This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] libgcc variant (was: [PATCH] Optionally emit SSE ABI libm calls on x86)


Richard Henderson <rth@redhat.com> writes:

> The compelling reason to distribute the new library with gcc is that
> we no longer have to do any configure checks.  We can treat the whole
> library as a set of compiler optimizations for sse code.

Yes.

This is in general an argument for moving all the simpler libm
functions from glibc into gcc, since they are the standard library
functions most susceptible to compiler optimization.  We've already
seen that discrepancies between glibc and gcc, with regard to
__NO_MATH_INLINES, etc., can give suboptimal results.

I think that if we do put libmsse in gcc, we should put it in a
framework which makes it straightforward to extend to other processors
and processor variants.

Ian


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]