Skip to content

Commit 5ab98ea

Browse files
JackAKirkJackAKirk
committed
Checking if single line is acceptable.
Signed-off-by: jack.kirk <[email protected]>
1 parent 29ed54d commit 5ab98ea

File tree

1 file changed

+30
-102
lines changed

1 file changed

+30
-102
lines changed

‎sycl/test/check_device_code/matrix/matrix-nvptx-int8-test.cpp

Lines changed: 30 additions & 102 deletions
Original file line numberDiff line numberDiff line change
@@ -39,27 +39,15 @@ int main() {
3939
joint_matrix<int8_t, matrix_use::b, 16, 16, matrix_layout::row_major>
4040
sub_b;
4141

42-
// CHECK: tail call { i32, i32, i32, i32, i32, i32, i32, i32 }
43-
// @llvm.nvvm.wmma.m16n16k16.load.c.row.stride.s32.p1i32(i32
44-
// addrspace(1)* %add.ptr.i, i32 16) #{{.*}}
42+
// CHECK: tail call { i32, i32, i32, i32, i32, i32, i32, i32 } @llvm.nvvm.wmma.m16n16k16.load.c.row.stride.s32.p1i32(i32 addrspace(1)* %_arg_, i32 16) #{{.*}}
4543
joint_matrix_load(sg, sub_c, accC.get_pointer(), stride);
46-
// CHECK: tail call { i32, i32 }
47-
// @llvm.nvvm.wmma.m16n16k16.load.a.row.stride.s8.p0i32(i32*
48-
// %call.ascast.i.i49.i, i32 16) #{{.*}}
44+
// CHECK: tail call { i32, i32 } @llvm.nvvm.wmma.m16n16k16.load.a.row.stride.s8.p0i32(i32* %call.ascast.i.i52.i, i32 16) #{{.*}}
4945
joint_matrix_load(sg, sub_a, accA.get_pointer(), stride);
50-
// CHECK: tail call { i32, i32 }
51-
// @llvm.nvvm.wmma.m16n16k16.load.b.row.stride.s8.p0i32(i32*
52-
// %call.ascast.i.i.i, i32 16) #{{.*}}
46+
// CHECK: tail call { i32, i32 } @llvm.nvvm.wmma.m16n16k16.load.b.row.stride.s8.p0i32(i32* %call.ascast.i.i.i, i32 16) #{{.*}}
5347
joint_matrix_load(sg, sub_b, accB.get_pointer(), stride);
54-
// CHECK: tail call { i32, i32, i32, i32, i32, i32, i32, i32 }
55-
// @llvm.nvvm.wmma.m16n16k16.mma.row.row.s8(i32 %19, i32 %20, i32 %23,
56-
// i32 %24, i32 %9, i32 %10, i32 %11, i32 %12, i32 %13, i32 %14, i32
57-
// %15, i32 %16) #{{.*}}
48+
// CHECK: tail call { i32, i32, i32, i32, i32, i32, i32, i32 } @llvm.nvvm.wmma.m16n16k16.mma.row.row.s8(i32 %11, i32 %12, i32 %15, i32 %16, i32 %1, i32 %2, i32 %3, i32 %4, i32 %5, i32 %6, i32 %7, i32 %8) #{{.*}}
5849
sub_c = joint_matrix_mad(sg, sub_a, sub_b, sub_c);
59-
// CHECK: tail call void
60-
// @llvm.nvvm.wmma.m16n16k16.store.d.row.stride.s32.p1i32(i32
61-
// addrspace(1)* %add.ptr.i76, i32 %26, i32 %27, i32 %28, i32 %29, i32
62-
// %30, i32 %31, i32 %32, i32 %33, i32 16) #{{.*}}
50+
// CHECK: tail call void @llvm.nvvm.wmma.m16n16k16.store.d.row.stride.s32.p1i32(i32 addrspace(1)* %_arg_14, i32 %18, i32 %19, i32 %20, i32 %21, i32 %22, i32 %23, i32 %24, i32 %25, i32 16) #{{.*}}
6351
joint_matrix_store(sg, sub_c, accD.get_pointer(), stride);
6452
});
6553

@@ -78,27 +66,15 @@ int main() {
7866
joint_matrix<int8_t, matrix_use::b, 16, 16, matrix_layout::col_major>
7967
sub_b;
8068

