[Mlir-commits] [mlir] [mlir][gpu] Add extra value types for gpu::ShuffleOp (PR #104605)

llvmlistbot at llvm.org llvmlistbot at llvm.org
Tue Aug 20 08:33:19 PDT 2024


================
@@ -1250,17 +1250,11 @@ def GPU_ShuffleMode : I32EnumAttr<"ShuffleMode",
 def GPU_ShuffleModeAttr : EnumAttr<GPU_Dialect, GPU_ShuffleMode,
                                    "shuffle_mode">;
 
-def GPUShuffleTypes : TypeConstraint<Or<[
-  SignlessIntOfWidths<[8, 16, 32, 64]>.predicate,
-  FloatOfWidths<[16, 32, 64]>.predicate
-  ]>,
-  "i8, i16, i32, i64, f16, f32 or f64">;
-
 def GPU_ShuffleOp : GPU_Op<
     "shuffle", [Pure, AllTypesMatch<["value", "shuffleResult"]>]>,
-    Arguments<(ins GPUShuffleTypes:$value, I32:$offset, I32:$width,
+    Arguments<(ins AnyIntegerOrFloatOr1DVector:$value, I32:$offset, I32:$width,
                GPU_ShuffleModeAttr:$mode)>,
-    Results<(outs GPUShuffleTypes:$shuffleResult, I1:$valid)> {
+    Results<(outs AnyIntegerOrFloatOr1DVector:$shuffleResult, I1:$valid)> {
----------------
FMarno wrote:

It matches the type constraints of `$value` of the `subgroup_reduce` op

https://github.com/llvm/llvm-project/pull/104605


More information about the Mlir-commits mailing list