[llvm] 35f7b60 - [AMDGPU] Extend permlane16, permlanex16 and permlane64 intrinsic lowering for generic types (#92725)

via llvm-commits llvm-commits at lists.llvm.org
Tue Jun 25 20:54:15 PDT 2024


Author: Vikram Hegde
Date: 2024-06-26T09:24:09+05:30
New Revision: 35f7b60aa6105753859bcccaf4a793aaf16b4acd

URL: https://github.com/llvm/llvm-project/commit/35f7b60aa6105753859bcccaf4a793aaf16b4acd
DIFF: https://github.com/llvm/llvm-project/commit/35f7b60aa6105753859bcccaf4a793aaf16b4acd.diff

LOG: [AMDGPU] Extend permlane16, permlanex16 and permlane64 intrinsic lowering for generic types (#92725)

These are incremental changes over #89217 , with core logic being the
same. This patch along with #89217 and #91190 should get us ready to enable 64
bit optimizations in atomic optimizer.

Added: 
    llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane.ptr.ll
    llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ptr.ll

Modified: 
    clang/lib/CodeGen/CGBuiltin.cpp
    clang/test/CodeGenOpenCL/builtins-amdgcn-gfx10.cl
    clang/test/CodeGenOpenCL/builtins-amdgcn-gfx11.cl
    llvm/docs/AMDGPUUsage.rst
    llvm/include/llvm/IR/IntrinsicsAMDGPU.td
    llvm/lib/Target/AMDGPU/AMDGPUAtomicOptimizer.cpp
    llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
    llvm/lib/Target/AMDGPU/SIISelLowering.cpp
    llvm/lib/Target/AMDGPU/VOP1Instructions.td
    llvm/lib/Target/AMDGPU/VOP3Instructions.td
    llvm/test/Analysis/UniformityAnalysis/AMDGPU/intrinsics.ll
    llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane.ll
    llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ll
    llvm/test/Transforms/InstCombine/AMDGPU/amdgcn-intrinsics.ll
    llvm/test/Verifier/AMDGPU/intrinsic-immarg.ll

Removed: 
    


################################################################################
diff  --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index 4edd8283aa03c..96dcf6283f9f8 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -18479,6 +18479,16 @@ 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:
+    return emitBuiltinWithOneOverloadedType<6>(
+        *this, E,
+        BuiltinID == AMDGPU::BI__builtin_amdgcn_permlane16
+            ? Intrinsic::amdgcn_permlane16
+            : Intrinsic::amdgcn_permlanex16);
+  case AMDGPU::BI__builtin_amdgcn_permlane64:
+    return emitBuiltinWithOneOverloadedType<1>(*this, E,
+                                               Intrinsic::amdgcn_permlane64);
   case AMDGPU::BI__builtin_amdgcn_readlane:
     return emitBuiltinWithOneOverloadedType<2>(*this, E,
                                                Intrinsic::amdgcn_readlane);

diff  --git a/clang/test/CodeGenOpenCL/builtins-amdgcn-gfx10.cl b/clang/test/CodeGenOpenCL/builtins-amdgcn-gfx10.cl
index f30776a8bb85b..3cf1056cf4f48 100644
--- a/clang/test/CodeGenOpenCL/builtins-amdgcn-gfx10.cl
+++ b/clang/test/CodeGenOpenCL/builtins-amdgcn-gfx10.cl
@@ -8,13 +8,13 @@ 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 868b5bed0c952..f444657463568 100644
--- a/clang/test/CodeGenOpenCL/builtins-amdgcn-gfx11.cl
+++ b/clang/test/CodeGenOpenCL/builtins-amdgcn-gfx11.cl
@@ -37,7 +37,7 @@ 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/docs/AMDGPUUsage.rst b/llvm/docs/AMDGPUUsage.rst
index 192df32229787..80dd619019e7b 100644
--- a/llvm/docs/AMDGPUUsage.rst
+++ b/llvm/docs/AMDGPUUsage.rst
@@ -1225,6 +1225,26 @@ 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, 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, 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, 
+                                                   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
                                                    with two v2i16 operands, summed with the third i32 operand. The

diff  --git a/llvm/include/llvm/IR/IntrinsicsAMDGPU.td b/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
index 11662ccc1a695..9eb1d7efb1067 100644
--- a/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
+++ b/llvm/include/llvm/IR/IntrinsicsAMDGPU.td
@@ -2355,16 +2355,16 @@ def int_amdgcn_pops_exiting_wave_id :
 //===----------------------------------------------------------------------===//
 
