0, v0
+; GFX11-GISEL-NEXT:    v_permlane64_b32 v1, v1
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+  %v = call i64 @llvm.amdgcn.permlane64.i64(i64 %src0)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @test_s_f64(ptr addrspace(1) %out, double %src0) {
+; GFX11-SDAG-LABEL: test_s_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v3, 0 :: v_dual_mov_b32 v0, s3
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, s2
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v1, v0
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v0, v2
+; GFX11-SDAG-NEXT:    global_store_b64 v3, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: test_s_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlane64_b32 v0, v0
+; GFX11-GISEL-NEXT:    v_permlane64_b32 v1, v1
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+  %v = call double @llvm.amdgcn.permlane64.f64(double %src0)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @test_i_i32(ptr addrspace(1) %out) {
+; GFX11-LABEL: test_i_i32:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11-NEXT:    v_dual_mov_b32 v0, 0x63 :: v_dual_mov_b32 v1, 0
@@ -36,13 +121,95 @@ define amdgpu_kernel void @test_i(ptr addrspace(1) %out) {
 ; GFX11-NEXT:    s_nop 0
 ; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
-  %v = call i32 @llvm.amdgcn.permlane64(i32 99)
+  %v = call i32 @llvm.amdgcn.permlane64.i32(i32 99)
   store i32 %v, ptr addrspace(1) %out
   ret void
 }
 
-define amdgpu_kernel void @test_v(ptr addrspace(1) %out, i32 %src0) #1 {
-; GFX11-SDAG-LABEL: test_v:
+define amdgpu_kernel void @test_i_f32(ptr addrspace(1) %out) {
+; GFX11-LABEL: test_i_f32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    v_dual_mov_b32 v0, 0x449a5000 :: v_dual_mov_b32 v1, 0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_permlane64_b32 v0, v0
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+  %v = call float @llvm.amdgcn.permlane64.f32(float 1234.5)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @test_i_i64(ptr addrspace(1) %out) {
+; GFX11-SDAG-LABEL: test_i_i64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, 0x63
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v1, v2
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v0, v0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: test_i_i64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v0, 0x63
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlane64_b32 v0, v0
+; GFX11-GISEL-NEXT:    v_permlane64_b32 v1, v2
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+  %v = call i64 @llvm.amdgcn.permlane64.i64(i64 99)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @test_i_f64(ptr addrspace(1) %out) {
+; GFX11-SDAG-LABEL: test_i_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v0, 0x40934a00
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v1, v0
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v0, v2
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: test_i_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, 0x40934a00
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-GISEL-NEXT:    v_permlane64_b32 v0, v2
+; GFX11-GISEL-NEXT:    v_permlane64_b32 v1, v1
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+  %v = call double @llvm.amdgcn.permlane64.f64(double 1234.5)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @test_v_i32(ptr addrspace(1) %out, i32 %src0) #1 {
+; GFX11-SDAG-LABEL: test_v_i32:
 ; GFX11-SDAG:       ; %bb.0:
 ; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11-SDAG-NEXT:    v_mov_b32_e32 v1, 0
@@ -53,7 +220,7 @@ define amdgpu_kernel void @test_v(ptr addrspace(1) %out, i32 %src0) #1 {
 ; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-SDAG-NEXT:    s_endpgm
 ;
-; GFX11-GISEL-LABEL: test_v:
+; GFX11-GISEL-LABEL: test_v_i32:
 ; GFX11-GISEL:       ; %bb.0:
 ; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
 ; GFX11-GISEL-NEXT:    v_permlane64_b32 v0, v0
@@ -64,7 +231,221 @@ define amdgpu_kernel void @test_v(ptr addrspace(1) %out, i32 %src0) #1 {
 ; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-GISEL-NEXT:    s_endpgm
   %tidx = call i32 @llvm.amdgcn.workitem.id.x()
-  %v = call i32 @llvm.amdgcn.permlane64(i32 %tidx)
+  %v = call i32 @llvm.amdgcn.permlane64.i32(i32 %tidx)
   store i32 %v, ptr addrspace(1) %out
   ret void
 }
+
+define amdgpu_kernel void @test_v_f32(ptr addrspace(1) %out, float %src0) #1 {
+; GFX11-SDAG-LABEL: test_v_f32:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v0, v0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: test_v_f32:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_permlane64_b32 v0, v0
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v1, 0
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_f32 = bitcast i32 %tidx to float
+  %v = call float @llvm.amdgcn.permlane64.f32(float %tidx_f32)
+  store float %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @test_v_i64(ptr addrspace(1) %out, i64 %src0) #1 {
+; GFX11-LABEL: test_v_i64:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-NEXT:    v_permlane64_b32 v0, v0
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2)
+; GFX11-NEXT:    v_permlane64_b32 v1, v2
+; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT:    s_nop 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_i64 = zext i32 %tidx to i64
+  %v = call i64 @llvm.amdgcn.permlane64.i64(i64 %tidx_i64)
+  store i64 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @test_v_f64(ptr addrspace(1) %out, double %src0) #1 {
+; GFX11-SDAG-LABEL: test_v_f64:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v1, v1
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v0, v0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+;
+; GFX11-GISEL-LABEL: test_v_f64:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    v_cvt_f64_f32_e32 v[0:1], v0
+; GFX11-GISEL-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-GISEL-NEXT:    v_mov_b32_e32 v2, 0
+; GFX11-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-GISEL-NEXT:    v_permlane64_b32 v0, v0
+; GFX11-GISEL-NEXT:    v_permlane64_b32 v1, v1
+; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-GISEL-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-GISEL-NEXT:    s_nop 0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-GISEL-NEXT:    s_endpgm
+  %tidx = call i32 @llvm.amdgcn.workitem.id.x()
+  %tidx_f32 = bitcast i32 %tidx to float
+  %tidx_f64 = fpext float %tidx_f32 to double
+  %v = call double @llvm.amdgcn.permlane64.f64(double %tidx_f64)
+  store double %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @test_half(ptr addrspace(1) %out, half %src0) {
+; GFX11-LABEL: test_half:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    v_permlane64_b32 v2, v2
+; GFX11-NEXT:    global_store_b16 v[0:1], v2, off
+; GFX11-NEXT:    s_setpc_b64 s[30:31]
+  %v = call half @llvm.amdgcn.permlane64.f16(half %src0)
+  store half %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @test_bfloat(ptr addrspace(1) %out, bfloat %src0) {
+; GFX11-LABEL: test_bfloat:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    v_permlane64_b32 v2, v2
+; GFX11-NEXT:    global_store_b16 v[0:1], v2, off
+; GFX11-NEXT:    s_setpc_b64 s[30:31]
+  %v = call bfloat @llvm.amdgcn.permlane64.bf16(bfloat %src0)
+  store bfloat %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @test_i16(ptr addrspace(1) %out, i16 %src0) {
+; GFX11-LABEL: test_i16:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    v_permlane64_b32 v2, v2
+; GFX11-NEXT:    global_store_b16 v[0:1], v2, off
+; GFX11-NEXT:    s_setpc_b64 s[30:31]
+  %v = call i16 @llvm.amdgcn.permlane64.i16(i16 %src0)
+  store i16 %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @test_v2f16(ptr addrspace(1) %out, <2 x half> %src0) {
+; GFX11-LABEL: test_v2f16:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-NEXT:    v_permlane64_b32 v2, v2
+; GFX11-NEXT:    global_store_b32 v[0:1], v2, off
+; GFX11-NEXT:    s_setpc_b64 s[30:31]
+  %v = call <2 x half> @llvm.amdgcn.permlane64.v2f16(<2 x half> %src0)
+  store <2 x half> %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @test_v2f32(ptr addrspace(1) %out, <2 x float> %src0) {
+; GFX11-SDAG-LABEL: test_v2f32:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v3, v3
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v2, v2
+; GFX11-SDAG-NEXT:    global_store_b64 v[0:1], v[2:3], off
+; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-GISEL-LABEL: test_v2f32:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_permlane64_b32 v2, v2
+; GFX11-GISEL-NEXT:    v_permlane64_b32 v3, v3
+; GFX11-GISEL-NEXT:    global_store_b64 v[0:1], v[2:3], off
+; GFX11-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %v = call <2 x float> @llvm.amdgcn.permlane64.v2f32(<2 x float> %src0)
+  store <2 x float> %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @test_v7i32(ptr addrspace(1) %out, <7 x i32> %src0) {
+; GFX11-SDAG-LABEL: test_v7i32:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v8, v8
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v7, v7
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v6, v6
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v5, v5
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v4, v4
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v3, v3
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v2, v2
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    global_store_b96 v[0:1], v[6:8], off offset:16
+; GFX11-SDAG-NEXT:    global_store_b128 v[0:1], v[2:5], off
+; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-GISEL-LABEL: test_v7i32:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_permlane64_b32 v2, v2
+; GFX11-GISEL-NEXT:    v_permlane64_b32 v3, v3
+; GFX11-GISEL-NEXT:    v_permlane64_b32 v4, v4
+; GFX11-GISEL-NEXT:    v_permlane64_b32 v5, v5
+; GFX11-GISEL-NEXT:    v_permlane64_b32 v6, v6
+; GFX11-GISEL-NEXT:    v_permlane64_b32 v7, v7
+; GFX11-GISEL-NEXT:    v_permlane64_b32 v8, v8
+; GFX11-GISEL-NEXT:    s_clause 0x1
+; GFX11-GISEL-NEXT:    global_store_b128 v[0:1], v[2:5], off
+; GFX11-GISEL-NEXT:    global_store_b96 v[0:1], v[6:8], off offset:16
+; GFX11-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %v = call <7 x i32> @llvm.amdgcn.permlane64.v7i32(<7 x i32> %src0)
+  store <7 x i32> %v, ptr addrspace(1) %out
+  ret void
+}
+
+define void @test_v8i16(ptr addrspace(1) %out, <8 x i16> %src0) {
+; GFX11-SDAG-LABEL: test_v8i16:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v5, v5
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v4, v4
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v3, v3
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v2, v2
+; GFX11-SDAG-NEXT:    global_store_b128 v[0:1], v[2:5], off
+; GFX11-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX11-GISEL-LABEL: test_v8i16:
+; GFX11-GISEL:       ; %bb.0:
+; GFX11-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX11-GISEL-NEXT:    v_permlane64_b32 v2, v2
+; GFX11-GISEL-NEXT:    v_permlane64_b32 v3, v3
+; GFX11-GISEL-NEXT:    v_permlane64_b32 v4, v4
+; GFX11-GISEL-NEXT:    v_permlane64_b32 v5, v5
+; GFX11-GISEL-NEXT:    global_store_b128 v[0:1], v[2:5], off
+; GFX11-GISEL-NEXT:    s_setpc_b64 s[30:31]
+  %v = call <8 x i16> @llvm.amdgcn.permlane64.v8i16(<8 x i16> %src0)
+  store <8 x i16> %v, ptr addrspace(1) %out
+  ret void
+}
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ptr.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ptr.ll
new file mode 100644
index 0000000000000..2070a832e0fcd
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ptr.ll
@@ -0,0 +1,180 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
+; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX11-SDAG %s
+
+define amdgpu_kernel void @test_p0(ptr addrspace(1) %out, ptr %src0) {
+; GFX11-SDAG-LABEL: test_p0:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_load_b128 s[0:3], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v3, 0 :: v_dual_mov_b32 v0, s3
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v2, s2
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v1, v0
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v0, v2
+; GFX11-SDAG-NEXT:    global_store_b64 v3, v[0:1], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+  %v = call ptr @llvm.amdgcn.permlane64.p0(ptr %src0)
+  store ptr %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @test_v3p0(ptr addrspace(1) %out, <3 x ptr> %src0) {
+; GFX11-SDAG-LABEL: test_v3p0:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x2
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x44
+; GFX11-SDAG-NEXT:    s_load_b64 s[2:3], s[0:1], 0x54
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v6, 0 :: v_dual_mov_b32 v1, s6
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v4, s5 :: v_dual_mov_b32 v5, s3
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v8, s2
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v0, s7 :: v_dual_mov_b32 v7, s4
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4)
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v2, v1
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v1, v4
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v5, v5
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v4, v8
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v3, v0
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v0, v7
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    global_store_b64 v6, v[4:5], s[0:1] offset:16
+; GFX11-SDAG-NEXT:    global_store_b128 v6, v[0:3], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+  %v = call <3 x ptr> @llvm.amdgcn.permlane64.v3p0(<3 x ptr> %src0)
+  store <3 x ptr> %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @test_p3(ptr addrspace(1) %out, ptr addrspace(3) %src0) {
+; GFX11-SDAG-LABEL: test_p3:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b32 s2, s[0:1], 0x2c
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v0, v0
+; GFX11-SDAG-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+  %v = call ptr addrspace(3) @llvm.amdgcn.permlane64.v3p0(ptr addrspace(3) %src0)
+  store ptr addrspace(3) %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @test_v3p3(ptr addrspace(1) %out, <3 x ptr addrspace(3)> %src0) {
+; GFX11-SDAG-LABEL: test_v3p3:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x34
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v4, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s5
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v3, s4
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v2, v0
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v1, v1
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_3)
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v0, v3
+; GFX11-SDAG-NEXT:    global_store_b96 v4, v[0:2], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+  %v = call <3 x ptr addrspace(3)> @llvm.amdgcn.permlane64.v3p3(<3 x ptr addrspace(3)> %src0)
+  store <3 x ptr addrspace(3)> %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @test_p5(ptr addrspace(1) %out, ptr addrspace(5) %src0) {
+; GFX11-SDAG-LABEL: test_p5:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b32 s2, s[0:1], 0x2c
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v0, v0
+; GFX11-SDAG-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+  %v = call ptr addrspace(5) @llvm.amdgcn.permlane64.p5(ptr addrspace(5) %src0)
+  store ptr addrspace(5) %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @test_v3p5(ptr addrspace(1) %out, <3 x ptr addrspace(5)> %src0) {
+; GFX11-SDAG-LABEL: test_v3p5:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x34
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v4, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s5
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v3, s4
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v2, v0
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v1, v1
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_3)
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v0, v3
+; GFX11-SDAG-NEXT:    global_store_b96 v4, v[0:2], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+  %v = call <3 x ptr addrspace(5)> @llvm.amdgcn.permlane64.v3p5(<3 x ptr addrspace(5)> %src0)
+  store <3 x ptr addrspace(5)> %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @test_p6(ptr addrspace(1) %out, ptr addrspace(6) %src0) {
+; GFX11-SDAG-LABEL: test_p6:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b32 s2, s[0:1], 0x2c
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v0, v0
+; GFX11-SDAG-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+  %v = call ptr addrspace(6) @llvm.amdgcn.permlane64.p6(ptr addrspace(6) %src0)
+  store ptr addrspace(6) %v, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_kernel void @test_v3p6(ptr addrspace(1) %out, <3 x ptr addrspace(6)> %src0) {
+; GFX11-SDAG-LABEL: test_v3p6:
+; GFX11-SDAG:       ; %bb.0:
+; GFX11-SDAG-NEXT:    s_clause 0x1
+; GFX11-SDAG-NEXT:    s_load_b128 s[4:7], s[0:1], 0x34
+; GFX11-SDAG-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v4, 0
+; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-SDAG-NEXT:    v_dual_mov_b32 v0, s6 :: v_dual_mov_b32 v1, s5
+; GFX11-SDAG-NEXT:    v_mov_b32_e32 v3, s4
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3)
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v2, v0
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v1, v1
+; GFX11-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_3)
+; GFX11-SDAG-NEXT:    v_permlane64_b32 v0, v3
+; GFX11-SDAG-NEXT:    global_store_b96 v4, v[0:2], s[0:1]
+; GFX11-SDAG-NEXT:    s_nop 0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-SDAG-NEXT:    s_endpgm
+  %v = call <3 x ptr addrspace(6)> @llvm.amdgcn.permlane64.v3p6(<3 x ptr addrspace(6)> %src0)
+  store <3 x ptr addrspace(6)> %v, ptr addrspace(1) %out
+  ret void
+}

