[LowerToAIE] Handle linalg ops like other ops in switch statement #693
Conversation
newling
requested review from
MaheshRavishankar,
nirvedhmeshram,
yzhang93,
Abhishek-Varma and
jtuyls
as code owners
| @@ -396,9 +384,6 @@ LogicalResult coreToAIE(IRRewriter &rewriter, AMDAIE::CoreOp coreOp, | |||
| return coreReleaseOpToAIE(rewriter, releaseOp, mapper, | |||
| toBeErased); | |||
| }) | |||
| .Case<linalg::LinalgOp>([&](auto linalgOp) { | |||
There was a problem hiding this comment.
Just reading the description and PR, this seems like it is dropping general handling of LinalgOp. Is that to mean that we dont expect LinalgOps to survive till here?
There was a problem hiding this comment.
Instead of handling them as a special case:
They are now handled like other not-special ops:
We (I) do expect LinalgOps to survive til here.
|
I thought this was a fix for the non-deterministic crash I'm seeing in this pass with convolution, but it's not. I think I've found the actual problem (IRMapping not being updated when operations are erased) I'll post a fix for this tomorrow |
really? where? |
|
newling
changed the title
newling
changed the title
newling
force-pushed
the
linalg_demining
branch
from
4fbbf74 to
c4f4afe
Compare
|
Bundled with #705 |
This is (now) just a code simplification. Linalg Ops don't need special handling.