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
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
|
;=========================== begin_copyright_notice ============================
;
; Copyright (C) 2024 Intel Corporation
;
; SPDX-License-Identifier: MIT
;
;============================ end_copyright_notice =============================
;
; RUN: igc_opt --typed-pointers -S -o - -igc-joint-matrix-resolution --platformpvc 2>&1 < %s | FileCheck %s --implicit-check-not error:
; ------------------------------------------------
; JointMatrixFuncsResolutionPass
; ------------------------------------------------
%spirv.CooperativeMatrixKHR._float_3_1_64_2 = type opaque
%spirv.CooperativeMatrixKHR._short_3_1_16_0 = type opaque
%spirv.CooperativeMatrixKHR._short_3_16_64_1 = type opaque
define spir_kernel void @mad_builtin_bfloat16_1x64x16(i8 addrspace(1)* %src, i64 %stride, float addrspace(1)* %dst) {
; CHECK-LABEL: @mad_builtin_bfloat16_1x64x16(
; CHECK-NEXT: [[TMP1:%.*]] = alloca <4 x float>
; CHECK-NEXT: [[TMP2:%.*]] = alloca i16
; CHECK-NEXT: [[TMP3:%.*]] = alloca <32 x i32>
; CHECK-NEXT: [[TMP4:%.*]] = alloca <4 x float>
; CHECK-NEXT: [[TMP5:%.*]] = alloca <4 x float>
; CHECK-NEXT: store i16 1, i16* [[TMP2]]
; CHECK-NEXT: store <32 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1>, <32 x i32>* [[TMP3]]
; CHECK-NEXT: store <4 x float> zeroinitializer, <4 x float>* [[TMP4]]
; CHECK-NEXT: [[TMP6:%.*]] = bitcast i16* [[TMP2]] to i8*
; CHECK-NEXT: [[TMP7:%.*]] = bitcast <32 x i32>* [[TMP3]] to i8*
; CHECK-NEXT: [[TMP8:%.*]] = bitcast <4 x float>* [[TMP4]] to i8*
; CHECK-NEXT: [[TMP9:%.*]] = bitcast <4 x float>* [[TMP5]] to i8*
; CHECK-NEXT: call void @__builtin_spriv_OpJointMatrixMadINTEL_1x64x16_bf16_bf16_fp32(i8* [[TMP6]], i8* [[TMP7]], i8* [[TMP8]], i8* [[TMP9]])
; CHECK-NEXT: [[TMP10:%.*]] = load <4 x float>, <4 x float>* [[TMP5]]
; CHECK-NEXT: store <4 x float> [[TMP10]], <4 x float>* [[TMP1]]
; CHECK-NEXT: [[TMP11:%.*]] = bitcast <4 x float>* [[TMP1]] to i8*
; CHECK-NEXT: call void @__builtin_spriv_OpJointMatrixStoreINTEL_Accumulator_RowMajor_SG16_1x64_i32_4_global_pi64_v8i8(float addrspace(1)* [[DST:%.*]], i8* [[TMP11]], i64 [[STRIDE:%.*]], i32 0)
; CHECK-NEXT: ret void
;
%1 = call spir_func %spirv.CooperativeMatrixKHR._float_3_1_64_2 addrspace(1)* @_Z26__spirv_CompositeConstructf_c1x64(float 0.0)
%2 = call spir_func %spirv.CooperativeMatrixKHR._short_3_1_16_0 addrspace(1)* @_Z76__spirv_CompositeConstructs_a1x16(i16 1)
%3 = call spir_func %spirv.CooperativeMatrixKHR._short_3_16_64_1 addrspace(1)* @_Z80__spirv_CompositeConstructs_b16x64(i16 -1)
