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
|
;=========================== begin_copyright_notice ============================
;
; Copyright (C) 2024 Intel Corporation
;
; SPDX-License-Identifier: MIT
;
;============================ end_copyright_notice =============================
; REQUIRES: llvm-16-plus
; RUN: igc_opt --opaque-pointers -platformpvc -igc-joint-matrix-resolution -S 2>&1 < %s | FileCheck %s
; ------------------------------------------------
; JointMatrixFuncsResolutionPass
; ------------------------------------------------
; A shorten version of Matrix/element_wise_all_ops_half.cpp (only compare OP remains)
target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v16:16:16-v24:32:32-v32:32:32-v48:64:64-v64:64:64-v96:128:128-v128:128:128-v192:256:256-v256:256:256-v512:512:512-v1024:1024:1024-n8:16:32"
target triple = "spir64-unknown-unknown"
%"class.sycl::_V1::range" = type { %"class.sycl::_V1::detail::array" }
%"class.sycl::_V1::detail::array" = type { [2 x i64] }
%"class.sycl::_V1::detail::half_impl::half" = type { half }
; Function Attrs: nounwind
define spir_kernel void @_ZTS5logicILm8ELm16EE(ptr addrspace(1) align 2 %_arg_accA, ptr byval(%"class.sycl::_V1::range") align 8 %_arg_accA2, ptr byval(%"class.sycl::_V1::range") align 8 %_arg_accA3, i64 %_arg_sg_size) #0 !kernel_arg_addr_space !7 !kernel_arg_access_qual !8 !kernel_arg_type !9 !kernel_arg_type_qual !10 !kernel_arg_base_type !9 !kernel_arg_name !374 !spirv.ParameterDecorations !375 {
entry:
%agg.tmp.i.i = alloca %"class.sycl::_V1::detail::half_impl::half", align 2, !spirv.Decorations !376
%sub_a.i = alloca target("spirv.CooperativeMatrixKHR", half, 3, 8, 16, 0), align 8, !spirv.Decorations !381
%0 = bitcast ptr %_arg_accA2 to ptr
%agg.tmp5.sroa.0.sroa.2.0._arg_accA2.ascast.sroa_idx = getelementptr inbounds i8, ptr %0, i64 8
%1 = bitcast ptr %agg.tmp5.sroa.0.sroa.2.0._arg_accA2.ascast.sroa_idx to ptr
%agg.tmp5.sroa.0.sroa.2.0.copyload = load i64, ptr %1, align 8
%2 = bitcast ptr %_arg_accA3 to ptr
%agg.tmp6.sroa.0.sroa.0.0.copyload = load i64, ptr %2, align 8
%3 = bitcast ptr %_arg_accA3 to ptr
%agg.tmp6.sroa.0.sroa.2.0._arg_accA3.ascast.sroa_idx = getelementptr inbounds i8, ptr %3, i64 8
%4 = bitcast ptr %agg.tmp6.sroa.0.sroa.2.0._arg_accA3.ascast.sroa_idx to ptr
%agg.tmp6.sroa.0.sroa.2.0.copyload = load i64, ptr %4, align 8
%sub_a.ascast.i = addrspacecast ptr %sub_a.i to ptr addrspace(4)
%5 = call spir_func i64 @_Z33__spirv_BuiltInGlobalInvocationIdi(i32 1) #3
%cmp.i16.i = icmp ult i64 %5, 2147483648
call void @llvm.assume(i1 %cmp.i16.i)
%6 = call spir_func i64 @_Z33__spirv_BuiltInGlobalInvocationIdi(i32 0) #3
%cmp.i.i = icmp ult i64 %6, 2147483648
call void @llvm.assume(i1 %cmp.i.i)
%7 = call spir_func i64 @_Z32__spirv_BuiltInLocalInvocationIdi(i32 1) #3
