Closed
[Requirement|需求建议]: (add_row_ranges): RegBase 寄存器驻留 Kahan 归约优化长尾性能 #8558
白昊8801Add on 11 days agoClosed at 9 days ago
11 days ago Added label: requirement
11 days ago Modify description of issue
白昊8801
11 days ago Comments:
11 days ago Comments:
/assign


11 days ago Assign qq_41311449 as responsible person
9 days ago Close issue
9 days ago changed custom state from 待办的 to 已完成
9 days ago Added label: resolved
Backgroud(背景信息)
AddRowRanges 长尾性能问题:原实现用单个 SIMT VF 按输出元素 grid-stride 遍历,每个线程对
[start,end) 逐行跨列串行累加。小 M 大 K 的用例只有 1~2 个核在工作,且按列跨步读 src 完全
非合并。
Origin(信息来源)
海思