Repository navigation

Optimized PFPL Performance for ROCm - #7

Open

Seth-Wolfgang wants to merge 8 commits into

szcompressor:mainfrom

Seth-Wolfgang:main

Open

Optimized PFPL Performance for ROCm #7
Seth-Wolfgang wants to merge 8 commits into
szcompressor:mainfrom
Seth-Wolfgang:main

Commits on Nov 4, 2026
- HIP: two-array PFPL stream, AMD kernels, and an FMA-safe ABS bound check
  
  Seth-Wolfgang
  and
  claude
  committed
  f826460
- Restore the HIP backend on main
  
  Seth-Wolfgang
  and
  claude
  committed
  6d22386
- fz::pfpl: clear a new Plan on the calling stream; add tests and notes
  
  Seth-Wolfgang
  and
  claude
  committed
  07919bf
Commits on Nov 5, 2026
- Link fzgmod_modules against fzgmod_mem
  
  Seth-Wolfgang
  and
  claude
  committed
  6ce75f5
- adds FP64 for ROCm
  Seth-Wolfgang
  committed
  eef094a
- fz::pfpl: record doubles in the changelog and notes
  
  Seth-Wolfgang
  and
  claude
  committed
  1811d45
- fz::pfpl: a host coder that writes the kernels' stream
  
  Seth-Wolfgang
  and
  claude
  committed
  c7b048a
Commits on Nov 6, 2026
- docs: state the host coder's reason in the library's terms
  
  Seth-Wolfgang
  and
  claude
  committed
  74855a4