Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[Op] Semi-affine modulo operation is not suppoted in MLIR #127

Open
chhzh123 opened this issue Aug 8, 2022 · 2 comments
Open

[Op] Semi-affine modulo operation is not suppoted in MLIR #127

chhzh123 opened this issue Aug 8, 2022 · 2 comments
Assignees
@chhzh123
Labels
MLIR Limitation MLIR limitations

Comments

@chhzh123
Copy link
Member

chhzh123 commented Aug 8, 2022

As the title says, sometimes we may get the following indices after we outline and parameterize the operation, but it is not yet supported by MLIR.

%0 = affine.load %arg0[%arg6, %arg7 mod symbol(%arg5), %arg8, %arg9] {from = "layer2_0_avgpool"} : memref<1x32x16x16xf32, #map0>

It throws the error.

error: semi-affine expressions (modulo by non-const) are not supported
@chhzh123 chhzh123 added the MLIR Limitation MLIR limitations label Aug 8, 2022
@chhzh123 chhzh123 self-assigned this Aug 8, 2022
@chhzh123
Copy link
Member Author

chhzh123 commented Aug 8, 2022

Similar issue in IREE project:
iree-org/iree#9279

@chhzh123
Copy link
Member Author

chhzh123 commented Aug 9, 2022

Currently we can only rewrite this kind of affine.load to memref.load

chhzh123 added a commit that referenced this issue Aug 10, 2022
@chhzh123
[Pass] Support ModOp in .outline() (#127)
20964ca
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@chhzh123 chhzh123

Labels
MLIR Limitation MLIR limitations
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
1 participant
@chhzh123