efriedma added a comment.
> setRoundingMode definitely should not call getAllowFEnvAccess() and it does
> not
Yes, it does?
// C2x: 7.6.2p3 If the FE_DYNAMIC mode is specified and FENV_ACCESS is "off",
// the translator may assume that the default rounding mode is in effect.
if (FPR == llvm::RoundingMode::Dynamic &&
!CurFPFeatures.getAllowFEnvAccess() &&
CurFPFeatures.getFPExceptionMode() == LangOptions::FPE_Ignore)
FPR = llvm::RoundingMode::NearestTiesToEven;
> As for ActOnPragmaFEnvAccess, it make sense to set the most general mode.
Shouldn't the default mode just be "dynamic"? The fact that we emit non-strict
floating-point ops if FENV_ACCESS is disabled is an implementation detail. I
don't see the point of switching the rounding mode between "Dynamic" and
"NearestTiesToEven" depending on whether FENV_ACCESS is enabled. Making the
required state transitions complicated like this just makes the code harder to
understand.
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D126364/new/
https://reviews.llvm.org/D126364
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
