RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-20 19:42:54 +02:00
Code Issues Projects Releases Wiki Activity

[MemCpyOpt] Only replace memcpy with bitcast if address spaces match

Browse Source 
Patch by James Price

llvm-svn: 299866
This commit is contained in:
Matt Arsenault 2017-04-10 19:00:25 +00:00
parent 463bc213fe
commit 4fcdff5130
2 changed files with 18 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
lib/Transforms/Scalar/MemCpyOptimizer.cpp
View File

@ -1335,6 +1335,11 @@ bool MemCpyOptPass::processByValArgument(CallSite CS, unsigned ArgNo) {
CS.getInstruction(), &AC, &DT) < ByValAlign) CS.getInstruction(), &AC, &DT) < ByValAlign)
return false; return false;
// The address space of the memcpy source must match the byval argument
if (MDep->getSource()->getType()->getPointerAddressSpace() !=
ByValArg->getType()->getPointerAddressSpace())
return false;
// Verify that the copied-from memory doesn't change in between the memcpy and // Verify that the copied-from memory doesn't change in between the memcpy and
// the byval call. // the byval call.
// memcpy(a <- b) // memcpy(a <- b)

13
test/Transforms/MemCpyOpt/memcpy.ll
View File

@ -76,8 +76,21 @@ define void @test4(i8 *%P) {
; CHECK-NEXT: call void @test4a( ; CHECK-NEXT: call void @test4a(
} }
; Make sure we don't remove the memcpy if the source address space doesn't match the byval argument
define void @test4_addrspace(i8 addrspace(1)* %P) {
%A = alloca %1
%a = bitcast %1* %A to i8*
call void @llvm.memcpy.p0i8.p1i8.i64(i8* %a, i8 addrspace(1)* %P, i64 8, i32 4, i1 false)
call void @test4a(i8* align 1 byval %a)
ret void
; CHECK-LABEL: @test4_addrspace(
; CHECK: call void @llvm.memcpy.p0i8.p1i8.i64(
; CHECK-NEXT: call void @test4a(
}
declare void @test4a(i8* align 1 byval) declare void @test4a(i8* align 1 byval)
declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture, i64, i32, i1) nounwind declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture, i64, i32, i1) nounwind
declare void @llvm.memcpy.p0i8.p1i8.i64(i8* nocapture, i8 addrspace(1)* nocapture, i64, i32, i1) nounwind
declare void @llvm.memcpy.p1i8.p1i8.i64(i8 addrspace(1)* nocapture, i8 addrspace(1)* nocapture, i64, i32, i1) nounwind declare void @llvm.memcpy.p1i8.p1i8.i64(i8 addrspace(1)* nocapture, i8 addrspace(1)* nocapture, i64, i32, i1) nounwind
%struct.S = type { i128, [4 x i8]} %struct.S = type { i128, [4 x i8]}