perf(fused_cross_entropy_loss_with_max_sum): 小V场景性能优化（平坦打包快路+预取流水+裸vdiv）-ops-nn-AtomGit

Star1431

Star1431

Merged

perf(fused_cross_entropy_loss_with_max_sum): 小V场景性能优化（平坦打包快路+预取流水+裸vdiv） #14270

AlfengYuanAdd on 2 days agoMerged fromalfengyuan/ops-nn： issue225-fce-perf into cann/ops-nn： master

perf(fused_cross_entropy_loss_with_max_sum): 小V场景性能优化（平坦打包快路+预取流水+裸vdiv） #14270

Merged

AlfengYuanAdd on 2 days ago

Merged fromalfengyuan/ops-nn issue225-fce-perfintocann/ops-nn master

All

2026-11-12

style(fused_cross_entropy_loss_with_max_sum): clang-format 格式化（pre-commit 钩子自动修复）

alfengyuanSubmitted on 2 days ago

3aaaf95e

fix(fused_cross_entropy_loss_with_max_sum): 修复小V快路队列深度越约与末组向量读越界

alfengyuanSubmitted on 2 days ago

1024dce7

2026-11-11

perf(fused_cross_entropy_loss_with_max_sum): 小V快路裸vdiv+b16内联cast

alfengyuanSubmitted on 3 days ago

9d3757ea

perf(fused_cross_entropy_loss_with_max_sum): 小V快路改预取流水隐藏MTE

alfengyuanSubmitted on 3 days ago

04fdbcbd

perf(fused_cross_entropy_loss_with_max_sum): 小V场景多行平坦打包快路

alfengyuanSubmitted on 3 days ago

9f350c6b