diff options
| author | Chris Lattner <sabre@nondot.org> | 2010-11-21 00:28:59 +0000 |
|---|---|---|
| committer | Chris Lattner <sabre@nondot.org> | 2010-11-21 00:28:59 +0000 |
| commit | 2f5f90ad3e9b00cf21ae8e3f55b93f0be1d504c3 (patch) | |
| tree | 0368003df16ef9b625afc9cdca10357bf6f22268 /test/Transforms/MemCpyOpt | |
| parent | a6fd81dd7f6039fbc1a55f6f4d45659fffdd81fb (diff) | |
| download | external_llvm-2f5f90ad3e9b00cf21ae8e3f55b93f0be1d504c3.zip external_llvm-2f5f90ad3e9b00cf21ae8e3f55b93f0be1d504c3.tar.gz external_llvm-2f5f90ad3e9b00cf21ae8e3f55b93f0be1d504c3.tar.bz2 | |
Implement PR8644: forwarding a memcpy value to a byval,
allowing the memcpy to be eliminated.
Unfortunately, the requirements on byval's without explicit
alignment are really weak and impossible to predict in the
mid-level optimizer, so this doesn't kick in much with current
frontends. The fix is to change clang to set alignment on all
byval arguments.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@119916 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'test/Transforms/MemCpyOpt')
| -rw-r--r-- | test/Transforms/MemCpyOpt/memcpy.ll | 15 | ||||
| -rw-r--r-- | test/Transforms/MemCpyOpt/sret.ll | 4 |
2 files changed, 17 insertions, 2 deletions
diff --git a/test/Transforms/MemCpyOpt/memcpy.ll b/test/Transforms/MemCpyOpt/memcpy.ll index 303c2fc..7309319 100644 --- a/test/Transforms/MemCpyOpt/memcpy.ll +++ b/test/Transforms/MemCpyOpt/memcpy.ll @@ -62,3 +62,18 @@ define void @test3({ x86_fp80, x86_fp80 }* noalias sret %agg.result) nounwind { ; CHECK-NEXT: call void @llvm.memcpy ; CHECK-NEXT: ret void } + + +; PR8644 +define void @test4(i8 *%P) { + %A = alloca {i32, i32} + %a = bitcast {i32, i32}* %A to i8* + call void @llvm.memcpy.p0i8.p0i8.i64(i8* %a, i8* %P, i64 8, i32 4, i1 false) + call void @test4a(i8* byval align 1 %a) + ret void +; CHECK: @test4 +; CHECK-NEXT: call void @test4a( +} + +declare void @test4a(i8* byval align 1) +declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture, i64, i32, i1) nounwind diff --git a/test/Transforms/MemCpyOpt/sret.ll b/test/Transforms/MemCpyOpt/sret.ll index d35ab91..ddfd0fd 100644 --- a/test/Transforms/MemCpyOpt/sret.ll +++ b/test/Transforms/MemCpyOpt/sret.ll @@ -3,7 +3,7 @@ target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128" target triple = "i686-apple-darwin9" -define void @ccosl({ x86_fp80, x86_fp80 }* noalias sret %agg.result, { x86_fp80, x86_fp80 }* byval %z) nounwind { +define void @ccosl({ x86_fp80, x86_fp80 }* noalias sret %agg.result, { x86_fp80, x86_fp80 }* byval align 8 %z) nounwind { entry: %iz = alloca { x86_fp80, x86_fp80 } ; <{ x86_fp80, x86_fp80 }*> [#uses=3] %memtmp = alloca { x86_fp80, x86_fp80 }, align 16 ; <{ x86_fp80, x86_fp80 }*> [#uses=2] @@ -16,7 +16,7 @@ entry: %tmp8 = load x86_fp80* %tmp7, align 16 ; <x86_fp80> [#uses=1] store x86_fp80 %tmp3, x86_fp80* %real, align 16 store x86_fp80 %tmp8, x86_fp80* %tmp4, align 16 - call void @ccoshl( { x86_fp80, x86_fp80 }* noalias sret %memtmp, { x86_fp80, x86_fp80 }* byval %iz ) nounwind + call void @ccoshl( { x86_fp80, x86_fp80 }* noalias sret %memtmp, { x86_fp80, x86_fp80 }* byval align 8 %iz ) nounwind %memtmp14 = bitcast { x86_fp80, x86_fp80 }* %memtmp to i8* ; <i8*> [#uses=1] %agg.result15 = bitcast { x86_fp80, x86_fp80 }* %agg.result to i8* ; <i8*> [#uses=1] call void @llvm.memcpy.i32( i8* %agg.result15, i8* %memtmp14, i32 32, i32 16 ) |