Merge of master into master_june24 and channelid fixes/reimplementation#882
Merged
valassi merged 778 commits intomadgraph5:master_june24from valassi:june24Sep 3, 2024
+1,312,263-180,326
Commits
This pull request is big! We're only showing the most recent 250 commits
Commits on Jul 17, 2024
- committed
[june24] in gg_ttgg.mad and CODEGEN, recreate the txt2 reference log after fixing the segfault madgraph5#917 in runTest
committed[june24] regenerate gg_ttggg.mad with all recent fixes including madgraph5#917, will now be able to regenerate the txt2 file
committed- committed
- committed
[june24] in gg_ttggg.mad runTest.cc, add some debug printouts of channel tests... this points to bug madgraph5#919
committed[june24] in gg_ttggg.mad and CODEGEN coloramps.h, add a variable nchannels and a sanity check madgraph5#910 that it equals CPPProcess::ndiagrams (but IT DOES NOT! bug madgraph5#919)
committed- committed
[june24] in gg_ttggg.mad and CODEGEN runTest.cc, fix madgraph5#920 (start idiagram loop at 0) and work around madgraph5#919 (loop until nchannels, not ndiagrams)
committed[june24] in CODEGEN generateAndCompare.sh, go back to WARP_SIZE=32 and NB_WARP=512 (total 16384) after fixing the crash madgraph5#885
committed- committed
[june24] at last, rerun a gg_tt tmad test and this succeeds normally (after fixing madgraph5#885 for NB_WARP_USED)
committed[june24] in gg_ttggg.mad and CODEGEN, recreate the txt2 reference log after fixing the assert madgraph5#920 in runTest
committed[june24] in tmad/madX.sh, export MG5AMC_CHANNELID_DEBUG=1 to debug channelid in MatrixElementKernelBase, and replace OMPFLAGS by USEOPENMP madgraph5#758
committed
Commits on Jul 18, 2024
[june24] regenerate all processes after fixes for madgraph5#917 madgraph5#920 and more - will rerun the CI and all manual tests
committed- committed
- committed
- committed
- committed
[june24] in gg_tt.mad (on LUMI) and CODEGEN, replace __CUDACC_ to support HIP where relevant (madgraph5#893), add a comment elsewhere when __CUDACC is needed
committed- committed
[june24] in gg_ttg.mad, recreate the txt and txt2 reference logs after fixing various issues madgraph5#922
committed- committed
- committed
- committed
- committed
- committed
- committed
- committed
- committed
- committed
- committed
- committed
- committed
[june24] in gg_ttggg.mad and CODEGEN coloramps, document that nchannels can be <= ndiagrams and add a sanity check madgraph5#910 madgraph5#919
committed[june24] in gg_ttggg.mad and CODEGEN CPPProcess.cc, add more sanity checks for nchannels and ndiagrams madgraph5#910 madgraph5#919
committed- committed
- committed
- committed
- committed
- committed
- committed
Commits on Jul 19, 2024
- committed
- committed
[june24] in gg_tt.mad and CODEGEN, MadgraphTest.h and runTest.cc, add comparison of selected helicity and color madgraph5#925
committed[june24] in gg_tt.mad and CODEGEN, MadgraphTest.h and runTest.cc, add also channelId comparison madgraph5#924 madgraph5#925
committed[june24] in gg_tt.mad and CODEGEN, create new reference files including selhel, selcol, channel madgraph5#925 madgraph5#924
committed- committed
[june24] in gg_tt.mad and CODEGEN runTest.cc, turn off debugging unless CUDACPP_RUNTEST_DEBUG is set
committed- committed
[june24] in gg_tt.mad and CODEGEN CPPProcess.cc, in mixed mode check that the 2nd neppV vector has the same channelId as the 1st SIMD vector madgraph5#924
committed- committed
- committed
- committed
- committed
- committed
[june24] in gg_tt.mad, temporarely go back to an older commit, to later show an old WIP related to madgraph5#926
committed[june24] in gg_tt.mad CPPProcess.cc madgraph5#924, determine a second channelId in the second event page... but calculate_wavefunction is called only once?
