[Mlir-commits] [mlir] [MLIR][Vector] Add warp distribution for `vector.step` op (PR #155425)

[Adam Siemieniuk]

================
@@ -1824,3 +1824,20 @@ func.func @warp_propagate_duplicated_operands_in_yield(%laneid: index)  {
 // CHECK-PROP       :   }
 // CHECK-PROP       :   %[T1:.*] = math.exp %[[W]] : vector<1xf32>
 // CHECK-PROP       :   "some_use"(%[[T1]]) : (vector<1xf32>) -> ()
+
+// -----
+
+func.func @warp_step_distribute(%laneid: index, %buffer: memref<128xindex>)  {
----------------
adam-smnk wrote:

Maybe this:
```mlir
func.func @warp_step_distribute(%laneid: index, %buffer: memref<128xindex>)  {
  %r = gpu.warp_execute_on_lane_0(%laneid)[32] -> (vector<2xindex>) {
    %seq = vector.step : vector<64xindex>
    gpu.yield %seq : vector<64xindex>
  }
  vector.transfer_write %r, %buffer[%laneid] : vector<2xindex>, memref<128xindex>
  return
}
```

> I do not see any failure tests for any other patterns. Did I miss them?

Maybe they don't use the new `@negative_...` naming scheme. Maybe there are none which is not ideal.
Either way, doesn't hurt to have extra coverage.

https://github.com/llvm/llvm-project/pull/155425