OSL ChatGPT Reviewer

NOTE: This is generated by an AI program, so some comments may not make sense.

src/irx/builders/llvmliteir.py

• (L.426) Correctness: Typo in API call. llvmlite IRBuilder uses set_fast_math_flags, not set_fastmath_flags. The current code will raise AttributeError.
  Suggested change:
  def set_fast_math(self, enable: bool) -> None:
  """Enable or disable fast math flags on the IR builder."""
  if enable:
  self._llvm.ir_builder.set_fast_math_flags(["fast"])
  else:
  self._llvm.ir_builder.clear_fast_math_flags()

• Behavioral risk: This unconditionally clobbers any existing fast-math configuration and then clears everything on disable, which can silently change semantics (NNAN/NINF/REASSOC, etc.). Consider preserving/restoring previous flags or offering a scoped/context-managed enable to avoid leaking state across codegen phases.

• Compatibility: Verify the llvmlite version’s expected type for flags. Some versions prefer a FastMathFlags object over a list[str]. Consider accepting a Sequence[str] or FastMathFlags to avoid future breakage.

@yuvimittal please review!

@yuvimittal & @xmnlab please review the pr!

@iihimanshuu , i dont understand in which file does we call self.set_fast_math() where it doesn't exist!?
Also we should add tests as well for the method, and can you push and empty commit!? i am not able re-run the jobs, as github disables re-running old workflow runs!

Thanks for looking at this. I opened this PR because set_fast_math() was being called but wasn’t defined at the time (Dec 15). I see you’ve since implemented a proper solution with tests on main, so my change is now redundant. I’ll close this PR—appreciate the quick fix!

@iihimanshuu , i dont understand in which file does we call self.set_fast_math() where it doesn't exist!? Also we should add tests as well for the method, and can you push and empty commit!? i am not able re-run the jobs, as github disables re-running old workflow runs!