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 ABIlibm calls on x86)



On Fri, 8 Jul 2005, Ian Lance Taylor wrote:

> Richard Guenther <rguenther@suse.de> writes:
> 
> > For decision-making on wether you like math routines creeping in
> > into libgcc, the attached patch is a variant of the -msselibm patch
> > with all (but pow) implementations in libgcc.
> > 
> > Q: Do we want these in libgcc?
> > Q: If not, where else?  (glibc was ruled out already)
> 
> I'm sure I'm missing some context here, but for these simple functions
> why do we want to call an external function at all?  If we care about
> getting optimized code, surely it would be better to inline them,
> since they are all so small.  If we don't care about optimized code,
> then why not just call the standard function in the standard way?

The missed context is that fast implementations will be a lot more
complicated and using SSE instructions.  They are supposed to called
for non-optimized code, too (i.e. w/o -ffast-math).  Just at the
moment, only Intel has such library.

Of course it remains to be investigated whether inlining a simplified
(for -ffast-math) implementation is beneficial, too.

Richard.


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