In OptimizeExprGPU, move the ReplaceUnitVarToZero pass before the EliminateCommonFactorOfLocalIndex pass. Otherwise, we may fail to eliminate some unit variables (variables that are in range [0, 1)) in local buffer's indices when there is transpose in the kernel.
你的PR提交成功,感谢你对开源项目的贡献!
请关注后续CI自动化测试结果,详情请参考Paddle-CI手册。
Your PR has been submitted. Thanks for your contribution!
Please wait for the result of CI firstly. See Paddle CI Manual for details.
PR Category
CINN
PR Types
Bug fixes
Description
In
OptimizeExprGPU
, move theReplaceUnitVarToZero
pass before theEliminateCommonFactorOfLocalIndex
pass. Otherwise, we may fail to eliminate some unit variables (variables that are in range[0, 1)
) in local buffer's indices when there is transpose in the kernel.Pcard-85711