File: llvm.amdgcn.mov.dpp8.ll

package info (click to toggle)
llvm-toolchain-20 1%3A20.1.6-1~exp1
  • links: PTS, VCS
  • area: main
  • in suites: experimental
  • size: 2,111,304 kB
  • sloc: cpp: 7,438,677; ansic: 1,393,822; asm: 1,012,926; python: 241,650; f90: 86,635; objc: 75,479; lisp: 42,144; pascal: 17,286; sh: 10,027; ml: 5,082; perl: 4,730; awk: 3,523; makefile: 3,349; javascript: 2,251; xml: 892; fortran: 672
file content (191 lines) | stat: -rw-r--r-- 8,772 bytes parent folder | download | duplicates (5) 
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
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
 
; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX10PLUS %s
; RUN: llc -global-isel=1 -global-isel-abort=2 -mtriple=amdgcn -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX10PLUS %s
; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1100 -amdgpu-enable-vopd=0 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX10PLUS %s
; RUN: llc -global-isel=1 -global-isel-abort=2 -mtriple=amdgcn -mcpu=gfx1100 -amdgpu-enable-vopd=0 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX10PLUS %s
; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1200 -amdgpu-enable-vopd=0 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX10PLUS %s
; RUN: llc -global-isel=1 -global-isel-abort=2 -mtriple=amdgcn -mcpu=gfx1200 -amdgpu-enable-vopd=0 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX10PLUS %s

; GFX10PLUS-LABEL: {{^}}dpp8_test:
; GFX10PLUS: v_mov_b32_e32 [[SRC:v[0-9]+]], s{{[0-9]+}}
; GFX10PLUS: v_mov_b32_dpp [[SRC]], [[SRC]]  dpp8:[1,0,0,0,0,0,0,0]{{$}}
define amdgpu_kernel void @dpp8_test(ptr addrspace(1) %out, i32 %in) {
  %tmp0 = call i32 @llvm.amdgcn.mov.dpp8.i32(i32 %in, i32 1) #0
  store i32 %tmp0, ptr addrspace(1) %out
  ret void
}

; GFX10PLUS-LABEL: {{^}}dpp8_wait_states:
; GFX10PLUS-NOOPT: v_mov_b32_e32 [[VGPR1:v[0-9]+]], s{{[0-9]+}}
; GFX10PLUS: v_mov_b32_e32 [[VGPR0:v[0-9]+]], s{{[0-9]+}}
; GFX10PLUS: v_mov_b32_dpp [[VGPR0]], [[VGPR0]] dpp8:[1,0,0,0,0,0,0,0]{{$}}
; GFX10PLUS: v_mov_b32_dpp [[VGPR0]], [[VGPR0]] dpp8:[5,0,0,0,0,0,0,0]{{$}}
define amdgpu_kernel void @dpp8_wait_states(ptr 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, ptr addrspace(1) %out
  ret void
}

; GFX10PLUS-LABEL: {{^}}dpp8_i64:
; GFX10PLUS-DAG: v_mov_b32_dpp v0, v0 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: v_mov_b32_dpp v1, v1 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: global_store_{{dwordx2|b64}} v[2:3], v[0:1], off
define amdgpu_ps void @dpp8_i64(i64 %in, ptr addrspace(1) %out) {
  %tmp0 = call i64 @llvm.amdgcn.mov.dpp8.i64(i64 %in, i32 1)
  store i64 %tmp0, ptr addrspace(1) %out
  ret void
}

; GFX10PLUS-LABEL: {{^}}dpp8_v2i32:
; GFX10PLUS-DAG: v_mov_b32_dpp v0, v0 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: v_mov_b32_dpp v1, v1 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: global_store_{{dwordx2|b64}} v[2:3], v[0:1], off
define amdgpu_ps void @dpp8_v2i32(<2 x i32> %in, ptr addrspace(1) %out) {
  %tmp0 = call <2 x i32> @llvm.amdgcn.mov.dpp8.v3i32(<2 x i32> %in, i32 1)
  store <2 x i32> %tmp0, ptr addrspace(1) %out
  ret void
}

