Skip to content

[Cherry-Pick][Optimization][#7687] optimize tritonmoe_preprocess op #7786

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
yongqiangma merged 4 commits into from
May 14, 2026
Merged

[Cherry-Pick][Optimization][#7687] optimize tritonmoe_preprocess op #7786

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
091f862
opt moe_align_kernel
yongqiangma May 12, 2026
7288323
Merge branch 'release/2.6' into moe_align_26
yongqiangma May 13, 2026
57e639f
Merge branch 'release/2.6' into moe_align_26
yongqiangma May 14, 2026
7281a4b
Merge branch 'release/2.6' into moe_align_26
yongqiangma May 14, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions custom_ops/gpu_ops/helper.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -73,6 +73,8 @@ namespace cub = hipcub;
using json = nlohmann::json;
#endif

#define CEILDIV(a, b) (((a + b - 1) / b))

This comment was marked as outdated.

Sign in to view

Copy link
Copy Markdown

@PaddlePaddle-bot PaddlePaddle-bot May 14, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

❓ 疑问 CEILDIV 宏存在两个潜在问题：

  1. 整数溢出：a + b - 1 在 a 接近 INT_MAX 时会溢出（例如 int32_t 场景）。
  2. 宏参数未加括号：若传入表达式（如 CEILDIV(x+1, y)），a 展开后无括号保护可能导致运算符优先级错误。

建议修改为：

#define CEILDIV(a, b) (((a) + (b) - 1) / (b))

或使用 inline 函数以避免宏的副作用：

template <typename T>
inline T ceildiv(T a, T b) { return (a + b - 1) / b; }


#define CUDA_CHECK(call) \
do { \
const cudaError_t error_code = call; \
Expand Down
Loading
Loading