>From 5a4c4c4dfbcbe85746947a6cfefa8dd48d793bd2 Mon Sep 17 00:00:00 2001
From: Vikram <Vikram.Hegde at amd.com>
Date: Fri, 31 May 2024 08:11:42 +0000
Subject: [PATCH 6/9] Take over recent changes from original patch

---
 llvm/docs/AMDGPUUsage.rst                     |  13 +-
 .../lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp | 186 +++---------
 llvm/lib/Target/AMDGPU/SIISelLowering.cpp     | 109 +++++--
 .../AMDGPU/llvm.amdgcn.readfirstlane.ll       |  58 ----
 .../AMDGPU/llvm.amdgcn.readfirstlane.ptr.ll   |  34 +++
 .../CodeGen/AMDGPU/llvm.amdgcn.readlane.ll    |  66 -----
 .../AMDGPU/llvm.amdgcn.readlane.ptr.ll        |  38 +++
 .../CodeGen/AMDGPU/llvm.amdgcn.writelane.ll   | 277 ------------------
 .../AMDGPU/llvm.amdgcn.writelane.ptr.ll       | 144 +++++++++
 9 files changed, 347 insertions(+), 578 deletions(-)

diff --git a/llvm/docs/AMDGPUUsage.rst b/llvm/docs/AMDGPUUsage.rst
index 97e2aecd614ed..2d24d9e26dbed 100644
--- a/llvm/docs/AMDGPUUsage.rst
+++ b/llvm/docs/AMDGPUUsage.rst
@@ -1194,20 +1194,21 @@ The AMDGPU backend implements the following LLVM IR intrinsics.
                                                    operation within a row (16 contiguous lanes) of the second input operand.
                                                    The third and fourth inputs must be scalar values. these are combined into
                                                    a single 64-bit value representing lane selects used to swizzle within each
-                                                   row. Currently implemented for i16, i32, float, half, bf16, v2i16, v2f16 and
-                                                   types whose sizes are multiples of 32-bit.
+                                                   row. Currently implemented for i16, i32, float, half, bfloat, <2 x i16>, 
+                                                   <2 x half>, <2 x bfloat>, i64, double, pointers, multiples of the 32-bit vectors.
 
   llvm.amdgcn.permlanex16                          Provides direct access to v_permlanex16_b32. Performs arbitrary gather-style
                                                    operation across two rows of the second input operand (each row is 16 contiguous
                                                    lanes). The third and fourth inputs must be scalar values. these are combined
                                                    into a single 64-bit value representing lane selects used to swizzle within each
-                                                   row. Currently implemented for i16, i32, float, half, bf16, v2i16, v2f16 and types
-                                                   whose sizes are multiples of 32-bit.
+                                                   row. Currently implemented for i16, i32, float, half, bfloat, <2 x i16>, <2 x half>, 
+                                                   <2 x bfloat>, i64, double, pointers, multiples of the 32-bit vectors.
 
   llvm.amdgcn.permlane64                           Provides direct access to v_permlane64_b32. Performs a specific permutation across
                                                    lanes of the input operand where the high half and low half of a wave64 are swapped.
-                                                   Performs no operation in wave32 mode. Currently implemented for i16, i32, float,
-                                                   half, bf16, v2i16, v2f16 and types whose sizes are multiples of 32-bit.
+                                                   Performs no operation in wave32 mode. Currently implemented for i16, i32, float, half, 
+                                                   bfloat, <2 x i16>, <2 x half>, <2 x bfloat>, i64, double, pointers, multiples of the 
+                                                   32-bit vectors.
 
   llvm.amdgcn.udot2                                Provides direct access to v_dot2_u32_u16 across targets which
                                                    support such instructions. This performs unsigned dot product
diff --git a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
index b28c3521d6336..aa0ea06c6092a 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
@@ -5394,15 +5394,12 @@ bool AMDGPULegalizerInfo::legalizeLaneOp(LegalizerHelper &Helper,
   MachineIRBuilder &B = Helper.MIRBuilder;
   MachineRegisterInfo &MRI = *B.getMRI();
 
-  Register DstReg = MI.getOperand(0).getReg();
-  Register Src0 = MI.getOperand(2).getReg();
-
   bool IsPermLane16 = IID == Intrinsic::amdgcn_permlane16 ||
                       IID == Intrinsic::amdgcn_permlanex16;
 
-  auto createLaneOp = [&](Register Src0, Register Src1,
-                          Register Src2) -> Register {
-    auto LaneOp = B.buildIntrinsic(IID, {S32}).addUse(Src0);
+  auto createLaneOp = [&IID, &B, &MI](Register Src0, Register Src1,
+                                      Register Src2, LLT VT) -> Register {
+    auto LaneOp = B.buildIntrinsic(IID, {VT}).addUse(Src0);
     switch (IID) {
     case Intrinsic::amdgcn_readfirstlane:
     case Intrinsic::amdgcn_permlane64:
@@ -5428,6 +5425,8 @@ bool AMDGPULegalizerInfo::legalizeLaneOp(LegalizerHelper &Helper,
     }
   };
 
+  Register DstReg = MI.getOperand(0).getReg();
+  Register Src0 = MI.getOperand(2).getReg();
   Register Src1, Src2;
   if (IID == Intrinsic::amdgcn_readlane || IID == Intrinsic::amdgcn_writelane ||
       IsPermLane16) {
@@ -5446,156 +5445,65 @@ bool AMDGPULegalizerInfo::legalizeLaneOp(LegalizerHelper &Helper,
   }
 
   if (Size < 32) {
-    Register Src0Cast = MRI.getType(Src0).isScalar()
-                            ? Src0
-                            : B.buildBitcast(LLT::scalar(Size), Src0).getReg(0);
-    Src0 = B.buildAnyExt(S32, Src0Cast).getReg(0);
-
-    if (IsPermLane16) {
-      Register Src1Cast =
-          MRI.getType(Src1).isScalar()
-              ? Src1
-              : B.buildBitcast(LLT::scalar(Size), Src2).getReg(0);
-      Src1 = B.buildAnyExt(LLT::scalar(32), Src1Cast).getReg(0);
-    }
+    Src0 = B.buildAnyExt(S32, Src0).getReg(0);
 
-    if (IID == Intrinsic::amdgcn_writelane) {
-      Register Src2Cast =
-          MRI.getType(Src2).isScalar()
-              ? Src2
-              : B.buildBitcast(LLT::scalar(Size), Src2).getReg(0);
-      Src2 = B.buildAnyExt(LLT::scalar(32), Src2Cast).getReg(0);
-    }
+    if (IsPermLane16)
+      Src1 = B.buildAnyExt(LLT::scalar(32), Src1).getReg(0);
 
-    Register LaneOpDst = createLaneOp(Src0, Src1, Src2);
-    if (Ty.isScalar())
-      B.buildTrunc(DstReg, LaneOpDst);
-    else {
-      auto Trunc = B.buildTrunc(LLT::scalar(Size), LaneOpDst);
-      B.buildBitcast(DstReg, Trunc);
-    }
+    if (IID == Intrinsic::amdgcn_writelane)
+      Src2 = B.buildAnyExt(LLT::scalar(32), Src2).getReg(0);
 
+    Register LaneOpDst = createLaneOp(Src0, Src1, Src2, S32);
+    B.buildTrunc(DstReg, LaneOpDst);
     MI.eraseFromParent();
     return true;
   }
 
-  if ((Size % 32) == 0) {
-    SmallVector<Register, 2> PartialRes;
-    unsigned NumParts = Size / 32;
-    auto IsS16Vec = Ty.isVector() && Ty.getElementType() == S16;
-    MachineInstrBuilder Src0Parts;
-
-    if (Ty.isPointer()) {
-      auto PtrToInt = B.buildPtrToInt(LLT::scalar(Size), Src0);
-      Src0Parts = B.buildUnmerge(S32, PtrToInt);
-    } else if (Ty.isPointerVector()) {
-      LLT IntVecTy = Ty.changeElementType(
-          LLT::scalar(Ty.getElementType().getSizeInBits()));
-      auto PtrToInt = B.buildPtrToInt(IntVecTy, Src0);
-      Src0Parts = B.buildUnmerge(S32, PtrToInt);
-    } else
-      Src0Parts =
-          IsS16Vec ? B.buildUnmerge(V2S16, Src0) : B.buildUnmerge(S32, Src0);
+  if (Size % 32 != 0)
+    return false;
 
-    switch (IID) {
-    case Intrinsic::amdgcn_readlane: {
-      Register Src1 = MI.getOperand(3).getReg();
-      for (unsigned i = 0; i < NumParts; ++i) {
-        Src0 = IsS16Vec ? B.buildBitcast(S32, Src0Parts.getReg(i)).getReg(0)
-                        : Src0Parts.getReg(i);
-        PartialRes.push_back(
-            (B.buildIntrinsic(Intrinsic::amdgcn_readlane, {S32})
-                 .addUse(Src0)
-                 .addUse(Src1))
-                .getReg(0));
-      }
+  LLT PartialResTy = S32;
+  if (Ty.isVector()) {
+    LLT EltTy = Ty.getElementType();
+    switch (EltTy.getSizeInBits()) {
+    case 16:
+      PartialResTy = Ty.changeElementCount(ElementCount::getFixed(2));
       break;
-    }
-    case Intrinsic::amdgcn_readfirstlane:
-    case Intrinsic::amdgcn_permlane64: {
-      for (unsigned i = 0; i < NumParts; ++i) {
-        Src0 = IsS16Vec ? B.buildBitcast(S32, Src0Parts.getReg(i)).getReg(0)
-                        : Src0Parts.getReg(i);
-        PartialRes.push_back(
-            (B.buildIntrinsic(IID, {S32}).addUse(Src0).getReg(0)));
-      }
-
+    case 32:
+      PartialResTy = EltTy;
       break;
-    }
-    case Intrinsic::amdgcn_writelane:
-    case Intrinsic::amdgcn_permlane16:
-    case Intrinsic::amdgcn_permlanex16: {
-      Register Src1 = MI.getOperand(3).getReg();
-      Register Src2 = MI.getOperand(4).getReg();
-
-      Register SrcX = IsPermLane16 ? Src1 : Src2;
-      MachineInstrBuilder SrcXParts;
-
-      if (Ty.isPointer()) {
-        auto PtrToInt = B.buildPtrToInt(S64, SrcX);
-        SrcXParts = B.buildUnmerge(S32, PtrToInt);
-      } else if (Ty.isPointerVector()) {
-        LLT IntVecTy = Ty.changeElementType(
-            LLT::scalar(Ty.getElementType().getSizeInBits()));
-        auto PtrToInt = B.buildPtrToInt(IntVecTy, SrcX);
-        SrcXParts = B.buildUnmerge(S32, PtrToInt);
-      } else
-        SrcXParts =
-            IsS16Vec ? B.buildUnmerge(V2S16, SrcX) : B.buildUnmerge(S32, SrcX);
-
-      for (unsigned i = 0; i < NumParts; ++i) {
-        Src0 = IsS16Vec ? B.buildBitcast(S32, Src0Parts.getReg(i)).getReg(0)
-                        : Src0Parts.getReg(i);
-        SrcX = IsS16Vec ? B.buildBitcast(S32, SrcXParts.getReg(i)).getReg(0)
-                        : SrcXParts.getReg(i);
-        PartialRes.push_back(IsPermLane16 ? createLaneOp(Src0, SrcX, Src2)
-                                          : createLaneOp(Src0, Src1, SrcX));
-      }
-
+    default:
+      // Handle all other cases via S32 pieces;
       break;
     }
-    }
+  }
 
-    if (Ty.isPointerVector()) {
-      unsigned PtrSize = Ty.getElementType().getSizeInBits();
-      SmallVector<Register, 2> PtrElements;
-      if (PtrSize == 32) {
-        // Handle 32 bit pointers
-        for (unsigned i = 0; i < NumParts; i++)
-          PtrElements.push_back(
-              B.buildIntToPtr(Ty.getElementType(), PartialRes[i]).getReg(0));
-      } else {
-        // Handle legalization of <? x [pointer type bigger than 32 bits]>
-        SmallVector<Register, 2> PtrParts;
-        unsigned NumS32Parts = PtrSize / 32;
-        unsigned PartIdx = 0;
-        for (unsigned i = 0, j = 1; i < NumParts; i += NumS32Parts, j++) {
-          // Merge S32 components of a pointer element first.
-          for (; PartIdx < (j * NumS32Parts); PartIdx++)
-            PtrParts.push_back(PartialRes[PartIdx]);
-
-          auto MergedPtr =
-              B.buildMergeLikeInstr(LLT::scalar(PtrSize), PtrParts);
-          PtrElements.push_back(
-              B.buildIntToPtr(Ty.getElementType(), MergedPtr).getReg(0));
-          PtrParts.clear();
-        }
-      }
+  SmallVector<Register, 2> PartialRes;
+  unsigned NumParts = Size / 32;
+  MachineInstrBuilder Src0Parts = B.buildUnmerge(PartialResTy, Src0);
+  MachineInstrBuilder Src1Parts, Src2Parts;
 
-      B.buildMergeLikeInstr(DstReg, PtrElements);
-    } else {
-      if (IsS16Vec) {
-        for (unsigned i = 0; i < NumParts; i++)
-          PartialRes[i] = B.buildBitcast(V2S16, PartialRes[i]).getReg(0);
-      }
-      B.buildMergeLikeInstr(DstReg, PartialRes);
-    }
+  if (IsPermLane16)
+    Src1Parts = B.buildUnmerge(PartialResTy, Src1);
 
-    MI.eraseFromParent();
-    return true;
+  if (IID == Intrinsic::amdgcn_writelane)
+    Src2Parts = B.buildUnmerge(PartialResTy, Src2);
+
+  for (unsigned i = 0; i < NumParts; ++i) {
+    Src0 = Src0Parts.getReg(i);
+
+    if (IsPermLane16)
+      Src1 = Src1Parts.getReg(i);
+
+    if (IID == Intrinsic::amdgcn_writelane)
+      Src2 = Src2Parts.getReg(i);
+
+    PartialRes.push_back(createLaneOp(Src0, Src1, Src2, PartialResTy));
   }
 
-  return false;
+  B.buildMergeLikeInstr(DstReg, PartialRes);
+  MI.eraseFromParent();
+  return true;
 }
 
 bool AMDGPULegalizerInfo::getImplicitArgPtr(Register DstReg,
diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index 5d34ed089f65d..acd290cc36441 100644
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -6093,30 +6093,36 @@ static SDValue lowerLaneOp(const SITargetLowering &TLI, SDNode *N,
   unsigned IntrinsicID = N->getConstantOperandVal(0);
   bool IsPermLane16 = IntrinsicID == Intrinsic::amdgcn_permlane16 ||
                       IntrinsicID == Intrinsic::amdgcn_permlanex16;
-  bool IsPermLane64 = IntrinsicID == Intrinsic::amdgcn_permlane64;
   SDValue Src0 = N->getOperand(1);
   SDLoc SL(N);
   MVT IntVT = MVT::getIntegerVT(ValSize);
 
-  auto createLaneOp = [&](SDValue Src0, SDValue Src1, SDValue Src2,
-                          MVT ValueT) -> SDValue {
-    if (IsPermLane16 || IsPermLane64) {
-      if (IsPermLane16) {
-        SDValue Src3 = N->getOperand(4);
-        SDValue Src4 = N->getOperand(5);
-        SDValue Src5 = N->getOperand(6);
-        return DAG.getNode(IntrinsicID == Intrinsic::amdgcn_permlane16
-                               ? AMDGPUISD::PERMLANE16
-                               : AMDGPUISD::PERMLANEX16,
-                           SL, ValueT, {Src0, Src1, Src2, Src3, Src4, Src5});
-      }
-      return DAG.getNode(AMDGPUISD::PERMLANE64, SL, ValueT, {Src0});
+  auto createLaneOp = [&DAG, &SL, N](SDValue Src0, SDValue Src1, SDValue Src2,
+                                     MVT ValueT) -> SDValue {
+    switch (unsigned IID = N->getConstantOperandVal(0)) {
+    case Intrinsic::amdgcn_readfirstlane:
+    case Intrinsic::amdgcn_permlane64:
+      return DAG.getNode(IID == Intrinsic::amdgcn_readfirstlane
+                             ? AMDGPUISD::READFIRSTLANE
+                             : AMDGPUISD::PERMLANE64,
+                         SL, ValueT, {Src0});
+    case Intrinsic::amdgcn_readlane:
+      return DAG.getNode(AMDGPUISD::READLANE, SL, ValueT, {Src0, Src1});
+    case Intrinsic::amdgcn_writelane:
+      return DAG.getNode(AMDGPUISD::WRITELANE, SL, ValueT, {Src0, Src1, Src2});
+    case Intrinsic::amdgcn_permlane16:
+    case Intrinsic::amdgcn_permlanex16: {
+      SDValue Src3 = N->getOperand(4);
+      SDValue Src4 = N->getOperand(5);
+      SDValue Src5 = N->getOperand(6);
+      return DAG.getNode(IID == Intrinsic::amdgcn_permlane16
+                             ? AMDGPUISD::PERMLANE16
+                             : AMDGPUISD::PERMLANEX16,
+                         SL, ValueT, {Src0, Src1, Src2, Src3, Src4, Src5});
+    }
+    default:
+      llvm_unreachable("unhandled lane op");
     }
-
-    return (
-        Src2 ? DAG.getNode(AMDGPUISD::WRITELANE, SL, ValueT, {Src0, Src1, Src2})
-        : Src1 ? DAG.getNode(AMDGPUISD::READLANE, SL, ValueT, {Src0, Src1})
-               : DAG.getNode(AMDGPUISD::READFIRSTLANE, SL, ValueT, {Src0}));
   };
 
   SDValue Src1, Src2;
@@ -6133,31 +6139,73 @@ static SDValue lowerLaneOp(const SITargetLowering &TLI, SDNode *N,
   }
 
   if (ValSize < 32) {
-    SDValue InitBitCast = DAG.getBitcast(IntVT, Src0);
-    Src0 = DAG.getAnyExtOrTrunc(InitBitCast, SL, MVT::i32);
+    bool IsFloat = VT.isFloatingPoint();
+    Src0 = DAG.getAnyExtOrTrunc(IsFloat ? DAG.getBitcast(IntVT, Src0) : Src0,
+                                SL, MVT::i32);
 
     if (IsPermLane16) {
-      SDValue Src1Cast = DAG.getBitcast(IntVT, Src1);
-      Src1 = DAG.getAnyExtOrTrunc(Src1Cast, SL, MVT::i32);
+      Src1 = DAG.getAnyExtOrTrunc(IsFloat ? DAG.getBitcast(IntVT, Src1) : Src1,
+                                  SL, MVT::i32);
     }
 
     if (IntrinsicID == Intrinsic::amdgcn_writelane) {
-      SDValue Src2Cast = DAG.getBitcast(IntVT, Src2);
-      Src2 = DAG.getAnyExtOrTrunc(Src2Cast, SL, MVT::i32);
+      Src2 = DAG.getAnyExtOrTrunc(IsFloat ? DAG.getBitcast(IntVT, Src2) : Src2,
+                                  SL, MVT::i32);
     }
 
     SDValue LaneOp = createLaneOp(Src0, Src1, Src2, MVT::i32);
     SDValue Trunc = DAG.getAnyExtOrTrunc(LaneOp, SL, IntVT);
-    return DAG.getBitcast(VT, Trunc);
+    return IsFloat ? DAG.getBitcast(VT, Trunc) : Trunc;
+  }
+
+  if (ValSize % 32 != 0)
+    return SDValue();
+
+  if (VT.isVector()) {
+    switch (MVT::SimpleValueType EltTy =
+                VT.getVectorElementType().getSimpleVT().SimpleTy) {
+    case MVT::i32:
+    case MVT::f32: {
+      SDValue LaneOp = createLaneOp(Src0, Src1, Src2, VT.getSimpleVT());
+      return DAG.UnrollVectorOp(LaneOp.getNode());
+    }
+    case MVT::i16:
+    case MVT::f16:
+    case MVT::bf16: {
+      MVT SubVecVT = MVT::getVectorVT(EltTy, 2);
+      SmallVector<SDValue, 4> Pieces;
+      SDValue Src0SubVec, Src1SubVec, Src2SubVec;
+      for (unsigned i = 0, EltIdx = 0; i < ValSize / 32; i++) {
+        Src0SubVec = DAG.getNode(ISD::EXTRACT_SUBVECTOR, SL, SubVecVT, Src0,
+                                 DAG.getConstant(EltIdx, SL, MVT::i32));
+
+        if (IsPermLane16)
+          Src1SubVec = DAG.getNode(ISD::EXTRACT_SUBVECTOR, SL, SubVecVT, Src1,
+                                   DAG.getConstant(EltIdx, SL, MVT::i32));
+
+        if (IntrinsicID == Intrinsic::amdgcn_writelane)
+          Src2SubVec = DAG.getNode(ISD::EXTRACT_SUBVECTOR, SL, SubVecVT, Src2,
+                                   DAG.getConstant(EltIdx, SL, MVT::i32));
+
+        Pieces.push_back(
+            IsPermLane16
+                ? createLaneOp(Src0SubVec, Src1SubVec, Src2, SubVecVT)
+                : createLaneOp(Src0SubVec, Src1, Src2SubVec, SubVecVT));
+        EltIdx += 2;
+      }
+      return DAG.getNode(ISD::CONCAT_VECTORS, SL, VT, Pieces);
+    }
+    default:
+      // Handle all other cases by bitcasting to i32 vectors
+      break;
+    }
   }
 
-  if ((ValSize % 32) == 0) {
     MVT VecVT = MVT::getVectorVT(MVT::i32, ValSize / 32);
     Src0 = DAG.getBitcast(VecVT, Src0);
 
