Increasing utilization - gdrcopy_copybw

Open
#288 3 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
4/5
Estimated time
3-5 days
Newbie friendliness
25/100
Issue type
Bug
Clarity
Needs clarification
Activity status
Stale
Tech stack
c, linux
Domain
performance

Research direction

Start with the gdrcopy_copybw benchmark and the memcpy_uncached_store_avx implementation mentioned in the issue. Compare the existing _mm256_load_pd and _mm256_stream_pd path with the _mm512 equivalents on the stated NVIDIA A100 80GB PCIe and Gen 4 PCIe setup. Done means documenting the reason for the current width and whether 512-bit operations improve utilization for 32kB-8MB buffers.

Written by the indexing model from the issue text.

Description

Hi,

I am running the gdrcopy_copybw benchmark on NVIDIA A100 80GB PCIe with Gen 4 PCIe.
It appears that the utilization doesn't reach its maximal possible value, getting about 20 GB/s out of the possible 32 GB/s, for buffers of sizes 32kB-8MB.
Upon looking into the code, it appears that in your implementation of memcpy_uncached_store_avx you are using 256 bit functions _mm256_load_pd and _mm256_stream_pd.
What could be the reason for that? Is there a reason for not using 512 bit functions _mm512_load_pd and _mm512_stream_pd instead? Could using the 512 bit functions increase the utilization?

Thanks!

Dominant language
C
Stars
1.4k
Forks
198
PR merge metrics
No merged PRs in 30d

Getting set up

This project ships no dev container, Dockerfile or contributing guide, so setting up is up to you: start from its README, and see our first-contribution guide for the general steps.

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from NVIDIA/gdrcopy

All issues in NVIDIA/gdrcopy

Similar issues

More C issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.