Choose timezone
Your profile timezone:
Powered by Indico
v3.3.3-pre
Attendance: Axel, David, Oksana, Vassil
Vassil: Added a few performance fixes wrt runtime. Some workflows are 30% faster. Investigating https://sft.its.cern.ch/jira/browse/ROOT-10514 . The memory footprint is a little high but expected. Oksana has more details. In the meantime I have send Richard Smith (from Google) some information and maybe he has some directions for us to optimize. I resumed working on the global module index. Preparing for the release.
Oksana: Preparing for the release. Busy with a lot of tiny things. There was an issue with the installation of modules. For the next release we will have a better way of doing things. We received an email v6.20 vs v6.18 from Alice. The results showed 200Mb more for v6.20. I tried to remeasure the numbers in the context of CMSSW. For v6.18 we have 0 pcms (as expected) for v6.20 we have 95 pcms and the ongoing work is around 200 in module-aware CMSSW. CMSSW with 6.18 vs v6.20 consumes 135mb more for 100 modules. In rootbench (we don't build with -Dall=On) we have 75 modules the overhead is around 40 Mb. The overhead seems to depend on the N of modules we are loading. I will generate memory flamegraphs for ROOT with 75 modules and 100 modules. Alice's 200Mb overhead should be ~80. It should not be that high. We should ask Peter for clarification. I am working on the CLHEP modulemap.
David: We have compilation errors in the CMSSW IB in DataFormats/BTauReco: https://cmssdt.cern.ch/SDT/cgi-bin/buildlogs/slc7_amd64_gcc820/CMSSW_11_1_CXXMODULE_X_2020-01-21-2300/DataFormats/BTauReco I have started an igprof job for measuring the ROOT startup.
Axel: Spent couple of hours with Philippe discussing the diagnostic client with recursive parsing: https://sft.its.cern.ch/jira/browse/ROOT-10504 . We will come up in with a temporary patch in local clang. I am hunting down a few non-module-related issues.
AOB: Next meeting in two weeks.