committed[june24] in gg_tt.mad, undo the previous two commits: this shows the difference to a partial implementation of madgraph5#926
committed[june24] in gg_tt.mad try to remove the hack in counters.cc madgraph5#891, but this still crashes - will need to revert
committed[june24] TEMPORARELY add back (again) the hack in counters.cc, while the issue is better understood madgraph5#891
committed
Commits on Jul 22, 2024
- committed
- committed
- committed
[june24] in gg_tt.mad and CODEGEN MadgraphTest.h, fix printout of failing tests for SelCol madgraph5#931 on LUMI
committed[june24] in gg_tt.mad and CODEGEN CPPProcess.cc, set SelCol=0 if channeldId==0 (fix nomultichannel test on LUMI madgraph5#931)
committed- committed
[june24] in gg_tt.mad and CODEGEN runTest.cc, add the computation of random numbers for helicities and colors (fix madgraph5#931)
committed- committed
[june24] in gg_tt.mad and CODEGEN, recreate new reference files including selhel, selcol, channel madgraph5#925 madgraph5#924 after fixing madgraph5#931
committed[june24] regenerate all processes with fixes for madgraph5#931 color/helicity selection causing test failures on LUMI/HIP
committed[june24] in *.mad and CODEGEN, recreate all new reference files including selhel, selcol, channel madgraph5#925 madgraph5#924 after fixing madgraph5#931
committed[june24] regenerate all processes - this propagates new reference logs to all .sa after fixing madgraph5#931
committed[june24] in gg_tt.mad and CODEGEN, remove the hack in counters.cc madgraph5#891 (this still crashes now but I will shortly fix the issue)
committed[june24] in gg_tt.mad auto_dsig1.f fix counters.cc madgraph5#891 (remove the unnecessary second call to COUNTERS_SMATRIX1MULTI_START outside #ifdef MG5AMC_MEEXPORTER_CUDACPP)
committed[june24] in CODEGEN smatrix_multi.f (backport gg_tt.mad auto_dsig1.f) add the first part of the fix for counters.cc madgraph5#891
committed[june24] in CODEGEN, recreate patches from gg_tt.mad to complete the fix for counters.cc/auto_dsig1.f bug madgraph5#891
committed[june24] in CODEGEN output.py, improve the fix for madgraph5#891 so that a smaller patch.P1 is needed (and auto_dsig1.f can be removed from patch.P1)
committed[june24] in gg_tt.mad auto_dsig1.f, improve the fix for madgraph5#891 so that a smaller patch.P1 is needed (and auto_dsig1.f can be removed from patch.P1)
committed[june24] in CODEGEN, recreate patches from gg_tt.mad after improving the madgraph5#891 fix: now patch.P1 *no longer includes auto_dsig1.f*
committed[june24] regenerate gg_tt.mad, check that all is ok after the madgraph5#891 fixes and patch.P1 improvements
committed- committed
- committed
[june24] in tput scripts, add -nocuda in throughputX.sh and propagate it from allTees.s -hip (part of madgraph5#933)
committed- committed
[june24] in tmad scripts, add -hip option (skipping cuda) and skip ggttggg builds in that case (workaround for madgraph5#933)
committed[susy2] check that issue madgraph5#826 is still pending (./tmad/teeMadX.sh -susyggt1t1 -makeclean) - yes indeed will revert
committed- committed
Commits on Jul 23, 2024
[june24] rerun 102 tput tests on itscrd90 - again all ok in the june24 branch after the latest fixes
committed- committed
- committed
- committed
- committed
- committed
- committed
- committed
[june24] rerun 96 tput tests on LUMI - ok (with known errors on gqttq; previous failure madgraph5#843 in smeft was fixed)
committed- committed
- committed
- committed
- committed
- committed
- committed
- committed
[susy2] regenerate all processes with the fix for madgraph5#826 from PR madgraph5#918 - confirms that only susy_gg_t1t1 code has been affected
committed- committed
- committed
- committed
[susy2] ** COMPLETE SUSY2 ** rerun 3 tmad tests for susyggt1t1, now all ok, confirming madgraph5#826 is fixed
committed
Commits on Jul 24, 2024
- committed
- committed
- committed
- committed
Merge branch 'june24' (with LUMI logs) of https://gitlab.cern.ch:8443/valassi/madgraph4gpu into june24 (with itgold91 logs)
committedMerge branch 'june24' (with LUMI and itgold91 logs) of https://gitlab.cern.ch:8443/valassi/madgraph4gpu into june24 (with new itscrd90 logs)
committed[actions] in .github/workflows, TEMPORARELY reduce CI tests to only one gg_tt "new ci" test while debugging madgraph5#937
committed[actions] in .github/workflows/testsuite_oneprocess.sh, fix the tput_test CI issue madgraph5#937 and add debug printouts
committed- committed
- committed
- committed
- authored
([june24] ** COMPLETE JUNE24 **) Merge remote-tracking branch 'upstream/master' (including CI fixes PR madgraph5#938) into june24
committed- committed
Merge remote-tracking branch 'upstream/less_patch' (PR madgraph5#849) into patchMadsh (current upstream/master)
committed[patchMadsh] ** COMPLETE PATCHMADSH ** regenerate all processes with Olivier's patch in PR madgraph5#849
committed
Commits on Jul 29, 2024
- authored
- authored
- committed
Merge remote-tracking branch 'upstream/master' (including smaller patch PRs madgraph5#849 and madgraph5#939) into susy2
