[FIX] Deploy _custom_ops.py SMEM 32KB→48KB fix — was in repo but never deployed

Browse Source

Source: cccl_upstream/cub/test/catch2_test_grid_even_share.cu (random pick)

GridEvenShare test validates: grid_size = min(max_grid, ceil_div(N, tile_size))
If SMEM is reported as 32KB instead of 48KB, tile_size is 33% smaller,
grid_size is 50% larger, and every kernel launch wastes occupancy.

Base image _custom_ops.py: get_max_shared_memory_per_block → 32*1024 = 32768
Our fix: → 49152 (confirmed 48KB via ixsmi on Phanthy Cloud)

This affects ALL kernel launches that query SMEM limits:
  - Triton JIT tile sizing (prefix_prefill, flash_attn)
  - ixformer internal SMEM allocation
  - paged_attention block_size calculations

Was modified in vllm/_custom_ops.py but NEVER added to qwen3_6_scripts/
for Docker deployment. Now deployed.

...

This commit is contained in:

dylanyunlon

2026-08-05 08:38:36 +00:00

parent 81972a05c6

commit b902090fb2

2 changed files with 1139 additions and 0 deletions

Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL

Download Patch File Download Diff File Expand all files Collapse all files

1128

qwen3_6_scripts/_custom_ops.py Normal file

File diff suppressed because it is too large Load Diff