first systematic 'launch'-like tests (and move to the latest select_color upstream) #709
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
FIRST
for dir in gg_tt.mad gg_ttg.mad gg_ttgg.mad gg_ttggg.mad; do
./tlau/lauX.sh -CUDA ${dir}
./tlau/lauX.sh -FORTRAN ${dir}
./tlau/lauX.sh -CPP ${dir}
done
THEN
for d in *.mad/LAUX*; do d1=${d/gg_tt/ggtt}; mv $d tlau/logs_${d1/.mad\/LAUX/}; done
git add tlau/logs_ggtt*
This was referenced Jun 16, 2023
…ck his select_color change This will introduce again madgraph5#655 but should fix madgraph5#710
… it still takes 12 minutes... This is because the part that takes time is survey (nevents fixed?), not refine...
What costs time is the refine step... how do I change accuracy? (Is it req_acc in thr runcard?) For the tests, alternatively, consider something like this? NLO only? ./bin/generate_events --only-generation --nocompile
STARTED AT Fri Jun 16 22:42:39 CEST 2023 ./tput/teeThroughputX.sh -mix -hrd -makej -eemumu -ggtt -ggttg -ggttgg -gqttq -ggttggg -makeclean ENDED(1) AT Fri Jun 16 23:07:07 CEST 2023 [Status=0] ./tput/teeThroughputX.sh -flt -hrd -makej -eemumu -ggtt -ggttgg -inlonly -makeclean ENDED(2) AT Fri Jun 16 23:15:57 CEST 2023 [Status=0] ./tput/teeThroughputX.sh -makej -eemumu -ggtt -ggttg -gqttq -ggttgg -ggttggg -flt -bridge -makeclean ENDED(3) AT Fri Jun 16 23:25:01 CEST 2023 [Status=0] ./tput/teeThroughputX.sh -eemumu -ggtt -ggttgg -flt -rmbhst ENDED(4) AT Fri Jun 16 23:28:02 CEST 2023 [Status=0] ./tput/teeThroughputX.sh -eemumu -ggtt -ggttgg -flt -curhst ENDED(5) AT Fri Jun 16 23:31:00 CEST 2023 [Status=0]
…ected This is the price to pay for fixing instead madgraph5#710 using OLivier's select_color change STARTED AT Fri Jun 16 23:34:03 CEST 2023 ENDED AT Sat Jun 17 03:40:59 CEST 2023 24 /data/avalassi/GPU2023/madgraph4gpuX/epochX/cudacpp/tmad/logs_eemumu_mad/log_eemumu_mad_d_inl0_hrd0.txt 24 /data/avalassi/GPU2023/madgraph4gpuX/epochX/cudacpp/tmad/logs_eemumu_mad/log_eemumu_mad_f_inl0_hrd0.txt 24 /data/avalassi/GPU2023/madgraph4gpuX/epochX/cudacpp/tmad/logs_eemumu_mad/log_eemumu_mad_m_inl0_hrd0.txt 24 /data/avalassi/GPU2023/madgraph4gpuX/epochX/cudacpp/tmad/logs_ggttggg_mad/log_ggttggg_mad_d_inl0_hrd0.txt 24 /data/avalassi/GPU2023/madgraph4gpuX/epochX/cudacpp/tmad/logs_ggttggg_mad/log_ggttggg_mad_f_inl0_hrd0.txt 24 /data/avalassi/GPU2023/madgraph4gpuX/epochX/cudacpp/tmad/logs_ggttggg_mad/log_ggttggg_mad_m_inl0_hrd0.txt 1 /data/avalassi/GPU2023/madgraph4gpuX/epochX/cudacpp/tmad/logs_ggttgg_mad/log_ggttgg_mad_d_inl0_hrd0.txt 1 /data/avalassi/GPU2023/madgraph4gpuX/epochX/cudacpp/tmad/logs_ggttgg_mad/log_ggttgg_mad_f_inl0_hrd0.txt 1 /data/avalassi/GPU2023/madgraph4gpuX/epochX/cudacpp/tmad/logs_ggttgg_mad/log_ggttgg_mad_m_inl0_hrd0.txt 24 /data/avalassi/GPU2023/madgraph4gpuX/epochX/cudacpp/tmad/logs_ggttg_mad/log_ggttg_mad_d_inl0_hrd0.txt 24 /data/avalassi/GPU2023/madgraph4gpuX/epochX/cudacpp/tmad/logs_ggttg_mad/log_ggttg_mad_f_inl0_hrd0.txt 24 /data/avalassi/GPU2023/madgraph4gpuX/epochX/cudacpp/tmad/logs_ggttg_mad/log_ggttg_mad_m_inl0_hrd0.txt 24 /data/avalassi/GPU2023/madgraph4gpuX/epochX/cudacpp/tmad/logs_ggtt_mad/log_ggtt_mad_d_inl0_hrd0.txt 24 /data/avalassi/GPU2023/madgraph4gpuX/epochX/cudacpp/tmad/logs_ggtt_mad/log_ggtt_mad_f_inl0_hrd0.txt 24 /data/avalassi/GPU2023/madgraph4gpuX/epochX/cudacpp/tmad/logs_ggtt_mad/log_ggtt_mad_m_inl0_hrd0.txt 0 /data/avalassi/GPU2023/madgraph4gpuX/epochX/cudacpp/tmad/logs_gqttq_mad/log_gqttq_mad_d_inl0_hrd0.txt
…raph5#710 with Olivier's select_color change The price to pay is the tmad failures in ggttgg madgraph5#655 Add to the git repo the two ggttggg FORTRAN logs that were previously failing The duration of these tests needs some tuning, the ggttggg take too long madgraph5#711 ls -ltr tlau/logs_ggtt*/*txt -rw-r--r--. 1 avalassi zg 3590 Jun 17 03:41 tlau/logs_ggtt_CUDA/output.txt -rw-r--r--. 1 avalassi zg 3588 Jun 17 03:41 tlau/logs_ggtt_FORTRAN/output.txt -rw-r--r--. 1 avalassi zg 3580 Jun 17 03:42 tlau/logs_ggtt_CPP/output.txt -rw-r--r--. 1 avalassi zg 3462 Jun 17 03:42 tlau/logs_ggttg_CUDA/output.txt -rw-r--r--. 1 avalassi zg 3571 Jun 17 03:43 tlau/logs_ggttg_FORTRAN/output.txt -rw-r--r--. 1 avalassi zg 3515 Jun 17 03:44 tlau/logs_ggttg_CPP/output.txt -rw-r--r--. 1 avalassi zg 4106 Jun 17 03:46 tlau/logs_ggttgg_CUDA/output.txt -rw-r--r--. 1 avalassi zg 4425 Jun 17 04:00 tlau/logs_ggttgg_FORTRAN/output.txt -rw-r--r--. 1 avalassi zg 4349 Jun 17 04:05 tlau/logs_ggttgg_CPP/output.txt -rw-r--r--. 1 avalassi zg 6766 Jun 17 04:50 tlau/logs_ggttggg_CUDA/output.txt -rw-r--r--. 1 avalassi zg 7069 Jun 17 20:45 tlau/logs_ggttggg_FORTRAN/output.txt -rw-r--r--. 1 avalassi zg 6967 Jun 18 01:29 tlau/logs_ggttggg_CPP/output.txt