 // 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],
+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">,
-  Intrinsic<[llvm_i32_ty],
-            [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, llvm_i1_ty, llvm_i1_ty],
+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,
              ImmArg<ArgIndex<4>>, ImmArg<ArgIndex<5>>, IntrNoCallback, IntrNoFree]>;
 
@@ -2407,8 +2407,7 @@ 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 0fdf3e468a71b..d7ef6f3c5dc43 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUAtomicOptimizer.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUAtomicOptimizer.cpp
@@ -404,7 +404,7 @@ 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));
@@ -416,7 +416,7 @@ 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));
   }
@@ -472,7 +472,7 @@ 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/AMDGPULegalizerInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
index 4b48091b7143e..f1254b2e9e1d2 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
@@ -5438,16 +5438,32 @@ bool AMDGPULegalizerInfo::legalizeLaneOp(LegalizerHelper &Helper,
   MachineIRBuilder &B = Helper.MIRBuilder;
   MachineRegisterInfo &MRI = *B.getMRI();
 
-  auto createLaneOp = [&IID, &B](Register Src0, Register Src1, Register Src2,
-                                 LLT VT) -> Register {
+  bool IsPermLane16 = IID == Intrinsic::amdgcn_permlane16 ||
+                      IID == Intrinsic::amdgcn_permlanex16;
+
+  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:
       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");
     }
@@ -5456,9 +5472,10 @@ 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) {
+  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();
     }
   }
@@ -5473,12 +5490,15 @@ bool AMDGPULegalizerInfo::legalizeLaneOp(LegalizerHelper &Helper,
 
   if (Size < 32) {
     Src0 = B.buildAnyExt(S32, Src0).getReg(0);
-    if (Src2.isValid())
+
+    if (IsPermLane16)
+      Src1 = B.buildAnyExt(LLT::scalar(32), Src1).getReg(0);
+
+    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;
   }
@@ -5505,15 +5525,23 @@ bool AMDGPULegalizerInfo::legalizeLaneOp(LegalizerHelper &Helper,
   SmallVector<Register, 2> PartialRes;
   unsigned NumParts = Size / 32;
   MachineInstrBuilder Src0Parts = B.buildUnmerge(PartialResTy, Src0);
-  MachineInstrBuilder Src2Parts;
+  MachineInstrBuilder Src1Parts, Src2Parts;
+
+  if (IsPermLane16)
+    Src1Parts = B.buildUnmerge(PartialResTy, Src1);
 
-  if (Src2.isValid())
+  if (IID == Intrinsic::amdgcn_writelane)
     Src2Parts = B.buildUnmerge(PartialResTy, Src2);
 
   for (unsigned i = 0; i < NumParts; ++i) {
     Src0 = Src0Parts.getReg(i);
-    if (Src2.isValid())
+
+    if (IsPermLane16)
+      Src1 = Src1Parts.getReg(i);
+
+    if (IID == Intrinsic::amdgcn_writelane)
       Src2 = Src2Parts.getReg(i);
+
     PartialRes.push_back(createLaneOp(Src0, Src1, Src2, PartialResTy));
   }
 
