1; RUN: opt -basic-aa -loop-idiom -S < %s -march=r600 -mcpu=redwood | FileCheck --check-prefix=R600 --check-prefix=FUNC %s
2; RUN: opt -basic-aa -loop-idiom -S < %s -march=amdgcn -mcpu=tahiti -verify-machineinstrs| FileCheck --check-prefix=SI --check-prefix=FUNC %s
3; RUN: opt -basic-aa -loop-idiom -S < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs| FileCheck --check-prefix=SI --check-prefix=FUNC %s
4
5target datalayout = "e-p:64:64-p1:64:64-p2:32:32-p3:32:32-p4:64:64-p5:32:32-p6:32:32-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-v2048:2048-n32:64-S32-A5"
6
7; Make sure loop-idiom doesn't create memcpy or memset.  There are no library
8; implementations of these for R600.
9
10; FUNC: @no_memcpy
11; R600-NOT: {{^}}llvm.memcpy
12; SI-NOT: {{^}}llvm.memcpy
13define amdgpu_kernel void @no_memcpy(i8 addrspace(3)* %in, i32 %size) {
14entry:
15  %dest = alloca i8, i32 32, addrspace(5)
16  br label %for.body
17
18for.body:
19  %0 = phi i32 [0, %entry], [%4, %for.body]
20  %1 = getelementptr i8, i8 addrspace(3)* %in, i32 %0
21  %2 = getelementptr i8, i8 addrspace(5)* %dest, i32 %0
22  %3 = load i8, i8 addrspace(3)* %1
23  store i8 %3, i8 addrspace(5)* %2
24  %4 = add i32 %0, 1
25  %5 = icmp eq i32 %4, %size
26  br i1 %5, label %for.end, label %for.body
27
28for.end:
29  ret void
30}
31
32; FUNC: @no_memset
33; R600-NOT: {{^}}llvm.memset
34; R600-NOT: {{^}}memset_pattern16:
35; SI-NOT: {{^}}llvm.memset
36; SI-NOT: {{^}}memset_pattern16:
37define amdgpu_kernel void @no_memset(i32 %size) {
38entry:
39  %dest = alloca i8, i32 32, addrspace(5)
40  br label %for.body
41
42for.body:
43  %0 = phi i32 [0, %entry], [%2, %for.body]
44  %1 = getelementptr i8, i8 addrspace(5)* %dest, i32 %0
45  store i8 0, i8 addrspace(5)* %1
46  %2 = add i32 %0, 1
47  %3 = icmp eq i32 %2, %size
48  br i1 %3, label %for.end, label %for.body
49
50for.end:
51  ret void
52}
53