benchmark for octave 2.0.5

classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|

benchmark for octave 2.0.5

Francesco Potorti`-9
Sorry if you receive multiple copies, I am writing to both help-octave
and octave-sources.

Since octave 2 is now available in binary distribution form and looks
stable, I would like to upgrade the octave benchmark in order to get
reference times relative to octave 2.0.5 for the Sun Sparc 10 / super
40, which I selected as the reference machine, in analogy with the
reference used for the Spec95 numbers (suggestion by Rick Niles).

So, if someone with a Sun Sparc 10 is willing to run the benchmark on
an unloaded machine, I will send to them the benchmark modified to run
under octave 2.  Once I get the results, I'll code them inside
benchmark.m and publish it, so people can run it on their machines and
we can have an updated performance list of the machines Octave runs
on.

If someone has good reasons why we should change the reference
machine, let us discuss them on octave-sources.

--
Francesco Potorti` (researcher)        Voice:    +39-50-593203
Computer Network Division              Operator: +39-50-593211
CNUCE-CNR, Via Santa Maria 36          Fax:      +39-50-904052
56126 Pisa - Italy                     Email:    [hidden email]

Reply | Threaded
Open this post in threaded view
|

benchmark for octave 2.0.5

John W. Eaton-6
On  7-Mar-1997, Francesco Potorti` <[hidden email]> wrote:

| Since octave 2 is now available in binary distribution form and looks
| stable, I would like to upgrade the octave benchmark in order to get
| reference times relative to octave 2.0.5 for the Sun Sparc 10 / super
| 40, which I selected as the reference machine, in analogy with the
| reference used for the Spec95 numbers (suggestion by Rick Niles).
|
| So, if someone with a Sun Sparc 10 is willing to run the benchmark on
| an unloaded machine, I will send to them the benchmark modified to run
| under octave 2.  Once I get the results, I'll code them inside
| benchmark.m and publish it, so people can run it on their machines and
| we can have an updated performance list of the machines Octave runs
| on.

If we need to have a reference machine, I think it would be best to
use something that is more widely available (sorry Rick).  Something
like a Pentium NNN running Linux (I vote for NNN == 133, since that's
what I happen to have, but maybe it would be better to use something a
bit faster).

Also, how about keeping the results obtained with older versions of
Octave (at least for runs on the reference machine) so that we can
easily see if Octave is getting faster or slower as it gets older.

| If someone has good reasons why we should change the reference
| machine, let us discuss them on octave-sources.

I would prefer to only have source code posted to octave-sources.
The help-octave group is fine for discussions as well as requests for
help.

Thanks,

jwe


Reply | Threaded
Open this post in threaded view
|

Re: benchmark for octave 2.0.5

Dave Comer-2
Not sure if my response made it through so I appologize if this is a
duplicate.

I would like to see the SUN Ultra (in addition to a Pentium) as a
benchmark machine. Octave is a very good substitute for MATLAB
on a SUN. The SPARC 5, 10, and 20 series are obsolete with respect
to SUNs current offering. The Ultra is, IMHO, best suited for such
a task.

Regards,

Dave Comer

-----------------------------------------
David M. Comer (KC5QNU)
Senior Test Engineer
Mass Storage Group
Philips Semiconductors

E-MAIL:     mailto:[hidden email]
            mailto:[hidden email] (alternate)
WWW:        http://www.rt66.com/~dcomer
-----------------------------------------

Reply | Threaded
Open this post in threaded view
|

Re: benchmark for octave 2.0.5

Francesco Potorti`-9
In reply to this post by John W. Eaton-6
"John W. Eaton" <[hidden email]> wrote:
   
   If we need to have a reference machine, I think it would be best to
   use something that is more widely available (sorry Rick).
   Something like a Pentium NNN running Linux (I vote for NNN == 133,
   since that's what I happen to have, but maybe it would be better to
   use something a bit faster).

The problem with Intel's machines is that they are not standard, but
differ, even a lot, because of the bus architecture, cache, memory
speed and so on.  For example, the benchmark (octave 1.1.1) results
list 2.0 as the performance for a Pentium 133, but I just obtained 2.5
on a recent Pentium 133.  

Sun machines are much less available, but they always have the same
architecture, given the model number.  My hope is that availability
will not be an issue, as long as we are on the Internet.

However the problem exists.  Opinions?
   
   Also, how about keeping the results obtained with older versions of
   Octave (at least for runs on the reference machine) so that we can
   easily see if Octave is getting faster or slower as it gets older.

Sure.

--
Francesco Potorti` (researcher)        Voice:    +39-50-593203
Computer Network Division              Operator: +39-50-593211
CNUCE-CNR, Via Santa Maria 36          Fax:      +39-50-904052
56126 Pisa - Italy                     Email:    [hidden email]

Reply | Threaded
Open this post in threaded view
|

Re: benchmark for octave 2.0.5

Dave Comer-2
Well...what ever the choice, I'd be happy to run the benchmark on
either an unloaded SPARC20, Ultra I (143), or Ultra II (humm, forgot
the model - I'm at home). Francesco , if you e-mail me on how to
run the benchmark I can do it. I would assume that the preferred
benchmark would be run using John's pre-compiled version. I can
compile Octave under Solaris. The only difference between my
compilation and John's (I think) is that I do not use the
"--enable-lite"
switch.

Regards,

Dave Comer

-----------------------------------------
David M. Comer (KC5QNU)
Senior Test Engineer
Mass Storage Group
Philips Semiconductors

E-MAIL:     mailto:[hidden email]
            mailto:[hidden email] (alternate)
WWW:        http://www.rt66.com/~dcomer
-----------------------------------------