[Mlir-commits] [mlir] [mlir][linalg][elementwise] Fold broadcast into new elementwise (PR #167626)

[Javed Absar]

================
@@ -28,16 +28,212 @@ func.func @unary_transpose(%A : tensor<16x8x32xf32>, %B: tensor<8x16x32xf32>) ->
 // CHECK-SAME:              ins(%[[A]], %[[B]] : tensor<?x?xf32>, tensor<?x?xf32>) outs(%[[C]] : tensor<?x?xf32>) -> tensor<?x?xf32>
 // CHECK-NEXT:  return %[[RES]] : tensor<?x?xf32>
 //
-func.func @binary_transposed(%A : tensor<?x?xf32>, %B: tensor<?x?xf32>, %C: tensor<?x?xf32>) ->  tensor<?x?xf32> {
+func.func @binary_transposed(%A: tensor<?x?xf32>, %B: tensor<?x?xf32>, %C: tensor<?x?xf32>) -> tensor<?x?xf32> {
   %c0 = arith.constant 0 : index
   %c1 = arith.constant 1 : index
   %dim0 = tensor.dim %A, %c0 : tensor<?x?xf32>
   %dim1 = tensor.dim %A, %c1 : tensor<?x?xf32>
 
   %empty = tensor.empty(%dim1, %dim0) : tensor<?x?xf32>
-  %transposed_B = linalg.transpose ins(%B : tensor<?x?xf32>) outs(%empty :  tensor<?x?xf32>) permutation = [1, 0]
+  %transposed_B = linalg.transpose ins(%B : tensor<?x?xf32>) outs(%empty : tensor<?x?xf32>) permutation = [1, 0]
   %result = linalg.elementwise kind=#linalg.elementwise_kind<add>
-                          ins(%A, %transposed_B : tensor<?x?xf32>,  tensor<?x?xf32>)
-                          outs(%C: tensor<?x?xf32>) -> tensor<?x?xf32>
+                          ins(%A, %transposed_B : tensor<?x?xf32>, tensor<?x?xf32>)
+                          outs(%C : tensor<?x?xf32>) -> tensor<?x?xf32>
   return %result : tensor<?x?xf32>
 }
+
+// -----
----------------
javedabsar1 wrote:

Nice examples.

https://github.com/llvm/llvm-project/pull/167626