biondizzle 2bbbead984 P3: CUDA RoPE kernel — single launch per call (vs 5-6 PyTorch ops) ...

New files:
- dsv4/kernels/cuda/rope_cuda.cu: GPT-J interleaved RoPE kernel (forward+inverse)
- dsv4/ops/rope_cuda.py: Python bridge with ctypes loading
- tests/unit/test_rope_cuda.py: correctness test (cos >= 0.999998)

Savings: ~915 launches/token → 183 launches/token

2026-06-02 09:05:22 +00:00

..

attention

perf: skip MQA GQA expansion in FMHA (stride=0, no 128x K/V copy)

2026-06-02 03:54:03 +00:00

cache

fix: correct gather.py kernel_dir path

2026-05-30 21:12:09 +00:00

compressor

fix: import torch.utils.cpp_extension explicitly in production_compress

2026-06-01 05:20:44 +00:00

cuda

P3: CUDA RoPE kernel — single launch per call (vs 5-6 PyTorch ops)

2026-06-02 09:05:22 +00:00

gemm

fix: use cute.where() directly for clamp in fused SwiGLU

2026-06-02 08:16:41 +00:00

indexer

P0 COMPLETE: Eliminate ALL .item() CPU-GPU syncs from NVFP4 activation path

2026-06-01 21:05:03 +00:00

router

Switch router to Nvfp4Linear production GEMM (custom CuTeDSL kernel crashes MLIR)

2026-06-01 11:17:54 +00:00

__init__.py

Restructure: cutedsl/ -> dsv4/ with proper layering

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