Page MenuHomePhabricator

[mlir][tosa] Add conv2d lowering to linalg.conv2d operator for FP
ClosedPublic

Authored by rsuderman on Apr 7 2021, 12:20 PM.

Details

Summary

Handles lowering conv2d to linalg's convolution operator. This implementation
only supports floating point values but handles all strides, dilations, and
padding values.

Diff Detail

Event Timeline

rsuderman created this revision.Apr 7 2021, 12:20 PM
rsuderman requested review of this revision.Apr 7 2021, 12:20 PM
silvas accepted this revision.Mon, Apr 12, 2:52 PM
silvas added inline comments.
mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp
821

need to check for static shapes above for this to work.

This revision is now accepted and ready to land.Mon, Apr 12, 2:52 PM
hanchung added inline comments.
mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp
788

I think this can be args[0].

rsuderman updated this revision to Diff 337245.Tue, Apr 13, 1:17 PM

Updated for review comments.

rsuderman marked 2 inline comments as done.Tue, Apr 13, 1:18 PM
This revision was landed with ongoing or failed builds.Tue, Apr 13, 1:28 PM
This revision was automatically updated to reflect the committed changes.