File: slp-v2f32.ll

package info (click to toggle)
llvm-toolchain-19 1%3A19.1.7-3~deb12u1
  • links: PTS, VCS
  • area: main
  • in suites: bookworm-proposed-updates
  • size: 1,998,492 kB
  • sloc: cpp: 6,951,680; ansic: 1,486,157; asm: 913,598; python: 232,024; f90: 80,126; objc: 75,281; lisp: 37,276; pascal: 16,990; sh: 10,009; ml: 5,058; perl: 4,724; awk: 3,523; makefile: 3,167; javascript: 2,504; xml: 892; fortran: 664; cs: 573
file content (66 lines) | stat: -rw-r--r-- 2,703 bytes parent folder | download | duplicates (12)
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 -passes=slp-vectorizer < %s | FileCheck -check-prefixes=GCN,GFX908 %s
; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=gfx90a -passes=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(ptr addrspace(1) %arg) {
bb:
  %tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
  %tmp1 = zext i32 %tmp to i64
  %tmp2 = getelementptr inbounds float, ptr addrspace(1) %arg, i64 %tmp1
  %tmp3 = load float, ptr addrspace(1) %tmp2, align 4
  %tmp4 = fadd float %tmp3, 1.000000e+00
  store float %tmp4, ptr addrspace(1) %tmp2, align 4
  %tmp5 = add nuw nsw i64 %tmp1, 1
  %tmp6 = getelementptr inbounds float, ptr addrspace(1) %arg, i64 %tmp5
  %tmp7 = load float, ptr addrspace(1) %tmp6, align 4
  %tmp8 = fadd float %tmp7, 1.000000e+00
  store float %tmp8, ptr 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(ptr addrspace(1) %arg) {
bb:
  %tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
  %tmp1 = zext i32 %tmp to i64
  %tmp2 = getelementptr inbounds float, ptr addrspace(1) %arg, i64 %tmp1
  %tmp3 = load float, ptr addrspace(1) %tmp2, align 4
  %tmp4 = fmul float %tmp3, 1.000000e+00
  store float %tmp4, ptr addrspace(1) %tmp2, align 4
  %tmp5 = add nuw nsw i64 %tmp1, 1
  %tmp6 = getelementptr inbounds float, ptr addrspace(1) %arg, i64 %tmp5
  %tmp7 = load float, ptr addrspace(1) %tmp6, align 4
  %tmp8 = fmul float %tmp7, 1.000000e+00
  store float %tmp8, ptr 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(ptr addrspace(1) %arg) {
bb:
  %tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
  %tmp1 = zext i32 %tmp to i64
  %tmp2 = getelementptr inbounds float, ptr addrspace(1) %arg, i64 %tmp1
  %tmp3 = load float, ptr 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, ptr addrspace(1) %tmp2, align 4
  %tmp5 = add nuw nsw i64 %tmp1, 1
  %tmp6 = getelementptr inbounds float, ptr addrspace(1) %arg, i64 %tmp5
  %tmp7 = load float, ptr 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, ptr addrspace(1) %tmp6, align 4
  ret void
}

declare i32 @llvm.amdgcn.workitem.id.x()
declare float @llvm.fma.f32(float, float, float)