[Mvapich-discuss] Announcing the release of OSU Micro-Benchmarks (OMB) 7.5

Chen, Chen Chun chen.10252 at buckeyemail.osu.edu
Wed Apr 16 21:33:42 EDT 2025 

Hi Adam,

OMB supports Intel GPUs through SYCL. Users can enable this support by adding the following configuration flags:
--enable-sycl --with-sycl=/PATH/TO/ONEAPI

For example:
./configure --enable-sycl --with-sycl=/opt/intel/oneapi CC=mpicc CXX=mpicxx

You’ll also find more detailed setup instructions in the README (https://mvapich.cse.ohio-state.edu/static/media/mvapich/README-OMB.txt).
Some SYCL-related content can be found in the README for reference:

SYCL extensions can be enabled by configuring OMB with --enable-sycl
option as shown below.

    ./configure CC=/path/to/mpicc
                CXX=/path/to/mpicxx
                --enable-sycl
                --with-sycl=/path/to/sycl/install
    make
    make install
…
If both CUDA and OpenACC support is enabled you can switch between the modes
using the -d [cuda|openacc] option to the benchmarks. If ROCm support is
enabled, you need to use -d rocm option to make the benchmarks use this feature.
If SYCL support is enabled, you need to use -d sycl option to make the
benchmarks use this feature. Whether a process allocates its communication
buffers on the GPU device or on the host can be controlled at run-time.
Use the -h option for more help.

If you have any questions, feel free to reach out—we’re always happy to help!

Best,
Chen-Chun

From: Mvapich-discuss <mvapich-discuss-bounces+chen.10252=osu.edu at lists.osu.edu> on behalf of Goldman, Adam via Mvapich-discuss <mvapich-discuss at lists.osu.edu>
Date: Wednesday, April 16, 2025 at 1:42 PM
To: mvapich-discuss at lists.osu.edu <mvapich-discuss at lists.osu.edu>
Subject: Re: [Mvapich-discuss] Announcing the release of OSU Micro-Benchmarks (OMB) 7.5
!-------------------------------------------------------------------|
  This Message Is From an External Sender
  This message came from outside your organization.
|-------------------------------------------------------------------!

Hello,

I noticed that the CHANGES list "Intel GPU support." Is this correct for v7.5?

I am not able to find a way to compile with the level-zero library and the README does not mention "Intel GPU" anywhere.

Thank you,
Adam Goldman


-----Original Message-----
From: Mvapich-discuss <mvapich-discuss-bounces+adam.goldman=intel.com at lists.osu.edu> On Behalf Of Panda, Dhabaleswar via Mvapich-discuss
Sent: Friday, November 1, 2024 7:07 PM
To: Mvapich-discuss at lists.osu.edu
Cc: Panda, Dhabaleswar <panda at cse.ohio-state.edu>
Subject: [Mvapich-discuss] Announcing the release of OSU Micro-Benchmarks (OMB) 7.5

The MVAPICH team is pleased to announce the release of OSU Micro-Benchmarks (OMB) 7.5.

Please note that OMB is also available through the Spack package manager. Now the system administrators and users of OSU Micro-Benchmarks (OMB) will be able to install these libraries on their systems using Spack.

The new features, enhancements, and bug fixes for OSU Micro-Benchmarks
(OMB) 7.5 are listed here:

* New Features & Enhancements (since 7.4)

