llvm dev - Apr 2019 - Question regarding X86::getCondFromBranch()

Hi,

I notice that the following recent addition

X86::CondCode X86::getCondFromBranch(const MachineInstr &MI) {
  switch (MI.getOpcode()) {
  default: return X86::COND_INVALID;
  case X86::JCC_1:
    return static_cast<X86::CondCode>(
        MI.getOperand(MI.getDesc().getNumOperands() - 1).getImm());
  }
}

returns an invalid condition for JCC_2 and JCC_4 conditional opcodes.

What is the suggested way to figure out the condition code for JCC_2 and JCC_4?

Should I just roll one up for myself to handle such opcodes based on
the above? Am I missing something conceptually?

Thanks,

Bharadwaj

Nothing in LLVM today creates a MachineInstr that has JCC_2 or JCC_4
opcode. Those only get created by assembler relaxation after everything has
been converted to MCInst. Do you have your own code that is creating
JCC_2/JCC_4?

~Craig


On Thu, Apr 11, 2019 at 9:47 AM S. Bharadwaj Yadavalli <bharadwajy at
gmail.com>
wrote:
> Hi,
>
> I notice that the following recent addition
>
> X86::CondCode X86::getCondFromBranch(const MachineInstr &MI) {
>   switch (MI.getOpcode()) {
>   default: return X86::COND_INVALID;
>   case X86::JCC_1:
>     return static_cast<X86::CondCode>(
>         MI.getOperand(MI.getDesc().getNumOperands() - 1).getImm());
>   }
> }
>
> returns an invalid condition for JCC_2 and JCC_4 conditional opcodes.
>
> What is the suggested way to figure out the condition code for JCC_2 and
> JCC_4?
>
> Should I just roll one up for myself to handle such opcodes based on
> the above? Am I missing something conceptually?
>
> Thanks,
>
> Bharadwaj
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20190411/f91945c8/attachment.html>

Thanks for your quick reply, Craig.

On Thu, Apr 11, 2019 at 1:18 PM Craig Topper <craig.topper at gmail.com>
wrote:
>
> Nothing in LLVM today creates a MachineInstr that has JCC_2 or JCC_4
opcode. Those only get created by assembler relaxation after everything has been
converted to MCInst.
OK. That is good information.
> Do you have your own code that is creating JCC_2/JCC_4?
I see a need for it in my LLVM-based tool that parses binaries. So
far, handling JCC_2 and JCC_4 opcodes as is done in the function,
seems to do the job for me. Hopefully there are no subtleties that I
am ignoring.

Bharadwaj