madgraph5/madgraph4gpu

Issues

• Port clang-format config to clang-format-18 and upgrade the github CI to ubuntu-22.04

  #1023 opened 3 months ago by valassi
  0

• Library path to AMD libraries may not be correct in certain cases

  #1020 opened 3 months ago by Qubitol
  1

• Reminder: move default fptype to m in cudacpp makefiles and scripts

  #995 opened 3 months ago by valassi
  0

• cudacpp_backend card should be removed from output.py (it is only needed in launch_plugin.py)

  #1015 opened 3 months ago by valassi
  2

• Found unexpected entry in run_card: "cudacpp_backend" with value "cpp"

  #1009 opened 3 months ago by valassi
  2

• "leading zeros in decimal integer literals are not permitted; use an 0o prefix for octal integers" from '1.00.01' cudacpp version

  #1013 opened 3 months ago by valassi
  0

• FPE in vxxxxx during runTest.exe (testxxx) for HIP on LUMI

  #1011 opened 3 months ago by valassi
  2

• No such file "ee_mumu.mad/SubProcesses/P1_epem_mupmum/Hel/selection"

  #1010 opened 3 months ago by valassi
  2

• cxtype_ref problem for some gcc versions

  #1004 opened 3 months ago by roiser
  9

• FPEs in check.exe with clang16 and clang17

  #1005 opened 3 months ago by valassi
  3

• Crash (FPE) in check_hip.exe on LUMI

  #1003 opened 3 months ago by valassi
  3

• Test the latest master on AMD GPUs on LUMI with ROCm 6.0

  #998 opened 3 months ago by valassi
  6

• Question for Olivier: does run.sh support more than one CPU core? Can it? Should it?

  #1001 opened 3 months ago by valassi
  4

• Cuda time profiles for DY+3j have high non-ME component

  #994 opened 4 months ago by valassi
  2

• Time profiles for DY+4j (and DY+3j) have high 'python/bash' component - especially for cuda

  #1000 opened 3 months ago by valassi
  0

• Cuda time profiles for DY+4j have very high 'HEL' component for helicity filtering?

  #999 opened 3 months ago by valassi
  0

• Move LIMHEL=0 to madgraph4gpu only and add a runcard allowing LIMHEL=0

  #950 opened 3 months ago by valassi
  7

• Permille cross section discrepancy fortran vs cudacpp (gg_tt.mad) after merging june24 and goodhel together

  #991 opened 3 months ago by valassi
  4

• GPU CI failing on itscrd-v100 "Failed to initialize NVML: Driver/library version mismatch"

  #996 opened 3 months ago by valassi
  1

• LHE file mismatch fortran vs cudacpp in tlau (for multi backend builds, possibly for normal builds too) - also: "fail to reach target"

  #993 opened 4 months ago by valassi
  1

• (june24) check if anything needs to be changed in dsample.f in the use of vecsize_used or nb_warp_used

  #983 opened 4 months ago by valassi
  4

• Support for multi-GPU nodes (3: orchestrate different madevent executables to target different GPUs)

  #990 opened 4 months ago by valassi
  0

• Support for multi-GPU nodes (2: send MEs to more than one GPU from the same madevent executable?)

  #989 opened 4 months ago by valassi
  0

• script and generate script does not work with bash3.2

  #982 opened 4 months ago by oliviermattelaer
  2

• Allow LIMHEL>0 in cudacpp

  #988 opened 4 months ago by valassi
  1

• (low priority) improve passcuts interface?

  #987 opened 4 months ago by valassi
  0

• Disable comparison of channelid in runTest for SA builds

  #976 opened 4 months ago by valassi
  2

• Port (or disable) rdtsc timers on ARM/MAC

  #977 opened 4 months ago by valassi
  3

• change name of variable warp/wrap ...

  #961 opened 5 months ago by oliviermattelaer
  1

• Reduce madevent 'Fortran overhead' by restricting cudacpp helicity calculation (which scales with SIMD) to only 16 events

  #958 opened 4 months ago by valassi
  9

• Use channelid in cudacpp (and fortran) helicity filtering?

  #975 opened 4 months ago by valassi
  0

• Compilation Error in cudacpp_backend=cpp for CMS DY+nj and TT+nj (nvcc exists but not NVTX or curand)

  #965 opened 4 months ago by choij1589
  6

• Fortran missing again on Mac CI nodes?

  #971 opened 4 months ago by valassi
  5

• Faster timers based on rdtsc instead of chrono

  #972 opened 4 months ago by valassi
  4

• Instrument Fortran code with additional profiling counters

  #973 opened 4 months ago by valassi
  0

• Remove htuple.f from generated code

  #967 opened 4 months ago by valassi
  4

• DY+3 jets cross section decreases by a factor 10 when changing vector size from 16384 to 32?

  #959 opened 5 months ago by valassi
  9

• Understand why CMS sees cross section discrepancy between fortran and cuda/cpp for DY+4 jets

  #944 opened 5 months ago by valassi
  7

• Trivial improvements for xbin_min and xbin_max may lead to speedups in sample_get_x

  #969 opened 4 months ago by valassi
  3

• Vectorise phase space sampling (port x_to_f_arg to cudacpp with SIMD and GPU support - starting with sample_get_x?)

  #963 opened 4 months ago by valassi
  7

• Performance slowdown in sample_get_x from the checks for warnings at the end

  #968 opened 4 months ago by valassi
  0

• Understand why CMS sees a speedup in DY+4jets but not DY+3 jets

  #943 opened 5 months ago by valassi
  21

• Vectorise running coupling scale updates (port update_scale_coupling_vec to cudacpp with SIMD/GPU)?? Maybe not

  #964 opened 4 months ago by valassi
  0

• Add support for nprocesses>2 (i.e. beyond mirror processes) in cudacpp to speed up directory handling?

  #951 opened 5 months ago by valassi
  3

• Fix clang-format version and eventually upgrade from v15 to v17

  #952 opened 5 months ago by valassi
  0

• instrument python code in gridpacks to provide timing profiles for event generation

  #957 opened 5 months ago by valassi
  0

• Add events.lhe comparison in tlau tests

  #956 opened 5 months ago by valassi
  0

• Improve usability of cacche by experiment users?

  #954 opened 5 months ago by valassi
  0

• generate_events tries to remove GpuAbstraction.h

  #947 opened 5 months ago by valassi
  7

• Option to produce multi-build (fortran, cuda, cpp) gridpack

  #945 opened 5 months ago by valassi
  0