diff options
Diffstat (limited to 'test/Transforms/MemCpyOpt/memcpy.ll')
| -rw-r--r-- | test/Transforms/MemCpyOpt/memcpy.ll | 93 |
1 files changed, 91 insertions, 2 deletions
diff --git a/test/Transforms/MemCpyOpt/memcpy.ll b/test/Transforms/MemCpyOpt/memcpy.ll index 724acfa..b387d32 100644 --- a/test/Transforms/MemCpyOpt/memcpy.ll +++ b/test/Transforms/MemCpyOpt/memcpy.ll @@ -1,9 +1,9 @@ -; RUN: opt < %s -memcpyopt -dse -S | grep {call.*memcpy} | count 1 +; RUN: opt < %s -basicaa -memcpyopt -dse -S | FileCheck %s 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 }* sret %agg.result, x86_fp80 %z.0, x86_fp80 %z.1) nounwind { +define void @test1({ x86_fp80, x86_fp80 }* sret %agg.result, x86_fp80 %z.0, x86_fp80 %z.1) nounwind { entry: %tmp2 = alloca { x86_fp80, x86_fp80 } ; <{ x86_fp80, x86_fp80 }*> [#uses=1] %memtmp = alloca { x86_fp80, x86_fp80 }, align 16 ; <{ x86_fp80, x86_fp80 }*> [#uses=2] @@ -14,9 +14,98 @@ entry: call void @llvm.memcpy.i32( i8* %tmp219, i8* %memtmp20, i32 32, i32 16 ) %agg.result21 = bitcast { x86_fp80, x86_fp80 }* %agg.result to i8* ; <i8*> [#uses=1] call void @llvm.memcpy.i32( i8* %agg.result21, i8* %tmp219, i32 32, i32 16 ) + +; Check that one of the memcpy's are removed. +;; FIXME: PR 8643 We should be able to eliminate the last memcpy here. + +; CHECK: @test1 +; CHECK: call void @ccoshl +; CHECK: call void @llvm.memcpy +; CHECK-NOT: llvm.memcpy +; CHECK: ret void ret void } declare void @ccoshl({ x86_fp80, x86_fp80 }* sret , x86_fp80, x86_fp80) nounwind declare void @llvm.memcpy.i32(i8*, i8*, i32, i32) nounwind + + +; The intermediate alloca and one of the memcpy's should be eliminated, the +; other should be related with a memmove. +define void @test2(i8* %P, i8* %Q) nounwind { + %memtmp = alloca { x86_fp80, x86_fp80 }, align 16 + %R = bitcast { x86_fp80, x86_fp80 }* %memtmp to i8* + call void @llvm.memcpy.i32( i8* %R, i8* %P, i32 32, i32 16 ) + call void @llvm.memcpy.i32( i8* %Q, i8* %R, i32 32, i32 16 ) + ret void + +; CHECK: @test2 +; CHECK-NEXT: call void @llvm.memmove{{.*}}(i8* %Q, i8* %P +; CHECK-NEXT: ret void +} + + + + +@x = external global { x86_fp80, x86_fp80 } + +define void @test3({ x86_fp80, x86_fp80 }* noalias sret %agg.result) nounwind { + %x.0 = alloca { x86_fp80, x86_fp80 } + %x.01 = bitcast { x86_fp80, x86_fp80 }* %x.0 to i8* + call void @llvm.memcpy.i32( i8* %x.01, i8* bitcast ({ x86_fp80, x86_fp80 }* @x to i8*), i32 32, i32 16 ) + %agg.result2 = bitcast { x86_fp80, x86_fp80 }* %agg.result to i8* + call void @llvm.memcpy.i32( i8* %agg.result2, i8* %x.01, i32 32, i32 16 ) + ret void +; CHECK: @test3 +; CHECK-NEXT: %agg.result2 = bitcast +; 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 + +%struct.S = type { i128, [4 x i8]} + +@sS = external global %struct.S, align 16 + +declare void @test5a(%struct.S* byval align 16) nounwind ssp + +declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture, i64, i32, i1) nounwind + +; rdar://8713376 - This memcpy can't be eliminated. +define i32 @test5(i32 %x) nounwind ssp { +entry: + %y = alloca %struct.S, align 16 + %tmp = bitcast %struct.S* %y to i8* + call void @llvm.memcpy.p0i8.p0i8.i64(i8* %tmp, i8* bitcast (%struct.S* @sS to i8*), i64 32, i32 16, i1 false) + %a = getelementptr %struct.S* %y, i64 0, i32 1, i64 0 + store i8 4, i8* %a + call void @test5a(%struct.S* byval align 16 %y) + ret i32 0 + ; CHECK: @test5( + ; CHECK: store i8 4 + ; CHECK: call void @test5a(%struct.S* byval align 16 %y) +} + +;; Noop memcpy should be zapped. +define void @test6(i8 *%P) { + call void @llvm.memcpy.p0i8.p0i8.i64(i8* %P, i8* %P, i64 8, i32 4, i1 false) + ret void +; CHECK: @test6 +; CHECK-NEXT: ret void +} + |
