MadGraph5_aMC@NLO

speed up Z+bb generation

Asked by Inês Ochoa

Dear developers,

I am interested in generating NLO Z+bb events with aMC@NLO. I've noticed it takes ~50 minutes to generate 100 events while 100 W+bb events take only 5 minutes (including the integration step), on the same machine. I understand that the Z+bb process has extra initial states and helicity configurations, but I wanted to ask if there is anything that can be done to speed up the process, e.g., any effect/correction that can be turned off.

Thanks,
Ines

Question information

Language:
English Edit question
Status:
Answered
For:
MadGraph5_aMC@NLO Edit question
Assignee:
Rikkert Frederix Edit question
Last query:
Last reply:

This question was reopened

Revision history for this message
Rikkert Frederix (frederix) said : 		#1

Dear Inês,

The glue-glue initial state processes is a lot slower than the qqbar ones. The Wbb does not have the glue-glue ones, which makes it a lot faster.

But I'm a bit surprised by your timings. Note that the scaling is not at all linear with the number of events. If you request only 100 events, nearly all the time is spend in setting up the integration grids and all that. It's much more efficient to set-up the grids with high precision only once, and then use this to generate any number of events. Like explained here

https://answers.launchpad.net/mg5amcnlo/+faq/2613

Best,
Rikkert

Revision history for this message
Inês Ochoa (inesochoa) said : 		#2

Dear Rikkert,

I was aware of the non-linearity, which obviously helps, but I don't think I was aware of the possibility of skipping the integration step, and in particular of the flags --nocompile --only_generation.

Thanks a lot!
Ines

Revision history for this message
Inês Ochoa (inesochoa) said : 		#3

Dear Rikkert,

I have a follow up question on this issue: I would like to add a generation-level cut before setting up the integration grids. Is this possible? Alternatively, having the original grids setup (which is the case), I would like to add a generation-level cut and extract the new cross-section. What would be the best way to proceed? So far I have added a cut manually to Subprocesses/cuts.f but this is not possible before the grid is setup as far as I know.

Thank you,
Ines

Revision history for this message
Rikkert Frederix (frederix) said : 		#4

Dear Ines,

Why do you say that it is not correct to put a generation cut before the grid setup? It is the correct thing to do. Only be aware that the generation cut has to be significantly more inclusive than the analysis cuts (The generation cut is obviously applied before parton showering, while the analysis cuts are after parton showering. The parton shower changes momentum of the particles in the process, so to not bias the analysis cuts, the generation cuts should be significantly more relax).

Best,
Rikkert

Revision history for this message
Inês Ochoa (inesochoa) said : 		#5

Dear Rikkert,

I didn't mean to imply incorrect, just not possible *as far as I know*, given the way that I am setting up the grid. The issue is simply that the Subprocesses folder is not yet created before I run the integration. Is there an obvious way to do it that I'm missing?

Cheers
Ines

Revision history for this message
Rikkert Frederix (frederix) said : 		#6

Dear Ines,

./bin/mg5_aMC
generate p p > e+ ve [QCD]
output W_boson
exit
cd W_boson
emacs SubProcesses/cuts.f
./bin/generate_events

Cheers,
Rik

Can you help with this problem?

Provide an answer of your own, or ask Inês Ochoa for more information if necessary.

To post a message you must log in.