81-
// CHECK: tail call { i32, i32, i32, i32, i32, i32, i32, i32 }
82-
// @llvm.nvvm.wmma.m16n16k16.load.c.col.stride.s32.p1i32(i32
83-
// addrspace(1)* %add.ptr.i, i32 16) #{{.*}}
69+
// CHECK: tail call { i32, i32, i32, i32, i32, i32, i32, i32 } @llvm.nvvm.wmma.m16n16k16.load.c.col.stride.s32.p1i32(i32 addrspace(1)* %_arg_, i32 16) #{{.*}}
8470
joint_matrix_load(sg, sub_c, accC.get_pointer(), stride);
85-
// CHECK: tail call { i32, i32 }
86-
// @llvm.nvvm.wmma.m16n16k16.load.a.col.stride.s8.p0i32(i32*
87-
// %call.ascast.i.i49.i, i32 16) #{{.*}}
71+
// CHECK: tail call { i32, i32 } @llvm.nvvm.wmma.m16n16k16.load.a.col.stride.s8.p0i32(i32* %call.ascast.i.i52.i, i32 16) #{{.*}}
8872
joint_matrix_load(sg, sub_a, accA.get_pointer(), stride);
89-
// CHECK: tail call { i32, i32 }
90-
// @llvm.nvvm.wmma.m16n16k16.load.b.col.stride.s8.p0i32(i32*
91-
// %call.ascast.i.i.i, i32 16) #{{.*}}
73+
// CHECK: tail call { i32, i32 } @llvm.nvvm.wmma.m16n16k16.load.b.col.stride.s8.p0i32(i32* %call.ascast.i.i.i, i32 16) #{{.*}}
9274
joint_matrix_load(sg, sub_b, accB.get_pointer(), stride);
93-
// CHECK: tail call { i32, i32, i32, i32, i32, i32, i32, i32 }
94-
// @llvm.nvvm.wmma.m16n16k16.mma.col.col.s8(i32 %19, i32 %20, i32 %23,
95-
// i32 %24, i32 %9, i32 %10, i32 %11, i32 %12, i32 %13, i32 %14, i32
96-
// %15, i32 %16) #{{.*}}
75+
// CHECK: tail call { i32, i32, i32, i32, i32, i32, i32, i32 } @llvm.nvvm.wmma.m16n16k16.mma.col.col.s8(i32 %11, i32 %12, i32 %15, i32 %16, i32 %1, i32 %2, i32 %3, i32 %4, i32 %5, i32 %6, i32 %7, i32 %8) #{{.*}}
9776
sub_c = joint_matrix_mad(sg, sub_a, sub_b, sub_c);
98-
// CHECK: tail call void
99-
// @llvm.nvvm.wmma.m16n16k16.store.d.col.stride.s32.p1i32(i32
100-
// addrspace(1)* %add.ptr.i76, i32 %26, i32 %27, i32 %28, i32 %29, i32
101-
// %30, i32 %31, i32 %32, i32 %33, i32 16) #{{.*}}
77+
// CHECK: tail call void @llvm.nvvm.wmma.m16n16k16.store.d.col.stride.s32.p1i32(i32 addrspace(1)* %_arg_14, i32 %18, i32 %19, i32 %20, i32 %21, i32 %22, i32 %23, i32 %24, i32 %25, i32 16) #{{.*}}
10278
joint_matrix_store(sg, sub_c, accD.get_pointer(), stride);
10379
});
10480

@@ -117,27 +93,15 @@ int main() {
11793
joint_matrix<int8_t, matrix_use::b, 16, 8, matrix_layout::row_major>
11894
sub_b;
11995