@@ -7465,6 +7493,9 @@ 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 b8ff5ed35ac80..83bfb622ee525 100644
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -6119,28 +6119,38 @@ static SDValue lowerLaneOp(const SITargetLowering &TLI, SDNode *N,
   EVT VT = N->getValueType(0);
   unsigned ValSize = VT.getSizeInBits();
   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, IID](SDValue Src0, SDValue Src1,
                                           SDValue Src2, MVT ValT) -> SDValue {
     SmallVector<SDValue, 8> Operands;
-    Operands.push_back(DAG.getTargetConstant(IID, SL, MVT::i32));
     switch (IID) {
-    case Intrinsic::amdgcn_readfirstlane:
-      Operands.push_back(Src0);
-      break;
+    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(Src0);
       Operands.push_back(Src1);
-      break;
-    case Intrinsic::amdgcn_writelane:
+      [[fallthrough]];
+    case Intrinsic::amdgcn_readfirstlane:
+    case Intrinsic::amdgcn_permlane64:
       Operands.push_back(Src0);
-      Operands.push_back(Src1);
-      Operands.push_back(Src2);
       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();
@@ -6153,9 +6163,10 @@ 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) {
+  if (IID == Intrinsic::amdgcn_readlane || IID == Intrinsic::amdgcn_writelane ||
+      IsPermLane16) {
     Src1 = N->getOperand(2);
-    if (IID == Intrinsic::amdgcn_writelane)
+    if (IID == Intrinsic::amdgcn_writelane || IsPermLane16)
       Src2 = N->getOperand(3);
   }
 
@@ -6168,10 +6179,17 @@ static SDValue lowerLaneOp(const SITargetLowering &TLI, SDNode *N,
     bool IsFloat = VT.isFloatingPoint();
     Src0 = DAG.getAnyExtOrTrunc(IsFloat ? DAG.getBitcast(IntVT, Src0) : Src0,
                                 SL, MVT::i32);
-    if (Src2.getNode()) {
+
+    if (IsPermLane16) {
+      Src1 = DAG.getAnyExtOrTrunc(IsFloat ? DAG.getBitcast(IntVT, Src1) : Src1,
+                                  SL, MVT::i32);
+    }
+
+    if (IID == Intrinsic::amdgcn_writelane) {
       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 IsFloat ? DAG.getBitcast(VT, Trunc) : Trunc;
@@ -6233,17 +6251,23 @@ static SDValue lowerLaneOp(const SITargetLowering &TLI, SDNode *N,
     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++) {
-        SDValue Src0SubVec =
-            DAG.getNode(ISD::EXTRACT_SUBVECTOR, SL, SubVecVT, Src0,
-                        DAG.getConstant(EltIdx, SL, MVT::i32));
+        Src0SubVec = DAG.getNode(ISD::EXTRACT_SUBVECTOR, SL, SubVecVT, Src0,
+                                 DAG.getConstant(EltIdx, SL, MVT::i32));
 
-        SDValue Src2SubVec;
-        if (Src2)
+        if (IsPermLane16)
+          Src1SubVec = DAG.getNode(ISD::EXTRACT_SUBVECTOR, SL, SubVecVT, Src1,
+                                   DAG.getConstant(EltIdx, SL, MVT::i32));
+
+        if (IID == Intrinsic::amdgcn_writelane)
           Src2SubVec = DAG.getNode(ISD::EXTRACT_SUBVECTOR, SL, SubVecVT, Src2,
                                    DAG.getConstant(EltIdx, SL, MVT::i32));
 
-        Pieces.push_back(createLaneOp(Src0SubVec, Src1, Src2SubVec, SubVecVT));
+        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);
@@ -6257,7 +6281,10 @@ static SDValue lowerLaneOp(const SITargetLowering &TLI, SDNode *N,
   MVT VecVT = MVT::getVectorVT(MVT::i32, ValSize / 32);
   Src0 = DAG.getBitcast(VecVT, Src0);
 
-  if (Src2)
+  if (IsPermLane16)
+    Src1 = DAG.getBitcast(VecVT, Src1);
+
+  if (IID == Intrinsic::amdgcn_writelane)
     Src2 = DAG.getBitcast(VecVT, Src2);
 
