File: llvm.amdgcn.mov.dpp8.ll

package info (click to toggle)
llvm-toolchain-14 1%3A14.0.6-12
  • links: PTS, VCS
  • area: main
  • in suites: bookworm
  • size: 1,496,180 kB
  • sloc: cpp: 5,593,972; ansic: 986,872; asm: 585,869; python: 184,223; objc: 72,530; lisp: 31,119; f90: 27,793; javascript: 9,780; pascal: 9,762; sh: 9,482; perl: 7,468; ml: 5,432; awk: 3,523; makefile: 2,538; xml: 953; cs: 573; fortran: 567
file content (26 lines) | stat: -rw-r--r-- 1,132 bytes parent folder | download | duplicates (10) 
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
 
; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=GFX10 %s

; GFX10-LABEL: {{^}}dpp8_test:
; GFX10: v_mov_b32_e32 [[SRC:v[0-9]+]], s{{[0-9]+}}
; GFX10: v_mov_b32_dpp [[SRC]], [[SRC]]  dpp8:[1,0,0,0,0,0,0,0]{{$}}
define amdgpu_kernel void @dpp8_test(i32 addrspace(1)* %out, i32 %in) {
  %tmp0 = call i32 @llvm.amdgcn.mov.dpp8.i32(i32 %in, i32 1) #0
  store i32 %tmp0, i32 addrspace(1)* %out
  ret void
}

; GFX10-LABEL: {{^}}dpp8_wait_states:
; GFX10-NOOPT: v_mov_b32_e32 [[VGPR1:v[0-9]+]], s{{[0-9]+}}
; GFX10: v_mov_b32_e32 [[VGPR0:v[0-9]+]], s{{[0-9]+}}
; GFX10: v_mov_b32_dpp [[VGPR0]], [[VGPR0]] dpp8:[1,0,0,0,0,0,0,0]{{$}}
; GFX10: v_mov_b32_dpp [[VGPR0]], [[VGPR0]] dpp8:[5,0,0,0,0,0,0,0]{{$}}
define amdgpu_kernel void @dpp8_wait_states(i32 addrspace(1)* %out, i32 %in) {
  %tmp0 = call i32 @llvm.amdgcn.mov.dpp8.i32(i32 %in, i32 1) #0
  %tmp1 = call i32 @llvm.amdgcn.mov.dpp8.i32(i32 %tmp0, i32 5) #0
  store i32 %tmp1, i32 addrspace(1)* %out
  ret void
}

declare i32 @llvm.amdgcn.mov.dpp8.i32(i32, i32) #0

attributes #0 = { nounwind readnone convergent }