-
Notifications
You must be signed in to change notification settings - Fork 184
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Backport OpenBLAS fixes [10_2_X] #5289
Backport OpenBLAS fixes [10_2_X] #5289
Conversation
please test |
The tests are being triggered in jenkins. |
A new Pull Request was created by @kpedro88 (Kevin Pedro) for branch IB/CMSSW_10_2_X/gcc700. @cmsbuild, @smuzaffar, @mrodozov can you please review it and eventually sign? Thanks. |
-1 Tested at: c94a4b9
I found compilation error when building: /cvmfs/cms-ib.cern.ch/nweek-02599/slc6_amd64_gcc700/external/gcc/7.0.0-omkpbe2/include/c++/7.3.1/bits/unique_ptr.h:51:28: note: declared here template class auto_ptr; ^~~~~~~~ /build/cmsbld/jenkins/workspace/ib-run-pr-tests/testBuildDir/slc6_amd64_gcc700/external/gcc/7.0.0-omkpbe2/bin/g++ -pthread -shared -L/cvmfs/cms-ib.cern.ch/nweek-02599/slc6_amd64_gcc700/external/expat/2.1.0-omkpbe2/lib -L/cvmfs/cms-ib.cern.ch/nweek-02599/slc6_amd64_gcc700/external/expat/2.1.0-omkpbe2/lib64 -L/cvmfs/cms-ib.cern.ch/nweek-02599/slc6_amd64_gcc700/external/bz2lib/1.0.6-omkpbe2/lib -L/cvmfs/cms-ib.cern.ch/nweek-02599/slc6_amd64_gcc700/external/bz2lib/1.0.6-omkpbe2/lib64 -L/cvmfs/cms-ib.cern.ch/nweek-02599/slc6_amd64_gcc700/external/db6/6.0.30-omkpbe2/lib -L/cvmfs/cms-ib.cern.ch/nweek-02599/slc6_amd64_gcc700/external/db6/6.0.30-omkpbe2/lib64 -L/cvmfs/cms-ib.cern.ch/nweek-02599/slc6_amd64_gcc700/external/gdbm/1.10-omkpbe2/lib -L/cvmfs/cms-ib.cern.ch/nweek-02599/slc6_amd64_gcc700/external/gdbm/1.10-omkpbe2/lib64 -L/cvmfs/cms-ib.cern.ch/nweek-02599/slc6_amd64_gcc700/external/openssl/1.0.2d-omkpbe2/lib -L/cvmfs/cms-ib.cern.ch/nweek-02599/slc6_amd64_gcc700/external/openssl/1.0.2d-omkpbe2/lib64 -L/cvmfs/cms-ib.cern.ch/nweek-02599/slc6_amd64_gcc700/external/libffi/3.2.1-omkpbe2/lib -L/cvmfs/cms-ib.cern.ch/nweek-02599/slc6_amd64_gcc700/external/libffi/3.2.1-omkpbe2/lib64 -L/cvmfs/cms-ib.cern.ch/nweek-02599/slc6_amd64_gcc700/external/zlib-x86_64/1.2.11-omkpbe2/lib -L/cvmfs/cms-ib.cern.ch/nweek-02599/slc6_amd64_gcc700/external/zlib-x86_64/1.2.11-omkpbe2/lib64 -L/cvmfs/cms-ib.cern.ch/nweek-02599/slc6_amd64_gcc700/external/sqlite/3.22.0-omkpbe/lib -L/cvmfs/cms-ib.cern.ch/nweek-02599/slc6_amd64_gcc700/external/sqlite/3.22.0-omkpbe/lib64 build/temp.linux-x86_64-2.7/rivet/core.o -L/build/cmsbld/jenkins/workspace/ib-run-pr-tests/testBuildDir/BUILD/slc6_amd64_gcc700/external/rivet/2.5.4-d3edaa/Rivet-2.5.4/src/.libs -L/build/cmsbld/jenkins/workspace/ib-run-pr-tests/testBuildDir/slc6_amd64_gcc700/external/hepmc/2.06.07-omkpbe2/lib -L/build/cmsbld/jenkins/workspace/ib-run-pr-tests/testBuildDir/slc6_amd64_gcc700/external/fastjet/3.3.0-omkpbe/lib -L/build/cmsbld/jenkins/workspace/ib-run-pr-tests/testBuildDir/slc6_amd64_gcc700/external/yoda/1.6.7-d3edaa/lib -L/build/cmsbld/jenkins/workspace/ib-run-pr-tests/testBuildDir/slc6_amd64_gcc700/external/OpenBLAS/0.2.20-d3edaa/lib -L/cvmfs/cms-ib.cern.ch/nweek-02599/slc6_amd64_gcc700/external/python/2.7.14-omkpbe4/lib -Wl,-R/build/cmsbld/jenkins/workspace/ib-run-pr-tests/testBuildDir/slc6_amd64_gcc700/external/hepmc/2.06.07-omkpbe2/lib -Wl,-R/build/cmsbld/jenkins/workspace/ib-run-pr-tests/testBuildDir/slc6_amd64_gcc700/external/fastjet/3.3.0-omkpbe/lib -Wl,-R/build/cmsbld/jenkins/workspace/ib-run-pr-tests/testBuildDir/slc6_amd64_gcc700/external/yoda/1.6.7-d3edaa/lib -Wl,-R/build/cmsbld/jenkins/workspace/ib-run-pr-tests/testBuildDir/slc6_amd64_gcc700/external/OpenBLAS/0.2.20-d3edaa/lib -lgsl -lHepMC -lfastjet -lYODA -lRivet -lpython2.7 -o build/lib.linux-x86_64-2.7/rivet/core.so /cvmfs/cms-ib.cern.ch/nweek-02599/slc6_amd64_gcc700/external/gcc/7.0.0-omkpbe2/bin/../lib/gcc/x86_64-unknown-linux-gnu/7.3.1/../../../../x86_64-unknown-linux-gnu/bin/ld: cannot find -lgsl collect2: error: ld returned 1 exit status error: command '/build/cmsbld/jenkins/workspace/ib-run-pr-tests/testBuildDir/slc6_amd64_gcc700/external/gcc/7.0.0-omkpbe2/bin/g++' failed with exit status 1 make[2]: *** [all-local] Error 1 make[2]: Leaving directory `/build/cmsbld/jenkins/workspace/ib-run-pr-tests/testBuildDir/BUILD/slc6_amd64_gcc700/external/rivet/2.5.4-d3edaa/Rivet-2.5.4/pyext' make[1]: *** [all-recursive] Error 1 make[1]: Leaving directory `/build/cmsbld/jenkins/workspace/ib-run-pr-tests/testBuildDir/BUILD/slc6_amd64_gcc700/external/rivet/2.5.4-d3edaa/Rivet-2.5.4/pyext' You can see the results of the tests here: |
