[llvm-branch-commits] [llvm] ae25a39 - AMDGPU/GlobalISel: Enable sret demotion

Christudasan Devadasan via llvm-branch-commits llvm-branch-commits at lists.llvm.org
Thu Jan 7 21:31:23 PST 2021


Author: Christudasan Devadasan
Date: 2021-01-08T10:56:35+05:30
New Revision: ae25a397e9de833ffbd5d8e3b480086404625cb7

URL: https://github.com/llvm/llvm-project/commit/ae25a397e9de833ffbd5d8e3b480086404625cb7
DIFF: https://github.com/llvm/llvm-project/commit/ae25a397e9de833ffbd5d8e3b480086404625cb7.diff

LOG: AMDGPU/GlobalISel: Enable sret demotion

Added: 
    

Modified: 
    llvm/include/llvm/CodeGen/GlobalISel/CallLowering.h
    llvm/lib/CodeGen/GlobalISel/CallLowering.cpp
    llvm/lib/Target/AMDGPU/AMDGPUCallLowering.cpp
    llvm/lib/Target/AMDGPU/AMDGPUCallLowering.h
    llvm/test/CodeGen/AMDGPU/GlobalISel/function-returns.ll
    llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-return-values.ll

Removed: 
    


################################################################################
diff  --git a/llvm/include/llvm/CodeGen/GlobalISel/CallLowering.h b/llvm/include/llvm/CodeGen/GlobalISel/CallLowering.h
index dff73d185114..57ff3900ef25 100644
--- a/llvm/include/llvm/CodeGen/GlobalISel/CallLowering.h
+++ b/llvm/include/llvm/CodeGen/GlobalISel/CallLowering.h
@@ -358,8 +358,8 @@ class CallLowering {
   /// described by \p Outs can fit into the return registers. If false
   /// is returned, an sret-demotion is performed.
   virtual bool canLowerReturn(MachineFunction &MF, CallingConv::ID CallConv,
-                              SmallVectorImpl<BaseArgInfo> &Outs, bool IsVarArg,
-                              LLVMContext &Context) const {
+                              SmallVectorImpl<BaseArgInfo> &Outs,
+                              bool IsVarArg) const {
     return true;
   }
 

diff  --git a/llvm/lib/CodeGen/GlobalISel/CallLowering.cpp b/llvm/lib/CodeGen/GlobalISel/CallLowering.cpp
index e1591a4bf19b..6666a6d4ea76 100644
--- a/llvm/lib/CodeGen/GlobalISel/CallLowering.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/CallLowering.cpp
@@ -95,8 +95,7 @@ bool CallLowering::lowerCall(MachineIRBuilder &MIRBuilder, const CallBase &CB,
 
   SmallVector<BaseArgInfo, 4> SplitArgs;
   getReturnInfo(CallConv, RetTy, CB.getAttributes(), SplitArgs, DL);
-  Info.CanLowerReturn =
-      canLowerReturn(MF, CallConv, SplitArgs, IsVarArg, RetTy->getContext());
+  Info.CanLowerReturn = canLowerReturn(MF, CallConv, SplitArgs, IsVarArg);
 
   if (!Info.CanLowerReturn) {
     // Callee requires sret demotion.
@@ -592,8 +591,7 @@ bool CallLowering::checkReturnTypeForCallConv(MachineFunction &MF) const {
   SmallVector<BaseArgInfo, 4> SplitArgs;
   getReturnInfo(CallConv, ReturnType, F.getAttributes(), SplitArgs,
                 MF.getDataLayout());
-  return canLowerReturn(MF, CallConv, SplitArgs, F.isVarArg(),
-                        ReturnType->getContext());
+  return canLowerReturn(MF, CallConv, SplitArgs, F.isVarArg());
 }
 
 bool CallLowering::analyzeArgInfo(CCState &CCState,

diff  --git a/llvm/lib/Target/AMDGPU/AMDGPUCallLowering.cpp b/llvm/lib/Target/AMDGPU/AMDGPUCallLowering.cpp
index 3b6e263ee6d8..b86052e3a14e 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUCallLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUCallLowering.cpp
@@ -20,6 +20,7 @@
 #include "SIMachineFunctionInfo.h"
 #include "SIRegisterInfo.h"
 #include "llvm/CodeGen/Analysis.h"
+#include "llvm/CodeGen/FunctionLoweringInfo.h"
 #include "llvm/CodeGen/GlobalISel/MachineIRBuilder.h"
 #include "llvm/IR/IntrinsicsAMDGPU.h"
 
@@ -420,6 +421,22 @@ static void unpackRegsToOrigType(MachineIRBuilder &B,
   B.buildUnmerge(UnmergeResults, UnmergeSrc);
 }
 
