[PATCH 4.5 0/N]: Integrate GCC with the complex math library MPC

[Kaveh Ghazi]

From: "Joseph S. Myers" <joseph@codesourcery.com>

> Now, I don't really think -fmpc-optimize is a particularly desirable
> option to release with; I'd prefer simply making MPC a required library.
> But it may be useful as an interim step to develop the optimizations
> before MPC is more widely distributed and there is approval to make it a
> build requirement.

I understand your's and Richard points about debugging, etc.  However I also
agree with you that -fmpc-foo is not desirable.  In the long run my hope is
to make this library mandatory like MPFR which would avoid the need for
these flags and I wouldn't want to have them in forever as placebos ever
more simply because of this interim period where MPC was optional.  Note the
output of gcc -v will show whether MPC is installed and what version the
user has (just like it shows GMP/MPFR), so at least this info can be used to
replicate their bug conditions if relevant.

Also, MPC can be used for fixing bugs like the complex folding one I
mentioned.  In that case, it's not something you would want activated only
sometimes like an optimization.  Perhaps this use of MPC should wait until
we make it mandatory.  But it would be wierd to control this with a -f flag.

Lastly, I seem to recall that folding of fortran intrinsics is not optional.
I think the standard requires it, but I could be mistaken.  Long-term I
don't think we want to have a switch to turn this off either.

Would it be okay to put in the code now without any flags, and revisit the
necessity of them at the end of stage2?  I will commit to creating patch(es)
for these flags at that point if MPC is not ready for prime-time yet.  In
this case, "ready for prime-time" would mean ready to be mandatory for
building GCC.

        Thanks,
        --Kaveh