The excerpt cut out an important part of the log, which specifies that the issue occurred when building rivet:
There are a lot of changes between https://github.com/cms-sw/cmsdist/commits/IB/CMSSW_10_2_X/gcc700/rivet.spec and https://github.com/cms-sw/cmsdist/commits/IB/CMSSW_11_0_X/master/rivet.spec (note that this PR does not directly modify rivet.spec). @intrepid42 can you comment if there is a specific change that would solve this problem? |
On Oct 21, 2019, at 12:23 PM, Kevin Pedro ***@***.***> wrote:
This is a combined backport of #5063 and #5091. The goal is to propagate the DeepAK8 speedup to the current analysis release. (Since ultra legacy production in 10_6_X will not finish for quite a while, 10_2_X will continue to see active use by analyzers.)
attn: @hqucms @smuzaffar @slava77
A question for @davidlange6: would we gain anything by also backporting the OpenBLAS version update from #4897?
not that I can summarize.
|
Correction: small CMSSW changes are needed for both 2.6.1 -> cms-sw/cmssw#25817 and 2.7.2 -> cms-sw/cmssw#26936 You can also try to fix (like in Herwig++/ThePEG) or remove the GSL dependency. |
Pull request #5289 was updated. |
Well, that was a fun rabbit hole. I reran this build with and without my changes. The successful command (in the clean build) has these flags:
whereas the failing command (in the changed build) has these flags:
Since libgsl is not located in OpenBLAS, the command naturally fails. One might wonder why these arguments change when rivet does not even depend on OpenBLAS. Here's the relevant output from gsl-config, before:
and after:
So rivet is not using I have fixed the regex to handle this case correctly, picking the gsl lib directory, and propagated this fix to the |
please test |
The tests are being triggered in jenkins. |
-1 Tested at: 20cb24f You can see the results of the tests here: I found follow errors while testing this PR Failed tests: UnitTests
I found errors in the following unit tests: ---> test EcalDAQ_O2O_test had ERRORS |
Comparison job queued. |
Comparison is ready Comparison Summary:
|
please test |
The tests are being triggered in jenkins. |
-1 Tested at: 20cb24f You can see the results of the tests here: I found follow errors while testing this PR Failed tests: UnitTests
I found errors in the following unit tests: ---> test EcalDAQ_O2O_test had ERRORS |
Comparison job queued. |
Comparison is ready Comparison Summary:
|
+1 |
@smuzaffar the test failure is unrelated to this PR, due to an update of the conditions test by Giacomo (all the 10_2_X IBs are affected). SO I would just merge this in view of next build |
+externals |
This pull request is fully signed and it will be integrated in one of the next IB/CMSSW_10_2_X/gcc700 IBs (but tests are reportedly failing). |
merge as the test failed the bot did not acted on our approvals |
This is a combined backport of #5063 and #5091. The goal is to propagate the DeepAK8 speedup to the current analysis release. (Since ultra legacy production in 10_6_X will not finish for quite a while, 10_2_X will continue to see active use by analyzers.)
attn: @hqucms @smuzaffar @slava77
A question for @davidlange6: would we gain anything by also backporting the OpenBLAS version update from #4897?