Merged
#66: 新增 Reg 矢量接口单双发与理论吞吐分析卡 #163
bianqueAdd on 15 days ago
#66: 新增 Reg 矢量接口单双发与理论吞吐分析卡 #163
Merged
Pull Request has been successfully merged by CANN-robot.
（Thank you for bianque's contribution.）15 days ago Added label: cann-cla/yes
CANN-robot
15 days ago Comments：
15 days ago Comments：
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.
You can self-configure the PR merge rules for this repository. For more details, please refer to Here.
For more, you also can visit HICANN.
PR Approval Progress
✅ Congratulations! All modules have met the lgtm and approve requirements.
Module Approval Details
| module | lgtm status | approve status |
|---|---|---|
| repo-cann/cannbot-knowledge | ✅ wangsj0012, qq_41667743 (2/2) | ✅ wangsj0012 (1/1) |
💡 Tip:
- Committer can comment
/approveor/lgtm- Commenting
/approveimplies both code review (lgtm) and intent to merge (approve)
CLA Signature Pass
bianque, thanks for your pull request. All authors of the commits have signed the CLA. 👍
CANN-robot15 days ago Set qq_41667743,wangsj0012,jon_jin,zhanghz1,fafahan,zhengshencn_hwca,linghuis,OpSphere,xfwang_ecnu as approver
15 days ago Set qq_41667743,wangsj0012,jon_jin,zhanghz1,fafahan,zhengshencn_hwca,linghuis,OpSphere,xfwang_ecnu as approver
CANN-robot15 days ago Set qq_41667743,wangsj0012,jon_jin,zhanghz1,fafahan,zhengshencn_hwca,linghuis,OpSphere,xfwang_ecnu as approver
15 days ago Set qq_41667743,wangsj0012,jon_jin,zhanghz1,fafahan,zhengshencn_hwca,linghuis,OpSphere,xfwang_ecnu as approver
99 messages are collapsed here。 See more
10 days ago Added label: ci-pipeline-passed
wangshaojun
10 days ago Comments：
10 days ago Comments：
/approve
10 days ago Added label: lgtmapproved
10 days ago Merged pull request
Pull Request 说明
变更说明
knowledge/ops/ascendc/optimizations/instruction_model/reg_vector_issue_and_theoretical_perf.md（Optimization），仅面向 Ascend 950PR/950DT：对 Reg 矢量接口按指令形态、源/目的 dtype 和实际处理元素数选择理论并行度P（elements/cycle）；仅在同位宽、单个 256B VL 的假设下示例换算n/P。MaskReg 表中以 cycle 表示的行单列，不与 elements/cycle 混算。LoadAlign按单/双搬入和DIST_DS_B8/DIST_DS_B16模式区分；Gather的VGATHER2（离散搬入）与VSELR（离散操作）均为单发，分别按各自的源/目的类型与工作量查表。Pack、UnPack、Unsqueeze、ExpSub、Reduce SUM等跨位宽情形先核定元素数，不能直接套用单 VL 示例。instruction_model/index.md、上级optimizations/index.md和logs/2026-10-28.md。e680c34c09ba6919b3cebd2b7ec027f4ff6f0f22下三份docs/zh/api/appendix/Markdown 核对。不扩展至其他平台，不修改 API/Guide 镜像、治理或 Skill。变更类型
验证