120-
// CHECK: tail call { i32, i32, i32, i32, i32, i32, i32, i32 }
121-
// @llvm.nvvm.wmma.m32n8k16.load.c.row.stride.s32.p1i32(i32
122-
// addrspace(1)* %add.ptr.i, i32 16) #{{.*}}
96+
// CHECK: tail call { i32, i32, i32, i32, i32, i32, i32, i32 } @llvm.nvvm.wmma.m32n8k16.load.c.row.stride.s32.p1i32(i32 addrspace(1)* %_arg_, i32 16) #{{.*}}
12397
joint_matrix_load(sg, sub_c, accC.get_pointer(), stride);
124-
// CHECK: tail call { i32, i32, i32, i32 }
125-
// @llvm.nvvm.wmma.m32n8k16.load.a.row.stride.s8.p0i32(i32*
126-
// %call.ascast.i.i46.i, i32 16) #{{.*}}
98+
// CHECK: tail call { i32, i32, i32, i32 } @llvm.nvvm.wmma.m32n8k16.load.a.row.stride.s8.p0i32(i32* %call.ascast.i.i49.i, i32 16) #{{.*}}
12799
joint_matrix_load(sg, sub_a, accA.get_pointer(), stride);
128-
// CHECK: tail call i32
129-
// @llvm.nvvm.wmma.m32n8k16.load.b.row.stride.s8.p0i32(i32*
130-
// %call.ascast.i.i.i, i32 16) #{{.*}}
100+
// CHECK: tail call i32 @llvm.nvvm.wmma.m32n8k16.load.b.row.stride.s8.p0i32(i32* %call.ascast.i.i.i, i32 16) #{{.*}}
131101
joint_matrix_load(sg, sub_b, accB.get_pointer(), stride);
132-
// CHECK: tail call { i32, i32, i32, i32, i32, i32, i32, i32 }
133-
// @llvm.nvvm.wmma.m32n8k16.mma.row.row.s8(i32 %19, i32 %20, i32 %21,
134-
// i32 %22, i32 %24, i32 %9, i32 %10, i32 %11, i32 %12, i32 %13, i32
135-
// %14, i32 %15, i32 %16) #{{.*}}
102+
// CHECK: tail call { i32, i32, i32, i32, i32, i32, i32, i32 } @llvm.nvvm.wmma.m32n8k16.mma.row.row.s8(i32 %11, i32 %12, i32 %13, i32 %14, i32 %16, i32 %1, i32 %2, i32 %3, i32 %4, i32 %5, i32 %6, i32 %7, i32 %8) #{{.*}}
136103
sub_c = joint_matrix_mad(sg, sub_a, sub_b, sub_c);
137-
// CHECK: tail call void
138-
// @llvm.nvvm.wmma.m32n8k16.store.d.row.stride.s32.p1i32(i32
139-
// addrspace(1)* %add.ptr.i76, i32 %26, i32 %27, i32 %28, i32 %29, i32
140-
// %30, i32 %31, i32 %32, i32 %33, i32 16) #{{.*}}
104+
// CHECK: tail call void @llvm.nvvm.wmma.m32n8k16.store.d.row.stride.s32.p1i32(i32 addrspace(1)* %_arg_14, i32 %18, i32 %19, i32 %20, i32 %21, i32 %22, i32 %23, i32 %24, i32 %25, i32 16) #{{.*}}
141105
joint_matrix_store(sg, sub_c, accD.get_pointer(), stride);
142106
});
143107

@@ -156,27 +120,15 @@ int main() {
156120
joint_matrix<int8_t, matrix_use::b, 16, 8, matrix_layout::col_major>
157121
sub_b;
158122

159-
// CHECK: tail call { i32, i32, i32, i32, i32, i32, i32, i32 }
160-
// @llvm.nvvm.wmma.m32n8k16.load.c.col.stride.s32.p1i32(i32
161-
// addrspace(1)* %add.ptr.i, i32 16) #{{.*}}
123+
// CHECK: tail call { i32, i32, i32, i32, i32, i32, i32, i32 } @llvm.nvvm.wmma.m32n8k16.load.c.col.stride.s32.p1i32(i32 addrspace(1)* %_arg_, i32 16) #{{.*}}
162124
joint_matrix_load(sg, sub_c, accC.get_pointer(), stride);
163-
// CHECK: tail call { i32, i32, i32, i32 }
164-
// @llvm.nvvm.wmma.m32n8k16.load.a.col.stride.s8.p0i32(i32*
165-
// %call.ascast.i.i46.i, i32 16) #{{.*}}
125+
// CHECK: tail call { i32, i32, i32, i32 } @llvm.nvvm.wmma.m32n8k16.load.a.col.stride.s8.p0i32(i32* %call.ascast.i.i49.i, i32 16) #{{.*}}
166126
joint_matrix_load(sg, sub_a, accA.get_pointer(), stride);
167-
// CHECK: tail call i32
168-
// @llvm.nvvm.wmma.m32n8k16.load.b.col.stride.s8.p0i32(i32*
169-
// %call.ascast.i.i.i, i32 16) #{{.*}}
127+
// CHECK: tail call i32 @llvm.nvvm.wmma.m32n8k16.load.b.col.stride.s8.p0i32(i32* %call.ascast.i.i.i, i32 16) #{{.*}}
170128
joint_matrix_load(sg, sub_b, accB.get_pointer(), stride);
171-
// CHECK: tail call { i32, i32, i32, i32, i32, i32, i32, i32 }
172-
// @llvm.nvvm.wmma.m32n8k16.mma.col.col.s8(i32 %19, i32 %20, i32 %21,
173-
// i32 %22, i32 %24, i32 %9, i32 %10, i32 %11, i32 %12, i32 %13, i32
174-
// %14, i32 %15, i32 %16) #{{.*}}
129+
// CHECK: tail call { i32, i32, i32, i32, i32, i32, i32, i32 } @llvm.nvvm.wmma.m32n8k16.mma.col.col.s8(i32 %11, i32 %12, i32 %13, i32 %14, i32 %16, i32 %1, i32 %2, i32 %3, i32 %4, i32 %5, i32 %6, i32 %7, i32 %8) #{{.*}}
175130
sub_c = joint_matrix_mad(sg, sub_a, sub_b, sub_c);
176-
// CHECK: tail call void
177-
// @llvm.nvvm.wmma.m32n8k16.store.d.col.stride.s32.p1i32(i32
178-
// addrspace(1)* %add.ptr.i76, i32 %26, i32 %27, i32 %28, i32 %29, i32
179-
// %30, i32 %31, i32 %32, i32 %33, i32 16) #{{.*}}
131+
// CHECK: tail call void @llvm.nvvm.wmma.m32n8k16.store.d.col.stride.s32.p1i32(i32 addrspace(1)* %_arg_14, i32 %18, i32 %19, i32 %20, i32 %21, i32 %22, i32 %23, i32 %24, i32 %25, i32 16) #{{.*}}
180132
joint_matrix_store(sg, sub_c, accD.get_pointer(), stride);
181133
});
182134

