[MLIR] Linalg introduce BRGEMM

Author: chelini

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

@@ -3791,3 +3791,71 @@ structured_op: !LinalgStructuredOpConfig
                           scalar_const: '2.3283063999999999E-10 : f64'
             - !ScalarExpression
               scalar_arg: min
+--- !LinalgOpConfig
+metadata: !LinalgOpMetadata
+  name: reduce_batch_matmul
+  cpp_class_name: ReduceBatchMatmulOp
+  doc: |-
+    Performs a batched matrix multiplication of two 3D inputs.
+    Numeric casting is performed on the operands to the inner multiply, promoting
+    them to the same data type as the accumulator/output.
+  implements:
+  - LinalgContractionOpInterface
+structured_op: !LinalgStructuredOpConfig
+  args:
+  - !LinalgOperandDefConfig
+    name: A
+    kind: input_tensor
+    type_var: T1
+    shape_map: affine_map<()[s0, s1, s2, s3] -> (s0, s1, s3)>
+  - !LinalgOperandDefConfig
+    name: B
+    kind: input_tensor
+    type_var: T2
+    shape_map: affine_map<()[s0, s1, s2, s3] -> (s0, s3, s2)>
+  - !LinalgOperandDefConfig
+    name: C
+    kind: output_tensor
+    type_var: U
+    shape_map: affine_map<()[s0, s1, s2, s3] -> (s0, s1, s2)>
+  indexing_maps: !LinalgIndexingMapsConfig
+    static_indexing_maps:
+    - affine_map<(d0, d1, d2, d3)[s0, s1, s2, s3] -> (d0, d1, d3)>
+    - affine_map<(d0, d1, d2, d3)[s0, s1, s2, s3] -> (d0, d3, d2)>
+    - affine_map<(d0, d1, d2, d3)[s0, s1, s2, s3] -> (d1, d2)>
+  iterator_types:
+  - reduction
+  - parallel
+  - parallel
+  - reduction
+  assignments:
+  - !ScalarAssign
+    arg: C
+    value: !ScalarExpression
+      scalar_fn:
+        kind: binary
+        fn_name: add
+        operands:
+        - !ScalarExpression
+          scalar_arg: C
+        - !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: A
+            - !ScalarExpression
+              scalar_fn:
+                kind: type
+                fn_name: cast_signed
+                type_var: U
+                operands:
+                - !ScalarExpression
+                  scalar_arg: B

mlir/test/Dialect/Linalg/named-ops.mlir

@@ -762,3 +762,11 @@ func.func @conv_interface_wrong_num_operands(
     }) {dilations = dense<1> : tensor<2xi64>, linalg.memoized_indexing_maps = [#map0, #map1, #map2], operand_segment_sizes = array<i32: 2, 1>, strides = dense<1> : tensor<2xi64>} : (tensor<?x?x?x?xf32>, tensor<?x?x?x?x?xf32>, tensor<?x?x?x?xf32>) -> tensor<?x?x?x?xf32>
   return %0 : tensor<?x?x?x?xf32>
 }
+
+// -----
+
+func.func @brgemm_test(%arg0: tensor<8x128x256xf32>, %arg1: tensor<8x256x512xf32>, %arg2: tensor<128x512xf32>) -> tensor<128x512xf32> {
+  // CHECK: linalg.reduce_batch_matmul
+  %0 = linalg.reduce_batch_matmul ins(%arg0, %arg1 : tensor<8x128x256xf32>, tensor<8x256x512xf32>) outs(%arg2: tensor<128x512xf32>) -> tensor<128x512xf32>
+  return %0: tensor<128x512xf32>
+}