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
|
#!/usr/bin/env python
# SPDX-FileCopyrightText: Copyright (c) 2020-2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: LicenseRef-NvidiaProprietary
#
# NVIDIA CORPORATION, its affiliates and licensors retain all intellectual
# property and proprietary rights in and to this material, related
# documentation and any modifications thereto. Any use, reproduction,
# disclosure or distribution of this material and related documentation
# without an express license agreement from NVIDIA CORPORATION or
# its affiliates is strictly prohibited.
import nsysstats
class CudaMemcpySync(nsysstats.ExpertSystemsReport):
display_name = "CUDA Synchronous Memcpy"
usage = f"""{{SCRIPT}}[:<option>[:<option>]...] -- {{DISPLAY_NAME}}
Options:
rows=<limit> - Limit the number of rows returned by the query.
Default is {{ROW_LIMIT}}. Use -1 for no limit.
start=<time> - Display data recorded after the specified time in
nanoseconds.
end=<time> - Display data recorded before the specified time in
nanoseconds.
nvtx=<range[@domain]> - Display data only for the specified NVTX range.
Note that only the first matching record will be considered.
<domain> should only be specified when the range is not in the
default domain. If this option is used along with the 'start'
and/or 'end' options, the explicit times will override the NVTX
range times.
Output: All time values default to nanoseconds
Duration : Duration of memcpy on GPU
Start : Start time of memcpy on GPU
Src Kind : Memcpy source memory kind
Dst Kind : Memcpy destination memory kind
Bytes : Number of bytes transferred
PID : Process identifier
Device ID : GPU device identifier
Context ID : Context identifier
Green Context ID : Green context identifier
Stream ID : Stream identifier
API Name : Runtime API function name
This rule identifies memory transfers that are synchronous. It does not
include cudaMemcpy*() (no Async suffix) occurred within the same device as
well as H2D copy kind with a memory block of 64 KB or less.
"""
message_advice = ("The following are synchronous memory transfers that"
" block the host. This does not include host to device transfers of a"
" memory block of 64 KB or less.\n\n"
"Suggestion: Use cudaMemcpy*Async() APIs instead.")
message_noresult = ("There were no problems detected related to"
" synchronous memcpy operations.")
query_sync_memcpy = """
WITH
sid AS (
SELECT
*
FROM
StringIds
WHERE
value LIKE 'cudaMemcpy%'
AND value NOT LIKE '%Async%'
),
memcpy AS (
SELECT
*
FROM
CUPTI_ACTIVITY_KIND_MEMCPY
WHERE
NOT (bytes <= 64000 AND copyKind == 1)
AND NOT (srcDeviceId IS NOT NULL AND srcDeviceId == dstDeviceId)
)
SELECT
memcpy.end - memcpy.start AS "Duration:dur_ns",
memcpy.start AS "Start:ts_ns",
msrck.label AS "Src Kind",
mdstk.label AS "Dst Kind",
memcpy.bytes AS "Bytes:mem_B",
(memcpy.globalPid >> 24) & 0x00FFFFFF AS "PID",
memcpy.deviceId AS "Device ID",
memcpy.contextId AS "Context ID",
NULLIF(memcpy.greenContextId, 0) AS "Green Context ID",
memcpy.streamId AS "Stream ID",
sid.value AS "API Name",
memcpy.globalPid AS "_Global ID",
memcpy.copyKind AS "_Copy Kind",
'cuda' AS "_API"
FROM
memcpy
JOIN
sid
ON sid.id == runtime.nameId
JOIN
main.CUPTI_ACTIVITY_KIND_RUNTIME AS runtime
ON runtime.correlationId == memcpy.correlationId
LEFT JOIN
ENUM_CUDA_MEM_KIND AS msrck
ON srcKind == msrck.id
LEFT JOIN
ENUM_CUDA_MEM_KIND AS mdstk
ON dstKind == mdstk.id
ORDER BY
1 DESC
LIMIT {ROW_LIMIT}
"""
table_checks = {
'StringIds': '{DBFILE} file does not contain StringIds table.',
'CUPTI_ACTIVITY_KIND_RUNTIME':
"{DBFILE} could not be analyzed because it does not contain the required CUDA data."
" Does the application use CUDA runtime APIs?",
'CUPTI_ACTIVITY_KIND_MEMCPY':
"{DBFILE} could not be analyzed because it does not contain the required CUDA data."
" Does the application use CUDA memcpy APIs?",
'ENUM_CUDA_MEM_KIND':
"{DBFILE} does not contain ENUM_CUDA_MEM_KIND table."
}
table_col_checks = {
'CUPTI_ACTIVITY_KIND_MEMCPY': {
'greenContextId':
"{DBFILE} could not be analyzed due to missing 'greenContextId'."
" Please re-export the report file with a recent version of Nsight Systems."
}
}
def setup(self):
err = super().setup()
if err != None:
return err
self.query = self.query_sync_memcpy.format(ROW_LIMIT = self._row_limit)
if __name__ == "__main__":
CudaMemcpySync.Main()
|