    - New OpenSHMEM benchmarks
        * osu_oshm_get_bw, osu_oshm_get_nb_bw
        * osu_oshm_put_bw, osu_oshm_put_nb_bw
        * osu_oshm_get_overlap
    - New partitioned point-to-point benchmkark
        * osu_partitioned_latency
    - Add NVIDIA, AMD, and Intel GPU support for neighborhood collective
      benchmarks
        * osu_neighbor_allgather, osu_neighbor_allgatherv,
        * osu_neighbor_alltoall, osu_neighbor_alltoallv,
        * osu_neighbor_alltoallw, osu_ineighbor_allgatherv,
        * osu_ineighbor_allgatherv, osu_ineighbor_alltoall,
        * osu_ineighbor_alltoallv, osu_ineighbor_alltoallw
    - Add Intel GPU support for the following benchmarks
        * Point-to-point benchmarks
            * osu_bibw, osu_bw, osu_mbw_mr, osu_latency, osu_multi_lat,
            * osu_latency_mp, osu_latency_mt, osu_bw_persistent,
            * osu_bibw_persistent, osu_latency_persistent,
            * osu_partitioned_latency
        * Network congestion benchmarks
            * osu_bw_fan_in, osu_bw_fan_out
        * Blocking collective benchmarks
            * osu_allgather, osu_allgatherv, osu_alltoall, osu_allreduce,
            * osu_alltoallv, osu_alltoallw, osu_bcast, osu_barrier, osu_gather,
            * osu_gatherv, osu_reduce, osu_reduce_scatter, osu_scatter,
            * osu_scatterv, osu_reduce_scatter_block
        * Non-Blocking collective benchmarks
            * osu_iallgather, osu_iallgatherv, osu_iallreduce, osu_ialltoall,
            * osu_ialltoallv, osu_ialltoallw, osu_ibcast, osu_ibarrier,
            * osu_igather, osu_igatherv, osu_ireduce, osu_iscatter,
            * osu_iscatterv, osu_ireduce_scatter, osu_ireduce_scatter_block
        * Persistent collective benchmarks
            * osu_allgather_persistent, osu_allgatherv_persistent,
            * osu_allreduce_persistent, osu_alltoall_persistent,
            * osu_alltoallv_persistent, osu_alltoallw_persistent,
            * osu_barrier_persistent, osu_bcast_persistent,
            * osu_gather_persistent, osu_gatherv_persistent,
            * osu_reduce_persistent, osu_reduce_scatter_persistent,
            * osu_scatter_persistent, osu_scatterv_persistent
        * One-sided benchmarks
            * osu_acc_latency, osu_cas_latency, osu_fop_latency, osu_get_bw,
            * osu_get_latency, osu_put_bibw, osu_put_bw, osu_put_latency,
            * osu_get_acc_latency
    - Changed default MPI_TYPE of reduce benchmarks to MPI_INT.
            * osu_reduce, osu_allreduce, osu_reduce_scatter, osu_reduce_scatter_block

* Bug Fixes
    - Fixed RCCL build in ROCm 6 due to installation path changes.
    - Fixed a bug leading to nvcc command not found error when builing with NCCL
      support.
        * Thanks to Raghu Raja @Amazon for report.
    - Resolved a bug in the maximum memory limit check, which previously
      considered only the memory allocated by a single buffer. This check and
      command line option 'M' no longer exists.
        * Thanks to Van Man Nguyen @Eviden for report and patch.
    - Fixed bug while printing validation results in some collective benchmarks.
    - Updated reduction benchmarks to use MPI_SIGNED_CHAR
       instead of MPI_CHAR.
        * Thanks to Rémi Dehenne @CEA for report and patch.
    - Fixed a bug in Reduce, Reduce-Scatter benchmarks that caused validation
      to fail when running with ROCM support.
        * Thanks to Anton Daumen @Eviden for report and patch.
        * Thanks to Edgar Gabriel @AMD and Mustafa Abduljabbar @AMD for report.
    - Fixed RCCL check in configure script.
        * Thanks to Anton Daumen @Eviden for report.

For downloading OMB 7.5 and associated README instructions, please visit the following URL:

http://mvapich.cse.ohio-state.edu

All questions, feedback, bug reports, hints for performance tuning, patches, and enhancements are welcome. Please post it to the mvapich-discuss mailing list (mvapich-discuss at lists.osu.edu).

Thanks,

The MVAPICH Team

PS: We are also happy to inform you that the number of organizations using MVAPICH libraries (and registered at the MVAPICH site) has crossed 3,400 worldwide (in 92 countries). The number of downloads from the MVAPICH site has crossed 1,835,000 (1.835 million).  The MVAPICH team would like to thank all its users and organizations!!
_______________________________________________
Mvapich-discuss mailing list
Mvapich-discuss at lists.osu.edu
https://lists.osu.edu/mailman/listinfo/mvapich-discuss
_______________________________________________
Mvapich-discuss mailing list
Mvapich-discuss at lists.osu.edu
https://lists.osu.edu/mailman/listinfo/mvapich-discuss
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osu.edu/pipermail/mvapich-discuss/attachments/20250417/ed870f55/attachment-0002.html>

More information about the Mvapich-discuss mailing list