; GFX10PLUS-LABEL: {{^}}dpp8_v3i32:
; GFX10PLUS-DAG: v_mov_b32_dpp v0, v0 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: v_mov_b32_dpp v1, v1 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: v_mov_b32_dpp v2, v2 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: global_store_{{dwordx3|b96}} v[3:4], v[0:2], off
define amdgpu_ps void @dpp8_v3i32(<3 x i32> %in, ptr addrspace(1) %out) {
  %tmp0 = call <3 x i32> @llvm.amdgcn.mov.dpp8.v3i32(<3 x i32> %in, i32 1)
  store <3 x i32> %tmp0, ptr addrspace(1) %out
  ret void
}

; GFX10PLUS-LABEL: {{^}}dpp8_v4i32:
; GFX10PLUS-DAG: v_mov_b32_dpp v0, v0 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: v_mov_b32_dpp v1, v1 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: v_mov_b32_dpp v2, v2 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: v_mov_b32_dpp v3, v3 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: global_store_{{dwordx4|b128}} v[4:5], v[0:3], off
define amdgpu_ps void @dpp8_v4i32(<4 x i32> %in, ptr addrspace(1) %out) {
  %tmp0 = call <4 x i32> @llvm.amdgcn.mov.dpp8.v3i32(<4 x i32> %in, i32 1)
  store <4 x i32> %tmp0, ptr addrspace(1) %out
  ret void
}

; GFX10PLUS-LABEL: {{^}}dpp8_p0:
; GFX10PLUS-DAG: v_mov_b32_dpp v0, v0 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: v_mov_b32_dpp v1, v1 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: global_store_{{dwordx2|b64}} v[2:3], v[0:1], off
define amdgpu_ps void @dpp8_p0(ptr %in, ptr addrspace(1) %out) {
  %tmp0 = call ptr @llvm.amdgcn.mov.dpp8.p0(ptr %in, i32 1)
  store ptr %tmp0, ptr addrspace(1) %out
  ret void
}

; GFX10PLUS-LABEL: {{^}}dpp8_p3:
; GFX10PLUS-DAG: v_mov_b32_dpp v0, v0 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: global_store_{{dword|b32}} v[1:2], v0, off
define amdgpu_ps void @dpp8_p3(ptr addrspace(3) %in, ptr addrspace(1) %out) {
  %tmp0 = call ptr addrspace(3) @llvm.amdgcn.mov.dpp8.v3p3(ptr addrspace(3) %in, i32 1)
  store ptr addrspace(3) %tmp0, ptr addrspace(1) %out
  ret void
}

; GFX10PLUS-LABEL: {{^}}dpp8_v3p3:
; GFX10PLUS-DAG: v_mov_b32_dpp v0, v0 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: v_mov_b32_dpp v1, v1 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: v_mov_b32_dpp v2, v2 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: global_store_{{dwordx3|b96}} v[3:4], v[0:2], off
define amdgpu_ps void @dpp8_v3p3(<3 x ptr addrspace(3)> %in, ptr addrspace(1) %out) {
  %tmp0 = call <3 x ptr addrspace(3)> @llvm.amdgcn.mov.dpp8.v3p3(<3 x ptr addrspace(3)> %in, i32 1)
  store <3 x ptr addrspace(3)> %tmp0, ptr addrspace(1) %out
  ret void
}

; GFX10PLUS-LABEL: {{^}}dpp8_i16:
; GFX10PLUS-DAG: v_mov_b32_dpp v0, v0 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: global_store_{{short|b16}} v[1:2], v0, off
define amdgpu_ps void @dpp8_i16(i16 %in, ptr addrspace(1) %out) {
  %tmp0 = call i16 @llvm.amdgcn.mov.dpp8.i16(i16 %in, i32 1)
  store i16 %tmp0, ptr addrspace(1) %out
  ret void
}

; GFX10PLUS-LABEL: {{^}}dpp8_v4i16:
; GFX10PLUS-DAG: v_mov_b32_dpp v0, v0 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: v_mov_b32_dpp v1, v1 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: global_store_{{dwordx2|b64}} v[2:3], v[0:1], off
define amdgpu_ps void @dpp8_v4i16(<4 x i16> %in, ptr addrspace(1) %out) {
  %tmp0 = call <4 x i16> @llvm.amdgcn.mov.dpp8.v4i16(<4 x i16> %in, i32 1)
  store <4 x i16> %tmp0, ptr addrspace(1) %out
  ret void
}

