[AMDGPU][True16] fix a bug in codeGen causing e64 with wrong vgpr type to shrink (#102942)

This bug is introduced in
#102198

The previous path change to use realTrue16 flag, however, we have some
t16 instructions that are implemented with fake16, and has Lo128
registers types. Thus we should still using hasTrue16Bit flag for
shrinking check

---------

Co-authored-by: guochen2 <[email protected]>

Author: broxigarchen

llvm/lib/Target/AMDGPU/SIShrinkInstructions.cpp

@@ -1048,7 +1048,7 @@ bool SIShrinkInstructions::runOnMachineFunction(MachineFunction &MF) {
               MachineFunctionProperties::Property::NoVRegs))
         continue;
 
-      if (ST->useRealTrue16Insts() && AMDGPU::isTrue16Inst(MI.getOpcode()) &&
+      if (ST->hasTrue16BitInsts() && AMDGPU::isTrue16Inst(MI.getOpcode()) &&
           !shouldShrinkTrue16(MI))
         continue;
 

llvm/test/CodeGen/AMDGPU/shrink-true16.mir

@@ -0,0 +1,28 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -run-pass=si-shrink-instructions -verify-machineinstrs -o - %s | FileCheck -check-prefix=GFX1100 %s
+
+---
+name: 16bit_lo128_shrink
+tracksRegLiveness: true
+body: |
+  bb.0:
+    liveins: $vgpr127
+    ; GFX1100-LABEL: name: 16bit_lo128_shrink
+    ; GFX1100: liveins: $vgpr127
+    ; GFX1100-NEXT: {{  $}}
+    ; GFX1100-NEXT: V_CMP_EQ_U16_t16_e32 0, $vgpr127, implicit-def $vcc, implicit $exec, implicit $exec
+    $vcc_lo = V_CMP_EQ_U16_t16_e64 0, $vgpr127, implicit-def $vcc, implicit $exec
+...
+
+---
+name: 16bit_lo128_no_shrink
+tracksRegLiveness: true
+body: |
+  bb.0:
+    liveins: $vgpr128
+    ; GFX1100-LABEL: name: 16bit_lo128_no_shrink
+    ; GFX1100: liveins: $vgpr128
+    ; GFX1100-NEXT: {{  $}}
+    ; GFX1100-NEXT: $vcc_lo = V_CMP_EQ_U16_t16_e64 0, $vgpr128, implicit-def $vcc_lo, implicit $exec
+    $vcc_lo = V_CMP_EQ_U16_t16_e64 0, $vgpr128, implicit-def $vcc, implicit $exec
+...