   SDValue LaneOp = createLaneOp(Src0, Src1, Src2, VecVT);
@@ -8734,6 +8761,9 @@ 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 0794ba2f66f40..2c0d61ee4afa1 100644
--- a/llvm/lib/Target/AMDGPU/VOP1Instructions.td
+++ b/llvm/lib/Target/AMDGPU/VOP1Instructions.td
@@ -732,9 +732,7 @@ 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> {
     let IsInvalidSingleUseConsumer = 1;
     let IsInvalidSingleUseProducer = 1;
   }
@@ -744,6 +742,12 @@ let SubtargetPredicate = isGFX11Plus in {
   defm V_CVT_U32_U16    : VOP1Inst_t16<"v_cvt_u32_u16", VOP_I32_I16>;
 } // End SubtargetPredicate = isGFX11Plus
 
+foreach vt = Reg32Types.types in {
+  def : GCNPat<(int_amdgcn_permlane64 (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 ce12506c9e6f6..efa8e9c74d449 100644
--- a/llvm/lib/Target/AMDGPU/VOP3Instructions.td
+++ b/llvm/lib/Target/AMDGPU/VOP3Instructions.td
@@ -842,9 +842,9 @@ 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,
-            timm:$fi, timm:$bc),
+  Instruction inst, ValueType vt> : GCNPat<
+  (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)
 >;
@@ -868,8 +868,10 @@ let SubtargetPredicate = isGFX10Plus in {
     defm V_PERMLANEX16_B32 : VOP3Inst<"v_permlanex16_b32", VOP3_PERMLANE_Profile>;
   } // End $vdst = $vdst_in, DisableEncoding $vdst_in, IsInvalidSingleUseConsumer = 1, IsInvalidSingleUseProducer = 1
 
-  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<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>;
   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..1ae22c3eec185 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane.ll
@@ -11,8 +11,8 @@ 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 +24,7 @@ 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 +38,7 @@ 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 +51,252 @@ 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 +308,7 @@ 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 +322,7 @@ 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
@@ -94,596 +333,680 @@ 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_vvs:
+; 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_readfirstlane_b32 s4, v0
+; 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, 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_vvs:
+; 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_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, 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_vvs:
+; 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_readfirstlane_b32 s4, v0
+; 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, 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()
-  %v = call i32 @llvm.amdgcn.permlane16(i32 %src0, i32 %src0, i32 %tidx, i32 %src2, i1 false, i1 false)
-  store i32 %v, ptr addrspace(1) %out
+  %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
 }
 
-define amdgpu_kernel void @v_permlane16_b32_vsv(ptr addrspace(1) %out, i32 %src0, i32 %src1) {
-; GFX10-SDAG-LABEL: v_permlane16_b32_vsv:
-; 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]
+; 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_vsv:
+; 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_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_permlane16_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:    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_vsv:
+; 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:    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:    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_vsv:
+; 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_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_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_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_vsv:
+; 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:    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:    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_vsv:
+; 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_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_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_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
-  %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 4660, i32 49617, i1 false, i1 false)
+  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_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_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_permlane16_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_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_vss_fi:
+; GFX11-LABEL: v_permlane16_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_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_permlane16_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_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_vss_fi:
+; GFX12-LABEL: v_permlane16_b32_vll_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_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_permlane16_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_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 %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 4660, i32 49617, i1 false, i1 false)
+  store float %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:
-; 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
+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
 ;
-; GFX11-LABEL: v_permlane16_b32_vss_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_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
+; 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
 ;
-; GFX12-LABEL: v_permlane16_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_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 false, 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:
-; 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:[1,1]
-; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
-; GFX10-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-LABEL: v_permlane16_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_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
+; 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-LABEL: v_permlane16_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_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 true, i1 true)
-  store i32 %v, ptr addrspace(1) %out
+; 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_permlanex16_b32_vss(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-LABEL: v_permlanex16_b32_vss:
+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_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_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, s6
-; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s7, s2
-; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; 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-LABEL: v_permlanex16_b32_vss:
-; 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
-; 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
+; 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
 ;
-; GFX12-LABEL: v_permlanex16_b32_vss:
-; 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
-; 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 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:
-; 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-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
 ;
-; GFX11-LABEL: v_permlanex16_b32_vii:
-; 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-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-LABEL: v_permlanex16_b32_vii:
-; 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 %src0, i32 %src0, i32 1, i32 2, i1 false, i1 false)
+; 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
 }
 
