Skip to content

Conversation

dmikushin
Copy link
Contributor

@dmikushin dmikushin commented Sep 23, 2025

Summary

This PR improves the GUPS benchmark README documentation to more clearly distinguish between global memory and shared memory test modes, as well as static vs dynamic shared memory allocation methods, based on the discussion in issue #56.

Changes

  • Added clear introduction explaining the two distinct memory modes (global vs shared)
  • Separated build instructions to clarify static vs dynamic shared memory allocation
  • Provided explicit examples for running each test type with recommended configurations
  • Clarified performance implications of different shared memory allocation methods
  • Added "Performance Considerations" section explaining the differences between test modes

Context

Based on feedback from @rkarim2 in issue #56, where it was noted that:

  • The shared memory test modes need clearer documentation
  • Dynamic shared memory allocation results in suboptimal performance
  • Users need to understand when to use -s 0 (maximum static allocation) vs -s n (dynamic allocation)

Related Issues

Addresses documentation clarity issues discussed in #56

cc @rkarim2

- Add clear introduction explaining the two memory modes
- Separate build instructions for static vs dynamic shared memory
- Provide explicit examples for running each test type
- Clarify performance implications of different shared memory allocations
- Add performance considerations section

Based on feedback from issue NVIDIA-developer-blog#56 discussion
@rkarim2
Copy link

rkarim2 commented Sep 24, 2025

I'll check the changes in the next few days and comment on them.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants