search

madgraph5 / madgraph4gpu

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

Fix default run_card for GPU #816

Closed oliviermattelaer closed 5 months ago

oliviermattelaer commented 5 months ago

With Stefan we noticed that "output madevent_gpu" was producing a run_card with the default value set for SIMD mode... This patch is fixing that.

oliviermattelaer commented 5 months ago

Can i merge this?

roiser commented 5 months ago

IIUC this will fix the problem that Jin has reported?

valassi commented 5 months ago

Hi @oliviermattelaer sorry I will also look at this one in the afternoon. It changes the generated vector.inc, not a problem but I need to update a few scripts/tests.

valassi commented 5 months ago

Hi @oliviermattelaer can you please merge #813 first? You changed the gpucpp in that PR, I will analyse this one after that merge. Thanks

oliviermattelaer commented 5 months ago

Ok, I have therefore done a rebase of the master within this branch to avoid to have to do a merge (since I do not expect everyone to branch from this branch

valassi commented 5 months ago

Ok, I have therefore done a rebase of the master within this branch to avoid to have to do a merge (since I do not expect everyone to branch from this branch

Hi @oliviermattelaer very good, thanks.

After including the changes from #813, I now see NO changes in generated code, so this is good to go for me (I am not at this stage testing the full workflow as you do... well I should, but I rely on you for that). So this is good to go for me.

To keep it simple, I will not include the new codegen logs, it would not add anything.

Just one comment/question, I see that in the heft codegen log this warning is now disappearing

< WARNING: coupling GC_13=-(complex(0,1)*GH) has direct dependence in aS but has QCD order set to 0. Automatic computation of scale uncertainty can be wrong for such model. 
< WARNING: coupling GC_16=(complex(0,1)*Gphi)/8. has direct dependence in aS but has QCD order set to 0. Automatic computation of scale uncertainty can be wrong for such model.

Has the warning disappeared because the actual issue has been fixed? Or should this still give a warning?

Anyway, good to go for me.

valassi commented 5 months ago

I see that @roiser had already reacted with a thumbs-up to the review request, so I assume that he also approves this.

I will merge this directly, I think this makes it easier to progress on other fronts. Thanks @oliviermattelaer