-; 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
+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_permlanex16_b32_vll:
+; GFX11-LABEL: v_permlane16_b32_vvv_i64:
 ; 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[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_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:    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_permlanex16_b32_vll:
+; GFX12-LABEL: v_permlane16_b32_vvv_i64:
 ; GFX12:       ; %bb.0:
-; GFX12-NEXT:    s_load_b96 s[0:2], s[0:1], 0x24
+; 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_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:    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
-  %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
+  %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_permlanex16_b32_vvv(ptr addrspace(1) %out, i32 %src0) {
-; GFX10-LABEL: v_permlanex16_b32_vvv:
+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
@@ -694,11 +1017,11 @@ define amdgpu_kernel void @v_permlanex16_b32_vvv(ptr addrspace(1) %out, i32 %src
 ; 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:    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_permlanex16_b32_vvv:
+; 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
@@ -712,13 +1035,13 @@ define amdgpu_kernel void @v_permlanex16_b32_vvv(ptr addrspace(1) %out, i32 %src
 ; 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:    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_permlanex16_b32_vvv:
+; 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
@@ -731,13 +1054,13 @@ define amdgpu_kernel void @v_permlanex16_b32_vvv(ptr addrspace(1) %out, i32 %src
 ; 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:    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_permlanex16_b32_vvv:
+; 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
@@ -749,13 +1072,13 @@ define amdgpu_kernel void @v_permlanex16_b32_vvv(ptr addrspace(1) %out, i32 %src
 ; 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:    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_permlanex16_b32_vvv:
+; 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
@@ -766,42 +1089,117 @@ define amdgpu_kernel void @v_permlanex16_b32_vvv(ptr addrspace(1) %out, i32 %src
 ; 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:    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.permlanex16(i32 %src0, i32 %src0, i32 %tidx, i32 %tidy, i1 false, i1 false)
-  store i32 %v, ptr addrspace(1) %out
+  %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_permlanex16_b32_vvs(ptr addrspace(1) %out, i32 %src0, i32 %src2) {
-; GFX10-SDAG-LABEL: v_permlanex16_b32_vvs:
+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_permlanex16_b32 v1, v1, s2, s3
+; 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_permlanex16_b32_vvs:
+; 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_permlanex16_b32 v0, v0, s4, s3
+; 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_permlanex16_b32_vvs:
+; 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)
@@ -809,26 +1207,26 @@ define amdgpu_kernel void @v_permlanex16_b32_vvs(ptr addrspace(1) %out, i32 %src
 ; 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:    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_permlanex16_b32_vvs:
+; 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_permlanex16_b32 v0, v0, s4, s3
+; 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_permlanex16_b32_vvs:
+; 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
@@ -836,863 +1234,8803 @@ define amdgpu_kernel void @v_permlanex16_b32_vvs(ptr addrspace(1) %out, i32 %src
 ; 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:    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_permlanex16_b32_vvs:
+; 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_permlanex16_b32 v0, v0, s4, s3
+; 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.permlanex16(i32 %src0, i32 %src0, i32 %tidx, i32 %src2, i1 false, i1 false)
+  %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_permlanex16_b32_vsv(ptr addrspace(1) %out, i32 %src0, i32 %src1) {
-; GFX10-SDAG-LABEL: v_permlanex16_b32_vsv:
+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_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_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 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_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_permlanex16_b32_vsv:
+; GFX10-GISEL-LABEL: v_permlane16_b32_vvs_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_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, 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 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_permlanex16_b32_vsv:
+; GFX11-SDAG-LABEL: v_permlane16_b32_vvs_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_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_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, 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_permlanex16_b32_vsv:
+; GFX11-GISEL-LABEL: v_permlane16_b32_vvs_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_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_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, 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_permlanex16_b32_vsv:
+; GFX12-SDAG-LABEL: v_permlane16_b32_vvs_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_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_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:    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_permlanex16_b32_vsv:
+; 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(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_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_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_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:[1,0]
+; 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_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:[1,0]
+; 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 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_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_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_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:[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_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:[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_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:[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_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:[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_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:[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 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_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
+; 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_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:[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_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:[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.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_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
+; 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:[1,1]
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; 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
+; 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:[1,1]
+; 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_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:[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(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_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
+; 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:[1,1]
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; 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
+; 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:[1,1]
+; 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_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_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 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
+}
+
+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
+; 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
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; 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
+; 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
+; 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_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
+; 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 false)
+  store i32 %v, ptr addrspace(1) %out
+  ret void
+}
+
+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_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
+; GFX10-NEXT:    global_store_dword v1, v0, s[4:5]
