[Mlir-commits] [mlir] [mlir][spirv] Add OpExtension "SPV_INTEL_tensor_float32_conversion " (PR #151337)

[Ivan Butygin]

================
@@ -52,22 +53,29 @@ def SPIRV_INTELConvertFToBF16Op : SPIRV_IntelVendorOp<"ConvertFToBF16", []> {
   ];
 
   let arguments = (ins
-    SPIRV_ScalarOrVectorOf<SPIRV_Float32>:$operand
----------------
Hardcode84 wrote:

I think you can just update `SPIRV_ScalarOrVectorOf` directly to use fixed vector, I do not expect SPIR-V to support scalable ever.

https://github.com/llvm/llvm-project/pull/151337