Lunderberg commented on issue #12907:
URL: https://github.com/apache/tvm/issues/12907#issuecomment-1262311429
This looks like it might be a good case for `T.assume`. If I make an init
block that assumes the `C` has already been zeroed out, then the
`meta_schedule.tune_tir` call passes successfully.
```python
@T.prim_func
def matmul(
A: T.Buffer[(512, 512), "float32"],
B: T.Buffer[(512, 512), "float32"],
C: T.Buffer[(512, 512), "float32"],
) -> None:
for i in range(512):
for j in range(512):
for k in range(512):
with T.block("update"):
with T.init():
T.assume(C[i, j] == 0.0)
C[i, j] = C[i, j] + A[i, k] * B[j, k]
```
--
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]
