[Mlir-commits] [mlir] [mlir][vector] Decouple unrolling gather and gather to llvm lowering (PR #132206)

[Kunwar Grover]

================
@@ -1663,7 +1663,7 @@ func.func @flat_transpose(%arg0: vector<16xf32>) -> vector<16xf32> {
 
 func.func @gather_with_mask(%arg0: memref<?xf32>, %arg1: vector<2x3xi32>, %arg2: vector<2x3xf32>) -> vector<2x3xf32> {
   %0 = arith.constant 0: index
-  %1 = vector.constant_mask [1, 2] : vector<2x3xi1>
+  %1 = vector.constant_mask [2, 2] : vector<2x3xi1>
----------------
Groverkss wrote:

Hi, this is mentioned in the PR summary:

There are still tests for 2D vector.gather, but the constant mask for these test is modified. This is because with the updated lowering, one of the unrolled vector.gather disappears because it is masked off (also demonstrating why this is a better lowering path)

https://github.com/llvm/llvm-project/pull/132206