[llvm] [AMDGPU][True16][MC] 16bit vsrc and vdst support in MC (PR #104510)
Brox Chen via llvm-commits
llvm-commits at lists.llvm.org
Wed Sep 11 07:12:27 PDT 2024
================
@@ -1242,12 +1244,19 @@ class FPInputMods <FPInputModsMatchClass matchClass> : InputMods <matchClass> {
}
def FP16InputMods : FPInputMods<FP16InputModsMatchClass>;
-def FPT16InputMods : FPInputMods<FPT16InputModsMatchClass>;
+class FPT16InputMods<bit IsFake16> : FPInputMods<FPT16InputModsMatchClass<IsFake16>> {
+ let EncoderMethod = "getMachineOpValueT16";
+}
def FP32InputMods : FPInputMods<FP32InputModsMatchClass>;
+def FP32T16DstInputMods : FPInputMods<FP32InputModsMatchClass> {
+ let EncoderMethod = "getMachineOpValueT16";
+}
def FP64InputMods : FPInputMods<FP64InputModsMatchClass>;
-class FP16VCSrcInputMods<bit IsFake16>
- : FPInputMods<FP16VCSrcInputModsMatchClass<IsFake16>>;
+class FPT16VCSrcInputMods<bit IsFake16 = 1>
----------------
broxigarchen wrote:
I think a bit and I think the benefits is that if someone is using these classes but not testing true16, then they will not notice that we removed the fake16 in the short future, and it does not impact anything in their side.
If someone does use fake16 then it means they are actively using true16. I guess then they will have to be impacted anyway. So I will like to keep it.
https://github.com/llvm/llvm-project/pull/104510
More information about the llvm-commits
mailing list