xref: /llvm-project/llvm/test/CodeGen/AMDGPU/divergence-driven-bfe-isel.ll (revision 9e9907f1cfa424366fba58d9520f9305b537cec9)
1; RUN: llc -mtriple=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
2
3; GCN-LABEL: @bfe_uniform
4; GCN: s_bfe_u32 s{{[0-9]+}}, s{{[0-9]+}}, 0x40010
5define amdgpu_kernel void @bfe_uniform(i32 %val, ptr addrspace(1) %out) {
6  %hibits = lshr i32 %val, 16
7  %masked = and i32 %hibits, 15
8  store i32 %masked, ptr addrspace(1) %out
9  ret void
10}
11
12; GCN-LABEL: @bfe_divergent
13; GCN: v_bfe_u32 v{{[0-9]+}}, v{{[0-9]+}}, 16, 4
14define amdgpu_kernel void @bfe_divergent(i32 %val, ptr addrspace(1) %out) {
15  %tid = call i32 @llvm.amdgcn.workitem.id.x()
16  %divergent = add i32 %val, %tid
17  %hibits = lshr i32 %divergent, 16
18  %masked = and i32 %hibits, 15
19  store i32 %masked, ptr addrspace(1) %out
20  ret void
21}
22
23
24declare i32 @llvm.amdgcn.workitem.id.x()
25
26