Re: [Gcl-devel] RE: GCL bugs (decode-float, rationalize, expt)

[Camm Maguire]

Greetings!

"Mike Thomas" <address@hidden> writes:

> Hi Camm.
> 
> | Hi Mike!  Thanks for this!  Just to comment that at present, the
> | '-mcpu=i486 -march=i486' is not preventing any illegal ISA extensions
> | from being compiled in, as gcc doesn't output them in any case.
> 
> Nothing like a little pointless endeavour.   I thought afterward that it
> should really be a generalised name such as '--enable-common-binary' so that
> other architectures can take advantage of it for their own particular needs.
> I'll probably make it that way unless you object.
> 

This is a very good idea.  I would hope that there would be a way to
get such a lowest-common-denominator canonical hostname out of the
autoconf macros, but don't (yet) know enough about it.

> |  I
> | don't know if this will slow down the code on a >=586 or not.
> 
> It will I think, but probably not in such a way that anyone will care.  If
> the worst comes to the worst we can provide binaries built for other CPU's
> at release time.
> 

No big deal, I agree.  But in general, I'd suggest optimizing
performance for the latest commonly available hardware, except where
incompatibilities (i.e. illegal instructions) are introduced vis a vis
older hardware.  I.e. run on all boxes, run most optimally on latest
commonly available hardware.

> |  The
> | gmp3 configure step does the heavy lifting, as gmp3 has hand coded
> | *assember* routines which can use these instructions.
> 
> Yes, the 486 switch is passed to the GMP3 configure.  It saves having to do
> a pre-configure as you were doing.
> 

A much better solution, I agree.  Thanks! 


> Cheers
> 
> Mike Thomas.
> 
> 
> 
> 

-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah