Hzfengsy opened a new pull request #8585:
URL: https://github.com/apache/tvm/pull/8585
`match_buffer` inside a block allow users to mark a subregion of the source
buffer into a new target buffer. In the following example it does the same job
as `A[i, j] = 1`
```Python
with block([128, 128]) as [i, j]:
A0 = tir.match_buffer(A[i, j], ())
A0[()] = 1
```
A further step, it is useful for opaque access (e.g. Tensor intrinsic). It
can narrow the access region, also provide some additional information, such as
strides and shape for buffer after the pass `CompactBuffer`. Please see details
at a Tensorcore demo with `match_buffer` in file
`tests/python/integration/test_lower.py`
Please take a look. thanks a lot. @vinx13 @junrushao1994 @tqchen
--
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: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