%4 = call spir_func %spirv.CooperativeMatrixKHR._float_3_1_64_2 addrspace(1)* @_Z34__spirv_CooperativeMatrixMulAddKHRPU3AS141__spirv_CooperativeMatrixKHR__1x64x16(%spirv.CooperativeMatrixKHR._short_3_1_16_0 addrspace(1)* %2, %spirv.CooperativeMatrixKHR._short_3_16_64_1 addrspace(1)* %3, %spirv.CooperativeMatrixKHR._float_3_1_64_2 addrspace(1)* %1, i32 64)
call spir_func void @_Z33__spirv_CooperativeMatrixStoreKHRPU3AS1iPU3AS140__spirv_CooperativeMatrixKHR__float_3_1_64_2(float addrspace(1)* %dst, %spirv.CooperativeMatrixKHR._float_3_1_64_2 addrspace(1)* %4, i32 0, i64 %stride, i32 3)
ret void
}
declare spir_func %spirv.CooperativeMatrixKHR._float_3_1_64_2 addrspace(1)* @_Z26__spirv_CompositeConstructf_c1x64(float)
declare spir_func %spirv.CooperativeMatrixKHR._short_3_1_16_0 addrspace(1)* @_Z76__spirv_CompositeConstructs_a1x16(i16)
declare spir_func %spirv.CooperativeMatrixKHR._short_3_16_64_1 addrspace(1)* @_Z80__spirv_CompositeConstructs_b16x64(i16)
declare spir_func %spirv.CooperativeMatrixKHR._float_3_1_64_2 addrspace(1)* @_Z34__spirv_CooperativeMatrixMulAddKHRPU3AS141__spirv_CooperativeMatrixKHR__1x64x16(%spirv.CooperativeMatrixKHR._short_3_1_16_0 addrspace(1)*, %spirv.CooperativeMatrixKHR._short_3_16_64_1 addrspace(1)*, %spirv.CooperativeMatrixKHR._float_3_1_64_2 addrspace(1)*, i32)
declare spir_func void @_Z33__spirv_CooperativeMatrixStoreKHRPU3AS1iPU3AS140__spirv_CooperativeMatrixKHR__float_3_1_64_2(float addrspace(1)*, %spirv.CooperativeMatrixKHR._float_3_1_64_2 addrspace(1)*, i32, i64, i32)
%spirv.CooperativeMatrixKHR._short_3_1_32_0 = type opaque
%spirv.CooperativeMatrixKHR._short_3_32_64_1 = type opaque
define spir_kernel void @mad_builtin_bfloat16_1x64x32(i8 addrspace(1)* %src, i64 %stride, float addrspace(1)* %dst) {
; CHECK-LABEL: @mad_builtin_bfloat16_1x64x32(
; CHECK-NEXT: [[TMP1:%.*]] = alloca <4 x float>
; CHECK-NEXT: [[TMP2:%.*]] = alloca <2 x i16>
; CHECK-NEXT: [[TMP3:%.*]] = alloca <64 x i32>
; CHECK-NEXT: [[TMP4:%.*]] = alloca <4 x float>
; CHECK-NEXT: [[TMP5:%.*]] = alloca <4 x float>
; CHECK-NEXT: store <2 x i16> <i16 1, i16 1>, <2 x i16>* [[TMP2]]
; CHECK-NEXT: store <64 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1>, <64 x i32>* [[TMP3]]
; CHECK-NEXT: store <4 x float> zeroinitializer, <4 x float>* [[TMP4]]
; CHECK-NEXT: [[TMP6:%.*]] = bitcast <2 x i16>* [[TMP2]] to i8*
; CHECK-NEXT: [[TMP7:%.*]] = bitcast <64 x i32>* [[TMP3]] to i8*
; CHECK-NEXT: [[TMP8:%.*]] = bitcast <4 x float>* [[TMP4]] to i8*
; CHECK-NEXT: [[TMP9:%.*]] = bitcast <4 x float>* [[TMP5]] to i8*
; CHECK-NEXT: call void @__builtin_spriv_OpJointMatrixMadINTEL_1x64x32_bf16_bf16_fp32(i8* [[TMP6]], i8* [[TMP7]], i8* [[TMP8]], i8* [[TMP9]])