+bool AMDGPUCallLowering::canLowerReturn(MachineFunction &MF,
+                                        CallingConv::ID CallConv,
+                                        SmallVectorImpl<BaseArgInfo> &Outs,
+                                        bool IsVarArg) const {
+  // For shaders. Vector types should be explicitly handled by CC.
+  if (AMDGPU::isEntryFunctionCC(CallConv))
+    return true;
+
+  SmallVector<CCValAssign, 16> ArgLocs;
+  const SITargetLowering &TLI = *getTLI<SITargetLowering>();
+  CCState CCInfo(CallConv, IsVarArg, MF, ArgLocs,
+                 MF.getFunction().getContext());
+
+  return checkReturn(CCInfo, Outs, TLI.CCAssignFnForReturn(CallConv, IsVarArg));
+}
+
 /// Lower the return value for the already existing \p Ret. This assumes that
 /// \p B's insertion point is correct.
 bool AMDGPUCallLowering::lowerReturnVal(MachineIRBuilder &B,
@@ -533,7 +550,9 @@ bool AMDGPUCallLowering::lowerReturn(MachineIRBuilder &B, const Value *Val,
     Ret.addUse(ReturnAddrVReg);
   }
 
-  if (!lowerReturnVal(B, Val, VRegs, Ret))
+  if (!FLI.CanLowerReturn)
+    insertSRetStores(B, Val->getType(), VRegs, FLI.DemoteRegister);
+  else if (!lowerReturnVal(B, Val, VRegs, Ret))
     return false;
 
   if (ReturnOpc == AMDGPU::S_SETPC_B64_return) {
@@ -872,6 +891,11 @@ bool AMDGPUCallLowering::lowerFormalArguments(
   unsigned Idx = 0;
   unsigned PSInputNum = 0;
 
+  // Insert the hidden sret parameter if the return value won't fit in the
+  // return registers.
+  if (!FLI.CanLowerReturn)
+    insertSRetIncomingArgument(F, SplitArgs, FLI.DemoteRegister, MRI, DL);
+
   for (auto &Arg : F.args()) {
     if (DL.getTypeStoreSize(Arg.getType()) == 0)
       continue;
@@ -1327,7 +1351,10 @@ bool AMDGPUCallLowering::lowerCall(MachineIRBuilder &MIRBuilder,
     MIRBuilder.buildInstr(AMDGPU::ADJCALLSTACKDOWN);
 
   SmallVector<ArgInfo, 8> InArgs;
-  if (!Info.OrigRet.Ty->isVoidTy()) {
+  if (!Info.CanLowerReturn) {
+    insertSRetLoads(MIRBuilder, Info.OrigRet.Ty, Info.OrigRet.Regs,
+                    Info.DemoteRegister, Info.DemoteStackIndex);
+  } else if (!Info.OrigRet.Ty->isVoidTy()) {
     SmallVector<ArgInfo, 8> PreSplitRetInfos;
 
     splitToValueTypes(
@@ -1350,7 +1377,7 @@ bool AMDGPUCallLowering::lowerCall(MachineIRBuilder &MIRBuilder,
   // Finally we can copy the returned value back into its virtual-register. In
   // symmetry with the arguments, the physical register must be an
   // implicit-define of the call instruction.
-  if (!Info.OrigRet.Ty->isVoidTy()) {
+  if (Info.CanLowerReturn && !Info.OrigRet.Ty->isVoidTy()) {
     CCAssignFn *RetAssignFn = TLI.CCAssignFnForReturn(Info.CallConv,
                                                       Info.IsVarArg);
     CallReturnHandler Handler(MIRBuilder, MRI, MIB, RetAssignFn);

diff  --git a/llvm/lib/Target/AMDGPU/AMDGPUCallLowering.h b/llvm/lib/Target/AMDGPU/AMDGPUCallLowering.h
index 0026b5fbf24b..1312388e4a38 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUCallLowering.h
+++ b/llvm/lib/Target/AMDGPU/AMDGPUCallLowering.h
@@ -42,6 +42,10 @@ class AMDGPUCallLowering final : public CallLowering {
                         bool IsOutgoing,
                         SplitArgTy PerformArgSplit) const;
 
+  bool canLowerReturn(MachineFunction &MF, CallingConv::ID CallConv,
+                      SmallVectorImpl<BaseArgInfo> &Outs,
+                      bool IsVarArg) const override;
+
   bool lowerReturnVal(MachineIRBuilder &B, const Value *Val,
                       ArrayRef<Register> VRegs, MachineInstrBuilder &Ret) const;
 

diff  --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/function-returns.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/function-returns.ll
index b340be5ac2c8..3cd0e0fdc4ac 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/function-returns.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/function-returns.ll
@@ -1172,34 +1172,63 @@ define void @void_func_sret_struct_i8_i32({ i8, i32 } addrspace(5)* sret({ i8, i
 