-    if (IsPermLane16) {
+    if (IsPermLane16)
       Src1 = DAG.getBitcast(VecVT, Src1);
-    }
 
     if (IntrinsicID == Intrinsic::amdgcn_writelane)
       Src2 = DAG.getBitcast(VecVT, Src2);
@@ -6165,9 +6213,6 @@ static SDValue lowerLaneOp(const SITargetLowering &TLI, SDNode *N,
     SDValue LaneOp = createLaneOp(Src0, Src1, Src2, VecVT);
     SDValue UnrolledLaneOp = DAG.UnrollVectorOp(LaneOp.getNode());
     return DAG.getBitcast(VT, UnrolledLaneOp);
-  }
-
-  return SDValue();
 }
 
 void SITargetLowering::ReplaceNodeResults(SDNode *N,
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll
index 732489f22c36f..ed0da0d2a61a2 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll
@@ -566,64 +566,6 @@ define void @test_readfirstlane_v7i32(ptr addrspace(1) %out, <7 x i32> %src) {
   ret void
 }
 
-define void @test_readfirstlane_p0(ptr addrspace(1) %out, ptr %src) {
-; CHECK-SDAG-LABEL: test_readfirstlane_p0:
-; CHECK-SDAG:       ; %bb.0:
-; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s5, v3
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
-; CHECK-SDAG-NEXT:    ;;#ASMSTART
-; CHECK-SDAG-NEXT:    ; use s[4:5]
-; CHECK-SDAG-NEXT:    ;;#ASMEND
-; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
-;
-; CHECK-GISEL-LABEL: test_readfirstlane_p0:
-; CHECK-GISEL:       ; %bb.0:
-; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
-; CHECK-GISEL-NEXT:    ;;#ASMSTART
-; CHECK-GISEL-NEXT:    ; use s[4:5]
-; CHECK-GISEL-NEXT:    ;;#ASMEND
-; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
-  %x = call ptr @llvm.amdgcn.readfirstlane.p0(ptr %src)
-  call void asm sideeffect "; use $0", "s"(ptr %x)
-  ret void
-}
-
-define void @test_readfirstlane_v3p0(ptr addrspace(1) %out, <3 x ptr> %src) {
-; CHECK-SDAG-LABEL: test_readfirstlane_v3p0:
-; CHECK-SDAG:       ; %bb.0:
-; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s9, v7
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s8, v6
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s7, v5
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s6, v4
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s5, v3
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
-; CHECK-SDAG-NEXT:    ;;#ASMSTART
-; CHECK-SDAG-NEXT:    ; use s[4:9]
-; CHECK-SDAG-NEXT:    ;;#ASMEND
-; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
-;
-; CHECK-GISEL-LABEL: test_readfirstlane_v3p0:
-; CHECK-GISEL:       ; %bb.0:
-; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s5, v3
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s6, v4
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s7, v5
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s8, v6
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s9, v7
-; CHECK-GISEL-NEXT:    ;;#ASMSTART
-; CHECK-GISEL-NEXT:    ; use s[4:9]
-; CHECK-GISEL-NEXT:    ;;#ASMEND
-; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
-  %x = call <3 x ptr> @llvm.amdgcn.readfirstlane.v3p0(<3 x ptr> %src)
-  call void asm sideeffect "; use $0", "s"(<3 x ptr> %x)
-  ret void
-}
-
 define void @test_readfirstlane_v8i16(ptr addrspace(1) %out, <8 x i16> %src) {
 ; CHECK-SDAG-LABEL: test_readfirstlane_v8i16:
 ; CHECK-SDAG:       ; %bb.0:
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ptr.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ptr.ll
index 588f239606f52..3882a5f0f9f4f 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ptr.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ptr.ll
@@ -1,6 +1,40 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
 ; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=fiji -verify-machineinstrs < %s | FileCheck -check-prefix=CHECK-SDAG -enable-var-scope %s
 
+define void @test_readfirstlane_p0(ptr addrspace(1) %out, ptr %src) {
+; CHECK-SDAG-LABEL: test_readfirstlane_p0:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s5, v3
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[4:5]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %x = call ptr @llvm.amdgcn.readfirstlane.p0(ptr %src)
+  call void asm sideeffect "; use $0", "s"(ptr %x)
+  ret void
+}
+
+define void @test_readfirstlane_v3p0(ptr addrspace(1) %out, <3 x ptr> %src) {
+; CHECK-SDAG-LABEL: test_readfirstlane_v3p0:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s9, v7
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s8, v6
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s7, v5
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s6, v4
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s5, v3
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[4:9]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %x = call <3 x ptr> @llvm.amdgcn.readfirstlane.v3p0(<3 x ptr> %src)
+  call void asm sideeffect "; use $0", "s"(<3 x ptr> %x)
+  ret void
+}
+
 define void @test_readfirstlane_p3(ptr addrspace(1) %out, ptr addrspace(3) %src) {
 ; CHECK-SDAG-LABEL: test_readfirstlane_p3:
 ; CHECK-SDAG:       ; %bb.0:
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ll
index 71cd3db81addd..325a39abb588a 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ll
@@ -862,72 +862,6 @@ define void @test_readlane_v7i32(ptr addrspace(1) %out, <7 x i32> %src, i32 %src
   ret void
 }
 
-define void @test_readlane_p0(ptr addrspace(1) %out, ptr %src, i32 %src1) {
-; CHECK-SDAG-LABEL: test_readlane_p0:
-; CHECK-SDAG:       ; %bb.0:
-; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v4
-; CHECK-SDAG-NEXT:    s_nop 3
-; CHECK-SDAG-NEXT:    v_readlane_b32 s5, v3, s4
-; CHECK-SDAG-NEXT:    v_readlane_b32 s4, v2, s4
-; CHECK-SDAG-NEXT:    ;;#ASMSTART
-; CHECK-SDAG-NEXT:    ; use s[4:5]
-; CHECK-SDAG-NEXT:    ;;#ASMEND
-; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
-;
-; CHECK-GISEL-LABEL: test_readlane_p0:
-; CHECK-GISEL:       ; %bb.0:
-; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s5, v4
-; CHECK-GISEL-NEXT:    s_nop 3
-; CHECK-GISEL-NEXT:    v_readlane_b32 s4, v2, s5
-; CHECK-GISEL-NEXT:    v_readlane_b32 s5, v3, s5
-; CHECK-GISEL-NEXT:    ;;#ASMSTART
-; CHECK-GISEL-NEXT:    ; use s[4:5]
-; CHECK-GISEL-NEXT:    ;;#ASMEND
-; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
-  %x = call ptr @llvm.amdgcn.readlane.p0(ptr %src, i32 %src1)
-  call void asm sideeffect "; use $0", "s"(ptr %x)
-  ret void
-}
-
-define void @test_readlane_v3p0(ptr addrspace(1) %out, <3 x ptr> %src, i32 %src1) {
-; CHECK-SDAG-LABEL: test_readlane_v3p0:
-; CHECK-SDAG:       ; %bb.0:
-; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v8
-; CHECK-SDAG-NEXT:    s_nop 3
-; CHECK-SDAG-NEXT:    v_readlane_b32 s9, v7, s4
-; CHECK-SDAG-NEXT:    v_readlane_b32 s8, v6, s4
-; CHECK-SDAG-NEXT:    v_readlane_b32 s7, v5, s4
-; CHECK-SDAG-NEXT:    v_readlane_b32 s6, v4, s4
-; CHECK-SDAG-NEXT:    v_readlane_b32 s5, v3, s4
-; CHECK-SDAG-NEXT:    v_readlane_b32 s4, v2, s4
-; CHECK-SDAG-NEXT:    ;;#ASMSTART
-; CHECK-SDAG-NEXT:    ; use s[4:9]
-; CHECK-SDAG-NEXT:    ;;#ASMEND
-; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
-;
-; CHECK-GISEL-LABEL: test_readlane_v3p0:
-; CHECK-GISEL:       ; %bb.0:
-; CHECK-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; CHECK-GISEL-NEXT:    v_readfirstlane_b32 s9, v8
-; CHECK-GISEL-NEXT:    s_nop 3
-; CHECK-GISEL-NEXT:    v_readlane_b32 s4, v2, s9
-; CHECK-GISEL-NEXT:    v_readlane_b32 s5, v3, s9
-; CHECK-GISEL-NEXT:    v_readlane_b32 s6, v4, s9
-; CHECK-GISEL-NEXT:    v_readlane_b32 s7, v5, s9
-; CHECK-GISEL-NEXT:    v_readlane_b32 s8, v6, s9
-; CHECK-GISEL-NEXT:    v_readlane_b32 s9, v7, s9
-; CHECK-GISEL-NEXT:    ;;#ASMSTART
-; CHECK-GISEL-NEXT:    ; use s[4:9]
-; CHECK-GISEL-NEXT:    ;;#ASMEND
-; CHECK-GISEL-NEXT:    s_setpc_b64 s[30:31]
-  %x = call <3 x ptr> @llvm.amdgcn.readlane.v3p0(<3 x ptr> %src, i32 %src1)
-  call void asm sideeffect "; use $0", "s"(<3 x ptr> %x)
-  ret void
-}
-
 define void @test_readlane_v8i16(ptr addrspace(1) %out, <8 x i16> %src, i32 %src1) {
 ; CHECK-SDAG-LABEL: test_readlane_v8i16:
 ; CHECK-SDAG:       ; %bb.0:
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ptr.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ptr.ll
index 1b4ee84c75250..49f8ef391c230 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ptr.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ptr.ll
@@ -1,6 +1,44 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
 ; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=fiji -verify-machineinstrs < %s | FileCheck --check-prefix=CHECK-SDAG -enable-var-scope %s
 
+define void @test_readlane_p0(ptr addrspace(1) %out, ptr %src, i32 %src1) {
+; CHECK-SDAG-LABEL: test_readlane_p0:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v4
+; CHECK-SDAG-NEXT:    s_nop 3
+; CHECK-SDAG-NEXT:    v_readlane_b32 s5, v3, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s4, v2, s4
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[4:5]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %x = call ptr @llvm.amdgcn.readlane.p0(ptr %src, i32 %src1)
+  call void asm sideeffect "; use $0", "s"(ptr %x)
+  ret void
+}
+
+define void @test_readlane_v3p0(ptr addrspace(1) %out, <3 x ptr> %src, i32 %src1) {
+; CHECK-SDAG-LABEL: test_readlane_v3p0:
+; CHECK-SDAG:       ; %bb.0:
+; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v8
+; CHECK-SDAG-NEXT:    s_nop 3
+; CHECK-SDAG-NEXT:    v_readlane_b32 s9, v7, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s8, v6, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s7, v5, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s6, v4, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s5, v3, s4
+; CHECK-SDAG-NEXT:    v_readlane_b32 s4, v2, s4
+; CHECK-SDAG-NEXT:    ;;#ASMSTART
+; CHECK-SDAG-NEXT:    ; use s[4:9]
+; CHECK-SDAG-NEXT:    ;;#ASMEND
+; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %x = call <3 x ptr> @llvm.amdgcn.readlane.v3p0(<3 x ptr> %src, i32 %src1)
+  call void asm sideeffect "; use $0", "s"(<3 x ptr> %x)
+  ret void
+}
+
 define void @test_readlane_p3(ptr addrspace(1) %out, ptr addrspace(3) %src, i32 %src1) {
 ; CHECK-SDAG-LABEL: test_readlane_p3:
 ; CHECK-SDAG:       ; %bb.0:
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.writelane.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.writelane.ll
index d0a865f565eeb..31f1085dd76ee 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.writelane.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.writelane.ll
@@ -2651,283 +2651,6 @@ define void @test_writelane_v7i32(ptr addrspace(1) %out, <7 x i32> %src, i32 %sr
   ret void
 }
 