; CHECK-NEXT: [[TMP10:%.*]] = load <4 x float>, <4 x float>* [[TMP5]]
; CHECK-NEXT: store <4 x float> [[TMP10]], <4 x float>* [[TMP1]]
; CHECK-NEXT: [[TMP11:%.*]] = bitcast <4 x float>* [[TMP1]] to i8*
; CHECK-NEXT: call void @__builtin_spriv_OpJointMatrixStoreINTEL_Accumulator_RowMajor_SG16_1x64_i32_4_global_pi64_v8i8(float addrspace(1)* [[DST:%.*]], i8* [[TMP11]], i64 [[STRIDE:%.*]], i32 0)
; CHECK-NEXT: ret void
;
%1 = call spir_func %spirv.CooperativeMatrixKHR._float_3_1_64_2 addrspace(1)* @_Z26__spirv_CompositeConstructf_c1x64(float 0.0)
%2 = call spir_func %spirv.CooperativeMatrixKHR._short_3_1_32_0 addrspace(1)* @_Z76__spirv_CompositeConstructs_a1x32(i16 1)
%3 = call spir_func %spirv.CooperativeMatrixKHR._short_3_32_64_1 addrspace(1)* @_Z80__spirv_CompositeConstructs_b32x64(i16 -1)
%4 = call spir_func %spirv.CooperativeMatrixKHR._float_3_1_64_2 addrspace(1)* @_Z34__spirv_CooperativeMatrixMulAddKHRPU3AS141__spirv_CooperativeMatrixKHR__1x64x32(%spirv.CooperativeMatrixKHR._short_3_1_32_0 addrspace(1)* %2, %spirv.CooperativeMatrixKHR._short_3_32_64_1 addrspace(1)* %3, %spirv.CooperativeMatrixKHR._float_3_1_64_2 addrspace(1)* %1, i32 64)
call spir_func void @_Z33__spirv_CooperativeMatrixStoreKHRPU3AS1iPU3AS140__spirv_CooperativeMatrixKHR__float_3_1_64_2(float addrspace(1)* %dst, %spirv.CooperativeMatrixKHR._float_3_1_64_2 addrspace(1)* %4, i32 0, i64 %stride, i32 3)
ret void
}
declare spir_func %spirv.CooperativeMatrixKHR._short_3_1_32_0 addrspace(1)* @_Z76__spirv_CompositeConstructs_a1x32(i16)
declare spir_func %spirv.CooperativeMatrixKHR._short_3_32_64_1 addrspace(1)* @_Z80__spirv_CompositeConstructs_b32x64(i16)
declare spir_func %spirv.CooperativeMatrixKHR._float_3_1_64_2 addrspace(1)* @_Z34__spirv_CooperativeMatrixMulAddKHRPU3AS141__spirv_CooperativeMatrixKHR__1x64x32(%spirv.CooperativeMatrixKHR._short_3_1_32_0 addrspace(1)*, %spirv.CooperativeMatrixKHR._short_3_32_64_1 addrspace(1)*, %spirv.CooperativeMatrixKHR._float_3_1_64_2 addrspace(1)*, i32)
%spirv.CooperativeMatrixKHR._float_3_16_16_2 = type opaque
%spirv.CooperativeMatrixKHR._short_3_16_16_0 = type opaque
%spirv.CooperativeMatrixKHR._short_3_16_16_1 = type opaque
define spir_kernel void @mad_builtin_bfloat16(i8 addrspace(1)* %src, i64 %stride, float addrspace(1)* %dst) {
; CHECK-LABEL: @mad_builtin_bfloat16(
; CHECK-NEXT: [[TMP1:%.*]] = alloca <16 x float>
; CHECK-NEXT: [[TMP2:%.*]] = alloca <16 x i16>
; CHECK-NEXT: [[TMP3:%.*]] = alloca <8 x i32>
; CHECK-NEXT: [[TMP4:%.*]] = alloca <16 x float>
; CHECK-NEXT: [[TMP5:%.*]] = alloca <16 x float>
; CHECK-NEXT: store <16 x i16> <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>, <16 x i16>* [[TMP2]]
