[llvm] [BPF] Remove 'may_goto 0' instructions (PR #123482)

[via llvm-commits]

yonghong-song wrote:

I checked other targets (X86, Arm) etc. They also parse and processes asm strings. For example, in ARM/ARMISelLowering.cpp,

```
ARM/ARMISelLowering.cpp:

bool ARMTargetLowering::ExpandInlineAsm(CallInst *CI) const {
  // Looking for "rev" which is V6+.
  if (!Subtarget->hasV6Ops())
    return false;

  InlineAsm *IA = cast<InlineAsm>(CI->getCalledOperand());
  StringRef AsmStr = IA->getAsmString();
  SmallVector<StringRef, 4> AsmPieces;
  SplitString(AsmStr, AsmPieces, ";\n");

  switch (AsmPieces.size()) {
  default: return false;
  case 1:
    AsmStr = AsmPieces[0];
    AsmPieces.clear();
    SplitString(AsmStr, AsmPieces, " \t,");

    // rev $0, $1
    if (AsmPieces.size() == 3 &&
        AsmPieces[0] == "rev" && AsmPieces[1] == "$0" && AsmPieces[2] == "$1" &&
        IA->getConstraintString().compare(0, 4, "=l,l") == 0) {
      IntegerType *Ty = dyn_cast<IntegerType>(CI->getType());
      if (Ty && Ty->getBitWidth() == 32)
        return IntrinsicLowering::LowerToByteSwap(CI);
    }
    break;
  }

  return false;
}
```

So use one of backend passes might be the better choice.

https://github.com/llvm/llvm-project/pull/123482