File: reduce-store-width-alignment.ll

package info (click to toggle)
llvm-toolchain-11 1%3A11.0.1-2
  • links: PTS, VCS
  • area: main
  • in suites: bullseye
  • size: 995,808 kB
  • sloc: cpp: 4,767,656; ansic: 760,916; asm: 477,436; python: 170,940; objc: 69,804; lisp: 29,914; sh: 23,855; f90: 18,173; pascal: 7,551; perl: 7,471; ml: 5,603; awk: 3,489; makefile: 2,573; xml: 915; cs: 573; fortran: 503; javascript: 452
file content (53 lines) | stat: -rw-r--r-- 2,239 bytes parent folder | download | duplicates (14)
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
; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s

; GCN-LABEL: {{^}}store_v2i32_as_v4i16_align_4:
; GCN: s_load_dwordx2
; GCN: ds_write2_b32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset1:1{{$}}
define amdgpu_kernel void @store_v2i32_as_v4i16_align_4(<4 x i16> addrspace(3)* align 4 %out, <2 x i32> %x) #0 {
  %x.bc = bitcast <2 x i32> %x to <4 x i16>
  store <4 x i16> %x.bc, <4 x i16> addrspace(3)* %out, align 4
  ret void
}

; GCN-LABEL: {{^}}store_v4i32_as_v8i16_align_4:
; GCN: s_load_dwordx4
; GCN-DAG: ds_write2_b32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset0:2 offset1:3
; GCN-DAG: ds_write2_b32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset1:1{{$}}
define amdgpu_kernel void @store_v4i32_as_v8i16_align_4(<8 x i16> addrspace(3)* align 4 %out, <4 x i32> %x) #0 {
  %x.bc = bitcast <4 x i32> %x to <8 x i16>
  store <8 x i16> %x.bc, <8 x i16> addrspace(3)* %out, align 4
  ret void
}

; GCN-LABEL: {{^}}store_v2i32_as_i64_align_4:
; GCN: s_load_dwordx2
; GCN: ds_write2_b32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset1:1{{$}}
define amdgpu_kernel void @store_v2i32_as_i64_align_4(<4 x i16> addrspace(3)* align 4 %out, <2 x i32> %x) #0 {
  %x.bc = bitcast <2 x i32> %x to <4 x i16>
  store <4 x i16> %x.bc, <4 x i16> addrspace(3)* %out, align 4
  ret void
}

; GCN-LABEL: {{^}}store_v4i32_as_v2i64_align_4:
; GCN: s_load_dwordx4
; GCN-DAG: ds_write2_b32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset0:2 offset1:3
; GCN-DAG: ds_write2_b32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset1:1{{$}}
define amdgpu_kernel void @store_v4i32_as_v2i64_align_4(<2 x i64> addrspace(3)* align 4 %out, <4 x i32> %x) #0 {
  %x.bc = bitcast <4 x i32> %x to <2 x i64>
  store <2 x i64> %x.bc, <2 x i64> addrspace(3)* %out, align 4
  ret void
}

; GCN-LABEL: {{^}}store_v4i16_as_v2i32_align_4:
; GCN: s_load_dword s
; GCN-NEXT: s_load_dwordx2 s
; GCN-NOT: {{buffer|flat|global}}

; GCN: ds_write2_b32 v{{[0-9]+}}, v{{[0-9]+}}, v{{[0-9]+}} offset1:1{{$}}
define amdgpu_kernel void @store_v4i16_as_v2i32_align_4(<2 x i32> addrspace(3)* align 4 %out, <4 x i16> %x) #0 {
  %x.bc = bitcast <4 x i16> %x to <2 x i32>
  store <2 x i32> %x.bc, <2 x i32> addrspace(3)* %out, align 4
  ret void
}

attributes #0 = { nounwind }