[Mlir-commits] [mlir] [mlir][spirv] Re-enable bf16/fp8 for vector composite ops (PR #204848)

[Igor Wodiany]

================
@@ -4314,6 +4314,10 @@ def SPIRV_FloatVector : VectorOfRankAndLengthAndType<[1], [2, 3, 4, 8, 16],
 def SPIRV_ExtendedFloatVector : VectorOfRankAndLengthAndType<[1], [2, 3, 4, 8, 16],
                                        [SPIRV_BFloat16KHR, SPIRV_Float8E4M3EXT,
                                         SPIRV_Float8E5M2EXT]>;
+def SPIRV_ExtendedVector : VectorOfRankAndLengthAndType<[1], [2, 3, 4, 8, 16],
+                                       [SPIRV_Bool, SPIRV_Integer, SPIRV_Float,
+                                        SPIRV_BFloat16KHR, SPIRV_Float8E4M3EXT,
+                                        SPIRV_Float8E5M2EXT]>;
----------------
IgWod wrote:

I think `SPIRV_AnyFloatVector` does exactly this, unless I'm missing something. To be honest `SPIRV_ExtendedVector` may be a better name. Either way we should keep only one.

https://github.com/llvm/llvm-project/pull/204848