1; RUN: opt -basicaa -loop-idiom -S < %s -march=r600 -mcpu=redwood | FileCheck --check-prefix=R600 --check-prefix=FUNC %s 2; RUN: opt -basicaa -loop-idiom -S < %s -march=amdgcn -mcpu=SI -verify-machineinstrs| FileCheck --check-prefix=SI --check-prefix=FUNC %s 3; RUN: opt -basicaa -loop-idiom -S < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs| FileCheck --check-prefix=SI --check-prefix=FUNC %s 4 5target datalayout = "e-p:32:32-p1:64:64-p2:64:64-p3:32:32-p4:32:32-p5:64:64-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-v2048:2048-n32:64" 6target triple = "r600--" 7 8 9 10; Make sure loop-idiom doesn't create memcpy or memset. There are no library 11; implementations of these for R600. 12 13; FUNC: @no_memcpy 14; R600-NOT: {{^}}llvm.memcpy 15; SI-NOT: {{^}}llvm.memcpy 16define void @no_memcpy(i8 addrspace(3)* %in, i32 %size) { 17entry: 18 %dest = alloca i8, i32 32 19 br label %for.body 20 21for.body: 22 %0 = phi i32 [0, %entry], [%4, %for.body] 23 %1 = getelementptr i8, i8 addrspace(3)* %in, i32 %0 24 %2 = getelementptr i8, i8* %dest, i32 %0 25 %3 = load i8, i8 addrspace(3)* %1 26 store i8 %3, i8* %2 27 %4 = add i32 %0, 1 28 %5 = icmp eq i32 %4, %size 29 br i1 %5, label %for.end, label %for.body 30 31for.end: 32 ret void 33} 34 35; FUNC: @no_memset 36; R600-NOT: {{^}}llvm.memset 37; R600-NOT: {{^}}memset_pattern16: 38; SI-NOT: {{^}}llvm.memset 39; SI-NOT: {{^}}memset_pattern16: 40define void @no_memset(i32 %size) { 41entry: 42 %dest = alloca i8, i32 32 43 br label %for.body 44 45for.body: 46 %0 = phi i32 [0, %entry], [%2, %for.body] 47 %1 = getelementptr i8, i8* %dest, i32 %0 48 store i8 0, i8* %1 49 %2 = add i32 %0, 1 50 %3 = icmp eq i32 %2, %size 51 br i1 %3, label %for.end, label %for.body 52 53for.end: 54 ret void 55} 56