Repository navigation
Update RAJA Perf baselines and docs - #88
Conversation
pearce8
left a comment
There was a problem hiding this comment.
@rhornung67 Thank you for working through this! A couple of minor requests:
- Please specify the minimum version of RAJA to use
- Please specify the CUDA version you are using on matrix/vector. Hopefully it's CUDA 13?
- Please remove the scaling sections, we don't need to have them as they are irrelevant
The versions of RAJA and RAJA Perf are pinned to the version of the RAJAPerf-Benchmark repo folks are instructed to use in the docs here: https://github.com/llnl/benchmarks/blob/hornung1/RAJAPerf-update/docs/13_rajaperf/rajaperf.rst#getting-the-code The compiler, etc. used to build the code used to generate the baselines are also in the docs. For example, here is for CUDA: https://github.com/llnl/benchmarks/blob/hornung1/RAJAPerf-update/docs/13_rajaperf/rajaperf.rst#h100-architecture and similarly for HIP: https://github.com/llnl/benchmarks/blob/hornung1/RAJAPerf-update/docs/13_rajaperf/rajaperf.rst#mi300a-architecture In fact, the build scripts give all the details of the CMake args used to configure the code. Is this not sufficiently clear? Do you want all these details summarized earlier in the document? Please let me know exactly what you would like. |
|
@pearce8 I added a callout statement early in the docs about library versions, etc. and cross referenced the section where the details are provided. Please let me know if this is what you want or something else. I removed the scaling sections as requested. |
|
@rhornung67 Thank you! All good now, I will merge this. |
This PR cleans up the RAJA Perf Benchmark documentation (fixes some incorrect statements, clarifies some points, etc.) and replaces the baseline data and plots which with new ones generated using the v2026.07.0 RAJA Perf release.
I tried to verify that the sphinx docs were generated correctly, but I couldn't find a link to them.