[AMDGPU] Use isMetaInstruction for instruction size

[sebastian-ne]

Summary

Meta instructions have a size of 0. Use isMetaInstruction instead of
listing them explicitely.

[foad]

Looks good.

Typo in summary "explicitly".

Could put the isMetaInstruction check inside the default case of the switch, to save a nanosecond in the common case.

llvm/lib/Target/AMDGPU/SIInstrInfo.cpp

if (isMIMG(MI)) {
     int VAddr0Idx = AMDGPU::getNamedOperandIdx(Opc, AMDGPU::OpName::vaddr0);
     if (VAddr0Idx < 0)
       return 8;
 
     int RSrcIdx = AMDGPU::getNamedOperandIdx(Opc, AMDGPU::OpName::srsrc);
     return 8 + 4 * ((RSrcIdx - VAddr0Idx + 2) / 4);
   }
 
-  switch (Opc) {
-  case TargetOpcode::IMPLICIT_DEF:
-  case TargetOpcode::KILL:
-  case TargetOpcode::DBG_VALUE:
-  case TargetOpcode::EH_LABEL:
+  if (MI.isMetaInstruction())
     return 0;
+
+  switch (Opc) {
   case TargetOpcode::BUNDLE:
     return getInstBundleSize(MI);
   case TargetOpcode::INLINEASM:
   case TargetOpcode::INLINEASM_BR: {
     const MachineFunction *MF = MI.getParent()->getParent();
     const char *AsmStr = MI.getOperand(0).getSymbolName();
     return getInlineAsmLength(AsmStr, *MF->getTarget().getMCAsmInfo(), &ST);
   }