[llvm] [AMDGPU][True16][CodeGen] update VGPRimm for t16 (PR #131021)

[Matt Arsenault via llvm-commits]

================
@@ -2283,7 +2283,7 @@ let True16Predicate = UseRealTrue16Insts in {
   foreach vt = [f16, bf16] in {
     def : GCNPat <
       (VGPRImm<(vt fpimm)>:$imm),
-      (V_MOV_B16_t16_e64 0, $imm, 0)
+      (V_MOV_B16_t16_e64 0, (vt (bitcast_fpimm_to_i16 $imm)), 0)
----------------
arsenm wrote:

The type of immediate nodes in the selected dag doesn't really matter, they're all turning into sign extended int64_t in the end anyway. We could change all of these to just use i64 in the first place 

https://github.com/llvm/llvm-project/pull/131021