; CHECK-NEXT: store <8 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1>, <8 x i32>* [[TMP3]]
; CHECK-NEXT: store <16 x float> zeroinitializer, <16 x float>* [[TMP4]]
; CHECK-NEXT: [[TMP6:%.*]] = bitcast <16 x i16>* [[TMP2]] to i8*
; CHECK-NEXT: [[TMP7:%.*]] = bitcast <8 x i32>* [[TMP3]] to i8*
; CHECK-NEXT: [[TMP8:%.*]] = bitcast <16 x float>* [[TMP4]] to i8*
; CHECK-NEXT: [[TMP9:%.*]] = bitcast <16 x float>* [[TMP5]] to i8*
; CHECK-NEXT: call void @__builtin_spriv_OpJointMatrixMadINTEL_16x16x16_bf16_bf16_fp32(i8* [[TMP6]], i8* [[TMP7]], i8* [[TMP8]], i8* [[TMP9]])
; CHECK-NEXT: [[TMP10:%.*]] = load <16 x float>, <16 x float>* [[TMP5]]
; CHECK-NEXT: store <16 x float> [[TMP10]], <16 x float>* [[TMP1]]
; CHECK-NEXT: [[TMP11:%.*]] = bitcast <16 x float>* [[TMP1]] to i8*
; CHECK-NEXT: call void @__builtin_spriv_OpJointMatrixStoreINTEL_Accumulator_RowMajor_SG16_16x16_i32_16_global_pi64_v8i8(float addrspace(1)* [[DST:%.*]], i8* [[TMP11]], i64 [[STRIDE:%.*]], i32 0)
; CHECK-NEXT: ret void
;
%1 = call spir_func %spirv.CooperativeMatrixKHR._float_3_16_16_2 addrspace(1)* @_Z26__spirv_CompositeConstructf(float 0.0)
%2 = call spir_func %spirv.CooperativeMatrixKHR._short_3_16_16_0 addrspace(1)* @_Z76__spirv_CompositeConstructs(i16 1)
%3 = call spir_func %spirv.CooperativeMatrixKHR._short_3_16_16_1 addrspace(1)* @_Z80__spirv_CompositeConstructs(i16 -1)
%4 = call spir_func %spirv.CooperativeMatrixKHR._float_3_16_16_2 addrspace(1)* @_Z34__spirv_CooperativeMatrixMulAddKHRPU3AS141__spirv_CooperativeMatrixKHR__short_3_16_16_0PU3AS142__spirv_CooperativeMatrixKHR__short_3_16_16_1PU3AS140__spirv_CooperativeMatrixKHR__float_3_16_16_2i(%spirv.CooperativeMatrixKHR._short_3_16_16_0 addrspace(1)* %2, %spirv.CooperativeMatrixKHR._short_3_16_16_1 addrspace(1)* %3, %spirv.CooperativeMatrixKHR._float_3_16_16_2 addrspace(1)* %1, i32 64)
call spir_func void @_Z33__spirv_CooperativeMatrixStoreKHRPU3AS1iPU3AS140__spirv_CooperativeMatrixKHR__float_3_16_16_2ili(float addrspace(1)* %dst, %spirv.CooperativeMatrixKHR._float_3_16_16_2 addrspace(1)* %4, i32 0, i64 %stride, i32 3)
ret void
}
declare spir_func %spirv.CooperativeMatrixKHR._float_3_16_16_2 addrspace(1)* @_Z26__spirv_CompositeConstructf(float)
declare spir_func %spirv.CooperativeMatrixKHR._short_3_16_16_0 addrspace(1)* @_Z76__spirv_CompositeConstructs(i16)
declare spir_func %spirv.CooperativeMatrixKHR._short_3_16_16_1 addrspace(1)* @_Z80__spirv_CompositeConstructs(i16)
