[llvm] [AMDGPU] Expose RTZ version of f16 interpolation (PR #86614)

[Matt Arsenault via llvm-commits]

================
@@ -2066,6 +2066,24 @@ def int_amdgcn_interp_inreg_p2_f16 :
             [IntrNoMem, IntrSpeculatable,
              ImmArg<ArgIndex<3>>]>;
 
+// llvm.amdgcn.interp.inreg.p10.rtz.f16 <p>, <i>, <p0>, <high>
+// fp16 interpolation intrinsic, but round-toward-zero.
+// high selects whether high or low 16-bits are used for p and p0 operands
+def int_amdgcn_interp_inreg_p10_rtz_f16:
----------------
arsenm wrote:

inreg should go to the end, or gain a gfx11 suffix or something. Shouldn't split the real opcode name up 

https://github.com/llvm/llvm-project/pull/86614