1 ; RUN: opt -S -mtriple=amdgcn-unknown-amdhsa -mcpu=gfx900 -loop-vectorize -simplifycfg < %s | FileCheck -check-prefixes=GCN,GFX9 %s
2 ; RUN: opt -S -mtriple=amdgcn-unknown-amdhsa -mcpu=gfx900 -loop-vectorize -pass-remarks-analysis='loop-vectorize' < %s 2>&1 | FileCheck -check-prefixes=REMARK %s
4 ; GCN-LABEL: @runtime_check_divergent_target(
5 ; GCN-NOT: load <2 x half>
6 ; GCN-NOT: store <2 x half>
8 ; REMARK: remark: <unknown>:0:0: loop not vectorized: runtime pointer checks needed. Not enabled for divergent target
9 define amdgpu_kernel void @runtime_check_divergent_target(half addrspace(1)* nocapture %a, half addrspace(1)* nocapture %b) #0 {
13 for.body: ; preds = %entry, %for.body
14 %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
15 %arrayidx = getelementptr inbounds half, half addrspace(1)* %b, i64 %indvars.iv
16 %load = load half, half addrspace(1)* %arrayidx, align 4
17 %mul = fmul half %load, 3.0
18 %arrayidx2 = getelementptr inbounds half, half addrspace(1)* %a, i64 %indvars.iv
19 store half %mul, half addrspace(1)* %arrayidx2, align 4
20 %indvars.iv.next = add i64 %indvars.iv, 1
21 %lftr.wideiv = trunc i64 %indvars.iv.next to i32
22 %exitcond = icmp eq i32 %lftr.wideiv, 1024
23 br i1 %exitcond, label %for.end, label %for.body
25 for.end: ; preds = %for.body, %entry
29 attributes #0 = { nounwind }