declare spir_func %spirv.CooperativeMatrixKHR._float_3_16_16_2 addrspace(1)* @_Z34__spirv_CooperativeMatrixMulAddKHRPU3AS141__spirv_CooperativeMatrixKHR__short_3_16_16_0PU3AS142__spirv_CooperativeMatrixKHR__short_3_16_16_1PU3AS140__spirv_CooperativeMatrixKHR__float_3_16_16_2i(%spirv.CooperativeMatrixKHR._short_3_16_16_0 addrspace(1)*, %spirv.CooperativeMatrixKHR._short_3_16_16_1 addrspace(1)*, %spirv.CooperativeMatrixKHR._float_3_16_16_2 addrspace(1)*, i32)
declare spir_func void @_Z33__spirv_CooperativeMatrixStoreKHRPU3AS1iPU3AS140__spirv_CooperativeMatrixKHR__float_3_16_16_2ili(float addrspace(1)*, %spirv.CooperativeMatrixKHR._float_3_16_16_2 addrspace(1)*, i32, i64, i32)
%spirv.CooperativeMatrixKHR._float_3_32_64_2 = type opaque
%spirv.CooperativeMatrixKHR._short_3_32_16_0 = type opaque
define spir_kernel void @mad_builtin_bfloat16_32x64x16(i8 addrspace(1)* %src, i64 %stride, float addrspace(1)* %dst) {
; CHECK-LABEL: @mad_builtin_bfloat16_32x64x16(
; CHECK-NEXT: [[TMP1:%.*]] = alloca { <64 x float>, <64 x float> }
; CHECK-NEXT: [[TMP2:%.*]] = alloca <32 x i16>
; CHECK-NEXT: [[TMP3:%.*]] = alloca <32 x i32>
; CHECK-NEXT: [[TMP4:%.*]] = alloca { <64 x float>, <64 x float> }
; CHECK-NEXT: [[TMP5:%.*]] = alloca { <64 x float>, <64 x float> }
; CHECK-NEXT: store <32 x i16> <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>, <32 x i16>* [[TMP2]]
; CHECK-NEXT: store <32 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1>, <32 x i32>* [[TMP3]]
; CHECK-NEXT: store { <64 x float>, <64 x float> } zeroinitializer, { <64 x float>, <64 x float> }* [[TMP4]]
; CHECK-NEXT: [[TMP6:%.*]] = bitcast <32 x i16>* [[TMP2]] to i8*
; CHECK-NEXT: [[TMP7:%.*]] = bitcast <32 x i32>* [[TMP3]] to i8*
; CHECK-NEXT: [[TMP8:%.*]] = bitcast { <64 x float>, <64 x float> }* [[TMP4]] to i8*
; CHECK-NEXT: [[TMP9:%.*]] = bitcast { <64 x float>, <64 x float> }* [[TMP5]] to i8*
; CHECK-NEXT: call void @__builtin_spriv_OpJointMatrixMadINTEL_32x64x16_bf16_bf16_fp32(i8* [[TMP6]], i8* [[TMP7]], i8* [[TMP8]], i8* [[TMP9]])
; CHECK-NEXT: [[TMP10:%.*]] = load { <64 x float>, <64 x float> }, { <64 x float>, <64 x float> }* [[TMP5]]
; CHECK-NEXT: store { <64 x float>, <64 x float> } [[TMP10]], { <64 x float>, <64 x float> }* [[TMP1]]
; CHECK-NEXT: [[TMP11:%.*]] = bitcast { <64 x float>, <64 x float> }* [[TMP1]] to i8*
; CHECK-NEXT: call void @__builtin_spriv_OpJointMatrixStoreINTEL_Accumulator_RowMajor_SG16_32x64_i32_128_global_pi64_v8i8(float addrspace(1)* [[DST:%.*]], i8* [[TMP11]], i64 [[STRIDE:%.*]], i32 0)
; CHECK-NEXT: ret void
;
%1 = call spir_func %spirv.CooperativeMatrixKHR._float_3_32_64_2 addrspace(1)* @_Z26__spirv_CompositeConstructf_c32x64(float 0.0)
%2 = call spir_func %spirv.CooperativeMatrixKHR._short_3_32_16_0 addrspace(1)* @_Z76__spirv_CompositeConstructs_a32x16(i16 1)
%3 = call spir_func %spirv.CooperativeMatrixKHR._short_3_16_64_1 addrspace(1)* @_Z80__spirv_CompositeConstructs_b16x64(i16 -1)
