Pull Request已成功合入, 合并人@CANN-robot
(感谢 马琦钧 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 |
|---|---|---|
| */*/op_api/*.h | ✅ tang-lei01, wangrui_ (2/2) | ✅ tang-lei01 (1/1) |
| */*/op_graph/*_proto.h | ✅ tang-lei01, wangrui_ (2/2) | ✅ tang-lei01 (1/1) |
| */*/op_host/*_def.cpp | ✅ tang-lei01, wangrui_ (2/2) | ✅ tang-lei01 (1/1) |
| conversion/trans_data | ✅ songkai111, wangrui_ (2/2) | ✅ songkai111 (1/1) |
| experimental | ✅ songkai111, wangrui_ (2/2) | ✅ songkai111 (1/1) |
| repo-cann/ops-math | ✅ wangrui_, songkai111 (2/2) | ✅ songkai111 (1/1) |
💡 Tip:
- Committer can comment
/approveor/lgtm- Commenting
/approveimplies both code review (lgtm) and intent to merge (approve)


🔵 source code change are detected, tasks labels is removed in this pull request!


问题/功能描述
本次PR为Ascend 950平台新增了TransData算子,以支持将多种数据类型的张量从ND格式高效转换为FRACTAL_NZ及其变体格式(如FRACTAL_NZ_C0_16/32),解决了在特定硬件架构下进行高效数据格式转换的需求,为后续的矩阵运算等操作提供数据准备。同时,为提升算子性能,新增了基于SIMT(单指令多线程)架构的向量化内核实现。
修改方案描述
通过新增算子定义、内核实现及配置文件,完整实现了TransData算子的AscendC版本。具体包括:1)在算子定义层注册算子,声明支持的数据类型及源/目标格式;2)新增SIMT内核实现(trans_data_with_simt.h),利用向量化指令实现数据布局转换的高效并行计算;3)更新CI编译配置(ops_math_operator_list.yaml)和格式解析工具(parse_ini_to_json.py),确保新算子能被正确编译、识别和运行。


/approve


Review Guide
This pull-request passes review.
Committers who wrote a comment of /approve are: songkai111, tang-lei01.
Reviewers who wrote a comment of /lgtm are: tang-lei01, wangrui_, songkai111.


描述
调整TransData为开源目录接口,开源kernel等相关内容
关联的Issue
https://gitcode.com/cann/ops-math/issues/763
测试
二级冒烟
文档更新
不涉及
类型标签