; GFX10PLUS-LABEL: {{^}}dpp8_v4f16:
; GFX10PLUS-DAG: v_mov_b32_dpp v0, v0 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: v_mov_b32_dpp v1, v1 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: global_store_{{dwordx2|b64}} v[2:3], v[0:1], off
define amdgpu_ps void @dpp8_v4f16(<4 x half> %in, ptr addrspace(1) %out) {
  %tmp0 = call <4 x half> @llvm.amdgcn.mov.dpp8.v4f16(<4 x half> %in, i32 1)
  store <4 x half> %tmp0, ptr addrspace(1) %out
  ret void
}

; GFX10PLUS-LABEL: {{^}}dpp8_float:
; GFX10PLUS-DAG: v_mov_b32_dpp v0, v0 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: global_store_{{dword|b32}} v[1:2], v0, off
define amdgpu_ps void @dpp8_float(float %in, ptr addrspace(1) %out) {
  %tmp0 = call float @llvm.amdgcn.mov.dpp8.f32(float %in, i32 1)
  store float %tmp0, ptr addrspace(1) %out
  ret void
}

; GFX10PLUS-LABEL: {{^}}dpp8_v3f32:
; GFX10PLUS-DAG: v_mov_b32_dpp v0, v0 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: v_mov_b32_dpp v1, v1 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: v_mov_b32_dpp v2, v2 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: global_store_{{dwordx3|b96}} v[3:4], v[0:2], off
define amdgpu_ps void @dpp8_v3f32(<3 x float> %in, ptr addrspace(1) %out) {
  %tmp0 = call <3 x float> @llvm.amdgcn.mov.dpp8.v3f32(<3 x float> %in, i32 1)
  store <3 x float> %tmp0, ptr addrspace(1) %out
  ret void
}

; GFX10PLUS-LABEL: {{^}}dpp8_half:
; GFX10PLUS-DAG: v_mov_b32_dpp v0, v0 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: global_store_{{short|b16}} v[1:2], v0, off
define amdgpu_ps void @dpp8_half(half %in, ptr addrspace(1) %out) {
  %tmp0 = call half @llvm.amdgcn.mov.dpp8.f16(half %in, i32 1)
  store half %tmp0, ptr addrspace(1) %out
  ret void
}

; GFX10PLUS-LABEL: {{^}}dpp8_bfloat:
; GFX10PLUS-DAG: v_mov_b32_dpp v0, v0 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: global_store_{{short|b16}} v[1:2], v0, off
define amdgpu_ps void @dpp8_bfloat(bfloat %in, ptr addrspace(1) %out) {
  %tmp0 = call bfloat @llvm.amdgcn.mov.dpp8.bf16(bfloat %in, i32 1)
  store bfloat %tmp0, ptr addrspace(1) %out
  ret void
}

; GFX10PLUS-LABEL: {{^}}dpp8_v4bf16:
; GFX10PLUS-DAG: v_mov_b32_dpp v0, v0 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: v_mov_b32_dpp v1, v1 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: global_store_{{dwordx2|b64}} v[2:3], v[0:1], off
define amdgpu_ps void @dpp8_v4bf16(<4 x bfloat> %in, ptr addrspace(1) %out) {
  %tmp0 = call <4 x bfloat> @llvm.amdgcn.mov.dpp8.v4bf16(<4 x bfloat> %in, i32 1)
  store <4 x bfloat> %tmp0, ptr addrspace(1) %out
  ret void
}

; GFX10PLUS-LABEL: {{^}}dpp8_double:
; GFX10PLUS-DAG: v_mov_b32_dpp v0, v0 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: v_mov_b32_dpp v1, v1 dpp8:[1,0,0,0,0,0,0,0]
; GFX10PLUS-DAG: global_store_{{dwordx2|b64}} v[2:3], v[0:1], off
define amdgpu_ps void @dpp8_double(double %in, ptr addrspace(1) %out) {
  %tmp0 = call double @llvm.amdgcn.mov.dpp8.f64(double %in, i32 1)
  store double %tmp0, ptr addrspace(1) %out
  ret void
}

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

attributes #0 = { nounwind readnone convergent }