Home Resources for Co-Design Programs Gpu-affinity Gpu-affinity_default-binding

GPU SAXPY with default cpu binding

Version's name: GPU SAXPY with default cpu binding ; a version of the GPU SAXPY program.
Repository: [home] and version downloads: [.zip] [.tar.gz] [.tar.bz2] [.tar]
Patterns and behaviours:

Poor GPU data transfer rate due to affinity

Recommended best-practices:

Appropriate process/thread mapping to GPUs

- Available version(s):

GPU SAXPY with optimal cpu binding

This version of the GPU SAXPY kernel is launched with an srun command with no additional parameters.

srun ./kernel.exe 8000000000

The default CPU binding for MPI tasks configured on the system is then used. Depending on the system configuration this might not be optimal. The offloading call may be executed by a CPU core that is not on the same NUMA domain that the target GPU is connected to. This leads to higher latencies and lower bandwidth for data transfers between CPU and GPU. In the worst case the CPU handling the data transfer to the GPU is on a different socket which increases this effect. Depending on the configuration the processes might even get moved to other cores.

The following experiments have been registered:

This project has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreements No 676553 (POP1) and 824080 (POP2).

Currently, the project receives funding from the European High-Performance Computing Joint Undertaking (JU) under grant agreement No 101143931 (POP3). The JU receives support from the European Union's Horizon Europe research and innovation programme and Spain, Germany, France, Portugal and the Czech Republic.