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: v_permlanex16_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_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_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_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 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
+}
+
+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
+;
+; 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_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_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_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_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_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_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_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_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
+  %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_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_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
+;
+; GFX10-GISEL-LABEL: v_permlanex16_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_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_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_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_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_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_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_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_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_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
+  %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_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_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_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_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_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_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_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_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_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_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_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_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 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
+}
+
+; 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_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_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_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_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_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(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_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_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_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_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_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_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 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_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_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
+;
+; GFX10-GISEL-LABEL: v_permlanex16_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_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_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_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_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_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_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_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_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_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 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_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
+;
+; 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_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_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:    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_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
-  %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)
+  %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_vss_fi(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-LABEL: v_permlanex16_b32_vss_fi:
+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_dwordx4 s[4:7], s[0:1], 0x24
-; GFX10-NEXT:    s_load_dword s2, s[0:1], 0x34
+; 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_mov_b32_e32 v0, s6
-; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s7, s2 op_sel:[1,0]
+; 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_vss_fi:
+; GFX11-LABEL: v_permlanex16_b32_i_tid_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_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_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_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_vss_fi:
-; 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 %src0, i32 %src0, i32 %src1, i32 %src2, i1 true, i1 false)
-  store i32 %v, ptr addrspace(1) %out
+; 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_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_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_dwordx4 s[4:7], s[0:1], 0x24
-; GFX10-NEXT:    s_load_dword s2, s[0:1], 0x34
+; 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_mov_b32_e32 v0, s6
-; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s7, s2 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_permlanex16_b32_vss_bc:
+; GFX11-LABEL: v_permlanex16_b32_i_tid_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_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_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:    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_vss_bc:
+; GFX12-LABEL: v_permlanex16_b32_i_tid_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_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_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:    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
-  %v = call i32 @llvm.amdgcn.permlanex16(i32 %src0, i32 %src0, i32 %src1, i32 %src2, i1 false, i1 true)
+  %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 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:
+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_dwordx4 s[4:7], s[0:1], 0x24
-; GFX10-NEXT:    s_load_dword s2, s[0:1], 0x34
+; 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_mov_b32_e32 v0, s6
-; GFX10-NEXT:    v_permlanex16_b32 v0, v0, s7, s2 op_sel:[1,1]
+; 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_permlanex16_b32_vss_fi_bc:
+; GFX11-LABEL: v_permlanex16_b32_i_tid_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_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_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:    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_vss_fi_bc:
+; GFX12-LABEL: v_permlanex16_b32_i_tid_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_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_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:    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
-  %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
+  %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 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_tid_tid(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-LABEL: v_permlane16_b32_tid_tid:
+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
+; 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_tid_tid:
+; 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
+; 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_tid_tid:
+; 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
+; 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.permlane16(i32 %tidx, i32 %tidx, i32 %src1, i32 %src2, i1 false, i1 false)
+  %undef = freeze i32 poison
+  %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_permlane16_b32_undef_tid(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-LABEL: v_permlane16_b32_undef_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_permlane16_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_permlane16_b32_undef_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_permlane16_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_permlane16_b32_undef_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
+; 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 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.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()
-  %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
+  %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_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_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 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:    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_permlane16_b32_i_tid:
+; 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 v1, 0x3039
+; 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_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_permlane16_b32_i_tid:
+; 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_dual_mov_b32 v1, 0x3039 :: v_dual_mov_b32 v2, 0
+; 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, v0, s2, s3
-; GFX11-SDAG-NEXT:    global_store_b32 v2, v1, s[0:1]
+; 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_permlane16_b32_i_tid:
+; 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 v1, 0x3039
+; 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:    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_permlane16_b32_i_tid:
+; 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_dual_mov_b32 v1, 0x3039 :: v_dual_mov_b32 v2, 0
+; 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, v0, s2, s3
-; GFX12-SDAG-NEXT:    global_store_b32 v2, v1, s[0:1]
+; 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_permlane16_b32_i_tid:
+; 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 v1, 0x3039
+; 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:    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()
-  %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
+  %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_permlane16_b32_i_tid_fi(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-LABEL: v_permlane16_b32_i_tid_fi:
+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_permlane16_b32 v0, v0, s2, s3 op_sel:[1,0]
-; 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_b32_i_tid_fi:
+; GFX11-LABEL: v_permlane16_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_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
+; 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_b32_i_tid_fi:
+; GFX12-LABEL: v_permlane16_half:
 ; 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_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 %undef, i32 %tidx, i32 %src1, i32 %src2, i1 true, 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_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 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 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_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
+; 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_permlane16_b32_i_tid_bc:
+; 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_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
+; 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_permlane16_b32_i_tid_bc:
+; GFX12-LABEL: v_permlanex16_half:
 ; 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_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 %undef, i32 %tidx, i32 %src1, i32 %src2, i1 false, i1 true)
-  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_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 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 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_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
+; 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_b32_i_tid_fi_bc:
+; GFX11-LABEL: v_permlane16_bfloat:
 ; 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
+; 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_b32_i_tid_fi_bc:
+; GFX12-LABEL: v_permlane16_bfloat:
 ; 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_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 %undef, i32 %tidx, i32 %src1, i32 %src2, i1 true, i1 true)
-  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_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 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 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_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_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_b32_tid_tid:
+; GFX11-LABEL: v_permlanex16_bfloat:
 ; 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_b32_tid_tid:
+; GFX12-LABEL: v_permlanex16_bfloat:
 ; 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()
-  %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
+; 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 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_i16(ptr addrspace(1) %out, i16 %src0, i32 %src1, i32 %src2) {
+; GFX10-LABEL: v_permlane16_i16:
 ; 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_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_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_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 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_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_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_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]
+;
+; GFX11-GISEL-LABEL: v_permlane16_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_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_permlane16_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_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_permlane16_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_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 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]
 ;
-; GFX11-LABEL: v_permlanex16_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_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
+; 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]
 ;
-; GFX12-LABEL: v_permlanex16_b32_undef_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_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
+; 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-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-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(ptr addrspace(1) %out, i32 %src0, i32 %src1, i32 %src2) {
-; GFX10-SDAG-LABEL: v_permlanex16_b32_i_tid:
+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_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-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_permlanex16_b32_i_tid:
+; GFX10-GISEL-LABEL: v_permlane16_v7i32:
 ; 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
+; 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_permlanex16_b32_i_tid:
+; GFX11-SDAG-LABEL: v_permlane16_v7i32:
 ; 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, 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 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 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-GISEL-LABEL: v_permlanex16_b32_i_tid:
+; GFX11-GISEL-LABEL: v_permlane16_v7i32:
 ; 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, 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 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:    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-SDAG-LABEL: v_permlanex16_b32_i_tid:
+; GFX12-SDAG-LABEL: v_permlane16_v7i32:
 ; 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, 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 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 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_permlanex16_b32_i_tid:
+; GFX12-GISEL-LABEL: v_permlane16_v7i32:
 ; 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, 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 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:    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(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_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]
 ;
-; 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
+; 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]
 ;
-; 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
+; 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-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-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 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_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]
 ;
-; 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
+; 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]
 ;
-; 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
+; 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 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_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]
 ;
-; 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
+; 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]
 ;
-; 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
+; 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
+}

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 9ed62d3545f07..13d13bca56de6 100644
--- a/llvm/test/Verifier/AMDGPU/intrinsic-immarg.ll
+++ b/llvm/test/Verifier/AMDGPU/intrinsic-immarg.ll
@@ -499,31 +499,31 @@ 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
 }
 


        


More information about the llvm-commits mailing list