 define <33 x i32> @v33i32_func_void() #0 {
   ; CHECK-LABEL: name: v33i32_func_void
-  ; CHECK: bb.0:
-  ; CHECK:   successors: %bb.1(0x80000000)
-  ; CHECK:   liveins: $sgpr30_sgpr31
-  ; CHECK:   [[COPY:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
-  ; CHECK:   [[DEF:%[0-9]+]]:_(p4) = G_IMPLICIT_DEF
   ; CHECK: bb.1 (%ir-block.0):
+  ; CHECK:   liveins: $vgpr0, $sgpr30_sgpr31
+  ; CHECK:   [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+  ; CHECK:   [[COPY1:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
+  ; CHECK:   [[DEF:%[0-9]+]]:_(p4) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[DEF]](p4) :: (volatile load 8 from `<33 x i32> addrspace(1)* addrspace(4)* undef`, addrspace 4)
   ; CHECK:   [[LOAD1:%[0-9]+]]:_(<33 x s32>) = G_LOAD [[LOAD]](p1) :: (load 132 from %ir.ptr, align 256, addrspace 1)
-  ; CHECK:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32), [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32), [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32), [[UV16:%[0-9]+]]:_(s32), [[UV17:%[0-9]+]]:_(s32), [[UV18:%[0-9]+]]:_(s32), [[UV19:%[0-9]+]]:_(s32), [[UV20:%[0-9]+]]:_(s32), [[UV21:%[0-9]+]]:_(s32), [[UV22:%[0-9]+]]:_(s32), [[UV23:%[0-9]+]]:_(s32), [[UV24:%[0-9]+]]:_(s32), [[UV25:%[0-9]+]]:_(s32), [[UV26:%[0-9]+]]:_(s32), [[UV27:%[0-9]+]]:_(s32), [[UV28:%[0-9]+]]:_(s32), [[UV29:%[0-9]+]]:_(s32), [[UV30:%[0-9]+]]:_(s32), [[UV31:%[0-9]+]]:_(s32), [[UV32:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<33 x s32>)
+  ; CHECK:   G_STORE [[LOAD1]](<33 x s32>), [[COPY]](p5) :: (store 132, align 256, addrspace 5)
+  ; CHECK:   [[COPY2:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY1]]
+  ; CHECK:   S_SETPC_B64_return [[COPY2]]
   %ptr = load volatile <33 x i32> addrspace(1)*, <33 x i32> addrspace(1)* addrspace(4)* undef
   %val = load <33 x i32>, <33 x i32> addrspace(1)* %ptr
   ret <33 x i32> %val
 }
 
+define <33 x i32> @v33i32_func_v33i32_i32(<33 x i32> addrspace(1)* %p, i32 %idx) #0 {
+  ; CHECK-LABEL: name: v33i32_func_v33i32_i32
+  ; CHECK: bb.1 (%ir-block.0):
+  ; CHECK:   liveins: $vgpr0, $vgpr1, $vgpr2, $vgpr3, $sgpr30_sgpr31
+  ; CHECK:   [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+  ; CHECK:   [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
+  ; CHECK:   [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr2
+  ; CHECK:   [[COPY3:%[0-9]+]]:_(s32) = COPY $vgpr3
+  ; CHECK:   [[COPY4:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
+  ; CHECK:   [[MV:%[0-9]+]]:_(p1) = G_MERGE_VALUES [[COPY1]](s32), [[COPY2]](s32)
+  ; CHECK:   [[SEXT:%[0-9]+]]:_(s64) = G_SEXT [[COPY3]](s32)
+  ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 256
+  ; CHECK:   [[MUL:%[0-9]+]]:_(s64) = G_MUL [[SEXT]], [[C]]
+  ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[MV]], [[MUL]](s64)
+  ; CHECK:   [[COPY5:%[0-9]+]]:_(p1) = COPY [[PTR_ADD]](p1)
+  ; CHECK:   [[LOAD:%[0-9]+]]:_(<33 x s32>) = G_LOAD [[COPY5]](p1) :: (load 132 from %ir.gep, align 256, addrspace 1)
+  ; CHECK:   G_STORE [[LOAD]](<33 x s32>), [[COPY]](p5) :: (store 132, align 256, addrspace 5)
+  ; CHECK:   [[COPY6:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY4]]
+  ; CHECK:   S_SETPC_B64_return [[COPY6]]
+  %gep = getelementptr inbounds <33 x i32>, <33 x i32> addrspace(1)* %p, i32 %idx
+  %val = load <33 x i32>, <33 x i32> addrspace(1)* %gep
+  ret <33 x i32> %val
+}
+
 define { <32 x i32>, i32 } @struct_v32i32_i32_func_void() #0 {
   ; CHECK-LABEL: name: struct_v32i32_i32_func_void
-  ; CHECK: bb.0:
-  ; CHECK:   successors: %bb.1(0x80000000)
-  ; CHECK:   liveins: $sgpr30_sgpr31
-  ; CHECK:   [[COPY:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
-  ; CHECK:   [[DEF:%[0-9]+]]:_(p4) = G_IMPLICIT_DEF
   ; CHECK: bb.1 (%ir-block.0):
+  ; CHECK:   liveins: $vgpr0, $sgpr30_sgpr31
+  ; CHECK:   [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+  ; CHECK:   [[COPY1:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
+  ; CHECK:   [[DEF:%[0-9]+]]:_(p4) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[DEF]](p4) :: (volatile load 8 from `{ <32 x i32>, i32 } addrspace(1)* addrspace(4)* undef`, addrspace 4)
   ; CHECK:   [[LOAD1:%[0-9]+]]:_(<32 x s32>) = G_LOAD [[LOAD]](p1) :: (load 128 from %ir.ptr, addrspace 1)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 128
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[LOAD]], [[C]](s64)
   ; CHECK:   [[LOAD2:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p1) :: (load 4 from %ir.ptr + 128, align 128, addrspace 1)
-  ; CHECK:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32), [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32), [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32), [[UV16:%[0-9]+]]:_(s32), [[UV17:%[0-9]+]]:_(s32), [[UV18:%[0-9]+]]:_(s32), [[UV19:%[0-9]+]]:_(s32), [[UV20:%[0-9]+]]:_(s32), [[UV21:%[0-9]+]]:_(s32), [[UV22:%[0-9]+]]:_(s32), [[UV23:%[0-9]+]]:_(s32), [[UV24:%[0-9]+]]:_(s32), [[UV25:%[0-9]+]]:_(s32), [[UV26:%[0-9]+]]:_(s32), [[UV27:%[0-9]+]]:_(s32), [[UV28:%[0-9]+]]:_(s32), [[UV29:%[0-9]+]]:_(s32), [[UV30:%[0-9]+]]:_(s32), [[UV31:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD1]](<32 x s32>)
+  ; CHECK:   G_STORE [[LOAD1]](<32 x s32>), [[COPY]](p5) :: (store 128, addrspace 5)
+  ; CHECK:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 128
+  ; CHECK:   [[PTR_ADD1:%[0-9]+]]:_(p5) = G_PTR_ADD [[COPY]], [[C1]](s32)
+  ; CHECK:   G_STORE [[LOAD2]](s32), [[PTR_ADD1]](p5) :: (store 4, align 128, addrspace 5)
+  ; CHECK:   [[COPY2:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY1]]
+  ; CHECK:   S_SETPC_B64_return [[COPY2]]
   %ptr = load volatile { <32 x i32>, i32 } addrspace(1)*, { <32 x i32>, i32 } addrspace(1)* addrspace(4)* undef
   %val = load { <32 x i32>, i32 }, { <32 x i32>, i32 } addrspace(1)* %ptr
   ret { <32 x i32>, i32 }%val
@@ -1207,18 +1236,22 @@ define { <32 x i32>, i32 } @struct_v32i32_i32_func_void() #0 {
 