-define void @test_writelane_p0(ptr addrspace(1) %out, ptr %src, i32 %src1) {
-; GFX802-SDAG-LABEL: test_writelane_p0:
-; GFX802-SDAG:       ; %bb.0:
-; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX802-SDAG-NEXT:    flat_load_dwordx2 v[5:6], v[0:1]
-; GFX802-SDAG-NEXT:    v_readfirstlane_b32 m0, v4
-; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s4, v3
-; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s5, v2
-; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
-; GFX802-SDAG-NEXT:    s_nop 0
-; GFX802-SDAG-NEXT:    v_writelane_b32 v6, s4, m0
-; GFX802-SDAG-NEXT:    v_writelane_b32 v5, s5, m0
-; GFX802-SDAG-NEXT:    flat_store_dwordx2 v[0:1], v[5:6]
-; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
-; GFX802-SDAG-NEXT:    s_setpc_b64 s[30:31]
-;
-; GFX1010-SDAG-LABEL: test_writelane_p0:
-; GFX1010-SDAG:       ; %bb.0:
-; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1010-SDAG-NEXT:    global_load_dwordx2 v[5:6], v[0:1], off
-; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s4, v3
-; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s5, v4
-; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s6, v2
-; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0)
-; GFX1010-SDAG-NEXT:    v_writelane_b32 v6, s4, s5
-; GFX1010-SDAG-NEXT:    v_writelane_b32 v5, s6, s5
-; GFX1010-SDAG-NEXT:    global_store_dwordx2 v[0:1], v[5:6], off
-; GFX1010-SDAG-NEXT:    s_setpc_b64 s[30:31]
-;
-; GFX1100-SDAG-LABEL: test_writelane_p0:
-; GFX1100-SDAG:       ; %bb.0:
-; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-SDAG-NEXT:    global_load_b64 v[5:6], v[0:1], off
-; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s0, v3
-; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s1, v4
-; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s2, v2
-; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0)
-; GFX1100-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1100-SDAG-NEXT:    v_writelane_b32 v6, s0, s1
-; GFX1100-SDAG-NEXT:    v_writelane_b32 v5, s2, s1
-; GFX1100-SDAG-NEXT:    global_store_b64 v[0:1], v[5:6], off
-; GFX1100-SDAG-NEXT:    s_setpc_b64 s[30:31]
-;
-; GFX802-GISEL-LABEL: test_writelane_p0:
-; GFX802-GISEL:       ; %bb.0:
-; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX802-GISEL-NEXT:    flat_load_dwordx2 v[5:6], v[0:1]
-; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s5, v4
-; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
-; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s6, v3
-; GFX802-GISEL-NEXT:    s_mov_b32 m0, s5
-; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0)
-; GFX802-GISEL-NEXT:    v_writelane_b32 v5, s4, m0
-; GFX802-GISEL-NEXT:    v_writelane_b32 v6, s6, m0
-; GFX802-GISEL-NEXT:    flat_store_dwordx2 v[0:1], v[5:6]
-; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0)
-; GFX802-GISEL-NEXT:    s_setpc_b64 s[30:31]
-;
-; GFX1010-GISEL-LABEL: test_writelane_p0:
-; GFX1010-GISEL:       ; %bb.0:
-; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1010-GISEL-NEXT:    global_load_dwordx2 v[5:6], v[0:1], off
-; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
-; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s5, v4
-; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s6, v3
-; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0)
-; GFX1010-GISEL-NEXT:    v_writelane_b32 v5, s4, s5
-; GFX1010-GISEL-NEXT:    v_writelane_b32 v6, s6, s5
-; GFX1010-GISEL-NEXT:    global_store_dwordx2 v[0:1], v[5:6], off
-; GFX1010-GISEL-NEXT:    s_setpc_b64 s[30:31]
-;
-; GFX1100-GISEL-LABEL: test_writelane_p0:
-; GFX1100-GISEL:       ; %bb.0:
-; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-GISEL-NEXT:    global_load_b64 v[5:6], v[0:1], off
-; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s0, v2
-; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s1, v4
-; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s2, v3
-; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0)
-; GFX1100-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1100-GISEL-NEXT:    v_writelane_b32 v5, s0, s1
-; GFX1100-GISEL-NEXT:    v_writelane_b32 v6, s2, s1
-; GFX1100-GISEL-NEXT:    global_store_b64 v[0:1], v[5:6], off
-; GFX1100-GISEL-NEXT:    s_setpc_b64 s[30:31]
-  %oldval = load ptr, ptr addrspace(1) %out
-  %writelane = call ptr @llvm.amdgcn.writelane.p0(ptr %src, i32 %src1, ptr %oldval)
-  store ptr %writelane, ptr addrspace(1) %out, align 4
-  ret void
-}
-
-define void @test_writelane_v3p0(ptr addrspace(1) %out, <4 x ptr> %src, i32 %src1) {
-; GFX802-SDAG-LABEL: test_writelane_v3p0:
-; GFX802-SDAG:       ; %bb.0:
-; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX802-SDAG-NEXT:    v_add_u32_e32 v19, vcc, 16, v0
-; GFX802-SDAG-NEXT:    flat_load_dwordx4 v[11:14], v[0:1]
-; GFX802-SDAG-NEXT:    v_addc_u32_e32 v20, vcc, 0, v1, vcc
-; GFX802-SDAG-NEXT:    flat_load_dwordx4 v[15:18], v[19:20]
-; GFX802-SDAG-NEXT:    v_readfirstlane_b32 m0, v10
-; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s8, v5
-; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s9, v4
-; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s10, v3
-; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s11, v2
-; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s4, v9
-; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s5, v8
-; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s6, v7
-; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s7, v6
-; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(1)
-; GFX802-SDAG-NEXT:    v_writelane_b32 v14, s8, m0
-; GFX802-SDAG-NEXT:    v_writelane_b32 v13, s9, m0
-; GFX802-SDAG-NEXT:    v_writelane_b32 v12, s10, m0
-; GFX802-SDAG-NEXT:    v_writelane_b32 v11, s11, m0
-; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
-; GFX802-SDAG-NEXT:    v_writelane_b32 v18, s4, m0
-; GFX802-SDAG-NEXT:    v_writelane_b32 v17, s5, m0
-; GFX802-SDAG-NEXT:    v_writelane_b32 v16, s6, m0
-; GFX802-SDAG-NEXT:    v_writelane_b32 v15, s7, m0
-; GFX802-SDAG-NEXT:    flat_store_dwordx4 v[0:1], v[11:14]
-; GFX802-SDAG-NEXT:    flat_store_dwordx4 v[19:20], v[15:18]
-; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
-; GFX802-SDAG-NEXT:    s_setpc_b64 s[30:31]
-;
-; GFX1010-SDAG-LABEL: test_writelane_v3p0:
-; GFX1010-SDAG:       ; %bb.0:
-; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1010-SDAG-NEXT:    s_clause 0x1
-; GFX1010-SDAG-NEXT:    global_load_dwordx4 v[11:14], v[0:1], off offset:16
-; GFX1010-SDAG-NEXT:    global_load_dwordx4 v[15:18], v[0:1], off
-; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s5, v10
-; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s9, v5
-; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s10, v4
-; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s11, v3
-; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s12, v2
-; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s4, v9
-; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s6, v8
-; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s7, v7
-; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s8, v6
-; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(1)
-; GFX1010-SDAG-NEXT:    v_writelane_b32 v14, s4, s5
-; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0)
-; GFX1010-SDAG-NEXT:    v_writelane_b32 v18, s9, s5
-; GFX1010-SDAG-NEXT:    v_writelane_b32 v17, s10, s5
-; GFX1010-SDAG-NEXT:    v_writelane_b32 v16, s11, s5
-; GFX1010-SDAG-NEXT:    v_writelane_b32 v15, s12, s5
-; GFX1010-SDAG-NEXT:    v_writelane_b32 v13, s6, s5
-; GFX1010-SDAG-NEXT:    v_writelane_b32 v12, s7, s5
-; GFX1010-SDAG-NEXT:    v_writelane_b32 v11, s8, s5
-; GFX1010-SDAG-NEXT:    global_store_dwordx4 v[0:1], v[15:18], off
-; GFX1010-SDAG-NEXT:    global_store_dwordx4 v[0:1], v[11:14], off offset:16
-; GFX1010-SDAG-NEXT:    s_setpc_b64 s[30:31]
-;
-; GFX1100-SDAG-LABEL: test_writelane_v3p0:
-; GFX1100-SDAG:       ; %bb.0:
-; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-SDAG-NEXT:    s_clause 0x1
-; GFX1100-SDAG-NEXT:    global_load_b128 v[11:14], v[0:1], off offset:16
-; GFX1100-SDAG-NEXT:    global_load_b128 v[15:18], v[0:1], off
-; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s1, v10
-; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s5, v5
-; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s6, v4
-; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s7, v3
-; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s8, v2
-; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s0, v9
-; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s2, v8
-; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s3, v7
-; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s4, v6
-; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(1)
-; GFX1100-SDAG-NEXT:    v_writelane_b32 v14, s0, s1
-; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0)
-; GFX1100-SDAG-NEXT:    v_writelane_b32 v18, s5, s1
-; GFX1100-SDAG-NEXT:    v_writelane_b32 v17, s6, s1
-; GFX1100-SDAG-NEXT:    v_writelane_b32 v16, s7, s1
-; GFX1100-SDAG-NEXT:    v_writelane_b32 v15, s8, s1
-; GFX1100-SDAG-NEXT:    v_writelane_b32 v13, s2, s1
-; GFX1100-SDAG-NEXT:    v_writelane_b32 v12, s3, s1
-; GFX1100-SDAG-NEXT:    v_writelane_b32 v11, s4, s1
-; GFX1100-SDAG-NEXT:    s_clause 0x1
-; GFX1100-SDAG-NEXT:    global_store_b128 v[0:1], v[15:18], off
-; GFX1100-SDAG-NEXT:    global_store_b128 v[0:1], v[11:14], off offset:16
-; GFX1100-SDAG-NEXT:    s_setpc_b64 s[30:31]
-;
-; GFX802-GISEL-LABEL: test_writelane_v3p0:
-; GFX802-GISEL:       ; %bb.0:
-; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX802-GISEL-NEXT:    v_add_u32_e32 v19, vcc, 16, v0
-; GFX802-GISEL-NEXT:    flat_load_dwordx4 v[11:14], v[0:1]
-; GFX802-GISEL-NEXT:    v_addc_u32_e32 v20, vcc, 0, v1, vcc
-; GFX802-GISEL-NEXT:    flat_load_dwordx4 v[15:18], v[19:20]
-; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s5, v10
-; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
-; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s6, v3
-; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s7, v4
-; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s8, v5
-; GFX802-GISEL-NEXT:    s_mov_b32 m0, s5
-; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s9, v6
-; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s10, v7
-; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s11, v8
-; GFX802-GISEL-NEXT:    v_readfirstlane_b32 s12, v9
-; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(1)
-; GFX802-GISEL-NEXT:    v_writelane_b32 v11, s4, m0
-; GFX802-GISEL-NEXT:    v_writelane_b32 v12, s6, m0
-; GFX802-GISEL-NEXT:    v_writelane_b32 v13, s7, m0
-; GFX802-GISEL-NEXT:    v_writelane_b32 v14, s8, m0
-; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0)
-; GFX802-GISEL-NEXT:    v_writelane_b32 v15, s9, m0
-; GFX802-GISEL-NEXT:    v_writelane_b32 v16, s10, m0
-; GFX802-GISEL-NEXT:    v_writelane_b32 v17, s11, m0
-; GFX802-GISEL-NEXT:    v_writelane_b32 v18, s12, m0
-; GFX802-GISEL-NEXT:    flat_store_dwordx4 v[0:1], v[11:14]
-; GFX802-GISEL-NEXT:    flat_store_dwordx4 v[19:20], v[15:18]
-; GFX802-GISEL-NEXT:    s_waitcnt vmcnt(0)
-; GFX802-GISEL-NEXT:    s_setpc_b64 s[30:31]
-;
-; GFX1010-GISEL-LABEL: test_writelane_v3p0:
-; GFX1010-GISEL:       ; %bb.0:
-; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1010-GISEL-NEXT:    s_clause 0x1
-; GFX1010-GISEL-NEXT:    global_load_dwordx4 v[11:14], v[0:1], off
-; GFX1010-GISEL-NEXT:    global_load_dwordx4 v[15:18], v[0:1], off offset:16
-; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s4, v2
-; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s5, v10
-; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s6, v3
-; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s7, v4
-; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s8, v5
-; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s9, v6
-; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s10, v7
-; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s11, v8
-; GFX1010-GISEL-NEXT:    v_readfirstlane_b32 s12, v9
-; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(1)
-; GFX1010-GISEL-NEXT:    v_writelane_b32 v11, s4, s5
-; GFX1010-GISEL-NEXT:    v_writelane_b32 v12, s6, s5
-; GFX1010-GISEL-NEXT:    v_writelane_b32 v13, s7, s5
-; GFX1010-GISEL-NEXT:    v_writelane_b32 v14, s8, s5
-; GFX1010-GISEL-NEXT:    s_waitcnt vmcnt(0)
-; GFX1010-GISEL-NEXT:    v_writelane_b32 v15, s9, s5
-; GFX1010-GISEL-NEXT:    v_writelane_b32 v16, s10, s5
-; GFX1010-GISEL-NEXT:    v_writelane_b32 v17, s11, s5
-; GFX1010-GISEL-NEXT:    v_writelane_b32 v18, s12, s5
-; GFX1010-GISEL-NEXT:    global_store_dwordx4 v[0:1], v[11:14], off
-; GFX1010-GISEL-NEXT:    global_store_dwordx4 v[0:1], v[15:18], off offset:16
-; GFX1010-GISEL-NEXT:    s_setpc_b64 s[30:31]
-;
-; GFX1100-GISEL-LABEL: test_writelane_v3p0:
-; GFX1100-GISEL:       ; %bb.0:
-; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX1100-GISEL-NEXT:    s_clause 0x1
-; GFX1100-GISEL-NEXT:    global_load_b128 v[11:14], v[0:1], off
-; GFX1100-GISEL-NEXT:    global_load_b128 v[15:18], v[0:1], off offset:16
-; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s0, v2
-; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s1, v10
-; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s2, v3
-; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s3, v4
-; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s4, v5
-; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s5, v6
-; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s6, v7
-; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s7, v8
-; GFX1100-GISEL-NEXT:    v_readfirstlane_b32 s8, v9
-; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(1)
-; GFX1100-GISEL-NEXT:    v_writelane_b32 v11, s0, s1
-; GFX1100-GISEL-NEXT:    v_writelane_b32 v12, s2, s1
-; GFX1100-GISEL-NEXT:    v_writelane_b32 v13, s3, s1
-; GFX1100-GISEL-NEXT:    v_writelane_b32 v14, s4, s1
-; GFX1100-GISEL-NEXT:    s_waitcnt vmcnt(0)
-; GFX1100-GISEL-NEXT:    v_writelane_b32 v15, s5, s1
-; GFX1100-GISEL-NEXT:    v_writelane_b32 v16, s6, s1
-; GFX1100-GISEL-NEXT:    v_writelane_b32 v17, s7, s1
-; GFX1100-GISEL-NEXT:    v_writelane_b32 v18, s8, s1
-; GFX1100-GISEL-NEXT:    s_clause 0x1
-; GFX1100-GISEL-NEXT:    global_store_b128 v[0:1], v[11:14], off
-; GFX1100-GISEL-NEXT:    global_store_b128 v[0:1], v[15:18], off offset:16
-; GFX1100-GISEL-NEXT:    s_setpc_b64 s[30:31]
-  %oldval = load <4 x ptr>, ptr addrspace(1) %out
-  %writelane = call <4 x ptr> @llvm.amdgcn.writelane.v3p0(<4 x ptr> %src, i32 %src1, <4 x ptr> %oldval)
-  store <4 x ptr> %writelane, ptr addrspace(1) %out, align 4
-  ret void
-}
-
 define void @test_writelane_v8i16(ptr addrspace(1) %out, <8 x i16> %src, i32 %src1) {
 ; GFX802-SDAG-LABEL: test_writelane_v8i16:
 ; GFX802-SDAG:       ; %bb.0:
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.writelane.ptr.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.writelane.ptr.ll
index afc394627d356..c45c73dd88150 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.writelane.ptr.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.writelane.ptr.ll
@@ -3,6 +3,150 @@
 ; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX1010-SDAG %s
 ; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=gfx1100 -verify-machineinstrs -amdgpu-enable-vopd=0 < %s | FileCheck -check-prefixes=GFX1100-SDAG %s
 
