[mvapich-discuss] FW: Announcing the release of MVAPICH2-GDR 2.2a and OSU Micro-Benchmarks (OMB) 5.1
Panda, Dhabaleswar
panda at cse.ohio-state.edu
Tue Nov 10 22:24:28 EST 2015
The MVAPICH team is pleased to announce the release of MVAPICH2-GDR
2.2a and OSU Micro-Benchmarks (OMB) 5.1.
MVAPICH2-GDR 2.2a is based on the standard MVAPICH2 2.2a release and
incorporates designs that take advantage of the new GPUDirect RDMA
technology for inter-node data movement on NVIDIA GPUs clusters with
Mellanox InfiniBand interconnect. Further, MVAPICH2-GDR 2.2a provides
efficient support for CUDA-Aware Non-Blocking Collectives (NBC) that
deliver the maximal overlap by combining GDR and Core-Direct features
Features, Enhancements, and Bug Fixes for MVAPICH2-GDR 2.2a are listed
here.
* Features and Enhancements (since MVAPICH2-GDR 2.1)
- Based on MVAPICH2-2.2a
- Support for efficient Non-Blocking Collectives for Device buffers
- Exploiting Core-Direct and GPUDirect RDMA features
- Maximal overlap of communication and computation on both CPU and GPU
- Enabling Support on GPU-Clusters using regular OFED
(without GPUDirect RDMA)
- Capability to use IPC
- Capability to use GDRCOPY
- Tuning of IPC thresholds for multi-GPU nodes
* Bug Fixes (since MVAPICH2-GDR 2.1)
- Fix IPC interaction with RMA Lock synchronization
- Thanks to Akihiro Tabuchi at University of Tsukuba
MVAPICH2-GDR 2.2a release requires the following software to be
installed on your system:
- Mellanox OFED 2.1 or later
- NVIDIA Driver 331.20 or later
- NVIDIA CUDA Toolkit 6.0 or later
- Plugin module to enable GPUDirect RDMA
- (Strongly recommended) NVIDIA GDRCOPY module
Further, MVAPICH2-GDR 2.2a enables support on GPU-Cluster using
regular OFED (without GPUDirect RDMA)
New features, enhancements and bug Fixes for OSU Micro-Benchmarks
(OMB) 5.1 are listed here.
* New Features & Enhancements (since OMB 5.0)
- Introduce non-blocking collective v-variants as well as ialltoallw
* osu_iallgatherv
* osu_ialltoallv
* osu_igatherv
* osu_iscatterv
* osu_ialltoallw
- Add support for benchmarking GPU-Aware non-blocking collectives.
Overlap can be computed using either CPU or GPU kernels.
* osu_iallgather
* osu_iallgatherv
* osu_ialltoall
* osu_ialltoallv
* osu_ialltoallw
* osu_ibcast
* osu_igather
* osu_igatherv
* osu_iscatter
* osu_iscatterv
- Allow users the ability to specify zero warmup iterations
* Bug Fixes
- fix openacc pragma
For downloading MVAPICH2-GDR 2.2a, OMB 5.1, associated user guide, and
sample performance numbers please visit the following URL:
http://mvapich.cse.ohio-state.edu
All questions, feedback, bug reports, hints for performance tuning,
and enhancements are welcome. Please post it to the mvapich-discuss
mailing list (mvapich-discuss at cse.ohio-state.edu).
Thanks,
The MVAPICH Team
PS: The number of downloads from the MVAPICH site has crossed 0.3
million (currently stands at more than 303,000). The MVAPICH team
would like to thank all its users and organizations!!
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cse.ohio-state.edu/pipermail/mvapich-discuss/attachments/20151111/7c414748/attachment-0001.html>
More information about the mvapich-discuss
mailing list