Hi

Boinc 4.54 released on Alpha project.
I will test it, because some trouble in 4.53 as i wrote in other thread but you have to know thats only alpha and no official release.
Some bugs can damage your work use at your own risk.

greetz from Germany and nice weekend.
Mike

---

With each crime and every kindness we birth our future.

hi mike a little benchmark from the alpha will be nice

> hi mike a little benchmark from the alpha will be nice
> [url=http://setiweb.ssl.berkeley.edu/team_display.php?teamid=113956]
>
CC 4.54 has not been in Alpha for long enough to make an opinion yet.
AP 4.00 has major problems.
SETI 4.07 has been released to S@H.

---

BOINC WIKI

Hi

There is a great decrease in the drystones of the Benchmark on my XP 2800+
With 4.13 - 4.53 49xx Mips sec with 4.54 32xx mips.
--- - 2004-11-06 10:40:35 - Benchmark results:
--- - 2004-11-06 10:40:35 - Number of CPUs: 1
--- - 2004-11-06 10:40:35 - 1930 double precision MIPS (Whetstone) per CPU
--- - 2004-11-06 10:40:35 - 3287 integer MIPS (Dhrystone) per CPU
--- - 2004-11-06 10:40:35 - Finished CPU benchmarks

Much less.

greetz Mike

---

With each crime and every kindness we birth our future.

Measured floating point speed 2063.99 million ops/sec
Measured integer speed 3371.25 million ops/sec

these are the benchmark results of my linux box (athlon xp 2400+ oc 2.2ghz)
optimized version of boinc 4.55

---

V4.54 still have the fatal upload-bug there if doesn't manage to upload for any reason, example your internet-connection is down or the upload-server is down, instead of doing exponential backoff immediately says "Giving up on upload of ... too much elapsed time".

> V4.54 still have the fatal upload-bug ..
tkx for info.

does this happen on *nix versions or m$ versions or general?

> > V4.54 still have the fatal upload-bug ..
> tkx for info.
>
> does this happen on *nix versions or m$ versions or general?
>

V4.5x-alpha is for the moment windows-only. No idea if it will show up on other platforms if you compile your own client.

Hi

A lot to do.
Graphics going wired with 4.07.
Upload errors when not connected or failed.

greetz Mike

---

With each crime and every kindness we birth our future.

>
> V4.5x-alpha is for the moment windows-only. No idea if it will show up on
> other platforms if you compile your own client.

Still have problems to load parts of the source into GW-Basic compiler.

So I let do this by people knowing this. And wish them success. keeping the hope, they don't stop to go on. They are on the right way.
Take the time it needs to fine tune until public go.

Most of *my* m$ clients are running now without interruption 24h since more than 3 weeks. some even more. No problems. Perhaps related, for the moment, not attached to project seti.
Running 4.09. *I* have no reason (now) to change anything ;-)

But watching close the progresses and availabilty of new, rendering improoved clients.

> Hi
>
> There is a great decrease in the drystones of the Benchmark on my XP 2800+
> With 4.13 - 4.53 49xx Mips sec with 4.54 32xx mips.
> --- - 2004-11-06 10:40:35 - Benchmark results:
> --- - 2004-11-06 10:40:35 - Number of CPUs: 1
> --- - 2004-11-06 10:40:35 - 1930 double precision MIPS (Whetstone) per CPU
> --- - 2004-11-06 10:40:35 - 3287 integer MIPS (Dhrystone) per CPU
> --- - 2004-11-06 10:40:35 - Finished CPU benchmarks
>
> Much less.
>
> greetz Mike
>
>

Looks like they may have fixed the over-optimisation of Dhrystome benchmark code by Windows compiler bug. See here for some discussion:

http://www.ssl.berkeley.edu/pipermail/boinc_opt/2004-October/000107.html

Ned

---

*** My Guide to Compiling Optimised BOINC and SETI Clients ***
*** Download Optimised BOINC and SETI Clients for Linux Here ***

On the CVS log checking_notes was this:

"David 2 Nov 2004
- split Dhrystone source into 2 files (from Peter Smithson).
This supposedly prevents compiler optimizations on Win
that give inflated results (I don't see this). "

I guess it worked after all!

---

SETI@home classic workunits: 1,985 CPU time: 24,567 hours



Professor Desty Nova
Researching Karma the Hard Way

> On the CVS log checking_notes was this:
>
> "David 2 Nov 2004
> - split Dhrystone source into 2 files (from Peter Smithson).
> This supposedly prevents compiler optimizations on Win
> that give inflated results (I don't see this). "
>
> I guess it worked after all!
>

Well yes, both Whetstone and Dhrystone benchmark scores are now almost identical (within reproducible errors) for linux and windows on the same hardware, which is as it should be. The actual seti client is still about 10-20% faster at processing on windows though which must simply be down to the windows compiler producing a more optimised client.

Ned

---

*** My Guide to Compiling Optimised BOINC and SETI Clients ***
*** Download Optimised BOINC and SETI Clients for Linux Here ***

> Measured floating point speed 2063.99 million ops/sec
> Measured integer speed 3371.25 million ops/sec
>
> these are the benchmark results of my linux box (athlon xp 2400+ oc 2.2ghz)
> optimized version of boinc 4.55
>
Maybe a nooby question, but optimized version of all the boinc core program, were does the optimation lie exactly, for me it looks just like another copy of Seti@home page file's when i enter the page. There's not much info about it on the web page. Therefor i ask. hope i'm not asking to much :-)

---

Regards

a compiler has different optimizations for cpus (athlon, athlon-xp, athlon 64, pentium4 ...).
if you compile the code with these parameters the programm will run faster. thats why you get better benchmark results and higher claimed credits. lets call it tuning :)

but:
it will take the same time to finish a work unit because boinc doesn't calculate the unit (it is just the "manager"). same time - more credits

(i'll add some more infos on the site)

---

Question to Professor Desty Nova and John McLeod VII or anyone for that matter, i am involved in allmost all of the BOINC projects, i was wondering how to get the sig stats to come up just like Professor Desty Nova and John McLeod VII have.
Thanks

---

Kevin

> a compiler has different optimizations for cpus (athlon, athlon-xp, athlon 64,
> pentium4 ...).
> if you compile the code with these parameters the programm will run faster.
> thats why you get better benchmark results and higher claimed credits. lets
> call it tuning :)
>
> but:
> it will take the same time to finish a work unit because boinc doesn't
> calculate the unit (it is just the "manager"). same time - more credits
>
> (i'll add some more infos on the site)
>
Ok, hmm tunning. that's the name :-D. Thx, i'll look forward to see how much i can get out of it.

---

Regards

I might have just figured it out
Test Test

---

Kevin

do i have to put <img> around it or <a> around it?

---

Kevin