Fixes for coloramps and more (patches from the August 2023 hackathon) by valassi · Pull Request #742 · madgraph5/madgraph4gpu

valassi · 2023-08-10T08:04:59Z

This MR includes the first batch of patches from the August 2023 hackathon, stripped off the ongoing MR #741

…tran port of Stephan's unweighting (with fewer events written to disk)

…wgt.f as expected

…d has now disappeared

…pt that headers mention 3.5.1 instead of 3.5.0)

./tput/teeThroughputX.sh -ggtt -makej -makeclean

./tmad/teeMadX.sh -ggtt +10x

Revert "[unw] rerun tmad ggtt, all looks ok (but slower?)" This reverts commit 37f2c3c. Revert "[unw] rerun tput ggtt, all looks ok (but slower?)" This reverts commit ca15783.

…olors is still there (madgraph5#655) ./tmad/teeMadX.sh -ggttgg +10x

…es coloramps.h from coloramps.inc

…ramps.h changes for ggttgg and ggttggg

…ceeds, ie madgraph5#655 is fixed? will rerun also on ggttggg ./tmad/teeMadX.sh -ggttgg +10x

… colors are still correct! this confirms that madgraph5#655 is fixed by disabling the coloramps.h patch ./tmad/teeMadX.sh -ggttggg

…amps.h fixes the LHE color mismatch in ggttgg (madgraph5#655), while also removing the need for the coloramps.h patch (madgraph5#713)

Revert "[color] rerun tmad ggttggg test (short version without +10x), the LHE colors are still correct! this confirms that madgraph5#655 is fixed by disabling the coloramps.h patch" This reverts commit aef35b1. Revert "[color] rerun tmad ggttgg with the new coloramps.h - now the test succeeds, ie madgraph5#655 is fixed? will rerun also on ggttggg" This reverts commit b04e4a8.

…ing -O3 for counters.cc (madgraph5#740)

…adgraph5#740 ./CODEGEN/generateAndCompare.sh gg_tt --mad --nopatch sed -i 's/DEFAULT_F2PY_COMPILER=f2py3.*/DEFAULT_F2PY_COMPILER=f2py3/' gg_tt.mad/Source/make_opts git diff --no-ext-diff -R gg_tt.mad/Source/makefile gg_tt.mad/Source/dsample.f gg_tt.mad/Source/genps.inc gg_tt.mad/Source/vector.inc gg_tt.mad/SubProcesses/makefile > CODEGEN/MG5aMC_patches/PROD/patch.common git diff --no-ext-diff -R gg_tt.mad/Source/make_opts gg_tt.mad/bin/internal/banner.py gg_tt.mad/bin/internal/gen_ximprove.py gg_tt.mad/bin/internal/madevent_interface.py >> CODEGEN/MG5aMC_patches/PROD/patch.common git diff --no-ext-diff -R gg_tt.mad/SubProcesses/P1_gg_ttx/auto_dsig1.f gg_tt.mad/SubProcesses/P1_gg_ttx/driver.f gg_tt.mad/SubProcesses/P1_gg_ttx/matrix1.f > CODEGEN/MG5aMC_patches/PROD/patch.P1 git checkout gg_tt.mad

… for madgraph5#740

Revert "[color] regenerate all 7 mad processes - th eonly change is that coloramps.h changes for ggttgg and ggttggg" This reverts commit d5eb960.

