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 2022/09/28 15:31:25 UTC

[GitHub] [tvm] Lunderberg opened a new pull request, #12927: [Arith][IndexMap] Correct MapShape result for small vectorized dims

Lunderberg opened a new pull request, #12927:
URL: https://github.com/apache/tvm/pull/12927

   Prior to this commit, `IndexMap::MapShape` could produce incorrect results when the split factor is greater than the size of the dimension being split.  For example, a buffer of shape `[N]` mapped transformed with `lambda i: [i//4, i%4]` should result in shape `[ceildiv(N,4), 4]`.  However, for `N<4`, the transformed shape was instead `[1, N%4]`.  This results in unexpected shapes when attempting to prepare a buffer for vectorized access.
   
   This commit preferentially uses the result of `arith::DetectIterMap` to determine the mapped buffer shape, similar to what is done when computing the inverse.  The old method of `MapShape`, which relied on `arith::EvalSet`, is maintained for transformations that aren't recognized by `arith::DetectIterMap`.


-- 
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.

To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org

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


[GitHub] [tvm] vinx13 merged pull request #12927: [Arith][IndexMap] Correct MapShape result for small vectorized dims

Posted by GitBox <gi...@apache.org>.
vinx13 merged PR #12927:
URL: https://github.com/apache/tvm/pull/12927


-- 
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.

To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org

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