+define void @test_writelane_p0(ptr addrspace(1) %out, ptr %src, i32 %src1) {
+; GFX802-SDAG-LABEL: test_writelane_p0:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-SDAG-NEXT:    flat_load_dwordx2 v[5:6], v[0:1]
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 m0, v4
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s4, v3
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s5, v2
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_nop 0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v6, s4, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v5, s5, m0
+; GFX802-SDAG-NEXT:    flat_store_dwordx2 v[0:1], v[5:6]
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-SDAG-LABEL: test_writelane_p0:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    global_load_dwordx2 v[5:6], v[0:1], off
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s4, v3
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s5, v4
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s6, v2
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v6, s4, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v5, s6, s5
+; GFX1010-SDAG-NEXT:    global_store_dwordx2 v[0:1], v[5:6], off
+; GFX1010-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-SDAG-LABEL: test_writelane_p0:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    global_load_b64 v[5:6], v[0:1], off
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s0, v3
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s1, v4
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s2, v2
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v6, s0, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v5, s2, s1
+; GFX1100-SDAG-NEXT:    global_store_b64 v[0:1], v[5:6], off
+; GFX1100-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %oldval = load ptr, ptr addrspace(1) %out
+  %writelane = call ptr @llvm.amdgcn.writelane.p0(ptr %src, i32 %src1, ptr %oldval)
+  store ptr %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
+define void @test_writelane_v3p0(ptr addrspace(1) %out, <4 x ptr> %src, i32 %src1) {
+; GFX802-SDAG-LABEL: test_writelane_v3p0:
+; GFX802-SDAG:       ; %bb.0:
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX802-SDAG-NEXT:    v_add_u32_e32 v19, vcc, 16, v0
+; GFX802-SDAG-NEXT:    flat_load_dwordx4 v[11:14], v[0:1]
+; GFX802-SDAG-NEXT:    v_addc_u32_e32 v20, vcc, 0, v1, vcc
+; GFX802-SDAG-NEXT:    flat_load_dwordx4 v[15:18], v[19:20]
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 m0, v10
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s8, v5
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s9, v4
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s10, v3
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s11, v2
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s4, v9
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s5, v8
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s6, v7
+; GFX802-SDAG-NEXT:    v_readfirstlane_b32 s7, v6
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(1)
+; GFX802-SDAG-NEXT:    v_writelane_b32 v14, s8, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v13, s9, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v12, s10, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v11, s11, m0
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    v_writelane_b32 v18, s4, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v17, s5, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v16, s6, m0
+; GFX802-SDAG-NEXT:    v_writelane_b32 v15, s7, m0
+; GFX802-SDAG-NEXT:    flat_store_dwordx4 v[0:1], v[11:14]
+; GFX802-SDAG-NEXT:    flat_store_dwordx4 v[19:20], v[15:18]
+; GFX802-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX802-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1010-SDAG-LABEL: test_writelane_v3p0:
+; GFX1010-SDAG:       ; %bb.0:
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1010-SDAG-NEXT:    s_clause 0x1
+; GFX1010-SDAG-NEXT:    global_load_dwordx4 v[11:14], v[0:1], off offset:16
+; GFX1010-SDAG-NEXT:    global_load_dwordx4 v[15:18], v[0:1], off
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s5, v10
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s9, v5
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s10, v4
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s11, v3
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s12, v2
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s4, v9
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s6, v8
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s7, v7
+; GFX1010-SDAG-NEXT:    v_readfirstlane_b32 s8, v6
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(1)
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v14, s4, s5
+; GFX1010-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v18, s9, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v17, s10, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v16, s11, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v15, s12, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v13, s6, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v12, s7, s5
+; GFX1010-SDAG-NEXT:    v_writelane_b32 v11, s8, s5
+; GFX1010-SDAG-NEXT:    global_store_dwordx4 v[0:1], v[15:18], off
+; GFX1010-SDAG-NEXT:    global_store_dwordx4 v[0:1], v[11:14], off offset:16
+; GFX1010-SDAG-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX1100-SDAG-LABEL: test_writelane_v3p0:
+; GFX1100-SDAG:       ; %bb.0:
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX1100-SDAG-NEXT:    s_clause 0x1
+; GFX1100-SDAG-NEXT:    global_load_b128 v[11:14], v[0:1], off offset:16
+; GFX1100-SDAG-NEXT:    global_load_b128 v[15:18], v[0:1], off
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s1, v10
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s5, v5
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s6, v4
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s7, v3
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s8, v2
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s0, v9
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s2, v8
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s3, v7
+; GFX1100-SDAG-NEXT:    v_readfirstlane_b32 s4, v6
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(1)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v14, s0, s1
+; GFX1100-SDAG-NEXT:    s_waitcnt vmcnt(0)
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v18, s5, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v17, s6, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v16, s7, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v15, s8, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v13, s2, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v12, s3, s1
+; GFX1100-SDAG-NEXT:    v_writelane_b32 v11, s4, s1
+; GFX1100-SDAG-NEXT:    s_clause 0x1
+; GFX1100-SDAG-NEXT:    global_store_b128 v[0:1], v[15:18], off
+; GFX1100-SDAG-NEXT:    global_store_b128 v[0:1], v[11:14], off offset:16
+; GFX1100-SDAG-NEXT:    s_setpc_b64 s[30:31]
+  %oldval = load <4 x ptr>, ptr addrspace(1) %out
+  %writelane = call <4 x ptr> @llvm.amdgcn.writelane.v3p0(<4 x ptr> %src, i32 %src1, <4 x ptr> %oldval)
+  store <4 x ptr> %writelane, ptr addrspace(1) %out, align 4
+  ret void
+}
+
 define void @test_writelane_p3(ptr addrspace(1) %out, ptr addrspace(3) %src, i32 %src1) {
 ; GFX802-SDAG-LABEL: test_writelane_p3:
 ; GFX802-SDAG:       ; %bb.0:

>From 12155f59f4b08de99b05d8744857c68e8adfce85 Mon Sep 17 00:00:00 2001
From: Vikram <Vikram.Hegde at amd.com>
Date: Fri, 31 May 2024 09:38:25 +0000
Subject: [PATCH 7/9] add hepler to emit N-ary builtins

---
 clang/lib/CodeGen/CGBuiltin.cpp | 31 ++++++++++++++++---------------
 1 file changed, 16 insertions(+), 15 deletions(-)

diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index d9ab0051f0d23..a21aa85821375 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -614,6 +614,17 @@ static Value *emitTernaryBuiltin(CodeGenFunction &CGF,
   return CGF.Builder.CreateCall(F, { Src0, Src1, Src2 });
 }
 
+// Emit an intrinsic that has N operands of the same type as its result.
+static Value *emitNaryBuiltin(CodeGenFunction &CGF, const CallExpr *E,
+                              unsigned IntrinsicID) {
+  SmallVector<Value *, 16> Args;
+  for (unsigned i = 0, e = E->getNumArgs(); i != e; i++)
+    Args.push_back(CGF.EmitScalarExpr(E->getArg(i)));
+
+  Function *F = CGF.CGM.getIntrinsic(IntrinsicID, Args[0]->getType());
+  return CGF.Builder.CreateCall(F, Args);
+}
+
 // Emit an intrinsic that has 1 float or double operand, and 1 integer.
 static Value *emitFPIntBuiltin(CodeGenFunction &CGF,
                                const CallExpr *E,
@@ -18480,21 +18491,11 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID,
     return Builder.CreateCall(F, Args);
   }
   case AMDGPU::BI__builtin_amdgcn_permlane16:
-  case AMDGPU::BI__builtin_amdgcn_permlanex16: {
-    llvm::Value *Src0 = EmitScalarExpr(E->getArg(0));
-    llvm::Value *Src1 = EmitScalarExpr(E->getArg(1));
-    llvm::Value *Src2 = EmitScalarExpr(E->getArg(2));
-    llvm::Value *Src3 = EmitScalarExpr(E->getArg(3));
-    llvm::Value *Src4 = EmitScalarExpr(E->getArg(4));
-    llvm::Value *Src5 = EmitScalarExpr(E->getArg(5));
-
-    Intrinsic::ID IID = BuiltinID == AMDGPU::BI__builtin_amdgcn_permlane16
-                            ? Intrinsic::amdgcn_permlane16
-                            : Intrinsic::amdgcn_permlanex16;
-
-    llvm::Function *F = CGM.getIntrinsic(IID, Src0->getType());
-    return Builder.CreateCall(F, {Src0, Src1, Src2, Src3, Src4, Src5});
-  }
+  case AMDGPU::BI__builtin_amdgcn_permlanex16:
+    return emitNaryBuiltin(*this, E,
+                           BuiltinID == AMDGPU::BI__builtin_amdgcn_permlane16
+                               ? Intrinsic::amdgcn_permlane16
+                               : Intrinsic::amdgcn_permlanex16);
   case AMDGPU::BI__builtin_amdgcn_permlane64:
     return emitUnaryBuiltin(*this, E, Intrinsic::amdgcn_permlane64);
   case AMDGPU::BI__builtin_amdgcn_readlane:

>From 40381ca63182f7986a1bf723128cfdfa6b387557 Mon Sep 17 00:00:00 2001
From: Vikram <Vikram.Hegde at amd.com>
Date: Mon, 17 Jun 2024 12:09:20 +0000
Subject: [PATCH 8/9] update with latest changes from #89217

---
 llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp |   6 -
 llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h   |   7 --
 llvm/lib/Target/AMDGPU/AMDGPUInstrInfo.td     |  54 --------
 llvm/lib/Target/AMDGPU/SIISelLowering.cpp     | 115 +++++++++++++-----
 llvm/lib/Target/AMDGPU/SIInstructions.td      |   2 +-
 llvm/lib/Target/AMDGPU/VOP1Instructions.td    |   4 +-
 llvm/lib/Target/AMDGPU/VOP2Instructions.td    |   4 +-
 llvm/lib/Target/AMDGPU/VOP3Instructions.td    |   8 +-
 .../AMDGPU/llvm.amdgcn.readfirstlane.ll       |   5 +-
 .../CodeGen/AMDGPU/llvm.amdgcn.readlane.ll    |   7 +-
 10 files changed, 96 insertions(+), 116 deletions(-)