…adgraph5#740

… coloramps.h for ggttgg and ggttggg fixing madgraph5#655 and madgraph5#713

…de, as expected

(Some performance fluctuations - but generally no real change) STARTED AT Thu Aug 10 00:37:54 CEST 2023 ./tput/teeThroughputX.sh -mix -hrd -makej -eemumu -ggtt -ggttg -ggttgg -gqttq -ggttggg -makeclean ENDED(1) AT Thu Aug 10 02:43:06 CEST 2023 [Status=0] ./tput/teeThroughputX.sh -flt -hrd -makej -eemumu -ggtt -ggttgg -inlonly -makeclean ENDED(2) AT Thu Aug 10 03:09:38 CEST 2023 [Status=0] ./tput/teeThroughputX.sh -makej -eemumu -ggtt -ggttg -gqttq -ggttgg -ggttggg -flt -bridge -makeclean ENDED(3) AT Thu Aug 10 03:19:58 CEST 2023 [Status=0] ./tput/teeThroughputX.sh -eemumu -ggtt -ggttgg -flt -rmbhst ENDED(4) AT Thu Aug 10 03:23:12 CEST 2023 [Status=0] ./tput/teeThroughputX.sh -eemumu -ggtt -ggttgg -flt -curhst ENDED(5) AT Thu Aug 10 03:26:23 CEST 2023 [Status=0]

…colormaps madgraph5#655 and madgraph5#713 are fixed (Some performance fluctuations, maybe generally a bit slower? but no clear pattern) STARTED AT Thu Aug 10 03:29:39 CEST 2023 ENDED AT Thu Aug 10 07:47:39 CEST 2023 Status=0 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 24 /data/avalassi/GPU2023/madgraph4gpuX/epochX/cudacpp/tmad/logs_ggttgg_mad/log_ggttgg_mad_d_inl0_hrd0.txt 24 /data/avalassi/GPU2023/madgraph4gpuX/epochX/cudacpp/tmad/logs_ggttgg_mad/log_ggttgg_mad_f_inl0_hrd0.txt 24 /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

…generate pp_tt012j.mad including coloramps fixes for madgraph5#655 and madgraph5#713

valassi · 2023-08-10T08:07:44Z

The complete list of changes includes

move CODEGEN to Olivier's latest gpucpp, including his port to Fortran of Stephan's unweighting improvements, writing out fewer events to disk after the first unweighting ("unw" branch)
add to code generation and to the repo the pp_tt012j process ("pptt" branch)
Stephan's fix for compile counters with -O3 in the Fortran madevent makefile #740, adding -O3 to the build of counters.cc in fortran madevent
fix by Olivier and Andrea for Problems in ggttgg lhe files after the latest upstream changes (replacing channel by iconfig in select_color) #655 and Generate coloramps.h directly rather than via "sed" on coloramps.inc #713: removing the coloramps.h patch in patchMad.sh ensures that the same LHE files with the same color choices in Fortran and Cuda/C++ are found in ggttgg tests

valassi · 2023-08-10T08:10:42Z

The CI tests succeeded - self merging to master

valassi and others added 30 commits August 9, 2023 10:09

[unw] in CODEGEN upgrade to Olivier's latest gpucpp including the For…

6389f3a

…tran port of Stephan's unweighting (with fewer events written to disk)

[unw] regenerate ggtt.mad with the latest gpucpp - some changes in un…

329b073

…wgt.f as expected

[unw] in CODEGEN, mark the plugin as requiring MG5AMC 3.5.1 (was 3.5.0)

3da1ca1

[unw] regenerate ggtt.mad, the warning about 3.5.1 being not validate…

e31d249

…d has now disappeared

[unw] generate the other 6 mad processes

fe71414

[unw] regenerate 7 processes sa - no difference in cudacpp code (exce…

64502da

…pt that headers mention 3.5.1 instead of 3.5.0)

[unw] rerun tput ggtt, all looks ok (but slower?)

ca15783

./tput/teeThroughputX.sh -ggtt -makej -makeclean

[unw] rerun tmad ggtt, all looks ok (but slower?)

37f2c3c

./tmad/teeMadX.sh -ggtt +10x

[unw] revert to upstream/master logs

8c339a6

Revert "[unw] rerun tmad ggtt, all looks ok (but slower?)" This reverts commit 37f2c3c. Revert "[unw] rerun tput ggtt, all looks ok (but slower?)" This reverts commit ca15783.

