search

madgraph5 / madgraph4gpu

GPU development for the Madgraph5_aMC@NLO event generator software package
28 stars 33 forks source link

Test and complete the fix for 825 in PR 850 #860

Closed valassi closed 1 month ago

valassi commented 1 month ago

Hi @oliviermattelaer this is a PR where I test and complete your fix in PR #850 for #825 (xsec mismatch in susy_gg_tt).

Your patch is good and we should merge it. In these tests however I found that a few additions are needed, in particular the reference files for runTest.exe must be updated (the code changed after adding Ccoeff, so I would say that the old reference files were wrong, see #859).

Still in WIP for now but will be ready soon.

valassi commented 1 month ago

Note: after fixing #825 with PR #850, again a SIGFPE reappeared also in gqttq. This is what I had reported in #845. I think that most likely this is the same SIGFPE crash in rotxxx #855 and can be fixed by adding volatile in PR #857... but I will check that.

valassi commented 1 month ago

Hi Olivier, this is now also ready for merging.It includes your #850 and complements it with an update of the susy reference files for runTest (#859), plus various other tests. Can you please have a look and approve?

I would suggest to first merge your #850 and then this one. Or I think that one can merge this directly and it will automatically close your #850 as merged.

Reminder, there are two other PRs from me ready to be reviewed and merged: PR #854 (just adding matrix.pdf files for easier visualization than ps) and PR #857 (fixing the SIGFPE crash in rotxxx by adding volatile).

Instead, I think that issues #826 (no xsec in susy_gg_t1t1) and #845 (SIGFPE crash in gqttq) are still pedning to be debugged. I think that a combination of your PR #852 and my PR #853 might fix those, but as far as I am concerned this will wait for end June.

Thanks! Andrea

valassi commented 1 month ago

Thanks Olivier, merging now