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
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
|
; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=gfx908 -slp-vectorizer < %s | FileCheck -check-prefixes=GCN,GFX908 %s
; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=gfx90a -slp-vectorizer < %s | FileCheck -check-prefixes=GCN,GFX90A %s
; GCN-LABEL: @fadd_combine
; GFX908: fadd float
; GFX908: fadd float
; GFX90A: fadd <2 x float>
define amdgpu_kernel void @fadd_combine(float addrspace(1)* %arg) {
bb:
%tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = zext i32 %tmp to i64
%tmp2 = getelementptr inbounds float, float addrspace(1)* %arg, i64 %tmp1
%tmp3 = load float, float addrspace(1)* %tmp2, align 4
%tmp4 = fadd float %tmp3, 1.000000e+00
store float %tmp4, float addrspace(1)* %tmp2, align 4
%tmp5 = add nuw nsw i64 %tmp1, 1
%tmp6 = getelementptr inbounds float, float addrspace(1)* %arg, i64 %tmp5
%tmp7 = load float, float addrspace(1)* %tmp6, align 4
%tmp8 = fadd float %tmp7, 1.000000e+00
store float %tmp8, float addrspace(1)* %tmp6, align 4
ret void
}
; GCN-LABEL: @fmul_combine
; GFX908: fmul float
; GFX908: fmul float
; GFX90A: fmul <2 x float>
define amdgpu_kernel void @fmul_combine(float addrspace(1)* %arg) {
bb:
%tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = zext i32 %tmp to i64
%tmp2 = getelementptr inbounds float, float addrspace(1)* %arg, i64 %tmp1
%tmp3 = load float, float addrspace(1)* %tmp2, align 4
%tmp4 = fmul float %tmp3, 1.000000e+00
store float %tmp4, float addrspace(1)* %tmp2, align 4
%tmp5 = add nuw nsw i64 %tmp1, 1
%tmp6 = getelementptr inbounds float, float addrspace(1)* %arg, i64 %tmp5
%tmp7 = load float, float addrspace(1)* %tmp6, align 4
%tmp8 = fmul float %tmp7, 1.000000e+00
store float %tmp8, float addrspace(1)* %tmp6, align 4
ret void
}
; GCN-LABEL: @fma_combine
; GFX908: call float @llvm.fma.f32
; GFX908: call float @llvm.fma.f32
; GFX90A: call <2 x float> @llvm.fma.v2f32
define amdgpu_kernel void @fma_combine(float addrspace(1)* %arg) {
bb:
%tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
%tmp1 = zext i32 %tmp to i64
%tmp2 = getelementptr inbounds float, float addrspace(1)* %arg, i64 %tmp1
%tmp3 = load float, float addrspace(1)* %tmp2, align 4
%tmp4 = tail call float @llvm.fma.f32(float %tmp3, float 1.000000e+00, float 1.000000e+00)
store float %tmp4, float addrspace(1)* %tmp2, align 4
%tmp5 = add nuw nsw i64 %tmp1, 1
%tmp6 = getelementptr inbounds float, float addrspace(1)* %arg, i64 %tmp5
%tmp7 = load float, float addrspace(1)* %tmp6, align 4
%tmp8 = tail call float @llvm.fma.f32(float %tmp7, float 1.000000e+00, float 1.000000e+00)
store float %tmp8, float addrspace(1)* %tmp6, align 4
ret void
}
declare i32 @llvm.amdgcn.workitem.id.x()
declare float @llvm.fma.f32(float, float, float)
|
