gmp-devel Digest, Vol 4, Issue 6

Raffi Chaglassian raffi at siliconsoul.com
Fri Dec 19 23:56:15 CET 2003


Date: Tue, 16 Dec 2003 22:24:18 +0000
From: John Edward Scott <gmp-list@>
Subject: Re: GMPBench 0.1 on an AMD64 == 5000
To: gmp-devel at swox.com
Message-ID: <20031216222418.GA24249 at gandalf>
Content-Type: text/plain; charset=iso-8859-1

On Thu, Dec 11, 2003 at 08:02:16PM +0100, Torbjorn Granlund wrote
> :... Just "configure; make" should give absymal performance, since it
> doesn't use any assembly loops.
> --
> Torbjörn
> _______________________________________________
> gmp-devel mailing list
> gmp-devel at swox.com
> https://gmplib.org/mailman/listinfo/gmp-devel

Yes I'd love to know how you compiled it too, since I ran across
the "absymal performance" that Torbjorn mentioned.

------------------------------
 
Basically John, it's my GCC.  I compiled my gcc from source (gentoo rocks), not a snapshot, like most *nix installs.  My build detected that 64-bit insns were available, and some asm stuff in longlong.h.  
 
I had no idea my build was special until Torbjorn pointed it out to me in a private email.  It's thanks to the new gmp configure which detected and set up the gmp.h in the first place. 
 
But anyway, let's get to the important stuff:  x86_64 is a legitimate 64 bit architecture, and the speeds, price and performance are pretty staggering.  If you can put the registers to use, this is easily the best chip out there.  By a shitload, imo.
 
-Raffi
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/ms-tnef
Size: 4509 bytes
Desc: not available
Url : /list-archives/gmp-devel/attachments/20031219/00d9d895/attachment.bin


More information about the gmp-devel mailing list