[Mlir-commits] [mlir] [mlir][vector] Add `vector.from_elements` op (PR #95938)

[Diego Caballero]

================
@@ -2642,3 +2642,72 @@ func.func @extract_from_0d_splat_broadcast_regression(%a: f32, %b: vector<f32>,
   // CHECK:   return %[[a]], %[[a]], %[[extract1]], %[[a]], %[[a]], %[[extract2]], %[[extract3]]
   return %1, %3, %5, %7, %9, %10, %11 : f32, f32, f32, f32, f32, vector<6x7xf32>, vector<3xf32>
 }
+
+// -----
+
+// CHECK-LABEL: func @extract_scalar_from_from_elements(
+//  CHECK-SAME:     %[[a:.*]]: f32, %[[b:.*]]: f32)
+func.func @extract_scalar_from_from_elements(%a: f32, %b: f32) -> (f32, f32, f32, f32, f32, f32, f32) {
+  // Extract from 0D.
+  %0 = vector.from_elements %a : vector<f32>
+  %1 = vector.extract %0[] : f32 from vector<f32>
+
+  // Extract from 1D.
+  %2 = vector.from_elements %a : vector<1xf32>
+  %3 = vector.extract %2[0] : f32 from vector<1xf32>
+  %4 = vector.from_elements %a, %b, %a, %a, %b : vector<5xf32>
+  %5 = vector.extract %4[4] : f32 from vector<5xf32>
+
+  // Extract from 2D.
----------------
dcaballe wrote:

Is it worth adding some CHECK-NOTs?

https://github.com/llvm/llvm-project/pull/95938