[Mlir-commits] [mlir] [mlir][vector] add consistent stride verification to `masked load/store` and `gather/scatter` ops (PR #204842)

[Federico Bruzzone]

================
@@ -1413,6 +1413,15 @@ func.func @maskedload_memref_mismatch(%base: memref<?xf32>, %mask: vector<16xi1>
 
 // -----
 
+func.func @maskedload_negative_stride(%flip: memref<100x100xf32, strided<[-100, 1]>>, %mask: vector<8xi1>, %pass: vector<8xf32>) -> vector<8xf32> {
+  // expected-error @+2 {{'vector.maskedload' op memref strides must be non-negative}}
+  %c0 = arith.constant 0 : index
+  %0 = vector.maskedload %flip[%c0, %c0], %mask, %pass : memref<100x100xf32, strided<[-100, 1]>>, vector<8xi1>, vector<8xf32> into vector<8xf32>
+  return %0 : vector<8xf32>
+}
----------------
FedericoBruzzone wrote:

YOU'RE RIGHT! :(


https://github.com/llvm/llvm-project/pull/204842