-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathmatmul_gpu_info.py
More file actions
156 lines (130 loc) · 3.43 KB
/
Copy pathmatmul_gpu_info.py
File metadata and controls
156 lines (130 loc) · 3.43 KB
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
Test forward type: 2
Tolerance Rate: 0.050000
Open Model /home/kai.wang/MNN/onnx_model/test_matmul_2d/model.mnn
Create CPU Session:
CPU Group: [ 20 21 31 23 25 17 27 19 29 30 22 28 24 18 16 26 ], 800000 - 4300000
CPU Group: [ 14 6 13 1 15 3 4 5 2 7 12 0 ], 800000 - 5500000
CPU Group: [ 10 11 9 8 ], 800000 - 5800000
The device supports: i8sdot:0, fp16:0, i8mm: 0, sve2: 0, sme2: 0
c Run on 13
0.000000, before Resize: c - 0
before Resize 2, calling: c - 0
Resize 1 op for index: 0
Input: 1,1,4,3
=== Input Tensors Information ===
Tensor Name: default
Shape: [3, 4, 1, 1]
ElementSize: 12
DataType: 2
----------------------------
precision=0 in main, 365
modeNum=1 in main, 370
stopOp.c_str()=s in main, 375
=== Output Tensors Information ===
Op Name: c
Output Tensors:
- c
----------------------------
Input: a, Shape: [3, 4]
Input: b, Shape: [4, 3]
outputName[0]=c
Start Test 0, opName=c
c Run on 2
0.000000, before Resize: c - 0
before Resize 2, calling: c - 0
=== MatMul onResize Debug Info ===
Input Tensor Shapes:
- A: [3, 4]
- B: [4, 3]
Computed Parameters:
- Batch Size: 1
- Total Dimensions: 2
Matrix Dimensions:
- E (Output rows): 3
- L (Inner dimension): 4
- H (Output cols): 3
Optimization Flags:
- Large Batch Small GEMM: 0
Padding Information:
- E padded: 8
- L padded: 8
- H padded: 8
Memory Requirements:
- Use RR Layout: 0
- Need A Temp Buffer: 1
- Need B Temp Buffer: 1
- Need Convert Mat AB: 1
=== MatMul Setup Debug Info ===
Matrix Dimensions:
- M (rows of A): 3
- K (cols of A/rows of B): 4
- N (cols of B): 3
Batch size: 1
Precision Mode: FP16/FP32 Mixed
Layout: Row-Column
GPU Compute Capability: 0
Memory Configuration:
- Need Temp Buffer A: 1
- Need Temp Buffer B: 1
- Has Bias: 0
Tensor Addresses:
- Input A: 0x7fffc6c00000
- Input B: 0x7fffc6c00200
- Output: 0x7fffc6c00400
==============================
==============================
Resize 1 op for index: 0
c Run on 13
0.000000, before Resize: c - 0
before Resize 2, calling: c - 0
Resize 1 op for index: 0
CUDABackend::onExecuteBegin
Group: c - 0, type=MatMul, inputs: input group: [ 0 0 ], devices: input: [ 140736527859712 140736527860224 ] - output: [ 140736527860736 ]
=== MatMul onExecute Debug Info ===
Execution Configuration:
- Precision: FP16/FP32 Mixed
- Layout: Row-Column
- GPU Compute Cap: 89
Matrix Dimensions:
- M (rows): 3
- K (inner): 4
- N (cols): 3
- Batch Size: 1
Memory Addresses:
- Input A: 0x7fffc6c00000
- Input B: 0x7fffc6c00200
- Output: 0x7fffc6c00400
Conversion Status:
- Need Convert MatAB: 1
- Need A Temp Buffer: 1
- Need B Temp Buffer: 1
Preparing Kernel Execution...
==============================
CUDABackend::onExecuteEnd
Group: c - 0, type=MatMul, inputs: input group: [ 0 0 ], devices: input: [ 1 1 ] - output: [ 1 ]
Correct ! Run second pass
CUDABackend::onExecuteBegin
Group: c - 0, type=MatMul, inputs: input group: [ 0 0 ], devices: input: [ 140736527859712 140736527860224 ] - output: [ 140736527860736 ]
=== MatMul onExecute Debug Info ===
Execution Configuration:
- Precision: FP16/FP32 Mixed
- Layout: Row-Column
- GPU Compute Cap: 89
Matrix Dimensions:
- M (rows): 3
- K (inner): 4
- N (cols): 3
- Batch Size: 1
Memory Addresses:
- Input A: 0x7fffc6c00000
- Input B: 0x7fffc6c00200
- Output: 0x7fffc6c00400
Conversion Status:
- Need Convert MatAB: 1
- Need A Temp Buffer: 1
- Need B Temp Buffer: 1
Preparing Kernel Execution...
==============================
CUDABackend::onExecuteEnd
Correct for 0, name=c
Correct !