%4 = call spir_func %spirv.CooperativeMatrixKHR._float_3_32_64_2 addrspace(1)* @_Z34__spirv_CooperativeMatrixMulAddKHRPU3AS141__spirv_CooperativeMatrixKHR__32x64x16(%spirv.CooperativeMatrixKHR._short_3_32_16_0 addrspace(1)* %2, %spirv.CooperativeMatrixKHR._short_3_16_64_1 addrspace(1)* %3, %spirv.CooperativeMatrixKHR._float_3_32_64_2 addrspace(1)* %1, i32 64)
call spir_func void @_Z33__spirv_CooperativeMatrixStoreKHRPU3AS1iPU3AS140__spirv_CooperativeMatrixKHR__float_3_32_64_2(float addrspace(1)* %dst, %spirv.CooperativeMatrixKHR._float_3_32_64_2 addrspace(1)* %4, i32 0, i64 %stride, i32 3)
ret void
}
declare spir_func %spirv.CooperativeMatrixKHR._float_3_32_64_2 addrspace(1)* @_Z26__spirv_CompositeConstructf_c32x64(float)
declare spir_func %spirv.CooperativeMatrixKHR._short_3_32_16_0 addrspace(1)* @_Z76__spirv_CompositeConstructs_a32x16(i16)
declare spir_func %spirv.CooperativeMatrixKHR._float_3_32_64_2 addrspace(1)* @_Z34__spirv_CooperativeMatrixMulAddKHRPU3AS141__spirv_CooperativeMatrixKHR__32x64x16(%spirv.CooperativeMatrixKHR._short_3_32_16_0 addrspace(1)*, %spirv.CooperativeMatrixKHR._short_3_16_64_1 addrspace(1)*, %spirv.CooperativeMatrixKHR._float_3_32_64_2 addrspace(1)*, i32)
declare spir_func void @_Z33__spirv_CooperativeMatrixStoreKHRPU3AS1iPU3AS140__spirv_CooperativeMatrixKHR__float_3_32_64_2(float addrspace(1)*, %spirv.CooperativeMatrixKHR._float_3_32_64_2 addrspace(1)*, i32, i64, i32)
%spirv.CooperativeMatrixKHR._short_3_32_32_0 = type opaque
define spir_kernel void @mad_builtin_bfloat16_32x64x32(i8 addrspace(1)* %src, i64 %stride, float addrspace(1)* %dst) {
; CHECK-LABEL: @mad_builtin_bfloat16_32x64x32(
; CHECK-NEXT: [[TMP1:%.*]] = alloca { <64 x float>, <64 x float> }
; CHECK-NEXT: [[TMP2:%.*]] = alloca <64 x i16>
; CHECK-NEXT: [[TMP3:%.*]] = alloca <64 x i32>
; CHECK-NEXT: [[TMP4:%.*]] = alloca { <64 x float>, <64 x float> }
; CHECK-NEXT: [[TMP5:%.*]] = alloca { <64 x float>, <64 x float> }
; CHECK-NEXT: store <64 x i16> <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>, <64 x i16>* [[TMP2]]
; CHECK-NEXT: store <64 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1>, <64 x i32>* [[TMP3]]
; CHECK-NEXT: store { <64 x float>, <64 x float> } zeroinitializer, { <64 x float>, <64 x float> }* [[TMP4]]
; CHECK-NEXT: [[TMP6:%.*]] = bitcast <64 x i16>* [[TMP2]] to i8*
; CHECK-NEXT: [[TMP7:%.*]] = bitcast <64 x i32>* [[TMP3]] to i8*
; CHECK-NEXT: [[TMP8:%.*]] = bitcast { <64 x float>, <64 x float> }* [[TMP4]] to i8*
; CHECK-NEXT: [[TMP9:%.*]] = bitcast { <64 x float>, <64 x float> }* [[TMP5]] to i8*
