You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tvm.apache.org by GitBox <gi...@apache.org> on 2020/12/23 15:45:41 UTC

[GitHub] [tvm] ANSHUMAN87 opened a new pull request #7158: [Cuda] Slight optimize the default injective schedule

ANSHUMAN87 opened a new pull request #7158:
URL: https://github.com/apache/tvm/pull/7158


   We do not need additional thread extent when the tensor size is known.
   Below output shows the impact of the change, taken Transpose Op for illustration.
   
   Before:
   ```
   producer_realize T_transpose([0, 3], [0, 4]) {
     // attr [iter_var(blockIdx.x, , blockIdx.x)] thread_extent = 1
     // attr [iter_var(threadIdx.x, , threadIdx.x)] thread_extent = 1024
     if (tir.likely((floordiv(threadIdx.x, 4) < 3))) {
       if (tir.likely((threadIdx.x < 12))) {
         T_transpose[floordiv(threadIdx.x, 4), floormod(threadIdx.x, 4)] =placeholder[floormod(threadIdx.x, 4), floordiv(threadIdx.x, 4)]
       }
     }
   }
   ```
   
   
   After:
   ```
   producer_realize T_transpose([0, 3], [0, 4]) {
     // attr [iter_var(blockIdx.x, , blockIdx.x)] thread_extent = 1
     // attr [iter_var(threadIdx.x, , threadIdx.x)] thread_extent = 12
     T_transpose[floordiv(threadIdx.x, 4), floormod(threadIdx.x, 4)] =placeholder[floormod(threadIdx.x, 4), floordiv(threadIdx.x, 4)]
   }
   ```
   


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org

[GitHub] [tvm] ANSHUMAN87 commented on pull request #7158: [Cuda] Slight optimize the default injective schedule

Posted by GitBox <gi...@apache.org>.
ANSHUMAN87 commented on pull request #7158:
URL: https://github.com/apache/tvm/pull/7158#issuecomment-750869686


   Thanks @junrushao1994 ! Looks like the test case is flaky. Does not depend on this PR change. Will re-trigger CI.


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org

[GitHub] [tvm] comaniac commented on pull request #7158: [Cuda] Slight optimize the default injective schedule

Posted by GitBox <gi...@apache.org>.
comaniac commented on pull request #7158:
URL: https://github.com/apache/tvm/pull/7158#issuecomment-751037640


   Thanks @ANSHUMAN87 @junrushao1994 


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org

[GitHub] [tvm] comaniac merged pull request #7158: [Cuda] Slight optimize the default injective schedule

Posted by GitBox <gi...@apache.org>.
comaniac merged pull request #7158:
URL: https://github.com/apache/tvm/pull/7158


   


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org

[GitHub] [tvm] junrushao1994 commented on pull request #7158: [Cuda] Slight optimize the default injective schedule

Posted by GitBox <gi...@apache.org>.
junrushao1994 commented on pull request #7158:
URL: https://github.com/apache/tvm/pull/7158#issuecomment-750952358


   Thanks @ANSHUMAN87! I don’t have other comments. @comaniac would you like to take a look?


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org