Adding a new variant of DepthwiseConv2D

[gpetters94]

Summary

This is the same as the existing multiplier-1 variant of DepthwiseConv2D, but in PyTorch dim order.

[bondhugula]

Please add a test case as well for the new op.

[bondhugula]

In D128575#3611124, @bondhugula wrote:

Please add a test case as well for the new op.

I noticed the previous update to this yaml file also missed adding a test case: bc07634b5a762686b818932eb350b4fc84217e67

[gpetters94]

In D128575#3611126, @bondhugula wrote:

In D128575#3611124, @bondhugula wrote:

Please add a test case as well for the new op.

I noticed the previous update to this yaml file also missed adding a test case: bc07634b5a762686b818932eb350b4fc84217e67

Oh, right. Should I just add the test case for that op here, or should I make a separate patch for it?

[gpetters94]

Added tests for this op, as well as for my previous op (grouped conv) that has already landed.

[silvas]

@MaheshRavishankar -- is there a risk that this op will be optimized worse due to not being recognized as well as the regular depthwise?

[mravishankar]

In D128575#3620893, @silvas wrote:

@MaheshRavishankar -- is there a risk that this op will be optimized worse due to not being recognized as well as the regular depthwise?

Thats a separate concern (and if IIUC a concern more related to IREE and not really for MLIR). But in any case, IREE should be able to handle this variation by default, and even if it is doesnt should be easy to adapt IREE to account for this change.

[gpetters94]

@silvas @mravishankar Do you think this is good to go then?

mlir/include/mlir/Dialect/Linalg/IR/LinalgNamedStructuredOps.yaml

                 - !ScalarExpression
                   scalar_arg: K
 --- !LinalgOpConfig
+metadata: !LinalgOpMetadata
+  name: depthwise_conv_2d_nchw_chw
+  cpp_class_name: DepthwiseConv2DNchwChwOp
+  doc: |-
+    Performs depth-wise 2-D convolution.
+
+    Numeric casting is performed on the operands to the inner multiply, promoting
+    them to the same data type as the accumulator/output. Multiplier is set to 1
+    which is a special case for most depthwise convolutions.
+  implements:
+  - LinalgConvolutionOpInterface
+structured_op: !LinalgStructuredOpConfig
+  args:
+  - !LinalgOperandDefConfig
+    name: I
+    kind: input_tensor
+    type_var: T1
+    shape_map: affine_map<()[s0, s1, s2, s3, s4, s5, s6, s7, s8, s9] -> (s0, s9, s1 *
+      s2 + s3 * s4, s5 * s6 + s7 * s8)>
+  - !LinalgOperandDefConfig
+    name: K
+    kind: input_tensor
+    type_var: T2
+    shape_map: affine_map<()[s0, s1, s2, s3, s4, s5, s6, s7, s8, s9] -> (s9, s3, s7)>
+  - !LinalgOperandDefConfig
+    name: O
+    kind: output_tensor
+    type_var: U
+    shape_map: affine_map<()[s0, s1, s2, s3, s4, s5, s6, s7, s8, s9] -> (s0, s9, s1, s5)>
+  - !LinalgOperandDefConfig
+    name: strides
+    kind: index_attr
+    index_attr_map: affine_map<()[s0, s1, s2, s3, s4, s5, s6, s7, s8, s9] -> (s2,
+      s6)>
+    default_indices:
+    - 1
+    - 1
+  - !LinalgOperandDefConfig
+    name: dilations
+    kind: index_attr
+    index_attr_map: affine_map<()[s0, s1, s2, s3, s4, s5, s6, s7, s8, s9] -> (s4,
+      s8)>
+    default_indices:
+    - 1
+    - 1
+  indexing_maps: !LinalgIndexingMapsConfig
+    static_indexing_maps:
+    - affine_map<(d0, d1, d2, d3, d4, d5)[s0, s1, s2, s3, s4, s5, s6, s7, s8, s9]
+      -> (d0, d3, d1 * s2 + d4 * s4, d2 * s6 + d5 * s8)>
+    - affine_map<(d0, d1, d2, d3, d4, d5)[s0, s1, s2, s3, s4, s5, s6, s7, s8, s9]
+      -> (d3, d4, d5)>
+    - affine_map<(d0, d1, d2, d3, d4, d5)[s0, s1, s2, s3, s4, s5, s6, s7, s8, s9]
+      -> (d0, d3, d1, d2)>
+  iterator_types:
+  - parallel
+  - parallel
+  - parallel
+  - parallel
+  - reduction
+  - reduction
+  assignments:
+  - !ScalarAssign
+    arg: O
+    value: !ScalarExpression
+      scalar_fn:
+        kind: binary
+        fn_name: add
+        operands:
+        - !ScalarExpression
+          scalar_arg: O
+        - !ScalarExpression
+          scalar_fn:
+            kind: binary
+            fn_name: mul
+            operands:
+            - !ScalarExpression
+              scalar_fn:
+                kind: type
+                fn_name: cast_signed
+                type_var: U
+                operands:
+                - !ScalarExpression
+                  scalar_arg: I
+            - !ScalarExpression
+              scalar_fn:
+                kind: type
+                fn_name: cast_signed
+                type_var: U
+                operands:
+                - !ScalarExpression
+                  scalar_arg: K
+--- !LinalgOpConfig
 metadata: !LinalgOpMetadata
   name: depthwise_conv_2d_nhwc_hwc_q
   cpp_class_name: DepthwiseConv2DNhwcHwcQOp

mlir/python/mlir/dialects/linalg/opdsl/ops/core_named_ops.py

            D.ic]) * TypeFn.cast_signed(U, K[D.kh, D.kw, D.ic])
 
 
+@linalg_structured_op
+def depthwise_conv_2d_nchw_chw(I=TensorDef(T1, S.N, S.IC, S.OH * S.SH + S.KH * S.DH,
+                                           S.OW * S.SW + S.KW * S.DW),
+                               K=TensorDef(T2, S.IC, S.KH, S.KW),
+                               O=TensorDef(U,
+                                           S.N,
+                                           S.IC,
+                                           S.OH,
+                                           S.OW,
+                                           output=True),
+                               strides=IndexAttrDef(S.SH, S.SW, default=[1, 1]),
+                               dilations=IndexAttrDef(S.DH,
+                                                      S.DW,
+                                                      default=[1, 1])):
+  """Performs depth-wise 2-D convolution.
+
+  Numeric casting is performed on the operands to the inner multiply, promoting
+  them to the same data type as the accumulator/output. Multiplier is set to 1
+  which is a special case for most depthwise convolutions.
+  """
+  implements(ConvolutionOpInterface)
+  domain(D.n, D.oh, D.ow, D.ic, D.kh, D.kw)
+  O[D.n, D.ic, D.oh, D.ow] += TypeFn.cast_signed(
+      U, I[D.n, D.ic, D.oh * S.SH + D.kh * S.DH, D.ow * S.SW + D.kw * S.DW]) * TypeFn.cast_signed(U, K[D.ic, D.kh, D.kw])
+
+
 @linalg_structured_op
 def depthwise_conv_2d_nhwc_hwc_q(I=TensorDef(T1, S.N, S.OH * S.SH + S.KH * S.DH,
                                              S.OW * S.SW + S.KW * S.DW, S.IC),