Test inter process random state transfers #29384
Conversation
This is used to allow inter-process communication of the random number state.
This provides a RandomNumberGenerator object that can easily be used with the TestProcessor to drive the random numbers in a test or for a non-cmsRun process.
Created a test that uses SharedMemory package to have a module within cmsRun pass random state to and from another EDProducer run in an external process.
|
The code-checks are being triggered in jenkins. |
|
+code-checks Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-29384/14495
|
|
A new Pull Request was created by @Dr15Jones (Chris Jones) for master. It involves the following packages: DataFormats/Common @makortel, @smuzaffar, @cmsbuild, @Dr15Jones can you please review it and eventually sign? Thanks. cms-bot commands are listed here |
|
please test |
|
The tests are being triggered in jenkins. |
| RandomNumberGeneratorState& operator=(RandomNumberGeneratorState&&) = default; | ||
|
|
||
| std::vector<unsigned long> state_; | ||
| long seed_; |
There was a problem hiding this comment.
Would encapsulation of the member data make sense here?
There was a problem hiding this comment.
This isn't a 'real' data format since it never is put in a Principal. It is really an internal detail of how the processes communicate. I put it here since I needed a ROOT dictionary to be generated and this class will be used by the FWCore test as well as the upcoming generator code. That is why I just used a struct.
| edm::LuminosityBlockIndex iLumi) { | ||
| edm::Service<edm::RandomNumberGenerator> gen; | ||
| //NOTE: root serialize requires a `void*` not a `void const*` even though it doesn't modify the object | ||
| randSerializer_.serialize(const_cast<edm::RandomNumberGeneratorState&>(iState)); |
There was a problem hiding this comment.
Can we change ROOTSerializer to take void const*? (in a later PR)
There was a problem hiding this comment.
We can. The underlying problem is ROOT's code which uses the 'void*' instead of the 'void const*'.
There was a problem hiding this comment.
We can. The underlying problem is ROOT's code which uses the 'void*' instead of the 'void const*'.
I guessed that would be the case. At least the const_cast would then be isolated and hidden from the rest...
This follows how the RandomNumberGeneratorService treats RanecuEngine when doing replay.
cmsbuild
added
code-checks-pending
tests-pending
and removed
code-checks-approved
tests-started
labels
|
The code-checks are being triggered in jenkins. |
|
+code-checks Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-29384/14497
|
|
Pull request #29384 was updated. @makortel, @smuzaffar, @cmsbuild, @Dr15Jones can you please check and sign again. |
|
+1 |
|
Comparison job queued. |
|
Comparison is ready Comparison Summary:
|
|
please test |
|
The tests are being triggered in jenkins. |
|
+1 |
|
Comparison job queued. |
|
Comparison is ready Comparison Summary:
|
|
+1 |
cmsbuild
added
core-approved
fully-signed
and removed
core-pending
pending-signatures
labels
|
This pull request is fully signed and it will be integrated in one of the next master IBs (tests are also fine). This pull request will now be reviewed by the release team before it's merged. @silviodonato, @dpiparo (and backports should be raised in the release meeting by the corresponding L2) |
|
+1 |
Backport from cms-sw#29384. Additional changes: - FWCore/Integration/bin/interprocess_random.cc: L72,100: disable macro CMS_SA_ALLOW. - FWCore/Services/test/BuildFile.xml: L16-20: copy the tag for entire unit test "testFWCoreServicesCatch2". - FWCore/Services/test/test_catch2_ExternalRandomNumberGeneratorService.cc: L8: uncomment "#define CATCH_CONFIG_MAIN".
PR description:
Adds a test to show how to use a EDProducer in cmsRun to drive a different EDProducer in a stand alone process where random engine state is passed back and forth.
PR validation:
The code compiles and framework unit tests pass.