%cmp.i24.i = icmp ult i64 %7, 2147483648
call void @llvm.assume(i1 %cmp.i24.i)
%8 = call spir_func i64 @_Z32__spirv_BuiltInLocalInvocationIdi(i32 0) #3
%cmp.i20.i = icmp ult i64 %8, 2147483648
; CHECK: [[ALLOCA1:%.*]] = alloca <8 x i16>, align 8, !spirv.Decorations
; CHECK: [[ASCAST_1:%.*]] = addrspacecast ptr [[ALLOCA1]] to ptr addrspace(4)
call void @llvm.assume(i1 %cmp.i20.i)
%9 = bitcast ptr %sub_a.i to ptr
call void @llvm.lifetime.start.p0(i64 8, ptr %9)
%10 = bitcast ptr %agg.tmp.i.i to ptr
call void @llvm.lifetime.start.p0(i64 2, ptr %10)
%11 = bitcast ptr %agg.tmp.i.i to ptr
store half 0xH4500, ptr %11, align 2
%call.i28.i = call spir_func target("spirv.CooperativeMatrixKHR", half, 3, 8, 16, 0) @"_Z26__spirv_CompositeConstructP40class.sycl::_V1::detail::half_impl::half"(ptr %agg.tmp.i.i) #0
store target("spirv.CooperativeMatrixKHR", half, 3, 8, 16, 0) %call.i28.i, ptr %sub_a.i, align 8
%12 = bitcast ptr %agg.tmp.i.i to ptr
call void @llvm.lifetime.end.p0(i64 2, ptr %12)
br label %for.cond.i
for.cond.i: ; preds = %_ZZ4testIN4sycl3_V16detail9half_impl4halfEfLm8ELm16EEvvENKUlRS4_E3_clES5_.exit, %entry
%13 = phi target("spirv.CooperativeMatrixKHR", half, 3, 8, 16, 0) [ %call.i28.i, %entry ], [ %.pre, %_ZZ4testIN4sycl3_V16detail9half_impl4halfEfLm8ELm16EEvvENKUlRS4_E3_clES5_.exit ]
%i.0.i = phi i32 [ 0, %entry ], [ %inc.i, %_ZZ4testIN4sycl3_V16detail9half_impl4halfEfLm8ELm16EEvvENKUlRS4_E3_clES5_.exit ]
%conv.i = zext i32 %i.0.i to i64
%call.i23 = call spir_func i64 @_Z34__spirv_CooperativeMatrixLengthKHRPU3AS143__spirv_CooperativeMatrixKHR__half_3_8_16_0(target("spirv.CooperativeMatrixKHR", half, 3, 8, 16, 0) %13) #0
%cmp.i = icmp ugt i64 %call.i23, %conv.i
br i1 %cmp.i, label %for.body.i, label %_ZN4sycl3_V13ext6oneapi12experimental6matrix18joint_matrix_applyINS0_9sub_groupENS0_6detail9half_impl4halfELNS4_3useE0ELm8ELm16ELNS4_6layoutE0ERKZ4testIS9_fLm8ELm16EEvvEUlRS9_E3_EEvT_RNS4_12joint_matrixISH_T0_XT1_EXT2_EXT3_EXT4_EEEOT5_.exit
; CHECK: [[LOAD_1:%.*]] = load <8 x i16>, ptr addrspace(4) [[ASCAST_1]]
; CHECK: [[EXTRACT:%.*]] = extractelement <8 x i16> [[LOAD_1]], i64 %{{.*}}
; CHECK: [[BITCAST_1:%.*]] = bitcast i16 [[EXTRACT]] to half, !joint_matrix_apply !{{.*}}
; CHECK: fcmp une half [[BITCAST_1]], 0xH0000
for.body.i: ; preds = %for.cond.i
%call.i = call spir_func ptr addrspace(4) @_Z19__spirv_AccessChainPU3AS4PU3AS143__spirv_CooperativeMatrixKHR__half_3_8_16_0l(ptr addrspace(4) %sub_a.ascast.i, i64 %conv.i) #0
%14 = load half, ptr addrspace(4) %call.i, align 2
%tobool.i = fcmp une half %14, 0xH0000
br i1 %tobool.i, label %if.then.i, label %_ZZ4testIN4sycl3_V16detail9half_impl4halfEfLm8ELm16EEvvENKUlRS4_E3_clES5_.exit
if.then.i: ; preds = %for.body.i
%or.cond68.i = fcmp ord half %14, 0xH0000
%cmp.i52.i = fcmp ole half %14, 0xH4000
%or.cond69.i = or i1 %or.cond68.i, %cmp.i52.i
