Skip to content

Commit 9aa951e

Browse files
committed
[MemCpyOpt] Preserve address space
Preserve address space when generating the cast to i8*.
1 parent 010f0f0 commit 9aa951e

File tree

2 files changed

+20
-1
lines changed

2 files changed

+20
-1
lines changed

llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp

Lines changed: 3 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1214,9 +1214,11 @@ bool MemCpyOptPass::processMemSetMemCpyDependence(MemCpyInst *MemCpy,
12141214
Value *SizeDiff = Builder.CreateSub(DestSize, SrcSize);
12151215
Value *MemsetLen = Builder.CreateSelect(
12161216
Ule, ConstantInt::getNullValue(DestSize->getType()), SizeDiff);
1217+
unsigned DestAS = Dest->getType()->getPointerAddressSpace();
12171218
Instruction *NewMemSet = Builder.CreateMemSet(
12181219
Builder.CreateGEP(Builder.getInt8Ty(),
1219-
Builder.CreatePointerCast(Dest, Builder.getInt8PtrTy()),
1220+
Builder.CreatePointerCast(Dest,
1221+
Builder.getInt8PtrTy(DestAS)),
12201222
SrcSize),
12211223
MemSet->getOperand(1), MemsetLen, MaybeAlign(Align));
12221224

llvm/test/Transforms/MemCpyOpt/memset-memcpy-redundant-memset.ll

Lines changed: 17 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -323,6 +323,21 @@ define void @test_weird_element_type(i16* %src, i64 %src_size, i16* noalias %dst
323323
ret void
324324
}
325325

326+
define void @test_addrspace(i8 addrspace(1)* %src, i64 %src_size, i8 addrspace(1)* noalias %dst, i64 %dst_size, i8 %c) {
327+
; CHECK-LABEL: @test_addrspace(
328+
; CHECK-NEXT: [[TMP1:%.*]] = icmp ule i64 [[DST_SIZE:%.*]], [[SRC_SIZE:%.*]]
329+
; CHECK-NEXT: [[TMP2:%.*]] = sub i64 [[DST_SIZE]], [[SRC_SIZE]]
330+
; CHECK-NEXT: [[TMP3:%.*]] = select i1 [[TMP1]], i64 0, i64 [[TMP2]]
331+
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr i8, i8 addrspace(1)* [[DST:%.*]], i64 [[SRC_SIZE]]
332+
; CHECK-NEXT: call void @llvm.memset.p1i8.i64(i8 addrspace(1)* align 1 [[TMP4]], i8 [[C:%.*]], i64 [[TMP3]], i1 false)
333+
; CHECK-NEXT: call void @llvm.memcpy.p1i8.p1i8.i64(i8 addrspace(1)* [[DST]], i8 addrspace(1)* [[SRC:%.*]], i64 [[SRC_SIZE]], i1 false)
334+
; CHECK-NEXT: ret void
335+
;
336+
call void @llvm.memset.p1i8.i64(i8 addrspace(1)* %dst, i8 %c, i64 %dst_size, i1 false)
337+
call void @llvm.memcpy.p1i8.p1i8.i64(i8 addrspace(1)* %dst, i8 addrspace(1)* %src, i64 %src_size, i1 false)
338+
ret void
339+
}
340+
326341
declare void @llvm.memset.p0i8.i64(i8* nocapture, i8, i64, i1)
327342
declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture readonly, i64, i1)
328343
declare void @llvm.memset.p0i8.i32(i8* nocapture, i8, i32, i1)
@@ -333,4 +348,6 @@ declare void @llvm.memset.p0.i64(ptr nocapture, i8, i64, i1)
333348
declare void @llvm.memcpy.p0.p0.i64(ptr nocapture, ptr nocapture readonly, i64, i1)
334349
declare void @llvm.memset.p0i16.i64(i16* nocapture, i8, i64, i1)
335350
declare void @llvm.memcpy.p0i16.p0i16.i64(i16* nocapture, i16* nocapture readonly, i64, i1)
351+
declare void @llvm.memset.p1i8.i64(i8 addrspace(1)* nocapture, i8, i64, i1)
352+
declare void @llvm.memcpy.p1i8.p1i8.i64(i8 addrspace(1)* nocapture, i8 addrspace(1)* nocapture readonly, i64, i1)
336353
declare void @call()

0 commit comments

Comments
 (0)