[PATCH] D75722: [mlir][Linalg] Implement padding for linalg.conv and lowering to loops.

[Mahesh Ravishankar via Phabricator via llvm-commits]

mravishankar requested changes to this revision.
mravishankar added inline comments.
This revision now requires changes to proceed.


================
Comment at: mlir/lib/Dialect/Linalg/Transforms/LinalgToLoops.cpp:178
 public:
+  static ValueHandle getConvOpInput(ConvOp convOp, IndexedValueType I,
+                                    ArrayRef<ValueHandle> imIdx) {
----------------
Could you add some comments about whats happening here?


================
Comment at: mlir/lib/Dialect/Linalg/Transforms/LinalgToLoops.cpp:206
+
+      assert(shapes[idx] != -1 && "dynamic shape is not available");
+      ValueHandle shape = std_constant_index(shapes[idx]);
----------------
I am not sure we need to restrict this for the static shape case. You can always get the dimension dynamically and generate code for that. Just generate a dim operation?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D75722/new/

https://reviews.llvm.org/D75722