@@ -195,27 +147,15 @@ int main() {
195147
joint_matrix<int8_t, matrix_use::b, 16, 32, matrix_layout::row_major>
196148
sub_b;
197149

198-
// CHECK: tail call { i32, i32, i32, i32, i32, i32, i32, i32 }
199-
// @llvm.nvvm.wmma.m8n32k16.load.c.row.stride.s32.p1i32(i32
200-
// addrspace(1)* %add.ptr.i, i32 16) #{{.*}}
150+
// CHECK: tail call { i32, i32, i32, i32, i32, i32, i32, i32 } @llvm.nvvm.wmma.m8n32k16.load.c.row.stride.s32.p1i32(i32 addrspace(1)* %_arg_, i32 16) #{{.*}}
201151
joint_matrix_load(sg, sub_c, accC.get_pointer(), stride);
202-
// CHECK: tail call i32
203-
// @llvm.nvvm.wmma.m8n32k16.load.a.row.stride.s8.p0i32(i32*
204-
// %call.ascast.i.i46.i, i32 16) #{{.*}}
152+
// CHECK: tail call i32 @llvm.nvvm.wmma.m8n32k16.load.a.row.stride.s8.p0i32(i32* %call.ascast.i.i49.i, i32 16) #{{.*}}
205153
joint_matrix_load(sg, sub_a, accA.get_pointer(), stride);
206-
// CHECK: tail call { i32, i32, i32, i32 }
207-
// @llvm.nvvm.wmma.m8n32k16.load.b.row.stride.s8.p0i32(i32*
208-
// %call.ascast.i.i.i, i32 16) #{{.*}}
154+
// CHECK: tail call { i32, i32, i32, i32 } @llvm.nvvm.wmma.m8n32k16.load.b.row.stride.s8.p0i32(i32* %call.ascast.i.i.i, i32 16) #{{.*}}
209155
joint_matrix_load(sg, sub_b, accB.get_pointer(), stride);
210-
// CHECK: tail call { i32, i32, i32, i32, i32, i32, i32, i32 }
211-
// @llvm.nvvm.wmma.m8n32k16.mma.row.row.s8(i32 %18, i32 %21, i32 %22,
212-
// i32 %23, i32 %24, i32 %9, i32 %10, i32 %11, i32 %12, i32 %13, i32
213-
// %14, i32 %15, i32 %16) #{{.*}}
156+
// CHECK: tail call { i32, i32, i32, i32, i32, i32, i32, i32 } @llvm.nvvm.wmma.m8n32k16.mma.row.row.s8(i32 %10, i32 %13, i32 %14, i32 %15, i32 %16, i32 %1, i32 %2, i32 %3, i32 %4, i32 %5, i32 %6, i32 %7, i32 %8) #{{.*}}
214157
sub_c = joint_matrix_mad(sg, sub_a, sub_b, sub_c);
215-
// CHECK: tail call void
216-
// @llvm.nvvm.wmma.m8n32k16.store.d.row.stride.s32.p1i32(i32
217-
// addrspace(1)* %add.ptr.i76, i32 %26, i32 %27, i32 %28, i32 %29, i32
218-
// %30, i32 %31, i32 %32, i32 %33, i32 16) #{{.*}}
158+
// CHECK: tail call void @llvm.nvvm.wmma.m8n32k16.store.d.row.stride.s32.p1i32(i32 addrspace(1)* %_arg_14, i32 %18, i32 %19, i32 %20, i32 %21, i32 %22, i32 %23, i32 %24, i32 %25, i32 16) #{{.*}}
219159
joint_matrix_store(sg, sub_c, accD.get_pointer(), stride);
220160
});
221161