 define { i32, <32 x i32> } @struct_i32_v32i32_func_void() #0 {
   ; CHECK-LABEL: name: struct_i32_v32i32_func_void
-  ; CHECK: bb.0:
-  ; CHECK:   successors: %bb.1(0x80000000)
-  ; CHECK:   liveins: $sgpr30_sgpr31
-  ; CHECK:   [[COPY:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
-  ; CHECK:   [[DEF:%[0-9]+]]:_(p4) = G_IMPLICIT_DEF
   ; CHECK: bb.1 (%ir-block.0):
+  ; CHECK:   liveins: $vgpr0, $sgpr30_sgpr31
+  ; CHECK:   [[COPY:%[0-9]+]]:_(p5) = COPY $vgpr0
+  ; CHECK:   [[COPY1:%[0-9]+]]:sgpr_64 = COPY $sgpr30_sgpr31
+  ; CHECK:   [[DEF:%[0-9]+]]:_(p4) = G_IMPLICIT_DEF
   ; CHECK:   [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[DEF]](p4) :: (volatile load 8 from `{ i32, <32 x i32> } addrspace(1)* addrspace(4)* undef`, addrspace 4)
   ; CHECK:   [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[LOAD]](p1) :: (load 4 from %ir.ptr, align 128, addrspace 1)
   ; CHECK:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 128
   ; CHECK:   [[PTR_ADD:%[0-9]+]]:_(p1) = G_PTR_ADD [[LOAD]], [[C]](s64)
   ; CHECK:   [[LOAD2:%[0-9]+]]:_(<32 x s32>) = G_LOAD [[PTR_ADD]](p1) :: (load 128 from %ir.ptr + 128, addrspace 1)
-  ; CHECK:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32), [[UV2:%[0-9]+]]:_(s32), [[UV3:%[0-9]+]]:_(s32), [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32), [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32), [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32), [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32), [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32), [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32), [[UV16:%[0-9]+]]:_(s32), [[UV17:%[0-9]+]]:_(s32), [[UV18:%[0-9]+]]:_(s32), [[UV19:%[0-9]+]]:_(s32), [[UV20:%[0-9]+]]:_(s32), [[UV21:%[0-9]+]]:_(s32), [[UV22:%[0-9]+]]:_(s32), [[UV23:%[0-9]+]]:_(s32), [[UV24:%[0-9]+]]:_(s32), [[UV25:%[0-9]+]]:_(s32), [[UV26:%[0-9]+]]:_(s32), [[UV27:%[0-9]+]]:_(s32), [[UV28:%[0-9]+]]:_(s32), [[UV29:%[0-9]+]]:_(s32), [[UV30:%[0-9]+]]:_(s32), [[UV31:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD2]](<32 x s32>)
+  ; CHECK:   G_STORE [[LOAD1]](s32), [[COPY]](p5) :: (store 4, align 128, addrspace 5)
+  ; CHECK:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 128
+  ; CHECK:   [[PTR_ADD1:%[0-9]+]]:_(p5) = G_PTR_ADD [[COPY]], [[C1]](s32)
+  ; CHECK:   G_STORE [[LOAD2]](<32 x s32>), [[PTR_ADD1]](p5) :: (store 128, addrspace 5)
+  ; CHECK:   [[COPY2:%[0-9]+]]:ccr_sgpr_64 = COPY [[COPY1]]
+  ; CHECK:   S_SETPC_B64_return [[COPY2]]
   %ptr = load volatile { i32, <32 x i32> } addrspace(1)*, { i32, <32 x i32> } addrspace(1)* addrspace(4)* undef
   %val = load { i32, <32 x i32> }, { i32, <32 x i32> } addrspace(1)* %ptr
   ret { i32, <32 x i32> }%val

