[14_0_X] Introduce edm::Async service, and use it in CUDA and Alpaka modules #45143

makortel · 2024-06-05T14:15:06Z

PR description:

Backport of #44901. Tests done by @fwyzard in #44901 (comment) showed this way of synchronizing host and NVIDIA GPU would improve HLT throughput by about 1.5 % ± 0.3 %.

PR validation:

None beyond #44901

If this PR is a backport please specify the original PR and why you need to backport that PR. If this PR will be backported please specify to which release cycle the backport is meant for:

Backport of #44901

Pretty much only piece that could throw is the destructor of T, and bad things would happen if the while loop would terminate early, so we could just call std::terminate() in case of an exception.

cmsbuild · 2024-06-05T14:15:35Z

A new Pull Request was created by @makortel for CMSSW_14_0_X.

It involves the following packages:

FWCore/Concurrency (core)
FWCore/Framework (core)
FWCore/Services (core)
FWCore/TestModules (core)
FWCore/Utilities (core)
HeterogeneousCore/AlpakaCore (heterogeneous)
HeterogeneousCore/CUDACore (heterogeneous)
HeterogeneousCore/CUDAUtilities (heterogeneous)

@Dr15Jones, @smuzaffar, @makortel, @fwyzard, @cmsbuild can you please review it and eventually sign? Thanks.
@wddgit, @missirol, @felicepantaleo, @fwyzard, @rovere this is something you requested to watch as well.
@rappoccio, @sextonkennedy, @antoniovilela you are the release manager for this.

cms-bot commands are listed here

cmsbuild · 2024-06-05T14:15:36Z

cms-bot internal usage

makortel · 2024-06-05T14:16:17Z

enable gpu

makortel · 2024-06-05T14:16:22Z

@cmsbuild, please test

cmsbuild · 2024-06-05T19:58:24Z

-1

Failed Tests: GpuUnitTests
Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-3fe5da/39717/summary.html
COMMIT: 189d1f8
CMSSW: CMSSW_14_0_X_2024-06-05-1100/el8_amd64_gcc12
Additional Tests: GPU
User test area: For local testing, you can use /cvmfs/cms-ci.cern.ch/week0/cms-sw/cmssw/45143/39717/install.sh to create a dev area with all the needed externals and cmssw changes.

GPU Unit Tests

I found 1 errors in the following unit tests:

---> test testTFVisibleDevicesCUDA had ERRORS

Comparison Summary

Summary:

You potentially added 29 lines to the logs
Reco comparison results: 134 differences found in the comparisons
DQMHistoTests: Total files compared: 48
DQMHistoTests: Total histograms compared: 3338712
DQMHistoTests: Total failures: 2577
DQMHistoTests: Total nulls: 0
DQMHistoTests: Total successes: 3336115
DQMHistoTests: Total skipped: 20
DQMHistoTests: Total Missing objects: 0
DQMHistoSizes: Histogram memory added: 0.0 KiB( 47 files compared)
Checked 202 log files, 165 edm output root files, 48 DQM output files
TriggerResults: no differences found

GPU Comparison Summary

Summary:

No significant changes to the logs found
Reco comparison results: 0 differences found in the comparisons
DQMHistoTests: Total files compared: 3
DQMHistoTests: Total histograms compared: 39740
DQMHistoTests: Total failures: 19
DQMHistoTests: Total nulls: 0
DQMHistoTests: Total successes: 39721
DQMHistoTests: Total skipped: 0
DQMHistoTests: Total Missing objects: 0
DQMHistoSizes: Histogram memory added: 0.0 KiB( 2 files compared)
Checked 8 log files, 10 edm output root files, 3 DQM output files
TriggerResults: no differences found

