[
https://issues.apache.org/jira/browse/GEODE-8213?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17142736#comment-17142736
]
ASF GitHub Bot commented on GEODE-8213:
---------------------------------------
albertogpz edited a comment on pull request #621:
URL: https://github.com/apache/geode-native/pull/621#issuecomment-647987941
The shared_lock provided by boost offered a much worse performance than the
spinlock or the mutex (see https://github.com/apache/geode-native/pull/611).
There is a tiny gain according to the benchmark performance test in most
cases using a mutex:
```
Benchmark
Time CPU Time Old Time New CPU Old
CPU New
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------
SerializationRegistryBM_findDataSerializablePrimitive/real_time/threads:1
-0.0997 -0.0998 193 174 193
174
SerializationRegistryBM_findDataSerializablePrimitive/real_time/threads:2
-0.1107 -0.1107 100 89 199
177
SerializationRegistryBM_findDataSerializablePrimitive/real_time/threads:4
-0.0666 -0.0667 51 48 206
192
SerializationRegistryBM_findDataSerializablePrimitive/real_time/threads:8
-0.0151 -0.0160 32 32 259
255
SerializationRegistryBM_findDataSerializablePrimitive/real_time/threads:16
+0.0258 +0.0430 24 25 300
313
SerializationRegistryBM_findDataSerializablePrimitive/real_time/threads:32
+0.0843 +0.0349 22 24 309
320
SerializationRegistryBM_findDataSerializablePrimitive/real_time/threads:64
+0.0283 +0.0359 22 23 312
323
SerializationRegistryBM_findDataSerializablePrimitive/real_time/threads:96
-0.1669 +0.0336 25 21 315
326
SerializationRegistryBM_findDataSerializableFixedId/real_time/threads:1
-0.0925 -0.0925 190 173 190
173
SerializationRegistryBM_findDataSerializableFixedId/real_time/threads:2
-0.0528 -0.0529 98 93 197
186
SerializationRegistryBM_findDataSerializableFixedId/real_time/threads:4
-0.0803 -0.0803 53 49 213
195
SerializationRegistryBM_findDataSerializableFixedId/real_time/threads:8
-0.0801 -0.0802 35 32 279
256
SerializationRegistryBM_findDataSerializableFixedId/real_time/threads:16
-0.0189 -0.0052 26 25 318
316
SerializationRegistryBM_findDataSerializableFixedId/real_time/threads:32
-0.0000 +0.0145 25 25 325
330
SerializationRegistryBM_findDataSerializableFixedId/real_time/threads:64
-0.0363 +0.0179 24 24 327
332
SerializationRegistryBM_findDataSerializableFixedId/real_time/threads:96
-0.0465 +0.0151 24 23 328
333
SerializationRegistryBM_findDataSerializable/real_time/threads:1
-0.0557 -0.0557 193 183 193
183
SerializationRegistryBM_findDataSerializable/real_time/threads:2
-0.0190 -0.0190 98 97 197
193
SerializationRegistryBM_findDataSerializable/real_time/threads:4
-0.0716 -0.0722 56 52 222
206
SerializationRegistryBM_findDataSerializable/real_time/threads:8
-0.1733 -0.1726 37 30 295
244
SerializationRegistryBM_findDataSerializable/real_time/threads:16
-0.0343 -0.0065 26 25 322
319
SerializationRegistryBM_findDataSerializable/real_time/threads:32
+0.0158 +0.0069 25 25 327
329
SerializationRegistryBM_findDataSerializable/real_time/threads:64
-0.0160 +0.0143 25 25 326
331
SerializationRegistryBM_findDataSerializable/real_time/threads:96
+0.1475 +0.0089 22 25 329
332
SerializationRegistryBM_findPdxSerializable/real_time/threads:1
-0.0616 -0.0616 198 186 198
186
SerializationRegistryBM_findPdxSerializable/real_time/threads:2
-0.0794 -0.0795 101 93 202
186
SerializationRegistryBM_findPdxSerializable/real_time/threads:4
-0.0946 -0.0964 57 52 228
206
SerializationRegistryBM_findPdxSerializable/real_time/threads:8
-0.0001 -0.0016 38 38 303
302
SerializationRegistryBM_findPdxSerializable/real_time/threads:16
-0.0356 -0.0215 30 29 375
367
SerializationRegistryBM_findPdxSerializable/real_time/threads:32
-0.0635 -0.0182 29 27 383
376
SerializationRegistryBM_findPdxSerializable/real_time/threads:64
+0.0668 -0.0076 27 28 384
381
SerializationRegistryBM_findPdxSerializable/real_time/threads:96
+0.0098 -0.0051 28 28 386
384
```
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
> C++ native client performance bottleneck in access to serialization registry
> ----------------------------------------------------------------------------
>
> Key: GEODE-8213
> URL: https://issues.apache.org/jira/browse/GEODE-8213
> Project: Geode
> Issue Type: Improvement
> Components: native client
> Reporter: Alberto Gomez
> Assignee: Alberto Gomez
> Priority: Major
> Fix For: 1.14.0
>
>
> It's been observed that when the number of threads used in a Geode client
> using PdxSerialization is greater than 8, there is an important drop in
> performance.
> Analysing the client process behavior with perf, it has been observed a very
> high CPU consumption by a spinlock
> (apache::geode::util::concurrent::spinlock_mutex::lock) used when accessing
> the serialization registry .
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