diff  --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-return-values.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-return-values.ll
index 0b238c78486d..7699738afebb 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-return-values.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/irtranslator-call-return-values.ll
@@ -46,7 +46,10 @@ declare <5 x i32> @external_v5i32_func_void() #0
 declare <8 x i32> @external_v8i32_func_void() #0
 declare <16 x i32> @external_v16i32_func_void() #0
 declare <32 x i32> @external_v32i32_func_void() #0
+declare <33 x i32> @external_v33i32_func_void() #0
+declare <33 x i32> @external_v33i32_func_v33i32_i32(<33 x i32> addrspace(1)*, i32) #0
 declare { <32 x i32>, i32 } @external_v32i32_i32_func_void() #0
+declare { i32, <32 x i32> } @external_i32_v32i32_func_void() #0
 
 declare { i32, i64 } @external_i32_i64_func_void() #0
 
@@ -2751,6 +2754,259 @@ define amdgpu_kernel void @test_call_external_a5i8_func_void() #0 {
   ret void
 }
 
+define amdgpu_kernel void @test_call_external_v32i32_i32_func_void() #0 {
+  ; GCN-LABEL: name: test_call_external_v32i32_i32_func_void
+  ; GCN: bb.1 (%ir-block.0):
+  ; GCN:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
+  ; GCN:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
+  ; GCN:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
+  ; GCN:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
+  ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
+  ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
+  ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
+  ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
+  ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
+  ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
+  ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
+  ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
+  ; GCN:   [[COPY10:%[0-9]+]]:_(p1) = COPY [[DEF]](p1)
+  ; GCN:   [[FRAME_INDEX:%[0-9]+]]:_(p5) = G_FRAME_INDEX %stack.0
+  ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
+  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_v32i32_i32_func_void
+  ; GCN:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY8]]
+  ; GCN:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY7]]
+  ; GCN:   [[COPY13:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
+  ; GCN:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
+  ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY13]], [[C]](s64)
+  ; GCN:   [[COPY14:%[0-9]+]]:_(s64) = COPY [[COPY6]]
+  ; GCN:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY5]]
+  ; GCN:   [[COPY16:%[0-9]+]]:_(s32) = COPY [[COPY4]]
+  ; GCN:   [[COPY17:%[0-9]+]]:_(s32) = COPY [[COPY3]]
+  ; GCN:   [[COPY18:%[0-9]+]]:_(s32) = COPY [[COPY2]](s32)
+  ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
+  ; GCN:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 10
+  ; GCN:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[COPY19]], [[C1]](s32)
+  ; GCN:   [[OR:%[0-9]+]]:_(s32) = G_OR [[COPY18]], [[SHL]]
+  ; GCN:   [[COPY20:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
+  ; GCN:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
+  ; GCN:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY20]], [[C2]](s32)
+  ; GCN:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
+  ; GCN:   $vgpr0 = COPY [[FRAME_INDEX]](p5)
+  ; GCN:   [[COPY21:%[0-9]+]]:_(<4 x s32>) = COPY $private_rsrc_reg
+  ; GCN:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY21]](<4 x s32>)
+  ; GCN:   $sgpr4_sgpr5 = COPY [[COPY11]](p4)
+  ; GCN:   $sgpr6_sgpr7 = COPY [[COPY12]](p4)
+  ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
+  ; GCN:   $sgpr10_sgpr11 = COPY [[COPY14]](s64)
+  ; GCN:   $sgpr12 = COPY [[COPY15]](s32)
+  ; GCN:   $sgpr13 = COPY [[COPY16]](s32)
+  ; GCN:   $sgpr14 = COPY [[COPY17]](s32)
+  ; GCN:   $vgpr31 = COPY [[OR1]](s32)
+  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_v32i32_i32_func_void, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
+  ; GCN:   [[LOAD:%[0-9]+]]:_(<32 x s32>) = G_LOAD [[FRAME_INDEX]](p5) :: (load 128 from %stack.0, addrspace 5)
+  ; GCN:   [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 128
+  ; GCN:   [[PTR_ADD1:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C3]](s32)