@@ -234,27 +174,15 @@ int main() {
234174
joint_matrix<int8_t, matrix_use::b, 16, 32, matrix_layout::col_major>
235175
sub_b;
236176

237-
// CHECK: tail call { i32, i32, i32, i32, i32, i32, i32, i32 }
238-
// @llvm.nvvm.wmma.m8n32k16.load.c.col.stride.s32.p1i32(i32
239-
// addrspace(1)* %add.ptr.i, i32 16) #{{.*}}
177+
// CHECK: tail call { i32, i32, i32, i32, i32, i32, i32, i32 } @llvm.nvvm.wmma.m8n32k16.load.c.col.stride.s32.p1i32(i32 addrspace(1)* %_arg_, i32 16) #{{.*}}
240178
joint_matrix_load(sg, sub_c, accC.get_pointer(), stride);
241-
// CHECK: tail call i32
242-
// @llvm.nvvm.wmma.m8n32k16.load.a.col.stride.s8.p0i32(i32*
243-
// %call.ascast.i.i46.i, i32 16) #{{.*}}
179+
// CHECK: tail call i32 @llvm.nvvm.wmma.m8n32k16.load.a.col.stride.s8.p0i32(i32* %call.ascast.i.i49.i, i32 16) #{{.*}}
244180
joint_matrix_load(sg, sub_a, accA.get_pointer(), stride);
245-
// CHECK: tail call { i32, i32, i32, i32 }
246-
// @llvm.nvvm.wmma.m8n32k16.load.b.col.stride.s8.p0i32(i32*
247-
// %call.ascast.i.i.i, i32 16) #{{.*}}
181+
// CHECK: tail call { i32, i32, i32, i32 } @llvm.nvvm.wmma.m8n32k16.load.b.col.stride.s8.p0i32(i32* %call.ascast.i.i.i, i32 16) #{{.*}}
248182
joint_matrix_load(sg, sub_b, accB.get_pointer(), stride);
249-
// CHECK: tail call { i32, i32, i32, i32, i32, i32, i32, i32 }
250-
// @llvm.nvvm.wmma.m8n32k16.mma.col.col.s8(i32 %18, i32 %21, i32 %22,
251-
// i32 %23, i32 %24, i32 %9, i32 %10, i32 %11, i32 %12, i32 %13, i32
252-
// %14, i32 %15, i32 %16) #{{.*}}
183+
// CHECK: tail call { i32, i32, i32, i32, i32, i32, i32, i32 } @llvm.nvvm.wmma.m8n32k16.mma.col.col.s8(i32 %10, i32 %13, i32 %14, i32 %15, i32 %16, i32 %1, i32 %2, i32 %3, i32 %4, i32 %5, i32 %6, i32 %7, i32 %8) #{{.*}}
253184
sub_c = joint_matrix_mad(sg, sub_a, sub_b, sub_c);
254-
// CHECK: tail call void
255-
// @llvm.nvvm.wmma.m8n32k16.store.d.col.stride.s32.p1i32(i32
256-
// addrspace(1)* %add.ptr.i76, i32 %26, i32 %27, i32 %28, i32 %29, i32
257-
// %30, i32 %31, i32 %32, i32 %33, i32 16) #{{.*}}
185+
// CHECK: tail call void @llvm.nvvm.wmma.m8n32k16.store.d.col.stride.s32.p1i32(i32 addrspace(1)* %_arg_14, i32 %18, i32 %19, i32 %20, i32 %21, i32 %22, i32 %23, i32 %24, i32 %25, i32 16) #{{.*}}
258186
joint_matrix_store(sg, sub_c, accD.get_pointer(), stride);
259187
});
260188
});

0 commit comments

Comments
 (0)