Add CUDA decompression kernels for OnPair - #8920

Open

robert3005 wants to merge 14 commits into

Commits on Aug 10, 2026

Add CudaExecutionCtx::fork and wait_for for sibling-stream execution

robert3005
and
claude
committed
793707d
Add CUDA decompression kernels for OnPair
robert3005
committed
3f2796b
fuse
robert3005
committed
9f2ca03
fixes
robert3005
committed
26ef32d
less
robert3005
committed
8d9f14c
formatted
robert3005
committed
4c1d35e
less
robert3005
committed
4bcb481
less
robert3005
committed
0398d0f
fixes
robert3005
committed
d62b64f
fix more
robert3005
committed
dcf6df7
fix
robert3005
committed
ac21a4f
fixes
robert3005
committed
2a2e8c9
fewer roundtrips
robert3005
committed
da8bb1b
Remove CudaExecutionCtx::fork and run the OnPair pipeline on one stream
robert3005
committed
47ad0ab