1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 2; MemCpy optimizations should take place even in presence of invariant.start 3; RUN: opt < %s -passes=memcpyopt -S -verify-memoryssa | FileCheck %s 4 5target 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" 6 7target triple = "i686-apple-darwin9" 8 9%0 = type { x86_fp80, x86_fp80 } 10declare void @llvm.memcpy.p0.p0.i32(ptr nocapture, ptr nocapture, i32, i1) nounwind 11declare void @llvm.memcpy.p0.p0.i64(ptr nocapture, ptr nocapture readonly, i64, i1) 12declare void @llvm.memset.p0.i64(ptr nocapture, i8, i64, i1) 13 14declare ptr @llvm.invariant.start.p0(i64, ptr nocapture) nounwind readonly 15 16; The intermediate alloca and one of the memcpy's should be eliminated, the 17; other should be transformed to a memmove. 18define void @test1(ptr %P, ptr %Q) nounwind { 19; CHECK-LABEL: @test1( 20; CHECK-NEXT: [[MEMTMP:%.*]] = alloca [[TMP0:%.*]], align 16 21; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i32(ptr align 16 [[MEMTMP]], ptr align 16 [[P:%.*]], i32 32, i1 false) 22; CHECK-NEXT: [[I:%.*]] = call ptr @llvm.invariant.start.p0(i64 32, ptr [[P]]) 23; CHECK-NEXT: call void @llvm.memmove.p0.p0.i32(ptr align 16 [[Q:%.*]], ptr align 16 [[P]], i32 32, i1 false) 24; CHECK-NEXT: ret void 25; 26 %memtmp = alloca %0, align 16 27 call void @llvm.memcpy.p0.p0.i32(ptr align 16 %memtmp, ptr align 16 %P, i32 32, i1 false) 28 %i = call ptr @llvm.invariant.start.p0(i64 32, ptr %P) 29 call void @llvm.memcpy.p0.p0.i32(ptr align 16 %Q, ptr align 16 %memtmp, i32 32, i1 false) 30 ret void 31} 32 33 34; The invariant.start intrinsic does not inhibit tranforming the memcpy to a 35; memset. 36define void @test2(ptr %dst1, ptr %dst2, i8 %c) { 37; CHECK-LABEL: @test2( 38; CHECK-NEXT: call void @llvm.memset.p0.i64(ptr [[DST1:%.*]], i8 [[C:%.*]], i64 128, i1 false) 39; CHECK-NEXT: [[I:%.*]] = call ptr @llvm.invariant.start.p0(i64 32, ptr [[DST1]]) 40; CHECK-NEXT: call void @llvm.memset.p0.i64(ptr align 8 [[DST2:%.*]], i8 [[C]], i64 128, i1 false) 41; CHECK-NEXT: ret void 42; 43 call void @llvm.memset.p0.i64(ptr %dst1, i8 %c, i64 128, i1 false) 44 %i = call ptr @llvm.invariant.start.p0(i64 32, ptr %dst1) 45 call void @llvm.memcpy.p0.p0.i64(ptr align 8 %dst2, ptr align 8 %dst1, i64 128, i1 false) 46 ret void 47} 48