makortel · 2024-06-05T20:40:54Z

The test failure is

===== Test "testTFVisibleDevicesCUDA" ====
Running .WARNING:tensorflow:From /cvmfs/cms-ib.cern.ch/sw/x86_64/nweek-02840/el8_amd64_gcc12/external/py3-cmsml/0.1.2-db4c88fc7ceb7db3b81affeb6fe4685e/lib/python3.9/site-packages/cmsml/tensorflow/tools.py:120: convert_variables_to_constants (from tensorflow.python.framework.convert_to_constants) is deprecated and will be removed in a future version.
Instructions for updating:
This API was designed for TensorFlow v1. See https://www.tensorflow.org/guide/migrate for instructions on how to migrate your code to TensorFlow v2.
WARNING:tensorflow:From /cvmfs/cms-ib.cern.ch/sw/x86_64/nweek-02840/el8_amd64_gcc12/external/py3-tensorflow/2.12.0-b4ed11643f470f05efabc5bddd5c6e30/lib/python3.9/site-packages/tensorflow/python/framework/convert_to_constants.py:952: extract_sub_graph (from tensorflow.python.framework.graph_util_impl) is deprecated and will be removed in a future version.
Instructions for updating:
This API was designed for TensorFlow v1. See https://www.tensorflow.org/guide/migrate for instructions on how to migrate your code to TensorFlow v2.

46.0

CUDA service enabled: 1
Testing CUDA backend
2024-06-05 19:45:18.727320: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: SSE4.1 SSE4.2 AVX AVX2 FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
2024-06-05 19:45:18.741973: I tensorflow/compiler/mlir/mlir_graph_optimization_pass.cc:353] MLIR V1 optimization pass is not enabled
0 /job:localhost/replica:0/task:0/device:CPU:0 type: CPU
Available devices: 1
F

testVisibleDevicesCUDA.cc:83:Assertion
Test name: testVisibleDevicesCUDA::test
assertion failed
- Expression: response.size() == 2

Failures !!!
Run: 1   Failure total: 1   Failures: 1   Errors: 0

---> test testTFVisibleDevicesCUDA had ERRORS

makortel · 2024-06-05T20:42:34Z

#44375 not being merged (whereas its 14_1_X counterpart #44376 is in master) looks related.

fwyzard · 2024-06-20T08:08:15Z

please test with #44375

fwyzard · 2024-06-20T08:08:57Z

+heterogeneous

cmsbuild · 2024-06-20T15:03:29Z

+1

Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-3fe5da/39986/summary.html
COMMIT: 189d1f8
CMSSW: CMSSW_14_0_X_2024-06-19-2300/el8_amd64_gcc12
Additional Tests: GPU
User test area: For local testing, you can use /cvmfs/cms-ci.cern.ch/week0/cms-sw/cmssw/45143/39986/install.sh to create a dev area with all the needed externals and cmssw changes.

Comparison Summary

Summary:

You potentially added 4 lines to the logs
Reco comparison results: 58 differences found in the comparisons
DQMHistoTests: Total files compared: 48
DQMHistoTests: Total histograms compared: 3342458
DQMHistoTests: Total failures: 1199
DQMHistoTests: Total nulls: 0
DQMHistoTests: Total successes: 3341239
DQMHistoTests: Total skipped: 20
DQMHistoTests: Total Missing objects: 0
DQMHistoSizes: Histogram memory added: 0.0 KiB( 47 files compared)
Checked 202 log files, 165 edm output root files, 48 DQM output files
TriggerResults: no differences found

GPU Comparison Summary

Summary:

No significant changes to the logs found
Reco comparison results: 0 differences found in the comparisons
DQMHistoTests: Total files compared: 3
DQMHistoTests: Total histograms compared: 39740
DQMHistoTests: Total failures: 20
DQMHistoTests: Total nulls: 0
DQMHistoTests: Total successes: 39720
DQMHistoTests: Total skipped: 0
DQMHistoTests: Total Missing objects: 0
DQMHistoSizes: Histogram memory added: 0.0 KiB( 2 files compared)
Checked 8 log files, 10 edm output root files, 3 DQM output files
TriggerResults: no differences found

fwyzard · 2024-06-20T15:12:28Z

@smuzaffar could you sign this backport ?
he 14.1.x PR has been merged a couple of weeks ago.

smuzaffar · 2024-06-20T15:33:08Z

@fwyzard , @Dr15Jones is back from vacations and I hope he will review this

Dr15Jones · 2024-06-20T16:27:54Z

+1

seems to have all the changes to the PR done in 14_1

cmsbuild · 2024-06-20T16:28:18Z

This pull request is fully signed and it will be integrated in one of the next CMSSW_14_0_X IBs (tests are also fine) and once validation in the development release cycle CMSSW_14_1_X is complete. This pull request will now be reviewed by the release team before it's merged. @antoniovilela, @rappoccio, @sextonkennedy (and backports should be raised in the release meeting by the corresponding L2)
Notice This PR was tested with additional Pull Request(s), please also merge them if necessary: #44375

rappoccio · 2024-06-24T14:14:54Z

+1

makortel added 8 commits June 5, 2024 09:11

Add edm::WaitingThreadPool

3bfda4c

Spinloops in WaitingThreadPool test

d362b62

Mark ReusableObjectHolder destructor as noexcept

f12db62

Pretty much only piece that could throw is the destructor of T, and bad things would happen if the while loop would terminate early, so we could just call std::terminate() in case of an exception.

