Skip to content

Latest commit

 

History

History
24 lines (18 loc) · 871 Bytes

File metadata and controls

24 lines (18 loc) · 871 Bytes

Experiment 27: Cache Thrashing, Random vs Sequential

1. Focus

  • Compare a locality-friendly sequential pass against a cache-thrashing random pass.

2. Question

  • How much throughput is lost when the same useful work is reordered into a poor-locality access pattern?

3. Variants

  • sequential
  • random

4. Method

  • Run the same logical transform with identical output semantics and only change the access ordering.
  • Keep arithmetic and total payload fixed so the result isolates locality loss rather than extra work.

5. Outputs

  • Median GPU time by ordering.
  • Effective GB/s by ordering.
  • Slowdown of random relative to sequential.

6. Interpretation

  • This is a direct locality study, so the important result is the gap between the two orderings.
  • The measured penalty supports arguments for sorting, binning, or data-layout reordering.