PhD meeting

Monday 27 Feb 2023, 10:30 → 15:00 Europe/Zurich

513/R-070 - Openlab Space (CERN)

# PhD meeting Mon 27.02.2023

Present: ZW, SR, AV, RS

## Status

ZW progressing on LHE and SLHE (susy specific parameter card) parser for reading/writing.
Almost done, will keep Robert posted.

The way this is meant to work now is that you write a parameter card for each reweighting.
So you have a general driver card that says "do three reweightings", but then you need to
write three cards, and then the ME engine reads them in.

Not assuming the bridge software API,
but assuming the format of the parameter card used in our MEs.

SR: still issues in pp to 4t?
AV: yes need to fix it.
ZW/RS: can start with gluon-gluon collisions only

RS/ZW: discussion of a specific example of parameter card for Robert's EFT processes
In these cards generally all operators are 0, except a few with tiny values
But there are also points with all operators non zero
RS: there seems to be a bug, if two operator values are the same, then a reweighting 
always keeps them equal to each other even if you only want to vary one?
ZW: sounds like a bug, will look into this before the next meeting

AV: how many particles you have in the final state? 6 particles would be too much
ZW: should be 4 or 5 initially, tttt or ttttg (eventually ttttgg)

RS: then will probably ask the CMS GEN conveners for help on running the software
SR: do you know where you will run?
RS: we have GPUs (V100, A100) in Vienna but I need to look into that
AV: good they are V100/A100, instead T4 would be no good

## AOB

Next meetings?
ZW: from next time it will be Thu afternoon eg at 3pm
RS: will be at Fermilab for 6 weeks, time difference 7hours
ZW: next meeting Thu 9 March
AV: every two weeks Thu 3pm there is also HSF meetings, we should avoid clash

## NLO

ZW: also want to start profiling NLO in upstream madgraph
Want to understand how much time the external loops take
AV/SR: very good!
SR: what do you use?
ZW: using callgrind now
AV: that takes longer, you can also use perf
SR: you can use flamegraphs too from perf or callgrind