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
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx906 -verify-machineinstrs < %s | FileCheck --check-prefix=GFX906 %s
; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx1011 -verify-machineinstrs < %s | FileCheck --check-prefix=GFX10PLUS %s
; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx1012 -verify-machineinstrs < %s | FileCheck --check-prefix=GFX10PLUS %s
; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 -verify-machineinstrs < %s | FileCheck --check-prefix=GFX10PLUS %s
define i32 @v_udot8(i32 %a, i32 %b, i32 %c) {
; GFX906-LABEL: v_udot8:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_dot8_u32_u4 v0, v0, v1, v2
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; GFX10PLUS-LABEL: v_udot8:
; GFX10PLUS: ; %bb.0:
; GFX10PLUS-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10PLUS-NEXT: v_dot8_u32_u4 v0, v0, v1, v2
; GFX10PLUS-NEXT: s_setpc_b64 s[30:31]
%r = call i32 @llvm.amdgcn.udot8(i32 %a, i32 %b, i32 %c, i1 false)
ret i32 %r
}
define i32 @v_udot8_clamp(i32 %a, i32 %b, i32 %c) {
; GFX906-LABEL: v_udot8_clamp:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_dot8_u32_u4 v0, v0, v1, v2 clamp
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; GFX10PLUS-LABEL: v_udot8_clamp:
; GFX10PLUS: ; %bb.0:
; GFX10PLUS-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10PLUS-NEXT: v_dot8_u32_u4 v0, v0, v1, v2 clamp
; GFX10PLUS-NEXT: s_setpc_b64 s[30:31]
%r = call i32 @llvm.amdgcn.udot8(i32 %a, i32 %b, i32 %c, i1 true)
ret i32 %r
}
; FIXME: Fix argument do not let these casts expand
; define i32 @v_udot8_cast_v8i4(<8 x i4> %a, <8 x i4> %b, i32 %c) {
; %a.cast = bitcast <8 x i4> %a to i32
; %b.cast = bitcast <8 x i4> %b to i32
; %r = call i32 @llvm.amdgcn.udot8(i32 %a.cast, i32 %b.cast, i32 %c, i1 false)
; ret i32 %r
; }
define i32 @v_udot8_fnegf32_a(float %a, i32 %b, i32 %c) {
; GFX906-LABEL: v_udot8_fnegf32_a:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_xor_b32_e32 v0, 0x80000000, v0
; GFX906-NEXT: v_dot8_u32_u4 v0, v0, v1, v2
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; GFX10PLUS-LABEL: v_udot8_fnegf32_a:
; GFX10PLUS: ; %bb.0:
; GFX10PLUS-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10PLUS-NEXT: v_xor_b32_e32 v0, 0x80000000, v0
; GFX10PLUS-NEXT: v_dot8_u32_u4 v0, v0, v1, v2
; GFX10PLUS-NEXT: s_setpc_b64 s[30:31]
%neg.a = fneg float %a
%cast.neg.a = bitcast float %neg.a to i32
%r = call i32 @llvm.amdgcn.udot8(i32 %cast.neg.a, i32 %b, i32 %c, i1 false)
ret i32 %r
}
define i32 @v_udot8_fnegv2f16_a(<2 x half> %a, i32 %b, i32 %c) {
; GFX906-LABEL: v_udot8_fnegv2f16_a:
; GFX906: ; %bb.0:
; GFX906-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX906-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
; GFX906-NEXT: v_dot8_u32_u4 v0, v0, v1, v2
; GFX906-NEXT: s_setpc_b64 s[30:31]
;
; GFX10PLUS-LABEL: v_udot8_fnegv2f16_a:
; GFX10PLUS: ; %bb.0:
; GFX10PLUS-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
; GFX10PLUS-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
; GFX10PLUS-NEXT: v_dot8_u32_u4 v0, v0, v1, v2
; GFX10PLUS-NEXT: s_setpc_b64 s[30:31]
%neg.a = fneg <2 x half> %a
%cast.neg.a = bitcast <2 x half> %neg.a to i32
%r = call i32 @llvm.amdgcn.udot8(i32 %cast.neg.a, i32 %b, i32 %c, i1 false)
ret i32 %r
}
declare i32 @llvm.amdgcn.udot8(i32, i32, i32, i1 immarg) #0
attributes #0 = { nounwind readnone speculatable }
|
