1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
| ; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; Test that we correctly commute a sub instruction
; FUNC-LABEL: {{^}}sub_rev:
; SI-NOT: v_sub_i32_e32 v{{[0-9]+}}, vcc, s
; SI: v_subrev_{{[iu]}}32_e32 v{{[0-9]+}}, vcc, s
; ModuleID = 'vop-shrink.ll'
define amdgpu_kernel void @sub_rev(i32 addrspace(1)* %out, <4 x i32> %sgpr, i32 %cond) {
entry:
%vgpr = call i32 @llvm.amdgcn.workitem.id.x() #1
%tmp = icmp eq i32 %cond, 0
br i1 %tmp, label %if, label %else
if: ; preds = %entry
%tmp1 = getelementptr i32, i32 addrspace(1)* %out, i32 1
%tmp2 = extractelement <4 x i32> %sgpr, i32 1
store i32 %tmp2, i32 addrspace(1)* %out
br label %endif
else: ; preds = %entry
%tmp3 = extractelement <4 x i32> %sgpr, i32 2
%tmp4 = sub i32 %vgpr, %tmp3
store i32 %tmp4, i32 addrspace(1)* %out
br label %endif
endif: ; preds = %else, %if
ret void
}
; Test that we fold an immediate that was illegal for a 64-bit op into the
; 32-bit op when we shrink it.
; FUNC-LABEL: {{^}}add_fold:
; SI: v_add_f32_e32 v{{[0-9]+}}, 0x44800000
define amdgpu_kernel void @add_fold(float addrspace(1)* %out) {
entry:
%tmp = call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = uitofp i32 %tmp to float
%tmp2 = fadd float %tmp1, 1.024000e+03
store float %tmp2, float addrspace(1)* %out
ret void
}
; Function Attrs: nounwind readnone
declare i32 @llvm.amdgcn.workitem.id.x() #0
attributes #0 = { nounwind readnone }
attributes #1 = { readnone }
|