br i1 %or.cond69.i, label %if.then14.i, label %_ZZ4testIN4sycl3_V16detail9half_impl4halfEfLm8ELm16EEvvENKUlRS4_E3_clES5_.exit
if.then14.i: ; preds = %if.then.i
%cmp.i54.i = fcmp une half %14, 0xH4000
%15 = fadd half %14, 0xHBC00
%16 = fadd half %15, 0xH3C00
%add.i.i.i.i = select i1 %cmp.i54.i, half %16, half 0xH4000
%cmp.i58.i = fcmp oeq half %14, 0xH4000
%sub.i.i = fadd half %add.i.i.i.i, 0xHC000
%mul.i.i = fmul half %sub.i.i, 0xH4200
%div.i.i = fmul half %mul.i.i, 0xH3800
%add.i.i = fadd half %add.i.i.i.i, 0xH4000
%val.sroa.0.1.i = select i1 %cmp.i58.i, half %div.i.i, half %add.i.i
br label %_ZZ4testIN4sycl3_V16detail9half_impl4halfEfLm8ELm16EEvvENKUlRS4_E3_clES5_.exit
; CHECK: [[LOAD_2:%.*]] = load <8 x i16>, ptr addrspace(4) [[ASCAST_1]]
; CHECK: [[BITCAST_2:%.*]] = bitcast half %{{.*}} to i16
; CHECK: [[INSERT:%.*]] = insertelement <8 x i16> [[LOAD_2]], i16 [[BITCAST_2]], i64 %{{.*}}
; CHECK: store <8 x i16> [[INSERT]], ptr addrspace(4) %{{.*}}
_ZZ4testIN4sycl3_V16detail9half_impl4halfEfLm8ELm16EEvvENKUlRS4_E3_clES5_.exit: ; preds = %if.then14.i, %if.then.i, %for.body.i
%element.i.sroa.0.0 = phi half [ %val.sroa.0.1.i, %if.then14.i ], [ %14, %if.then.i ], [ %14, %for.body.i ]
%call.i26 = call spir_func ptr addrspace(4) @_Z19__spirv_AccessChainPU3AS4PU3AS143__spirv_CooperativeMatrixKHR__half_3_8_16_0l(ptr addrspace(4) %sub_a.ascast.i, i64 %conv.i) #0
store half %element.i.sroa.0.0, ptr addrspace(4) %call.i26, align 2
%inc.i = add nuw nsw i32 %i.0.i, 1, !spirv.Decorations !382
%.pre = load target("spirv.CooperativeMatrixKHR", half, 3, 8, 16, 0), ptr %sub_a.i, align 8
br label %for.cond.i
_ZN4sycl3_V13ext6oneapi12experimental6matrix18joint_matrix_applyINS0_9sub_groupENS0_6detail9half_impl4halfELNS4_3useE0ELm8ELm16ELNS4_6layoutE0ERKZ4testIS9_fLm8ELm16EEvvEUlRS9_E3_EEvT_RNS4_12joint_matrixISH_T0_XT1_EXT2_EXT3_EXT4_EEEOT5_.exit: ; preds = %for.cond.i
%sub5.i = sub nsw i64 %6, %8, !spirv.Decorations !385
%sub.i = sub nsw i64 %5, %7, !spirv.Decorations !385
%mul.i6.i.i.i.i = mul i64 %agg.tmp6.sroa.0.sroa.0.0.copyload, %agg.tmp5.sroa.0.sroa.2.0.copyload
%17 = getelementptr %"class.sycl::_V1::detail::half_impl::half", ptr addrspace(1) %_arg_accA, i64 %mul.i6.i.i.i.i
%add.ptr.i = getelementptr %"class.sycl::_V1::detail::half_impl::half", ptr addrspace(1) %17, i64 %agg.tmp6.sroa.0.sroa.2.0.copyload
%add.i7.i.i.i.i.i = add i64 %mul.i6.i.i.i.i, %agg.tmp6.sroa.0.sroa.2.0.copyload
%idx.neg.i.i = sub i64 0, %add.i7.i.i.i.i.i
%add.ptr.i.i = getelementptr inbounds %"class.sycl::_V1::detail::half_impl::half", ptr addrspace(1) %add.ptr.i, i64 %idx.neg.i.i
%add.ptr.i27.idx = shl nsw i64 %sub.i, 9, !spirv.Decorations !385
%18 = bitcast ptr addrspace(1) %add.ptr.i.i to ptr addrspace(1)
%add.ptr.i27 = getelementptr inbounds i8, ptr addrspace(1) %18, i64 %add.ptr.i27.idx
