dylanyunlong/project_6
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

[muh] gen_patch: expand VLLM_INJECTION_POINTS to full real injection surface

Browse Source 
- Replace DEAD csrc/*.cu targets with 11 confirmed Python/Triton injection points
- Add paged_attn.py: _PARTITION_SIZE, use_v1 (V1/V2 dispatch threshold)
- Add computility-run.yaml: max-num-seqs, max-num-batched-tokens, gpu-mem-utilization
- Preserve Triton autotune injection: flash_attn BLOCK_M/N, prefix_prefill BLOCK/NUM_WARPS
- Fix PARTITION_SIZE semantic: tile size (threads*items), not items_per_thread alone
- Document CCCL parallels for each injection point
- Validated: gen_patch --dry-run produces patch (reduce -> paged_attn.py)
- Validated: test_smem_safety.py 191/191 all safe
- Validated: scale_mem_bound CCCL parity 14/14 pass
This commit is contained in:
muh-bot
2026-08-06 04:01:40 +00:00
parent bf5d19991c
commit 5aba296eba
2 changed files with 88 additions and 30 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
.gitignore vendored
View File

@@ -5,3 +5,4 @@ muh/
baseline.muh
pkgs/
enginex_base/
__pycache__/