Thanks for sending an requirement! Please fill in the following template to help quickly solve your problem.
bucketizev2算子b32/b64在boundaries较大场景下性能低于H800
算子开发团队
boundaries较大场景使用simt vf可进行提前跳出,减少无效查找
/assign
Thanks for sending an requirement! Please fill in the following template to help quickly solve your problem.
Backgroud(背景信息)
bucketizev2算子b32/b64在boundaries较大场景下性能低于H800
Origin(信息来源)
算子开发团队
Benefit / Necessity (价值/作用)
Design(设计方案)
boundaries较大场景使用simt vf可进行提前跳出,减少无效查找