%div.i = udiv i64 %sub5.i, %_arg_sg_size
%add.ptr.i28.idx = shl i64 %div.i, 5
%add.ptr.i28 = getelementptr inbounds i8, ptr addrspace(1) %add.ptr.i27, i64 %add.ptr.i28.idx
%19 = load target("spirv.CooperativeMatrixKHR", half, 3, 8, 16, 0), ptr %sub_a.i, align 8
call spir_func void @_Z33__spirv_CooperativeMatrixStoreKHRPU3AS1cPU3AS143__spirv_CooperativeMatrixKHR__half_3_8_16_0ili(ptr addrspace(1) %add.ptr.i28, target("spirv.CooperativeMatrixKHR", half, 3, 8, 16, 0) %19, i32 0, i64 32, i32 0) #0
%20 = bitcast ptr %sub_a.i to ptr
call void @llvm.lifetime.end.p0(i64 8, ptr %20)
ret void
}
; Function Attrs: nounwind
declare spir_func void @_Z33__spirv_CooperativeMatrixStoreKHRPU3AS1cPU3AS143__spirv_CooperativeMatrixKHR__half_3_8_16_0ili(ptr addrspace(1), target("spirv.CooperativeMatrixKHR", half, 3, 8, 16, 0), i32, i64, i32) #0
; Function Attrs: nounwind
declare spir_func i64 @_Z34__spirv_CooperativeMatrixLengthKHRPU3AS143__spirv_CooperativeMatrixKHR__half_3_8_16_0(target("spirv.CooperativeMatrixKHR", half, 3, 8, 16, 0)) #0
; Function Attrs: nounwind
declare spir_func ptr addrspace(4) @_Z19__spirv_AccessChainPU3AS4PU3AS143__spirv_CooperativeMatrixKHR__half_3_8_16_0l(ptr addrspace(4), i64) #0
; Function Attrs: nounwind willreturn memory(none)
declare spir_func i64 @_Z33__spirv_BuiltInGlobalInvocationIdi(i32) #3
; Function Attrs: nounwind willreturn memory(none)
declare spir_func i64 @_Z32__spirv_BuiltInLocalInvocationIdi(i32) #3
; Function Attrs: nocallback nofree nosync nounwind willreturn memory(inaccessiblemem: readwrite)
declare void @llvm.assume(i1 noundef) #1
; Function Attrs: nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)
declare void @llvm.lifetime.start.p0(i64 immarg, ptr nocapture) #2
; Function Attrs: nounwind
declare spir_func target("spirv.CooperativeMatrixKHR", half, 3, 8, 16, 0) @"_Z26__spirv_CompositeConstructP40class.sycl::_V1::detail::half_impl::half"(ptr) #0
; Function Attrs: nocallback nofree nosync nounwind willreturn memory(argmem: readwrite)
declare void @llvm.lifetime.end.p0(i64 immarg, ptr nocapture) #2
!spirv.MemoryModel = !{!0}
!spirv.Source = !{!1}
!spirv.Generator = !{!2}
!igc.functions = !{}
!0 = !{i32 2, i32 2}
!1 = !{i32 4, i32 100000}
!2 = !{i16 6, i16 14}
!4 = !{!5}
!5 = !{!"function_type", i32 0}
!6 = !{i32 1, i32 0, i32 0, i32 0}
!7 = !{!"none", !"none", !"none", !"none"}
!8 = !{!"class.sycl::_V1::detail::half_impl::half*", !"class.sycl::_V1::range", !"class.sycl::_V1::range", !"long"}
!9 = !{!"", !"", !"", !""}
!10 = !{!"_arg_accA", !"_arg_accA2", !"_arg_accA4", !"_arg_sg_size"}
!374 = !{!375, !377, !377, !4}
!375 = !{!376}
!376 = !{i32 44, i32 2}
!377 = !{!378, !379}
!378 = !{i32 38, i32 2}
!379 = !{i32 44, i32 8}
!380 = !{!379}
!381 = !{!382, !383}
!382 = !{i32 4469}
!383 = !{i32 4470}
!384 = !{!382}
!385 = !{!383}
|