; CHECK-NEXT: call void @__builtin_spriv_OpJointMatrixMadINTEL_32x64x32_bf16_bf16_fp32(i8* [[TMP6]], i8* [[TMP7]], i8* [[TMP8]], i8* [[TMP9]])
; CHECK-NEXT: [[TMP10:%.*]] = load { <64 x float>, <64 x float> }, { <64 x float>, <64 x float> }* [[TMP5]]
; CHECK-NEXT: store { <64 x float>, <64 x float> } [[TMP10]], { <64 x float>, <64 x float> }* [[TMP1]]
; CHECK-NEXT: [[TMP11:%.*]] = bitcast { <64 x float>, <64 x float> }* [[TMP1]] to i8*
; CHECK-NEXT: call void @__builtin_spriv_OpJointMatrixStoreINTEL_Accumulator_RowMajor_SG16_32x64_i32_128_global_pi64_v8i8(float addrspace(1)* [[DST:%.*]], i8* [[TMP11]], i64 [[STRIDE:%.*]], i32 0)
; CHECK-NEXT: ret void
;
%1 = call spir_func %spirv.CooperativeMatrixKHR._float_3_32_64_2 addrspace(1)* @_Z26__spirv_CompositeConstructf_c32x64(float 0.0)
%2 = call spir_func %spirv.CooperativeMatrixKHR._short_3_32_32_0 addrspace(1)* @_Z76__spirv_CompositeConstructs_a32x32(i16 1)
%3 = call spir_func %spirv.CooperativeMatrixKHR._short_3_32_64_1 addrspace(1)* @_Z80__spirv_CompositeConstructs_b32x64(i16 -1)
%4 = call spir_func %spirv.CooperativeMatrixKHR._float_3_32_64_2 addrspace(1)* @_Z34__spirv_CooperativeMatrixMulAddKHRPU3AS141__spirv_CooperativeMatrixKHR__32x64x32(%spirv.CooperativeMatrixKHR._short_3_32_32_0 addrspace(1)* %2, %spirv.CooperativeMatrixKHR._short_3_32_64_1 addrspace(1)* %3, %spirv.CooperativeMatrixKHR._float_3_32_64_2 addrspace(1)* %1, i32 64)
call spir_func void @_Z33__spirv_CooperativeMatrixStoreKHRPU3AS1iPU3AS140__spirv_CooperativeMatrixKHR__float_3_32_64_2(float addrspace(1)* %dst, %spirv.CooperativeMatrixKHR._float_3_32_64_2 addrspace(1)* %4, i32 0, i64 %stride, i32 3)
ret void
}
declare spir_func %spirv.CooperativeMatrixKHR._short_3_32_32_0 addrspace(1)* @_Z76__spirv_CompositeConstructs_a32x32(i16)
declare spir_func %spirv.CooperativeMatrixKHR._float_3_32_64_2 addrspace(1)* @_Z34__spirv_CooperativeMatrixMulAddKHRPU3AS141__spirv_CooperativeMatrixKHR__32x64x32(%spirv.CooperativeMatrixKHR._short_3_32_32_0 addrspace(1)*, %spirv.CooperativeMatrixKHR._short_3_32_64_1 addrspace(1)*, %spirv.CooperativeMatrixKHR._float_3_32_64_2 addrspace(1)*, i32)
!igc.functions = !{!0, !4, !5, !6, !7}
!0 = !{void (i8 addrspace(1)*, i64, float addrspace(1)*)* @mad_builtin_bfloat16_1x64x16, !1}
!1 = !{!2, !3}
!2 = !{!"function_type", i32 0}
!3 = !{!"sub_group_size", i32 16}
!4 = !{void (i8 addrspace(1)*, i64, float addrspace(1)*)* @mad_builtin_bfloat16_1x64x32, !1}
!5 = !{void (i8 addrspace(1)*, i64, float addrspace(1)*)* @mad_builtin_bfloat16, !1}
!6 = !{void (i8 addrspace(1)*, i64, float addrspace(1)*)* @mad_builtin_bfloat16_32x64x16, !1}
!7 = !{void (i8 addrspace(1)*, i64, float addrspace(1)*)* @mad_builtin_bfloat16_32x64x32, !1}
|