+  ; GCN:   [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD1]](p5) :: (load 4 from %stack.0, align 128, addrspace 5)
+  ; GCN:   G_STORE [[LOAD]](<32 x s32>), [[DEF]](p1) :: (volatile store 128 into `<32 x i32> addrspace(1)* undef`, align 8, addrspace 1)
+  ; GCN:   G_STORE [[LOAD1]](s32), [[COPY10]](p1) :: (volatile store 4 into `i32 addrspace(1)* undef`, addrspace 1)
+  ; GCN:   S_ENDPGM 0
+  %val = call { <32 x i32>, i32 } @external_v32i32_i32_func_void()
+  %val0 = extractvalue { <32 x i32>, i32 } %val, 0
+  %val1 = extractvalue { <32 x i32>, i32 } %val, 1
+  store volatile <32 x i32> %val0, <32 x i32> addrspace(1)* undef, align 8
+  store volatile i32 %val1, i32 addrspace(1)* undef
+  ret void
+}
+
+define amdgpu_kernel void @test_call_external_i32_v32i32_func_void() #0 {
+  ; GCN-LABEL: name: test_call_external_i32_v32i32_func_void
+  ; GCN: bb.1 (%ir-block.0):
+  ; GCN:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
+  ; GCN:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
+  ; GCN:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
+  ; GCN:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
+  ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
+  ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
+  ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
+  ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
+  ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
+  ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
+  ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
+  ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
+  ; GCN:   [[COPY10:%[0-9]+]]:_(p1) = COPY [[DEF]](p1)
+  ; GCN:   [[FRAME_INDEX:%[0-9]+]]:_(p5) = G_FRAME_INDEX %stack.0
+  ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
+  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_i32_v32i32_func_void
+  ; GCN:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY8]]
+  ; GCN:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY7]]
+  ; GCN:   [[COPY13:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
+  ; GCN:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
+  ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY13]], [[C]](s64)
+  ; GCN:   [[COPY14:%[0-9]+]]:_(s64) = COPY [[COPY6]]
+  ; GCN:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY5]]
+  ; GCN:   [[COPY16:%[0-9]+]]:_(s32) = COPY [[COPY4]]
+  ; GCN:   [[COPY17:%[0-9]+]]:_(s32) = COPY [[COPY3]]
+  ; GCN:   [[COPY18:%[0-9]+]]:_(s32) = COPY [[COPY2]](s32)
+  ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
+  ; GCN:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 10
+  ; GCN:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[COPY19]], [[C1]](s32)
+  ; GCN:   [[OR:%[0-9]+]]:_(s32) = G_OR [[COPY18]], [[SHL]]
+  ; GCN:   [[COPY20:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
+  ; GCN:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
+  ; GCN:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY20]], [[C2]](s32)
+  ; GCN:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
+  ; GCN:   $vgpr0 = COPY [[FRAME_INDEX]](p5)
+  ; GCN:   [[COPY21:%[0-9]+]]:_(<4 x s32>) = COPY $private_rsrc_reg
+  ; GCN:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY21]](<4 x s32>)
+  ; GCN:   $sgpr4_sgpr5 = COPY [[COPY11]](p4)
+  ; GCN:   $sgpr6_sgpr7 = COPY [[COPY12]](p4)
+  ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
+  ; GCN:   $sgpr10_sgpr11 = COPY [[COPY14]](s64)
+  ; GCN:   $sgpr12 = COPY [[COPY15]](s32)
+  ; GCN:   $sgpr13 = COPY [[COPY16]](s32)
+  ; GCN:   $sgpr14 = COPY [[COPY17]](s32)
+  ; GCN:   $vgpr31 = COPY [[OR1]](s32)
+  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_i32_v32i32_func_void, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
+  ; GCN:   [[LOAD:%[0-9]+]]:_(s32) = G_LOAD [[FRAME_INDEX]](p5) :: (load 4 from %stack.0, align 128, addrspace 5)
+  ; GCN:   [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 128
+  ; GCN:   [[PTR_ADD1:%[0-9]+]]:_(p5) = G_PTR_ADD [[FRAME_INDEX]], [[C3]](s32)
+  ; GCN:   [[LOAD1:%[0-9]+]]:_(<32 x s32>) = G_LOAD [[PTR_ADD1]](p5) :: (load 128 from %stack.0, addrspace 5)
+  ; GCN:   G_STORE [[LOAD]](s32), [[DEF]](p1) :: (volatile store 4 into `i32 addrspace(1)* undef`, addrspace 1)
