Stephen R Marenka wrote:
On Fri, Jun 15, 2007 at 03:18:44PM +, Bill Allombert wrote:
Here the result with current aranym CVS on the same host:
The result are much better all around. It is about twice as fast
as crest.
Sweet! It be nice to get that packaged. The archive only has
0.9.4beta2
On Fri, Jun 15, 2007 at 03:18:44PM +, Bill Allombert wrote:
> Here the result with current aranym CVS on the same host:
> The result are much better all around. It is about twice as fast
> as crest.
Sweet! It be nice to get that packaged. The archive only has
0.9.4beta2-1.
Thanks,
Stephen
On Wed, Oct 11, 2006 at 11:37:36PM +0200, Bill Allombert wrote:
> Hello Debian M68k,
>
> I have made a very small benchmark of aranym and crest:
> Here the results:
>
> The host is a Pentium 4 at 3.20GHz with 1Gb RAM.
> aranym version is 0.9.4beta-3
>
> The benchmark has been run three time on e
On Thu, Oct 12, 2006 at 12:57:37PM +0200, Michael Schmitz wrote:
> > I have made a very small benchmark of aranym and crest:
> > Here the results:
> >
> > wget tar zxf rm -r MPQS galois
> > =
> > host-i386
On Wed, 18 Oct 2006, Kars de Jong wrote:
> On do, 2006-10-12 at 18:37 +0200, Geert Uytterhoeven wrote:
> > On Thu, 12 Oct 2006, Michael Schmitz wrote:
> > > > So we see that aranym-amd64 and aranym-i386 are about the same
> > > > speed, that crest is between 3 and 4 times faster and that the
> > >
On do, 2006-10-12 at 18:37 +0200, Geert Uytterhoeven wrote:
> On Thu, 12 Oct 2006, Michael Schmitz wrote:
> > > So we see that aranym-amd64 and aranym-i386 are about the same
> > > speed, that crest is between 3 and 4 times faster and that the
> > > host is between 250 and 400 times faster.
> >
>
On Thu, Oct 12, 2006 at 03:58:33PM +0200, Bill Allombert wrote:
> On Thu, Oct 12, 2006 at 12:57:37PM +0200, Michael Schmitz wrote:
> > Thanks for doing such an extensive benchmark. For testing package builds,
> > we really need Petr's disk access speedups ...
>
> Well before doing packages builds,
> > it just throws random kernel panics from the interrupt handler,
> > and the soft lockup watchdog. Any ideas on this?
>
> No idea, doesn't happen to me. I am still using kernel 2.4.27 but I
> almost started compiling 2.4.33 with the eth patch for Bill as I can't
> believe that the gcc ICE is ara
On Mon, 16 Oct 2006, Michael Schmitz wrote:
> > > we really need Petr's disk access speedups ...
> >
> > BTW, I have just tried the tar zxf test on ramdisk (to eliminate the IDE
> > bottleneck) and it finished in 43 seconds (Athlon XP 2500+ = 1833 MHz).
> > Compare with 26 seconds on crest...
>
>
Michael Schmitz wrote:
BTW: something seems utterly broken in the current testing (-3) aranym
package. I seem to recall you mentioned something went wrong there
already;
Bill found that current sid with its gcc-4.1.1-14 does not compile
anything due to ICE. But that shouldn't be related to -3
> > we really need Petr's disk access speedups ...
>
> BTW, I have just tried the tar zxf test on ramdisk (to eliminate the IDE
> bottleneck) and it finished in 43 seconds (Athlon XP 2500+ = 1833 MHz).
> Compare with 26 seconds on crest...
BTW: something seems utterly broken in the current testing
On Fri, Oct 13, 2006 at 02:23:52PM +0200, Petr Stehlik wrote:
> The gcc-4.1.1-14 does not compile anything. It always crashes here
> (strace output):
>
> 7314 open("/tmp/ccoNE7AY.s", O_RDWR|O_CREAT|O_TRUNC, 0666) = 3
> 7314 fstat64(3, {st_mode=S_IFREG|0600, st_size=0, ...}) = 0
> 7314 old_mma
Bill Allombert wrote:
[EMAIL PROTECTED]:~/pari-2.3.1$ cat config/ansi.c
int main(int argc, char **argv){ return 0; }
[EMAIL PROTECTED]:~/pari-2.3.1$ cc config/ansi.c
config/ansi.c: In function âmainâ:
config/ansi.c:1: internal compiler error: Illegal instruction
The system run sid with cc bein
On Thu, Oct 12, 2006 at 06:37:30PM +0200, Geert Uytterhoeven wrote:
> On Thu, 12 Oct 2006, Michael Schmitz wrote:
> > > So we see that aranym-amd64 and aranym-i386 are about the same
> > > speed, that crest is between 3 and 4 times faster and that the
> > > host is between 250 and 400 times faster.
Michael Schmitz wrote:
crest 5.325s 26.827s 3.693s 28,160 ms. 10,850 ms.
Thanks for doing such an extensive benchmark. For testing package builds,
we really need Petr's disk access speedups ...
BTW, I have just tried the tar zxf test on ramdisk (to eliminate the IDE
bottleneck)
On Thu, 12 Oct 2006, Michael Schmitz wrote:
> > So we see that aranym-amd64 and aranym-i386 are about the same
> > speed, that crest is between 3 and 4 times faster and that the
> > host is between 250 and 400 times faster.
>
> Thanks for doing such an extensive benchmark. For testing package buil
> I have made a very small benchmark of aranym and crest:
> Here the results:
>
> wget tar zxf rm -r MPQS galois
> =
> host-i386 3.313s0.371s 0.036s 249 ms. 93 ms.
> host-i386 6.022s0.
On Thu, Oct 12, 2006 at 12:57:37PM +0200, Michael Schmitz wrote:
> Thanks for doing such an extensive benchmark. For testing package builds,
> we really need Petr's disk access speedups ...
Well before doing packages builds, there is still some issues to fix:
[EMAIL PROTECTED]:~/pari-2.3.1$ cat c
Bill Allombert píše v St 11. 10. 2006 v 23:37 +0200:
> wget : downloading a 2Mb tarball with wget
> tar zxf: unpacking a 2Mb tarball
> rm -r : deleting the tree in the tarball
> MPQS : number crunching+file access
> galois : number crunching+data handling
All tests suffered from the extremely
Hello Debian M68k,
I have made a very small benchmark of aranym and crest:
Here the results:
The host is a Pentium 4 at 3.20GHz with 1Gb RAM.
aranym version is 0.9.4beta-3
The benchmark has been run three time on each hardware.
host-i386: host running debian-i386
host-amd64: host running debia
20 matches
Mail list logo