Add edm::Async service base class

e685dcd

Add AsyncService service implementation

cadbf90

Add AsyncService to the set of default services

465f1fe

Use edm::Async service in CUDA modules

5f83a64

Use edm::Async service in Alpaka modules

189d1f8

cmsbuild added this to the CMSSW_14_0_X milestone Jun 5, 2024

cmsbuild added core-pending pending-signatures tests-pending orp-pending heterogeneous-pending labels Jun 5, 2024

makortel mentioned this pull request Jun 5, 2024

Backport to 14_0_X cms-sw/framework-team#934

Closed

cmsbuild added tests-started and removed tests-pending labels Jun 5, 2024

cmsbuild added tests-rejected and removed tests-started labels Jun 5, 2024

cmsbuild added the requires-external label Jun 20, 2024

cmsbuild removed the tests-rejected label Jun 20, 2024

cmsbuild added tests-started heterogeneous-approved and removed heterogeneous-pending labels Jun 20, 2024

fwyzard mentioned this pull request Jun 20, 2024

[backport] Improve handling of TF CUDA tests for 14_0_X #44375

Merged

cmsbuild added tests-approved and removed tests-started labels Jun 20, 2024

cmsbuild added core-approved fully-signed and removed core-pending pending-signatures labels Jun 20, 2024

cmsbuild added orp-approved and removed orp-pending labels Jun 24, 2024

cmsbuild merged commit e551de3 into cms-sw:CMSSW_14_0_X Jun 24, 2024
13 checks passed

makortel deleted the edmAsync_140x branch July 10, 2024 20:25

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[14_0_X] Introduce edm::Async service, and use it in CUDA and Alpaka modules #45143

[14_0_X] Introduce edm::Async service, and use it in CUDA and Alpaka modules #45143

makortel commented Jun 5, 2024

cmsbuild commented Jun 5, 2024

cmsbuild commented Jun 5, 2024 •

edited

Loading

makortel commented Jun 5, 2024

makortel commented Jun 5, 2024

cmsbuild commented Jun 5, 2024

makortel commented Jun 5, 2024

makortel commented Jun 5, 2024

fwyzard commented Jun 20, 2024

fwyzard commented Jun 20, 2024

cmsbuild commented Jun 20, 2024

fwyzard commented Jun 20, 2024

smuzaffar commented Jun 20, 2024

Dr15Jones commented Jun 20, 2024

cmsbuild commented Jun 20, 2024

rappoccio commented Jun 24, 2024

[14_0_X] Introduce edm::Async service, and use it in CUDA and Alpaka modules #45143

[14_0_X] Introduce edm::Async service, and use it in CUDA and Alpaka modules #45143

Conversation

makortel commented Jun 5, 2024

PR description:

PR validation:

If this PR is a backport please specify the original PR and why you need to backport that PR. If this PR will be backported please specify to which release cycle the backport is meant for:

cmsbuild commented Jun 5, 2024

cmsbuild commented Jun 5, 2024 • edited Loading

makortel commented Jun 5, 2024

makortel commented Jun 5, 2024

cmsbuild commented Jun 5, 2024

GPU Unit Tests

Comparison Summary

GPU Comparison Summary

makortel commented Jun 5, 2024

makortel commented Jun 5, 2024

fwyzard commented Jun 20, 2024

fwyzard commented Jun 20, 2024

cmsbuild commented Jun 20, 2024

Comparison Summary

GPU Comparison Summary

fwyzard commented Jun 20, 2024

smuzaffar commented Jun 20, 2024

Dr15Jones commented Jun 20, 2024

cmsbuild commented Jun 20, 2024

rappoccio commented Jun 24, 2024

cmsbuild commented Jun 5, 2024 •

edited

Loading