[Mlir-commits] [mlir] [mlir][Vector] Remove uses of vector.extractelement/vector.insertelement (PR #113827)
Andrzej WarzyĆski
llvmlistbot at llvm.org
Mon Oct 28 10:57:56 PDT 2024
================
@@ -233,10 +233,9 @@ func.func @broadcast_vec2d_from_vec0d(%arg0: vector<f32>) -> vector<3x2xf32> {
// CHECK-LABEL: @broadcast_vec2d_from_vec0d(
// CHECK-SAME: %[[A:.*]]: vector<f32>)
// CHECK: %[[T0:.*]] = builtin.unrealized_conversion_cast %[[A]] : vector<f32> to vector<1xf32>
+// CHECK: %[[T5:.*]] = builtin.unrealized_conversion_cast %[[T0]] : vector<1xf32> to f32
----------------
banach-space wrote:
Hm, why wouldn't this be lowered to `llvm.extractelement`? And is `T5` used at all? In fact, would you mind sharing the output?
https://github.com/llvm/llvm-project/pull/113827
More information about the Mlir-commits
mailing list