[mlir] [MLIR][Linalg] Introduce transpose/broadcast semantic to linalg.batch… (PR #130944)

[Andrzej Warzyński]

================
@@ -1637,6 +1637,171 @@ func.func @batch_matmul_bcast_A_transpose_B(%arg0: memref<3x5xf32>, %arg1: memre
 
 // -----
 
+// CHECK: #[[$ATTR_0:.+]] = affine_map<(d0, d1, d2, d3) -> (d3)>
+// CHECK: #[[$ATTR_1:.+]] = affine_map<(d0, d1, d2, d3) -> (d0, d3, d2)>
+// CHECK: #[[$ATTR_2:.+]] = affine_map<(d0, d1, d2, d3) -> (d1, d2)>
+
+// CHECK-LABEL:   func.func @batch_reduce_matmul_bcast_k_to_fill_missing_dims_A(
+// CHECK-SAME:                                                                  %[[VAL_0:[0-9]+|[a-zA-Z$._-][a-zA-Z0-9$._-]*]]: memref<5xf32>,
+// CHECK-SAME:                                                                  %[[VAL_1:[0-9]+|[a-zA-Z$._-][a-zA-Z0-9$._-]*]]: memref<2x5x7xf32>,
+// CHECK-SAME:                                                                  %[[VAL_2:[0-9]+|[a-zA-Z$._-][a-zA-Z0-9$._-]*]]: memref<3x7xf32>) {
----------------
banach-space wrote:

These strict regex matchers `([0-9]+|[a-zA-Z$._-][a-zA-Z0-9$._-]*])` aren’t really required — they were only introduced relatively recently:

* https://github.com/llvm/llvm-project/pull/128083

I’ve proposed restoring the original default here:

* https://github.com/llvm/llvm-project/pull/136721

Also, could you reduce the indentation here and in the following tests? Some of these don’t fit on my screen when reviewing in “split” mode. I've just updated "generate-test-checks.py" to automate this:
* https://github.com/llvm/llvm-project/pull/136757

One other note: your LIT variables don’t match the MLIR variable names. You can control this via:
```bash
generate-test-checks.py --variable_names=ARG_0,ARG_1,ARG_2
```

Or even better, if you prefer cleaner names like A, B, C, you can run:

```bash
generate-test-checks.py --variable_names=A,B,C
```

:) 

https://github.com/llvm/llvm-project/pull/130944