committed[susy2] ** COMPLETE SUSY2 (again) ** regenerate all processes after merging madgraph5#849 and madgraph5#939 - no change in code, only in codegen logs
committed- authored
Merge remote-tracking branch 'upstream/master' (including couplings order madgraph5#918, and smaller patch madgraph5#939 and madgraph5#849) into susy2
committed- authored
- committed
- committed
- committed
- committed
- committed
Commits on Aug 2, 2024
[grid] in gg_tt.mad/bin/internal/madevent_interface.py, avoid deleting GpuAbstraction.h and GpuRuntime.h madgraph5#947
committed[grid] in CODEGEN (backport gg_tt.mad) madevent_interface.py, avoid deleting GpuAbstraction.h and GpuRuntime.h madgraph5#947
committed[grid] in gg_tt.mad/bin/internal/madevent_interface.py, improve madgraph5#947 fix: do not delete G* if it is not a directory (e.g. GpuAbstraction.h)
committed[grid] in CODEGEN (backport gg_tt.mad) madevent_interface.py, improve madgraph5#947 fix: do not delete G* if it is not a directory (e.g. GpuAbstraction.h)
committed[grid] update MG5AMC from ff76bb24b to c488004bd, including my fix for madgraph5#947 in madevent_interface.py (do not delete GpuAbstraction.h) via mg5amcnlo#125
committed[grid] in CODEGEN, recreate patch.common and patch.P1 *without including any python file* as suggested by Olivier, after moving the fix for madgraph5#947 to mg5amcnlo#125
committed[grid] regenerate gg_tt,mad, all is ok (there is one change as I added a comment in the fix for madgraph5#947)
committed[grid/patchMadsh] update MG5AMC from c488004bd to 8c70b1ba4 (HEAD of gpucpp after merging/squashing mg5amcnlo#125 fixing madgraph5#947, with an additional fix by Olivier)
committed- committed
[june24] upgrade mg5amcnlo from 2f4c46765 (previous valassi_gpucpp_june24) to f0b429915 (current valassi_gpucpp_june24 including merge of mg5amcnlo#125 to fix madgraph5#947)
committed- authored
Merge remote-tracking branch 'upstream/master' (including merge of madgraph5#949 fixing madgraph5#947 via mg5amcnlo#125) into june24
committed- committed
- committed
Commits on Aug 8, 2024
- committed
- committed
- committed
- committed
- committed
[hel] in gg_ttggg.mad auto_dsig1.f and counters.cc, add counters for CudaCpp helicities madgraph5#958
committed- authored
- committed
[hel] in gg_ttggg.mad and CODEGEN counters.cc, improve counters for CudaCpp helicities madgraph5#958 (remove event count and throughput)
committed[hel] in gg_ttggg.mad (and CODEGEN manually changing patch.P1) fbridge.inc interface, add parameter goodHelOnly as in Bridge to quit after few events in cudacpp helicity computation (fix madgraph5#958
committed- committed
- committed
- committed
- committed
[hel] manually fix matrix1.pdf in ggttggg.mad using the awk command I added to generateAndCompare.sh
committed- committed
- committed
- committed
- committed
Commits on Aug 14, 2024
- committed
- committed
[curand] in gg_tt.mad cudacpp.mk, disable CURAND if CURAND headers are not found (madgraph5#965), also clean up NVTX and debug printouts
committed[curand] in CODEGEN (backport gg_tt.mad) cudacpp.mk, disable NVTX/CURAND if NVTX/CURAND headers are not found (madgraph5#965)
committed- committed
- committed
Commits on Aug 21, 2024
[mac] in .github/workflows/c-cpp.yml, move from gfortran-11 to gfortran-14 on Mac githib-hosted runners (fix madgraph5#971)
committed- authored
- committed
Merge remote-tracking branch 'upstream/master' (including fixes for CMS nvcc without nvtx PR madgraph5#966) into hel
committed- committed
- authored
- authored
[june24] in gg_tt.sa and CODEGEN, disable comparison of channelId in runTest.exe if MULTICHANNEL is disabled madgraph5#976
committed- committed
- committed
[mac] in .github/workflows/c-cpp.yml, move from gfortran-11 to gfortran-14 on Mac githib-hosted runners (fix madgraph5#971)
committed- committed
- committed
- committed
- committed
- committed
[june24] in CODEGEN patchMad.sh, remove '-f' option from patch commands (introduced in 55b3e74): I prefer that users get and report an error if there is something wrong here...
committed- committed
- committed
- committed
- committed
Merge remote-tracking branch 'upstream/master' (with hel madgraph5#960, mac madgraph5#974, nvcc madgraph5#966) into june24
committed- committed
- committed
- committed
- committed
- committed
- committed
- committed
- committed
Commits on Aug 22, 2024
Commits on Aug 30, 2024
- committed
- committed
- committed
- committed
Commits on Aug 31, 2024
- committed
- committed
- committed
- committed
- committed
- committed
- committed
- committed
- committed
- committed
[gpucpp] regenerate all processes - this completes the add-ons to PR madgraph5#979 and madgraph5#980
committed
Commits on Sep 1, 2024
- committed
- committed
- committed
- committed
- committed
- committed
- committed
- authored
- authored
- authored
- committed
- committed
- committed
Merge remote-tracking branch 'upstream/master' (including madgraph5#979, madgraph5#980, madgraph5#984 patches for the new CI and Source/makefile) into june24
committed- committed
- committed
- committed
- committed