I tried as well to compile the source code but failed.
I am getting error messages during configure.

If somebody could post some detailed instructions on how to proceed with the compilation would be helpful.

Especialy the explanation that you need to copy the boinc and the seti_boinc source into the same directory are strange...

Thanks in advance!


Chris Bosshard
astroinfo Team
http://www.astroinfo.org

[/url] [/url]

What is meant by copied into the same directory is this:

/tmp/boinc/boinc [---- boinc goes here ]
/tmp/boinc/seti_boinc [---- seti_boinc goes here ]

Thanks Rom

I will have another try...



Chris Bosshard
astroinfo Team
http://www.astroinfo.org

Thanks again Rom...

The compilation of at least the boinc client worked without problem.
Here the results:

Original boinc client for Linux:
1081: Float
2418: Integer

Self compiled:
2367: Float
2998: Integer

Flags used:
export CFLAGS='-O3 -mcpu=athlon-xp -march=athlon-xp -funroll-loops -fomit-frame-pointer -ffast-math'

These are results per CPU since it is a two way box with two AMD Athlon MP 2400's. Quite a difference, even the integer value went up a bit.


Chris Bosshard
astroinfo Team
http://www.astroinfo.org

[/url] [/url]

I just tried to recompile the seti_boinc with the same settings that worked for the boinc client.

It failed to compile properly...

Did any of you guys had success compiling the seti_boinc client for an Athlon-XP?


Chris Bosshard
astroinfo Team
http://www.astroinfo.org

> I just tried to recompile the seti_boinc with the same settings that worked
> for the boinc client.
>
> It failed to compile properly...
>
> Did any of you guys had success compiling the seti_boinc client for an
> Athlon-XP?
>
>
> Chris Bosshard
> astroinfo Team
> http://www.astroinfo.org
>
Yeah I tried to compile seti_boinc too, but I got some errors about .xml files :[

I managed to compile the client, with a little help.
For some reason the thread in which I posted my problems has dissapeared.
But I was getting similar errors like this guy:
http://setiweb.ssl.berkeley.edu/forum_thread.php?id=1290#6655
Maybe it can be of help.

Edit:
More stuff here:
http://setiweb.ssl.berkeley.edu/forum_thread.php?id=136

comparing the Windows client on both Operating Systems.

Running Boinc GUI on WINE, knoppix 3.4
Measured floating point speed
1400 million ops/sec
Measured integer speed
2766.2 million ops/sec


Running Boinc on WinXP SP1.
Measured floating point speed
1333.33 million ops/sec
Measured integer speed
2962.07 million ops/sec

Hmm.. what gives better throughput for seti? Integer math or Floating point math?

---
Dell Poweredge 600SC
CPU type GenuineIntel Intel(R) Pentium(R) 4 CPU 2.40GHz (533FSB)
ATI Radeon 7000 PCI 60mb DDR Graphics card (so no using onboard 8mb rage XL)
Memory 640 MB
40 gig 7200rpm 8mb cache IDE HD
---

Fedora Core 2 - Linux Kernel 2.6.5-1.358
Measured floating point speed 684.53 million ops/sec
Measured integer speed 1975 million ops/sec
Measured memory bandwidth 953.67 MB/sec


Microsoft Windows 2003 Enterprise Server Edition
Measured floating point speed 1559.79 million ops/sec
Measured integer speed 3128.06 million ops/sec
Measured memory bandwidth 953.67 MB/sec