Repository navigation

Fix/dx12 shader f16 - #4

Merged

dafioram merged 12 commits into

mainfrom

fix/dx12-shader-f16

Nov 5, 2026

Merged

Fix/dx12 shader f16#4
dafioram merged 12 commits into
mainfrom
fix/dx12-shader-f16

Commits on Nov 3, 2026
- Explain why DX12 FP16 is skipped
  
  dafioram
  and
  claude
  committed
  e267553
- Add a DX12 FP16 experiment on WARP, with and without DXC
  
  dafioram
  and
  claude
  committed
  729658c
- Warn that dxcompiler.dll can hide DX12 GPUs; show the compiler in --list
  
  dafioram
  and
  claude
  committed
  1ff4a5f
Commits on Nov 4, 2026
- Name the K block behind a wrong matmul entry
  
  dafioram
  and
  claude
  committed
  f4bd5ae
- Map which matmul tiles were written when the check fails
  
  dafioram
  and
  claude
  committed
  cc1687c
- Create the matmul band offsets with their contents
  
  dafioram
  and
  claude
  committed
  18d5b11
- Retry a failed matmul check as one dispatch over all rows
  
  dafioram
  and
  claude
  committed
  b4678c3
- Use one matmul dispatch over all rows on software adapters
  
  dafioram
  and
  claude
  committed
  a8551d9
- Make each matmul band's bind group just before its dispatch
  
  dafioram
  and
  claude
  committed
  28830f1
- Run the WARP CI matmul at N=2048 and ship README with the builds
  
  dafioram
  and
  claude
  committed
  3b2d11c
Commits on Nov 5, 2026
- Make the WARP CI gate fail unless the matmul ran at N=2048
  
  dafioram
  and
  claude
  committed
  145befb
- Keep the requested matmul size on software adapters
  
  dafioram
  and
  claude
  committed
  d4cf924