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
index 47347ada935be..18193d8807597 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
@@ -5508,12 +5508,6 @@ const char* AMDGPUTargetLowering::getTargetNodeName(unsigned Opcode) const {
   NODE_NAME_CASE(LDS)
   NODE_NAME_CASE(FPTRUNC_ROUND_UPWARD)
   NODE_NAME_CASE(FPTRUNC_ROUND_DOWNWARD)
-  NODE_NAME_CASE(READLANE)
-  NODE_NAME_CASE(READFIRSTLANE)
-  NODE_NAME_CASE(WRITELANE)
-  NODE_NAME_CASE(PERMLANE16)
-  NODE_NAME_CASE(PERMLANEX16)
-  NODE_NAME_CASE(PERMLANE64)
   NODE_NAME_CASE(DUMMY_CHAIN)
   case AMDGPUISD::FIRST_MEM_OPCODE_NUMBER: break;
   NODE_NAME_CASE(LOAD_D16_HI)
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h
index 293fa6259fef7..71c4334029b43 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h
+++ b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.h
@@ -558,13 +558,6 @@ enum NodeType : unsigned {
   FPTRUNC_ROUND_UPWARD,
   FPTRUNC_ROUND_DOWNWARD,
 
-  READLANE,
-  READFIRSTLANE,
-  WRITELANE,
-  PERMLANE16,
-  PERMLANEX16,
-  PERMLANE64,
-
   DUMMY_CHAIN,
   FIRST_MEM_OPCODE_NUMBER = ISD::FIRST_TARGET_MEMORY_OPCODE,
   LOAD_D16_HI,
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUInstrInfo.td b/llvm/lib/Target/AMDGPU/AMDGPUInstrInfo.td
index a82f48950f493..702f6e67c5527 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUInstrInfo.td
+++ b/llvm/lib/Target/AMDGPU/AMDGPUInstrInfo.td
@@ -342,34 +342,6 @@ def AMDGPUfdot2_impl : SDNode<"AMDGPUISD::FDOT2",
 
 def AMDGPUperm_impl : SDNode<"AMDGPUISD::PERM", AMDGPUDTIntTernaryOp, []>;
 
-def AMDGPUReadfirstlaneOp : SDTypeProfile<1, 1, [
-  SDTCisSameAs<0, 1>
-]>;
-
-def AMDGPUReadlaneOp : SDTypeProfile<1, 2, [
-  SDTCisSameAs<0, 1>, SDTCisInt<2>
-]>;
-
-def AMDGPUDWritelaneOp : SDTypeProfile<1, 3, [
-  SDTCisSameAs<0, 1>, SDTCisInt<2>, SDTCisSameAs<0, 3>
-]>;
-
-def AMDGPUDPermlane16Op : SDTypeProfile<1, 6, [
-  SDTCisSameAs<0, 1>,  // old
-  SDTCisSameAs<0, 2>,  // src0
-  SDTCisInt<3>,        // src1
-  SDTCisInt<4>,        // src2
-  SDTCisInt<5>,        // i1 fi
-  SDTCisInt<6>         // i1 bound_ctrl
-]>;
-
-def AMDGPUreadlane_impl : SDNode<"AMDGPUISD::READLANE", AMDGPUReadlaneOp>;
-def AMDGPUreadfirstlane_impl : SDNode<"AMDGPUISD::READFIRSTLANE", AMDGPUReadfirstlaneOp>;
-def AMDGPUwritelane_impl : SDNode<"AMDGPUISD::WRITELANE", AMDGPUDWritelaneOp>;
-def AMDGPUpermlane16_impl : SDNode<"AMDGPUISD::PERMLANE16", AMDGPUDPermlane16Op>;
-def AMDGPUpermlanex16_impl : SDNode<"AMDGPUISD::PERMLANEX16", AMDGPUDPermlane16Op>;
-def AMDGPUpermlane64_impl : SDNode<"AMDGPUISD::PERMLANE64", AMDGPUReadfirstlaneOp>;
-
 // SI+ export
 def AMDGPUExportOp : SDTypeProfile<0, 8, [
   SDTCisInt<0>,       // i8 tgt
@@ -534,29 +506,3 @@ def AMDGPUdiv_fmas : PatFrags<(ops node:$src0, node:$src1, node:$src2, node:$vcc
 def AMDGPUperm : PatFrags<(ops node:$src0, node:$src1, node:$src2),
   [(int_amdgcn_perm node:$src0, node:$src1, node:$src2),
    (AMDGPUperm_impl node:$src0, node:$src1, node:$src2)]>;
-
-def AMDGPUreadlane : PatFrags<(ops node:$src0, node:$src1),
-  [(int_amdgcn_readlane node:$src0, node:$src1),
-   (AMDGPUreadlane_impl node:$src0, node:$src1)]>;
-
-def AMDGPUreadfirstlane : PatFrags<(ops node:$src),
-  [(int_amdgcn_readfirstlane node:$src),
-   (AMDGPUreadfirstlane_impl node:$src)]>;
-
-def AMDGPUwritelane : PatFrags<(ops node:$src0, node:$src1, node:$src2),
-  [(int_amdgcn_writelane node:$src0, node:$src1, node:$src2),
-   (AMDGPUwritelane_impl node:$src0, node:$src1, node:$src2)]>;
-
-def AMDGPUpermlane16 : PatFrags<(ops node:$src0, node:$src1, node:$src2, node:$src3, node:$src4, node:$src5),
-  [(int_amdgcn_permlane16 node:$src0, node:$src1, node:$src2, node:$src3, node:$src4, node:$src5),
-   (AMDGPUpermlane16_impl node:$src0, node:$src1, node:$src2, node:$src3, node:$src4, node:$src5)]>;
-
-def AMDGPUpermlanex16 : PatFrags<(ops node:$src0, node:$src1, node:$src2, node:$src3, node:$src4, node:$src5),
-  [(int_amdgcn_permlanex16 node:$src0, node:$src1, node:$src2, node:$src3, node:$src4, node:$src5),
-   (AMDGPUpermlanex16_impl node:$src0, node:$src1, node:$src2, node:$src3, node:$src4, node:$src5)]>;
-
-def AMDGPUpermlane64 : PatFrags<(ops node:$src),
-  [(int_amdgcn_permlane64 node:$src),
-   (AMDGPUpermlane64_impl node:$src)]>;
-
-
diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index d91ab31832070..f55025ec5e08d 100644
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -6095,46 +6095,55 @@ static SDValue lowerLaneOp(const SITargetLowering &TLI, SDNode *N,
                            SelectionDAG &DAG) {
   EVT VT = N->getValueType(0);
   unsigned ValSize = VT.getSizeInBits();
-  unsigned IntrinsicID = N->getConstantOperandVal(0);
-  bool IsPermLane16 = IntrinsicID == Intrinsic::amdgcn_permlane16 ||
-                      IntrinsicID == Intrinsic::amdgcn_permlanex16;
-  SDValue Src0 = N->getOperand(1);
+  unsigned IID = N->getConstantOperandVal(0);
+  bool IsPermLane16 = IID == Intrinsic::amdgcn_permlane16 ||
+                      IID == Intrinsic::amdgcn_permlanex16;
   SDLoc SL(N);
   MVT IntVT = MVT::getIntegerVT(ValSize);
 
-  auto createLaneOp = [&DAG, &SL, N](SDValue Src0, SDValue Src1, SDValue Src2,
-                                     MVT ValueT) -> SDValue {
-    switch (unsigned IID = N->getConstantOperandVal(0)) {
+  auto createLaneOp = [&DAG, &SL, N, IID](SDValue Src0, SDValue Src1, SDValue Src2,
+                                     MVT ValT) -> SDValue {
+    SmallVector<SDValue, 8> Operands;
+    switch (IID) {
+    case Intrinsic::amdgcn_permlane16:
+    case Intrinsic::amdgcn_permlanex16:
+      Operands.push_back(N->getOperand(6));
+      Operands.push_back(N->getOperand(5));
+      Operands.push_back(N->getOperand(4));
+      [[fallthrough]];
+    case Intrinsic::amdgcn_writelane:
+      Operands.push_back(Src2);
+      [[fallthrough]];
+    case Intrinsic::amdgcn_readlane:
+      Operands.push_back(Src1);
+      [[fallthrough]];
     case Intrinsic::amdgcn_readfirstlane:
     case Intrinsic::amdgcn_permlane64:
-      return DAG.getNode(IID == Intrinsic::amdgcn_readfirstlane
-                             ? AMDGPUISD::READFIRSTLANE
-                             : AMDGPUISD::PERMLANE64,
-                         SL, ValueT, {Src0});
-    case Intrinsic::amdgcn_readlane:
-      return DAG.getNode(AMDGPUISD::READLANE, SL, ValueT, {Src0, Src1});
-    case Intrinsic::amdgcn_writelane:
-      return DAG.getNode(AMDGPUISD::WRITELANE, SL, ValueT, {Src0, Src1, Src2});
-    case Intrinsic::amdgcn_permlane16:
-    case Intrinsic::amdgcn_permlanex16: {
-      SDValue Src3 = N->getOperand(4);
-      SDValue Src4 = N->getOperand(5);
-      SDValue Src5 = N->getOperand(6);
-      return DAG.getNode(IID == Intrinsic::amdgcn_permlane16
-                             ? AMDGPUISD::PERMLANE16
-                             : AMDGPUISD::PERMLANEX16,
-                         SL, ValueT, {Src0, Src1, Src2, Src3, Src4, Src5});
-    }
+      Operands.push_back(Src0);
+      break;
     default:
       llvm_unreachable("unhandled lane op");
     }
+
+    Operands.push_back(DAG.getTargetConstant(IID, SL, MVT::i32));
+    std::reverse(Operands.begin(), Operands.end());
+
+    if (SDNode *GL = N->getGluedNode()) {
+      assert(GL->getOpcode() == ISD::CONVERGENCECTRL_GLUE);
+      GL = GL->getOperand(0).getNode();
+      Operands.push_back(DAG.getNode(ISD::CONVERGENCECTRL_GLUE, SL, MVT::Glue,
+                                     SDValue(GL, 0)));
+    }
+
+    return DAG.getNode(ISD::INTRINSIC_WO_CHAIN, SL, ValT, Operands);
   };
 
+  SDValue Src0 = N->getOperand(1);
   SDValue Src1, Src2;
-  if (IntrinsicID == Intrinsic::amdgcn_readlane ||
-      IntrinsicID == Intrinsic::amdgcn_writelane || IsPermLane16) {
+  if (IID == Intrinsic::amdgcn_readlane ||
+      IID == Intrinsic::amdgcn_writelane || IsPermLane16) {
     Src1 = N->getOperand(2);
-    if (IntrinsicID == Intrinsic::amdgcn_writelane || IsPermLane16)
+    if (IID == Intrinsic::amdgcn_writelane || IsPermLane16)
       Src2 = N->getOperand(3);
   }
 
@@ -6153,7 +6162,7 @@ static SDValue lowerLaneOp(const SITargetLowering &TLI, SDNode *N,
                                   SL, MVT::i32);
     }
 
-    if (IntrinsicID == Intrinsic::amdgcn_writelane) {
+    if (IID == Intrinsic::amdgcn_writelane) {
       Src2 = DAG.getAnyExtOrTrunc(IsFloat ? DAG.getBitcast(IntVT, Src2) : Src2,
                                   SL, MVT::i32);
     }
@@ -6165,6 +6174,46 @@ static SDValue lowerLaneOp(const SITargetLowering &TLI, SDNode *N,
 
   if (ValSize % 32 != 0)
     return SDValue();
+  
+  auto unrollLaneOp = [&DAG, &SL](SDNode *N) -> SDValue {
+    EVT VT = N->getValueType(0);
+    unsigned NE = VT.getVectorNumElements();
+    EVT EltVT = VT.getVectorElementType();
+    SmallVector<SDValue, 8> Scalars;
+    unsigned NumOperands = N->getNumOperands();
+    SmallVector<SDValue, 4> Operands(NumOperands);
+    SDNode *GL = N->getGluedNode();
+
+    // only handle convergencectrl_glue
+    assert(!GL || GL->getOpcode() == ISD::CONVERGENCECTRL_GLUE);
+
+    for (unsigned i = 0; i != NE; ++i) {
+      for (unsigned j = 0, e = GL ? NumOperands - 1 : NumOperands; j != e;
+           ++j) {
+        SDValue Operand = N->getOperand(j);
+        EVT OperandVT = Operand.getValueType();
+        if (OperandVT.isVector()) {
+          // A vector operand; extract a single element.
+          EVT OperandEltVT = OperandVT.getVectorElementType();
+          Operands[j] = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, SL, OperandEltVT,
+                                    Operand, DAG.getVectorIdxConstant(i, SL));
+        } else {
+          // A scalar operand; just use it as is.
+          Operands[j] = Operand;
+        }
+      }
+
+      if (GL)
+        Operands[NumOperands - 1] =
+            DAG.getNode(ISD::CONVERGENCECTRL_GLUE, SL, MVT::Glue,
+                        SDValue(GL->getOperand(0).getNode(), 0));
+
+      Scalars.push_back(DAG.getNode(N->getOpcode(), SL, EltVT, Operands));
+    }
+
+    EVT VecVT = EVT::getVectorVT(*DAG.getContext(), EltVT, NE);
+    return DAG.getBuildVector(VecVT, SL, Scalars);
+  };
 
   if (VT.isVector()) {
     switch (MVT::SimpleValueType EltTy =
@@ -6172,7 +6221,7 @@ static SDValue lowerLaneOp(const SITargetLowering &TLI, SDNode *N,
     case MVT::i32:
     case MVT::f32: {
       SDValue LaneOp = createLaneOp(Src0, Src1, Src2, VT.getSimpleVT());
-      return DAG.UnrollVectorOp(LaneOp.getNode());
+      return unrollLaneOp(LaneOp.getNode());
     }
     case MVT::i16:
     case MVT::f16:
@@ -6188,7 +6237,7 @@ static SDValue lowerLaneOp(const SITargetLowering &TLI, SDNode *N,
           Src1SubVec = DAG.getNode(ISD::EXTRACT_SUBVECTOR, SL, SubVecVT, Src1,
                                    DAG.getConstant(EltIdx, SL, MVT::i32));
 
-        if (IntrinsicID == Intrinsic::amdgcn_writelane)
+        if (IID == Intrinsic::amdgcn_writelane)
           Src2SubVec = DAG.getNode(ISD::EXTRACT_SUBVECTOR, SL, SubVecVT, Src2,
                                    DAG.getConstant(EltIdx, SL, MVT::i32));
 
@@ -6212,11 +6261,11 @@ static SDValue lowerLaneOp(const SITargetLowering &TLI, SDNode *N,
     if (IsPermLane16)
       Src1 = DAG.getBitcast(VecVT, Src1);
 
-    if (IntrinsicID == Intrinsic::amdgcn_writelane)
+    if (IID == Intrinsic::amdgcn_writelane)
       Src2 = DAG.getBitcast(VecVT, Src2);
 
     SDValue LaneOp = createLaneOp(Src0, Src1, Src2, VecVT);
-    SDValue UnrolledLaneOp = DAG.UnrollVectorOp(LaneOp.getNode());
+    SDValue UnrolledLaneOp = unrollLaneOp(LaneOp.getNode());
     return DAG.getBitcast(VT, UnrolledLaneOp);
 }
 
diff --git a/llvm/lib/Target/AMDGPU/SIInstructions.td b/llvm/lib/Target/AMDGPU/SIInstructions.td
index 5941b264fdb21..a5e59fde107ea 100644
--- a/llvm/lib/Target/AMDGPU/SIInstructions.td
+++ b/llvm/lib/Target/AMDGPU/SIInstructions.td
@@ -3389,7 +3389,7 @@ def : GCNPat<
 // FIXME: Should also do this for readlane, but tablegen crashes on
 // the ignored src1.
 def : GCNPat<
-  (i32 (AMDGPUreadfirstlane (i32 imm:$src))),
+  (i32 (int_amdgcn_readfirstlane (i32 imm:$src))),
   (S_MOV_B32 SReg_32:$src)
 >;
 
diff --git a/llvm/lib/Target/AMDGPU/VOP1Instructions.td b/llvm/lib/Target/AMDGPU/VOP1Instructions.td
index 71c83e072b4b0..2c0d61ee4afa1 100644
--- a/llvm/lib/Target/AMDGPU/VOP1Instructions.td
+++ b/llvm/lib/Target/AMDGPU/VOP1Instructions.td
@@ -255,7 +255,7 @@ def V_READFIRSTLANE_B32 : VOP1_Pseudo <"v_readfirstlane_b32", VOP_READFIRSTLANE,
 }
 
 foreach vt = Reg32Types.types in {
-  def : GCNPat<(vt (AMDGPUreadfirstlane (vt VRegOrLdsSrc_32:$src0))),
+  def : GCNPat<(vt (int_amdgcn_readfirstlane (vt VRegOrLdsSrc_32:$src0))),
         (V_READFIRSTLANE_B32 (vt VRegOrLdsSrc_32:$src0))
   >;
 }
@@ -743,7 +743,7 @@ let SubtargetPredicate = isGFX11Plus in {
 } // End SubtargetPredicate = isGFX11Plus
 
 foreach vt = Reg32Types.types in {
-  def : GCNPat<(AMDGPUpermlane64 (vt VRegSrc_32:$src0)),
+  def : GCNPat<(int_amdgcn_permlane64 (vt VRegSrc_32:$src0)),
         (vt (V_PERMLANE64_B32 (vt VRegSrc_32:$src0)))
   >;
 }
diff --git a/llvm/lib/Target/AMDGPU/VOP2Instructions.td b/llvm/lib/Target/AMDGPU/VOP2Instructions.td
index 40106314cb0e6..ae4030e509e9c 100644
--- a/llvm/lib/Target/AMDGPU/VOP2Instructions.td
+++ b/llvm/lib/Target/AMDGPU/VOP2Instructions.td
@@ -789,11 +789,11 @@ def V_WRITELANE_B32 : VOP2_Pseudo<"v_writelane_b32", VOP_WRITELANE, []> {
 } // End isConvergent = 1
 
 foreach vt = Reg32Types.types in {
-  def : GCNPat<(vt (AMDGPUreadlane vt:$src0, i32:$src1)),
+  def : GCNPat<(vt (int_amdgcn_readlane vt:$src0, i32:$src1)),
         (V_READLANE_B32 VRegOrLdsSrc_32:$src0, SCSrc_b32:$src1)
   >;
 
-  def : GCNPat<(vt (AMDGPUwritelane vt:$src0, i32:$src1, vt:$src2)),
+  def : GCNPat<(vt (int_amdgcn_writelane vt:$src0, i32:$src1, vt:$src2)),
         (V_WRITELANE_B32 SCSrc_b32:$src0, SCSrc_b32:$src1, VGPR_32:$src2)
   >;
 }
diff --git a/llvm/lib/Target/AMDGPU/VOP3Instructions.td b/llvm/lib/Target/AMDGPU/VOP3Instructions.td
index ff098f8fe50f3..6ba54cca44958 100644
--- a/llvm/lib/Target/AMDGPU/VOP3Instructions.td
+++ b/llvm/lib/Target/AMDGPU/VOP3Instructions.td
@@ -841,8 +841,8 @@ def gi_opsel_i1timm : GICustomOperandRenderer<"renderOpSelTImm">,
 
 class PermlanePat<SDPatternOperator permlane,
   Instruction inst, ValueType vt> : GCNPat<
-  (permlane vt:$vdst_in, vt:$src0, i32:$src1, i32:$src2,
-            timm:$fi, timm:$bc),
+  (vt (permlane vt:$vdst_in, vt:$src0, i32:$src1, i32:$src2,
+            timm:$fi, timm:$bc)),
   (inst (opsel_i1timm $fi), VGPR_32:$src0, (opsel_i1timm $bc),
         SCSrc_b32:$src1, 0, SCSrc_b32:$src2, VGPR_32:$vdst_in)
 >;
@@ -867,8 +867,8 @@ let SubtargetPredicate = isGFX10Plus in {
   } // End $vdst = $vdst_in, DisableEncoding $vdst_in, IsInvalidSingleUseConsumer = 1, IsInvalidSingleUseProducer = 1
 
   foreach vt = Reg32Types.types in {
-    def : PermlanePat<AMDGPUpermlane16, V_PERMLANE16_B32_e64, vt>;
-    def : PermlanePat<AMDGPUpermlanex16, V_PERMLANEX16_B32_e64, vt>;
+    def : PermlanePat<int_amdgcn_permlane16, V_PERMLANE16_B32_e64, vt>;
+    def : PermlanePat<int_amdgcn_permlanex16, V_PERMLANEX16_B32_e64, vt>;
   }
 
   defm V_ADD_NC_U16 : VOP3Inst <"v_add_nc_u16", VOP3_Profile<VOP_I16_I16_I16, VOP3_OPSEL>, add>;
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll
index ed0da0d2a61a2..cc6c630ae6466 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll
@@ -463,10 +463,9 @@ define void @test_readfirstlane_i16(ptr addrspace(1) %out, i16 %src) {
 ; CHECK-SDAG:       ; %bb.0:
 ; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v2
-; CHECK-SDAG-NEXT:    v_mov_b32_e32 v0, 0xffff
-; CHECK-SDAG-NEXT:    v_and_b32_e32 v0, s4, v0
+; CHECK-SDAG-NEXT:    s_and_b32 s4, s4, 0xffff
 ; CHECK-SDAG-NEXT:    ;;#ASMSTART
-; CHECK-SDAG-NEXT:    ; use v0
+; CHECK-SDAG-NEXT:    ; use s4
 ; CHECK-SDAG-NEXT:    ;;#ASMEND
 ; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ll
index 325a39abb588a..66e1f9396de5a 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ll
@@ -743,12 +743,11 @@ define void @test_readlane_i16(ptr addrspace(1) %out, i16 %src, i32 %src1) {
 ; CHECK-SDAG:       ; %bb.0:
 ; CHECK-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; CHECK-SDAG-NEXT:    v_readfirstlane_b32 s4, v3
-; CHECK-SDAG-NEXT:    v_mov_b32_e32 v0, 0xffff
-; CHECK-SDAG-NEXT:    s_nop 2
+; CHECK-SDAG-NEXT:    s_nop 3
 ; CHECK-SDAG-NEXT:    v_readlane_b32 s4, v2, s4
-; CHECK-SDAG-NEXT:    v_and_b32_e32 v0, s4, v0
+; CHECK-SDAG-NEXT:    s_and_b32 s4, s4, 0xffff
 ; CHECK-SDAG-NEXT:    ;;#ASMSTART
-; CHECK-SDAG-NEXT:    ; use v0
+; CHECK-SDAG-NEXT:    ; use s4
 ; CHECK-SDAG-NEXT:    ;;#ASMEND
 ; CHECK-SDAG-NEXT:    s_setpc_b64 s[30:31]
 ;

>From 4e4cdd961e3daea74e2ae5689fea73c0b6685f57 Mon Sep 17 00:00:00 2001
From: Vikram <Vikram.Hegde at amd.com>
Date: Mon, 17 Jun 2024 08:16:39 -0400
Subject: [PATCH 9/9] clang format

---
 llvm/lib/Target/AMDGPU/SIISelLowering.cpp | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index f55025ec5e08d..1101de206a8fe 100644
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -6101,8 +6101,8 @@ static SDValue lowerLaneOp(const SITargetLowering &TLI, SDNode *N,
   SDLoc SL(N);
   MVT IntVT = MVT::getIntegerVT(ValSize);
 
-  auto createLaneOp = [&DAG, &SL, N, IID](SDValue Src0, SDValue Src1, SDValue Src2,
-                                     MVT ValT) -> SDValue {
+  auto createLaneOp = [&DAG, &SL, N, IID](SDValue Src0, SDValue Src1,
+                                          SDValue Src2, MVT ValT) -> SDValue {
     SmallVector<SDValue, 8> Operands;
     switch (IID) {
     case Intrinsic::amdgcn_permlane16:
@@ -6140,8 +6140,8 @@ static SDValue lowerLaneOp(const SITargetLowering &TLI, SDNode *N,
 
   SDValue Src0 = N->getOperand(1);
   SDValue Src1, Src2;
-  if (IID == Intrinsic::amdgcn_readlane ||
-      IID == Intrinsic::amdgcn_writelane || IsPermLane16) {
+  if (IID == Intrinsic::amdgcn_readlane || IID == Intrinsic::amdgcn_writelane ||
+      IsPermLane16) {
     Src1 = N->getOperand(2);
     if (IID == Intrinsic::amdgcn_writelane || IsPermLane16)
       Src2 = N->getOperand(3);
@@ -6174,7 +6174,7 @@ static SDValue lowerLaneOp(const SITargetLowering &TLI, SDNode *N,
 
   if (ValSize % 32 != 0)
     return SDValue();
-  
+
   auto unrollLaneOp = [&DAG, &SL](SDNode *N) -> SDValue {
     EVT VT = N->getValueType(0);
     unsigned NE = VT.getVectorNumElements();



More information about the cfe-commits mailing list