Lesley's Digital Garden

Home

❯

Notes

❯

CUDA performance

CUDA performance

Mar 14, 2024

  • parent : CUDA
  • computing/performance/gpu
  • computing/cuda
  • computing/gpu

Consider cache and CUDA memory coalescing.

One way to improve performance is to use shared memory to avoid hostile memory access pattern. However, bank conflicts can play a role in reducing performance when using shared memory.

See Also

  • compute overlap

Graph View

Backlinks

  • CUDA

Created with Quartz v4.5.2 © 2026. All notes are written in Obsidian.

  • GitHub Repository
  • Mastodon