Biweekly meeting: C++ Modules in ROOT

Europe/Zurich
Oksana Shadura (University of Nebraska Lincoln (US)), Vasil Georgiev Vasilev (Institute for Advanced Physical Studies (BG))

Attendance: Axel, David, Oksana, Vaibhav, Vassil

Vassil: Working on the global module index -- we are very close to converging, a couple of failures left. I will ask Arpitha to redo the performance comparisons. I have adjusted the boost modulemap and also fixed one of the blockers which I have unintentionally introduced for non-modules builds. There were minor fixes in rootcling fixing possible bugs. Those fixes allowed us to build modules for tinyxml2, clhep and a few from boost. The CXXMODULE build built in release looks pretty well -- 2 distinct failure kinds. Unfortunately, as we expected, it is not performing well. Oksana will maybe share some details. David will maybe tell us a little more later. We are also making substantial progress on Windows and Vaibhav will tell us more.

David: We package boost and clhep -- a new package external in cmssw and we build before other packages. We are able to pick those pcms up and use them. We have our first clean (compiling) release builds. The intent was to turn back debug build. We also believe that we fixed two problems with the incremental builds of cmssw.

Oksana: I checked the green cmssw ib to measure performance, before building the boost pcms. We wanted to have a baseline to compare when enabling new modules. After running couple of workflows (for 5 times) we see that we have similar runtime (the comparison was between 6.18 with pch) -- virtual memory is high and 1gb higher RSS; we have on average 5% better throughput. GMI should fix the memory overhead.

Vaibhav: We can build root with modules -- we have disabled 7 modules; the rest is working fine for windows. There are some issues with the ninja build system. We are trying to fix an issue when loading the modules from the ROOTSYS/lib as the modules are next to the dll files which are in ROOTSYS/bin

Axel: Nothing to report.

 

Priorities: We switch to release in cmssw after trying to comment the assert for the lazy spec info.  We want to focus on enabling more modules (being it boost and cmssw)

AOB: Next meeting in two weeks.

There are minutes attached to this event. Show them.
    • 14:00 14:30
      Updates 30m
      Speakers: Oksana Shadura (University of Nebraska Lincoln (US)), Vasil Georgiev Vasilev (Institute for Advanced Physical Studies (BG))
    • 14:30 15:00
      Round table 30m