[Mvapich-discuss] OSU Benchmark Regressions in 7.4

[Paniraja Guptha, Akshay]

Hi Thomas,

Thanks for contacting and yes, this is the right forum for OMB.

I have not come across data validation failure for MPI_CHAR. So, please provide the following information for debugging,


  *   OpenMPI build details
  *   Cluster details
  *   OMB config line
  *   Run command

Thanks,
Akshay

From: Mvapich-discuss <mvapich-discuss-bounces+panirajaguptha.1=osu.edu at lists.osu.edu> On Behalf Of Huber, Thomas via Mvapich-discuss
Sent: Tuesday, August 27, 2024 11:41 AM
To: mvapich-discuss at lists.osu.edu
Subject: [Mvapich-discuss] OSU Benchmark Regressions in 7.4

Good afternoon, I’ve begun utilizing the OSU 7. 4 benchmark suite release recently and have found data validation regressions introduced for the osu_reduce and osu_allreduce. To be clear, I am utilizing OpenMPI and libfabric (OFI). When I use
ZjQcmQRYFpfptBannerStart
This Message Is From an External Sender
This message came from outside your organization.
    Report Suspicious  <https://us-phishalarm-ewt.proofpoint.com/EWT/v1/KGKeukY!vIQf0sYND6YARRdwHy7kN1Im7qbxJ3rMk_uQ6P8Am4eD2AQKi_8Rd8eQY-jszxVTJbj1npdCOTTxv8YFj4_Kv4xx8MdlRc7PMdzzFT-s7msHBs1y09Ylmt8zvlN3LCi0tBcRGdtlE2TE53wp$>   ‌
ZjQcmQRYFpfptBannerEnd
Good afternoon,

I’ve begun utilizing the OSU 7.4 benchmark suite release recently and have found data validation regressions introduced for the osu_reduce and osu_allreduce. To be clear, I am utilizing OpenMPI and libfabric (OFI). When I use the 7.3 benchmarks, I do not hit a data validation issue. Also, if I take the OSU 7.3 .c files for the osu_reduce and osu_allreduce tests and overwrite the OSU 7.4 osu_reduce and osu_allreduce files with the older 7.3 files, I do not hit a data validation issue. It seems something fundamentally changed between these two releases, and by just diff’ing the files, it looks like no major changes occurred to the individual benchmark test files themselves.

These data validation errors only occur for MPI_TYPE of mpi_char. MPI_TYPE of mpi_int and mpi_float work just fine.

Is this issue known already? In the OSU 7.4 changelog, it states the following:

“  - Fixed data validation issue with Reduce, Allreduce, Reduce_scatter
      benchmarks

But it seems these issues still exist.

Thanks in advance for any help! If this is not the appropriate forum for reporting OSU benchmark issues, please refer me to wherever is best.
External recipient
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osu.edu/pipermail/mvapich-discuss/attachments/20240827/c00a7d4c/attachment-0002.html>