biondizzle 696462f07a feat: TMA async load infrastructure for FMHA kernel ...

- fmha_tma.cuh: TMA descriptor creation, mbarrier helpers, cp.async.bulk.tensor.2d wrappers
- fmha_6warp_tma.cuh: TMA-integrated multirow kernel with async GMEM→SMEM loads
  - TMA loads Q, K, V tiles to row-major SMEM
  - Transposes to canonical K-major layout for MMA
  - Same softmax/epilogue as non-TMA kernel
- test_fmha_tma.cu: Test harness for TMA FMHA (HD=64 first)

2026-05-29 04:36:52 +00:00

..

attention

feat: TMA async load infrastructure for FMHA kernel

2026-05-29 04:36:52 +00:00

cache

KV Cache: schema, allocator, pools, manager, append_swa kernel

2026-05-22 00:08:38 +00:00

compressor

Restructure: cutedsl/ -> dsv4/ with proper layering

2026-05-21 17:30:44 +00:00

cuda

Clean up E2M1 dequant: use LUT approach (consultant recommendation)

2026-05-28 16:17:47 +00:00

decode

Restructure: cutedsl/ -> dsv4/ with proper layering

2026-05-21 17:30:44 +00:00

gemm

NVFP4-1.1: Mark fp4_quant.py as toolchain-blocked, clean up test files

2026-05-28 04:59:01 +00:00

indexer

Clean up E2M1 dequant: use LUT approach (consultant recommendation)

2026-05-28 16:17:47 +00:00

router

Router: Blackwell-native fused decode kernel — real CuTeDSL implementation

2026-05-21 22:04:20 +00:00

__init__.py

Restructure: cutedsl/ -> dsv4/ with proper layering

2026-05-21 17:30:44 +00:00