On Wed, 23 May 2001, Anders Magnusson wrote:

> > On Wed, 23 May 2001, Anders Magnusson wrote:
> > >
> > > libc only uses a few of them because there are VAXen that do not have
> > > them in hardware that gets extremely slow if those isns's are used...
> >
> > 	Are there any compile time options to control the use of them?
> > 	I can't see anything in the gcc info page. Would it make sense
> > 	to add any so those with more 'instruction rich' CPUs could
> > 	elect to recompile to gain more speed, something akin to -mcpu=v8
> > 	on a sparc box. Hmmm, presumably that could also be useful if
> > 	the cost of scheduling instructions varies between different
> > 	vaxes. At the evry last it could be set automatically in kernel
> > 	compiles when targeting specific vax models.
> >
> gcc don't use them, and I don't know if there would be of any use to
> learn gcc about them.

	Depending on how much space they save it might be worth it
	under -Os, or in kernels?

>                      If something should be done then I think that
> it would be two different libc's and do something like the x86 libm
> changing.

	That would be assembler stubs for appropriate routines?

-- 
		David/absolute		-- www.netbsd.org: No hype required --