[Mlir-commits] [mlir] [MLIR][AMDGPU] Add OCP FP8 support for new hardware (PR #106160)

[Paul C Fuqua]

================
@@ -454,6 +454,20 @@ static void wmmaPushOutputOperand(ConversionPatternRewriter &rewriter,
   }
 }
 
+/// Return true if `type` is the E5M2 variant of an 8-bit float that is
+/// supported by the `_bf8` instructions on the given `chipset`.
+static bool isNativeBf8(Chipset chipset, Type type) {
----------------
pcf000 wrote:

Well, yes, but all chipsets have a native BF8 type -- FNUZ for the older ones, OCP for the newer -- so the answer to "has native BF8" is "yes".  Rather, we want to know "is this type the same as the chipset's native BF8 type?"  Also, in a context in which we're doing isa<T1,T2>() and isFloat(), "is" flows better.

https://github.com/llvm/llvm-project/pull/106160