[pptt] in CODEGEN, add pp_tt012j

ff64206

[pptt] generate pp_tt012j.mad and add it to the git repo

80b1fa3

[color] rerun ggttgg tmad test, confirm fortran/cpp mismatch in LHE c…

0344239

…olors is still there (madgraph5#655) ./tmad/teeMadX.sh -ggttgg +10x

[color] in CODEGEN patchMad.sh, try to disable the patch that recreat…

4241a94

…es coloramps.h from coloramps.inc

[color] regenerate all 7 mad processes - th eonly change is that colo…

d5eb960

…ramps.h changes for ggttgg and ggttggg

[color] rerun tmad ggttgg with the new coloramps.h - now the test suc…

b04e4a8

…ceeds, ie madgraph5#655 is fixed? will rerun also on ggttggg ./tmad/teeMadX.sh -ggttgg +10x

Enable optimisations for counters and OMP in Fortran executables.

3881d3e

[color] rerun tmad ggttggg test (short version without +10x), the LHE…

aef35b1

… colors are still correct! this confirms that madgraph5#655 is fixed by disabling the coloramps.h patch ./tmad/teeMadX.sh -ggttggg

[color] in CODEGEN, add a comment to confirm that disabling the color…

3b1b4b6

…amps.h fixes the LHE color mismatch in ggttgg (madgraph5#655), while also removing the need for the coloramps.h patch (madgraph5#713)

[pptt] in pptt012j.mad, add a comment to Stephan's previous patch add…

65e3c14

…ing -O3 for counters.cc (madgraph5#740)

[pptt] port Stephan's patch for madgraph5#740 to ggtt.mad

772ec1f

[pptt] regenerate ggtt.mad, all ok no change

22c317d

[pptt] regenerate the other 6 processes mad - including Stephan's fix…

c554933

… for madgraph5#740

[pptt] regenerate all 7 processes sa - no change, as expected

1e1e8f3

[color] revert generation of 7 mad processes to ease merging

be8d2ed

Revert "[color] regenerate all 7 mad processes - th eonly change is that coloramps.h changes for ggttgg and ggttggg" This reverts commit d5eb960.

Merge branch 'AUG2023' into color - this includes Stephan's fix for m…

f79a963

…adgraph5#740

[color] regenerate all 7 processes mad - this recovers the changes in…

3113334

… coloramps.h for ggttgg and ggttggg fixing madgraph5#655 and madgraph5#713

[color] regenerate all 7 processes sa - no changes anywhere in the co…

9262351

…de, as expected

valassi added 2 commits August 10, 2023 09:55

[color/pptt] ** COMPLETE COLOR/PPTT (1st part of AUG2023 patch) ** re…

0ed5928

…generate pp_tt012j.mad including coloramps fixes for madgraph5#655 and madgraph5#713

valassi assigned valassi, hageboeck and oliviermattelaer Aug 10, 2023

valassi merged commit 3df24c3 into madgraph5:master Aug 10, 2023

valassi mentioned this pull request Aug 10, 2023

Patches from the August 2023 hackathon #741

Closed

This was referenced Aug 10, 2023

Generate coloramps.h directly rather than via "sed" on coloramps.inc #713

Closed

Problems in ggttgg lhe files after the latest upstream changes (replacing channel by iconfig in select_color) #655

Closed

valassi mentioned this pull request Jun 28, 2024

Fixes for iconfig-channel mappings in coloramps.h (obsolete, will be closed) #873

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fixes for coloramps and more (patches from the August 2023 hackathon)#742

Fixes for coloramps and more (patches from the August 2023 hackathon)#742
valassi merged 32 commits intomadgraph5:masterfrom
valassi:color

valassi commented Aug 10, 2023

Uh oh!

valassi commented Aug 10, 2023

Uh oh!

valassi commented Aug 10, 2023

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

valassi commented Aug 10, 2023

Uh oh!

valassi commented Aug 10, 2023

Uh oh!

valassi commented Aug 10, 2023

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants