Skip to content

Commit

Permalink
Merging r259558:
Browse files Browse the repository at this point in the history 
------------------------------------------------------------------------
r259558 | Matthew.Arsenault | 2016-02-02 12:28:10 -0800 (Tue, 02 Feb 2016) | 4 lines

AMDGPU: Handle promoting memmove

Also add missing tests for the others.

------------------------------------------------------------------------

git-svn-id: https://llvm.org/svn/llvm-project/llvm/branches/release_38@271593 91177308-0d34-0410-b5e6-96231b3b80d8
  • Loading branch information
@tstellarAMD
tstellarAMD committed Jun 2, 2016
1 parent c7c7ff0 commit 5a2d99b
Show file tree
Hide file tree
Showing 2 changed files with 89 additions and 0 deletions.
24 changes: 24 additions & 0 deletions lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -268,12 +268,14 @@ static bool isCallPromotable(CallInst *CI) {

switch (II->getIntrinsicID()) {
case Intrinsic::memcpy:
case Intrinsic::memmove:
case Intrinsic::memset:
case Intrinsic::lifetime_start:
case Intrinsic::lifetime_end:
case Intrinsic::invariant_start:
case Intrinsic::invariant_end:
case Intrinsic::invariant_group_barrier:
case Intrinsic::objectsize:
return true;
default:
return false;
Expand Down Expand Up @@ -450,6 +452,14 @@ void AMDGPUPromoteAlloca::visitAlloca(AllocaInst &I) {
Intr->eraseFromParent();
continue;
}
case Intrinsic::memmove: {
MemMoveInst *MemMove = cast<MemMoveInst>(Intr);
Builder.CreateMemMove(MemMove->getRawDest(), MemMove->getRawSource(),
MemMove->getLength(), MemMove->getAlignment(),
MemMove->isVolatile());
Intr->eraseFromParent();
continue;
}
case Intrinsic::memset: {
MemSetInst *MemSet = cast<MemSetInst>(Intr);
Builder.CreateMemSet(MemSet->getRawDest(), MemSet->getValue(),
Expand All @@ -466,6 +476,20 @@ void AMDGPUPromoteAlloca::visitAlloca(AllocaInst &I) {
// but the intrinsics need to be changed to accept pointers with any
// address space.
continue;
case Intrinsic::objectsize: {
Value *Src = Intr->getOperand(0);
Type *SrcTy = Src->getType()->getPointerElementType();
Function *ObjectSize = Intrinsic::getDeclaration(Mod,
Intrinsic::objectsize,
{ Intr->getType(), PointerType::get(SrcTy, AMDGPUAS::LOCAL_ADDRESS) }
);

CallInst *NewCall
= Builder.CreateCall(ObjectSize, { Src, Intr->getOperand(1) });
Intr->replaceAllUsesWith(NewCall);
Intr->eraseFromParent();
continue;
}
default:
Intr->dump();
llvm_unreachable("Don't know how to promote alloca intrinsic use.");
Expand Down
65 changes: 65 additions & 0 deletions test/CodeGen/AMDGPU/promote-alloca-mem-intrinsics.ll
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,65 @@
; RUN: opt -S -mtriple=amdgcn-unknown-amdhsa -amdgpu-promote-alloca < %s | FileCheck %s

declare void @llvm.memcpy.p0i8.p1i8.i32(i8* nocapture, i8 addrspace(1)* nocapture, i32, i32, i1) #0
declare void @llvm.memcpy.p1i8.p0i8.i32(i8 addrspace(1)* nocapture, i8* nocapture, i32, i32, i1) #0

declare void @llvm.memmove.p0i8.p1i8.i32(i8* nocapture, i8 addrspace(1)* nocapture, i32, i32, i1) #0
declare void @llvm.memmove.p1i8.p0i8.i32(i8 addrspace(1)* nocapture, i8* nocapture, i32, i32, i1) #0

declare void @llvm.memset.p0i8.i32(i8* nocapture, i8, i32, i32, i1) #0

declare i32 @llvm.objectsize.i32.p0i8(i8*, i1) #1

; CHECK-LABEL: @promote_with_memcpy(
; CHECK: getelementptr [256 x [17 x i32]], [256 x [17 x i32]] addrspace(3)* @alloca, i32 0, i32 %{{[0-9]+}}
; CHECK: call void @llvm.memcpy.p3i8.p1i8.i32(i8 addrspace(3)* %alloca.bc, i8 addrspace(1)* %in.bc, i32 68, i32 4, i1 false)
; CHECK: call void @llvm.memcpy.p1i8.p3i8.i32(i8 addrspace(1)* %out.bc, i8 addrspace(3)* %alloca.bc, i32 68, i32 4, i1 false)
define void @promote_with_memcpy(i32 addrspace(1)* %out, i32 addrspace(1)* %in) #0 {
%alloca = alloca [17 x i32], align 16
%alloca.bc = bitcast [17 x i32]* %alloca to i8*
%in.bc = bitcast i32 addrspace(1)* %in to i8 addrspace(1)*
%out.bc = bitcast i32 addrspace(1)* %out to i8 addrspace(1)*
call void @llvm.memcpy.p0i8.p1i8.i32(i8* %alloca.bc, i8 addrspace(1)* %in.bc, i32 68, i32 4, i1 false)
call void @llvm.memcpy.p1i8.p0i8.i32(i8 addrspace(1)* %out.bc, i8* %alloca.bc, i32 68, i32 4, i1 false)
ret void
}

; CHECK-LABEL: @promote_with_memmove(
; CHECK: getelementptr [256 x [17 x i32]], [256 x [17 x i32]] addrspace(3)* @alloca.1, i32 0, i32 %{{[0-9]+}}
; CHECK: call void @llvm.memmove.p3i8.p1i8.i32(i8 addrspace(3)* %alloca.bc, i8 addrspace(1)* %in.bc, i32 68, i32 4, i1 false)
; CHECK: call void @llvm.memmove.p1i8.p3i8.i32(i8 addrspace(1)* %out.bc, i8 addrspace(3)* %alloca.bc, i32 68, i32 4, i1 false)
define void @promote_with_memmove(i32 addrspace(1)* %out, i32 addrspace(1)* %in) #0 {
%alloca = alloca [17 x i32], align 16
%alloca.bc = bitcast [17 x i32]* %alloca to i8*
%in.bc = bitcast i32 addrspace(1)* %in to i8 addrspace(1)*
%out.bc = bitcast i32 addrspace(1)* %out to i8 addrspace(1)*
call void @llvm.memmove.p0i8.p1i8.i32(i8* %alloca.bc, i8 addrspace(1)* %in.bc, i32 68, i32 4, i1 false)
call void @llvm.memmove.p1i8.p0i8.i32(i8 addrspace(1)* %out.bc, i8* %alloca.bc, i32 68, i32 4, i1 false)
ret void
}

; CHECK-LABEL: @promote_with_memset(
; CHECK: getelementptr [256 x [17 x i32]], [256 x [17 x i32]] addrspace(3)* @alloca.2, i32 0, i32 %{{[0-9]+}}
; CHECK: call void @llvm.memset.p3i8.i32(i8 addrspace(3)* %alloca.bc, i8 7, i32 68, i32 4, i1 false)
define void @promote_with_memset(i32 addrspace(1)* %out, i32 addrspace(1)* %in) #0 {
%alloca = alloca [17 x i32], align 16
%alloca.bc = bitcast [17 x i32]* %alloca to i8*
%in.bc = bitcast i32 addrspace(1)* %in to i8 addrspace(1)*
%out.bc = bitcast i32 addrspace(1)* %out to i8 addrspace(1)*
call void @llvm.memset.p0i8.i32(i8* %alloca.bc, i8 7, i32 68, i32 4, i1 false)
ret void
}

; CHECK-LABEL: @promote_with_objectsize(
; CHECK: [[PTR:%[0-9]+]] = getelementptr [256 x [17 x i32]], [256 x [17 x i32]] addrspace(3)* @alloca.3, i32 0, i32 %{{[0-9]+}}
; CHECK: call i32 @llvm.objectsize.i32.p3i8(i8 addrspace(3)* %alloca.bc, i1 false)
define void @promote_with_objectsize(i32 addrspace(1)* %out) #0 {
%alloca = alloca [17 x i32], align 16
%alloca.bc = bitcast [17 x i32]* %alloca to i8*
%size = call i32 @llvm.objectsize.i32.p0i8(i8* %alloca.bc, i1 false)
store i32 %size, i32 addrspace(1)* %out
ret void
}

attributes #0 = { nounwind }
attributes #1 = { nounwind readnone }

0 comments on commit 5a2d99b

Please sign in to comment.