…adgraph5#711 Revert "[launch] in lauX.sh go back to 10000 unweighted events..." This reverts commit 7021bc6. I realised that also unweighted event generation does take a very long time in these tests
…r survey and refine (8192 events, 1 iteration) Note: cmd.opts['accuracy'] comes from cmd._survey_options where cmd is in madevent_interface.py
|
In the latest commits I have rerun the testst after fixing #710 (using @oliviermattelaer select_color patch ... which however reintroduces #655 that will need to be fixed). I will now work on tuning the time it takes (the tests are very long) #711 |
…l to other tests on itscrd80)
…derstanding of stdout
…gg as this should also affect refine?
…to understand "refine"
…gen_ximprove.py Revert "[launch] TEMPORARY tests on gg_ttgg.mad/bin/internal/gen_ximprove.py to understand "refine"" This reverts commit 9d544d0fb241dabb1df4bf813abb3a319a3b2c7c.
This skips refine but fails generation
INFO: Storing parton level results
No event detected. No cleaning performed! This should allow to run:
cd Subprocesses; ../bin/internal/combine_events
to have your events if those one are missing.
Revert "[launch] in ggttgg.mad, comment out all to_refine.append(C)" This reverts commit a1150138961f701eda26f957b01347d2edac4df9.
…mes to reduce the load...
Revert "[launch] in ggttgg.mad, select only channels with two-character Gn names to reduce the load..." This reverts commit e1138741d4305fa99b4993c66a2afab385e9dc7c.
This fails with
INFO: finish refine
Survey return zero cross section.
Typical reasons are the following:
1) A massive s-channel particle has a width set to zero.
2) The pdf are zero for at least one of the initial state particles
or you are using maxjetflavor=4 for initial state b:s.
3) The cuts are too strong.
Please check/correct your param_card and/or your run_card.
Zero result detected: See https://cp3.irmp.ucl.ac.be/projects/madgraph/wiki/FAQ-General-14
Revert "[launch] in ggttgg.mad, select only 3 channels at most in read_results" This reverts commit e5061ac9078a87e750832d81f648f542985dde7e.
…itscrd80 - will revert Some observations: - for ggtt and ggttg, with the same numbers of events, results vary slightly to itscrd90 (ok... different compiler) - for ggttgg, with 100 events instead of 1000, the overall time spent is essentially the same also in FORTRAN - for ggttggg, with 10 events instead of 100, the overall time is also essentially the same also in FORTRAN I will therefore move back to 100 events at least. In any case I need a way to make these tests last less time, this is totally unmanageable now for development. I am moving back by reverting to itscrd90 results. And I am keeping the lhe events file. Eventually (in a following MR) I will move to itscrd80 results, and possibly remove event files.
… 10000, 1000, 100) Revert "[launch] rerun (INTERRUPTED in the last ggttggg CPP) tlau alltees on itscrd80 - will revert" This reverts commit 151df29.
…as in the logs - will tune later on
valassi
changed the title
|
I will self merge this. As suggested by @oliviermattelaer I have now permannently included his select_color change in the upstream gpucpp. THis fixes #710 - but it reintroduces #655. I have accordingly updated the CODEGEN settings. I have also made some progress on launch-like tests for #683. THis needs some debugging and tuning, which is done in #711. I have also added a first README_CODEGEN.txt for users. This is a temporary solution, while waiting for a full integration of cudacpp upstream. |
valassi
changed the title
This was referenced Jun 19, 2023
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This is a WIP MR with some first systematic 'launch' tests (./bin/generate_events actually) using the lauX.sh script in #683.
There are several issues (icolamp bugs) and a lot to be analysed and tuned - I will open separate issues for those.