Pull Request已成功合入, 合并人@CANN-robot
（感谢 jinpenghe 的贡献）
Thanks for your pull-request.
The full list of commands accepted by me can be found at here。
You can get sig-info at here
PR Approval Progress
✅ Congratulations! All modules have met the lgtm and approve requirements.
Module Approval Details
| module | lgtm status | approve status |
|---|---|---|
| index | ✅ chenqi317, fanqirui (2/2) | ✅ chenqi317 (1/1) |
💡 Tip:
- Committer can comment
/approveor/lgtm- Commenting
/approveimplies both code review (lgtm) and intent to merge (approve)
CANN-robotCANN-robot🔵 source code change are detected, tasks labels is removed in this pull request!
问题/功能描述
本次PR旨在优化“未排序段”操作核函数的分块（Tiling）策略，以提升计算效率和资源利用率。主要解决了两个问题：一是动态排序分块算法的关键参数（如基础块大小）未能充分发挥硬件性能；二是分列（SplitCol）策略的启用条件存在逻辑错误，导致在某些数据规模下无法选择更优策略。
修改方案描述
修改方案包含三部分核心内容：1. 提升动态排序分块算法的基础块大小（BASE_A_SIZE）和列限制大小（COL_LIMIT_SIZE）至2048，并调整AutoTiling函数参数，以扩大单核处理数据块，减少核间通信。2. 修复分列策略的启用条件，将判断比率大于阈值改为小于阈值，确保在数据行数远小于输出段数时能正确启用该策略。3. 调整通用分块算法AutoTiling中的排序逻辑，优先比较行分块数（M）而非列分块数（N），以适配不同策略对分块形状的偏好。
jinpenghe🔵 source code change are detected, tasks labels is removed in this pull request!
jinpenghejinpenghe🔵 source code change are detected, tasks labels is removed in this pull request!
jinpenghejinpenghe🔵 source code change are detected, tasks labels is removed in this pull request!
compile
流水线任务触发成功
任务链接 [2e4b4a2e3bbd4718aa0669562c85d242][流水线指导]
| 任务名称 | 状态 | 日志 | 下载链接 |
|---|---|---|---|
| 解析CI分支 | ✅ SUCCESS | ||
| codecheck | ✅ SUCCESS | >>>>> | |
| anti_virus | ✅ SUCCESS | >>>>> | |
| Check_Pr | ✅ SUCCESS | >>>>> | |
| Compile_Ascend_X86_ubuntu24 | ✅ SUCCESS | >>>>> | |
| Compile_Ascend_X86_mobile_station | ✅ SUCCESS | >>>>> | >>>>> |
| Compile_Ascend_ARM | ✅ SUCCESS | >>>>> | >>>>> |
| Compile_Ascend_single | ✅ SUCCESS | >>>>> | >>>>> |
| Compile_Ascend_experimental | ✅ SUCCESS | >>>>> | >>>>> |
| Compile_Ascend_X86_950 | ✅ SUCCESS | >>>>> | >>>>> |
| Compile_Ascend_ARM_950 | ✅ SUCCESS | >>>>> | >>>>> |
| API_Check | ✅ SUCCESS | >>>>> | |
| Pre_Compile | ✅ SUCCESS | ||
| pre_comment | ✅ SUCCESS | >>>>> | |
| Compile_Ascend_X86 | ✅ SUCCESS | >>>>> | >>>>> |
| Compile_Ascend_X86_mobile_station_ubuntu24 | ✅ SUCCESS | >>>>> | |
| Compile_Ascend_X86_950_ubuntu24 | ✅ SUCCESS | >>>>> | |
| UT_Test_ophost | ✅ SUCCESS | ||
| UT_Test_opapi | ✅ SUCCESS | ||
| UT_Test_kernel | ✅ SUCCESS | ||
| UT_Test_opgraph | ✅ SUCCESS | ||
| Smoke_A900_npupool | ✅ SUCCESS | >>>>> |
[2026-03-10 20:21:31] CI执行结束
| 检查项 | 状态 | 详情 |
|---|---|---|
| 接口管理 |
通过
|
/lgtm
描述
UnsortedSegmentMin算子的排序模板在进行Block切分时，过多切分列数，导致搬运资源浪费和vector计算间空隙较大；此外, 高位宽数据排序性能较差，根据情况将高位宽数据cast到低位宽数据进行排序。
关联的Issue
https://gitcode.com/cann/ops-nn/issues/1385
测试
文档更新
类型标签