+  ; GCN:   G_STORE [[LOAD1]](<32 x s32>), [[COPY10]](p1) :: (volatile store 128 into `<32 x i32> addrspace(1)* undef`, align 8, addrspace 1)
+  ; GCN:   S_ENDPGM 0
+  %val = call { i32, <32 x i32> } @external_i32_v32i32_func_void()
+  %val0 = extractvalue { i32, <32 x i32> } %val, 0
+  %val1 = extractvalue { i32, <32 x i32> } %val, 1
+  store volatile i32 %val0, i32 addrspace(1)* undef
+  store volatile <32 x i32> %val1, <32 x i32> addrspace(1)* undef, align 8
+  ret void
+}
+
+define amdgpu_kernel void @test_call_external_v33i32_func_void() #0 {
+  ; GCN-LABEL: name: test_call_external_v33i32_func_void
+  ; GCN: bb.1 (%ir-block.0):
+  ; GCN:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
+  ; GCN:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
+  ; GCN:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
+  ; GCN:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
+  ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
+  ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
+  ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
+  ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
+  ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
+  ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
+  ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
+  ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
+  ; GCN:   [[FRAME_INDEX:%[0-9]+]]:_(p5) = G_FRAME_INDEX %stack.0
+  ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
+  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_v33i32_func_void
+  ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
+  ; GCN:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
+  ; GCN:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
+  ; GCN:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
+  ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C]](s64)
+  ; GCN:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
+  ; GCN:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
+  ; GCN:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
+  ; GCN:   [[COPY16:%[0-9]+]]:_(s32) = COPY [[COPY3]]
+  ; GCN:   [[COPY17:%[0-9]+]]:_(s32) = COPY [[COPY2]](s32)
+  ; GCN:   [[COPY18:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
+  ; GCN:   [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 10
+  ; GCN:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[COPY18]], [[C1]](s32)
+  ; GCN:   [[OR:%[0-9]+]]:_(s32) = G_OR [[COPY17]], [[SHL]]
+  ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
+  ; GCN:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
+  ; GCN:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY19]], [[C2]](s32)
+  ; GCN:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
+  ; GCN:   $vgpr0 = COPY [[FRAME_INDEX]](p5)
+  ; GCN:   [[COPY20:%[0-9]+]]:_(<4 x s32>) = COPY $private_rsrc_reg
+  ; GCN:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY20]](<4 x s32>)
+  ; GCN:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
+  ; GCN:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
+  ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD]](p4)
+  ; GCN:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
+  ; GCN:   $sgpr12 = COPY [[COPY14]](s32)
+  ; GCN:   $sgpr13 = COPY [[COPY15]](s32)
+  ; GCN:   $sgpr14 = COPY [[COPY16]](s32)
+  ; GCN:   $vgpr31 = COPY [[OR1]](s32)
+  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_v33i32_func_void, csr_amdgpu_highregs, implicit $vgpr0, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
+  ; GCN:   [[LOAD:%[0-9]+]]:_(<33 x s32>) = G_LOAD [[FRAME_INDEX]](p5) :: (load 132 from %stack.0, align 256, addrspace 5)
+  ; GCN:   G_STORE [[LOAD]](<33 x s32>), [[DEF]](p1) :: (volatile store 132 into `<33 x i32> addrspace(1)* undef`, align 8, addrspace 1)
+  ; GCN:   S_ENDPGM 0
+  %val = call <33 x i32> @external_v33i32_func_void()
+  store volatile <33 x i32> %val, <33 x i32> addrspace(1)* undef, align 8
+  ret void
+}
+
+define amdgpu_kernel void @test_call_external_v33i32_func_v33i32_i32(<33 x i32> addrspace(1)* %p, i32 %idx) #0 {
+  ; GCN-LABEL: name: test_call_external_v33i32_func_v33i32_i32
+  ; GCN: bb.1 (%ir-block.0):
+  ; GCN:   liveins: $sgpr14, $sgpr15, $sgpr16, $vgpr0, $vgpr1, $vgpr2, $sgpr4_sgpr5, $sgpr6_sgpr7, $sgpr8_sgpr9, $sgpr10_sgpr11
+  ; GCN:   [[COPY:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr2
+  ; GCN:   [[COPY1:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr1
+  ; GCN:   [[COPY2:%[0-9]+]]:vgpr_32(s32) = COPY $vgpr0
+  ; GCN:   [[COPY3:%[0-9]+]]:sgpr_32 = COPY $sgpr16
+  ; GCN:   [[COPY4:%[0-9]+]]:sgpr_32 = COPY $sgpr15
+  ; GCN:   [[COPY5:%[0-9]+]]:sgpr_32 = COPY $sgpr14
+  ; GCN:   [[COPY6:%[0-9]+]]:sgpr_64 = COPY $sgpr10_sgpr11
+  ; GCN:   [[COPY7:%[0-9]+]]:sgpr_64 = COPY $sgpr6_sgpr7
+  ; GCN:   [[COPY8:%[0-9]+]]:sgpr_64 = COPY $sgpr4_sgpr5
+  ; GCN:   [[COPY9:%[0-9]+]]:_(p4) = COPY $sgpr8_sgpr9
+  ; GCN:   [[DEF:%[0-9]+]]:_(p1) = G_IMPLICIT_DEF
+  ; GCN:   [[INT:%[0-9]+]]:_(p4) = G_INTRINSIC intrinsic(@llvm.amdgcn.kernarg.segment.ptr)
+  ; GCN:   [[LOAD:%[0-9]+]]:_(p1) = G_LOAD [[INT]](p4) :: (dereferenceable invariant load 8 from %ir.p.kernarg.offset.cast, align 16, addrspace 4)
+  ; GCN:   [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 8
+  ; GCN:   [[PTR_ADD:%[0-9]+]]:_(p4) = G_PTR_ADD [[INT]], [[C]](s64)
+  ; GCN:   [[LOAD1:%[0-9]+]]:_(s32) = G_LOAD [[PTR_ADD]](p4) :: (dereferenceable invariant load 4 from %ir.idx.kernarg.offset.cast, align 8, addrspace 4)
+  ; GCN:   [[FRAME_INDEX:%[0-9]+]]:_(p5) = G_FRAME_INDEX %stack.0
+  ; GCN:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[LOAD]](p1)
+  ; GCN:   ADJCALLSTACKUP 0, 0, implicit-def $scc
+  ; GCN:   [[GV:%[0-9]+]]:sreg_64(p0) = G_GLOBAL_VALUE @external_v33i32_func_v33i32_i32
+  ; GCN:   [[COPY10:%[0-9]+]]:_(p4) = COPY [[COPY8]]
+  ; GCN:   [[COPY11:%[0-9]+]]:_(p4) = COPY [[COPY7]]
+  ; GCN:   [[COPY12:%[0-9]+]]:_(p4) = COPY [[COPY9]](p4)
+  ; GCN:   [[C1:%[0-9]+]]:_(s64) = G_CONSTANT i64 16
+  ; GCN:   [[PTR_ADD1:%[0-9]+]]:_(p4) = G_PTR_ADD [[COPY12]], [[C1]](s64)
+  ; GCN:   [[COPY13:%[0-9]+]]:_(s64) = COPY [[COPY6]]
+  ; GCN:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[COPY5]]
+  ; GCN:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[COPY4]]
+  ; GCN:   [[COPY16:%[0-9]+]]:_(s32) = COPY [[COPY3]]
+  ; GCN:   [[COPY17:%[0-9]+]]:_(s32) = COPY [[COPY2]](s32)
+  ; GCN:   [[COPY18:%[0-9]+]]:_(s32) = COPY [[COPY1]](s32)
+  ; GCN:   [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 10
+  ; GCN:   [[SHL:%[0-9]+]]:_(s32) = G_SHL [[COPY18]], [[C2]](s32)
+  ; GCN:   [[OR:%[0-9]+]]:_(s32) = G_OR [[COPY17]], [[SHL]]
+  ; GCN:   [[COPY19:%[0-9]+]]:_(s32) = COPY [[COPY]](s32)
+  ; GCN:   [[C3:%[0-9]+]]:_(s32) = G_CONSTANT i32 20
+  ; GCN:   [[SHL1:%[0-9]+]]:_(s32) = G_SHL [[COPY19]], [[C3]](s32)
+  ; GCN:   [[OR1:%[0-9]+]]:_(s32) = G_OR [[OR]], [[SHL1]]
+  ; GCN:   $vgpr0 = COPY [[FRAME_INDEX]](p5)
+  ; GCN:   $vgpr1 = COPY [[UV]](s32)
+  ; GCN:   $vgpr2 = COPY [[UV1]](s32)
+  ; GCN:   $vgpr3 = COPY [[LOAD1]](s32)
+  ; GCN:   [[COPY20:%[0-9]+]]:_(<4 x s32>) = COPY $private_rsrc_reg
+  ; GCN:   $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY20]](<4 x s32>)
+  ; GCN:   $sgpr4_sgpr5 = COPY [[COPY10]](p4)
+  ; GCN:   $sgpr6_sgpr7 = COPY [[COPY11]](p4)
+  ; GCN:   $sgpr8_sgpr9 = COPY [[PTR_ADD1]](p4)
+  ; GCN:   $sgpr10_sgpr11 = COPY [[COPY13]](s64)
+  ; GCN:   $sgpr12 = COPY [[COPY14]](s32)
+  ; GCN:   $sgpr13 = COPY [[COPY15]](s32)
+  ; GCN:   $sgpr14 = COPY [[COPY16]](s32)
+  ; GCN:   $vgpr31 = COPY [[OR1]](s32)
+  ; GCN:   $sgpr30_sgpr31 = SI_CALL [[GV]](p0), @external_v33i32_func_v33i32_i32, csr_amdgpu_highregs, implicit $vgpr0, implicit $vgpr1, implicit $vgpr2, implicit $vgpr3, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $vgpr31
+  ; GCN:   ADJCALLSTACKDOWN 0, 0, implicit-def $scc
+  ; GCN:   [[LOAD2:%[0-9]+]]:_(<33 x s32>) = G_LOAD [[FRAME_INDEX]](p5) :: (load 132 from %stack.0, align 256, addrspace 5)
+  ; GCN:   G_STORE [[LOAD2]](<33 x s32>), [[DEF]](p1) :: (volatile store 132 into `<33 x i32> addrspace(1)* undef`, align 8, addrspace 1)
+  ; GCN:   S_ENDPGM 0
+  %val = call <33 x i32> @external_v33i32_func_v33i32_i32(<33 x i32> addrspace(1)* %p, i32 %idx)
+  store volatile <33 x i32> %val, <33 x i32> addrspace(1)* undef, align 8
+  ret void
+}
+
 attributes #0 = { nounwind }
 attributes #1 = { nounwind readnone }
 attributes #2 = { nounwind noinline }


        


More information about the llvm-branch-commits mailing list