-frounding-math and llvm?

What exactly is the situation with support for -frounding-math in llvm-gcc/clang? I ask
because we are seeing regressions in the ppl 0.11.2 testsuite when compiling with either llvm-gcc
or clang from Xcode 4.1 as well as clang svn. The ppl developers seem to be seeing warnings to
the effect...

clang: warning: argument unused during compilation: '-frounding-math'

in their builds.

http://www.cs.unipr.it/pipermail/ppl-devel/2011-August/017559.html

Thanks in advance for any clarifications on the status of -frounding-math in the llvm
based compilers.
             Jack

Not supported, and I don't know of anyone who plans on working on it
(or FENV_ACCESS) in the near future.

-Eli

> What exactly is the situation with support for -frounding-math in llvm-gcc/clang? I ask
> because we are seeing regressions in the ppl 0.11.2 testsuite when compiling with either llvm-gcc
> or clang from Xcode 4.1 as well as clang svn. The ppl developers seem to be seeing warnings to
> the effect...
>
> clang: warning: argument unused during compilation: '-frounding-math'
>
> in their builds.
>
> http://www.cs.unipr.it/pipermail/ppl-devel/2011-August/017559.html
>
> Thanks in advance for any clarifications on the status of -frounding-math in the llvm
> based compilers.

Not supported, and I don't know of anyone who plans on working on it
(or FENV_ACCESS) in the near future.

Eli,
   Is this the same situation with llvm-gcc? I don't see the warnings from llvm-gcc
but we do see the same failures in ppl's testsuite. Is this a case of llvm-gcc silently not supporting
-frounding-math?
              Jack

Yep, llvm-gcc just accepts and ignores the flag.

-Chris

Yikes. That seems rather evil. Shouldn't llvm-gcc at least emit that same warning
as clang to give the user notice that -frounding-math is not really implemented?
         Jack

*shrug*, "fixed in clang" :slight_smile:

-Chris