[llvm] [AMDGPU][True16] Don't use the VGPR_LO/HI16 register classes. (PR #76440)

[Ivan Kosarev via llvm-commits]

================
@@ -2808,8 +2810,6 @@ getAlignedVectorSuperClassForBitWidth(unsigned BitWidth) {
 
 const TargetRegisterClass *
 SIRegisterInfo::getVectorSuperClassForBitWidth(unsigned BitWidth) const {
-  if (BitWidth == 16)
-    return &AMDGPU::VGPR_LO16RegClass;
----------------
kosarev wrote:

Looks like this `if` and the removed `case`s below were always dead code.

https://github.com/llvm/llvm-project/pull/76440