Hi all,

I've been working with Chris Bosshard on an optimized SETI app for LINUX to accompany my optimized boinc clients, and I wanted to guage interest to see if it's worth compiling a whole bunch of these for different processor architectures and making them available for download.

So far we've only tested on AMD processors and are seeing about 13% improvements for AthlonXP and slightly more (15-20% ??) for 64-bit systems (x86_64). Given these improvements I wondered what sort of interest there would be.

If you would be interested, please post your processor architecture (Athlon XP, PIII, P4 etc) and I might make some optimized clients available for public download :)

And thanks again to Chris Bosshard :)

Regards,

Ned

---

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

I'd be interested in an Athlon XP SETI client. I am already running your optimized boinc client...
> Hi all,
>
> I've been working with Chris Bosshard on an optimized SETI app for LINUX to
> accompany my optimized boinc clients, and I wanted to guage interest to see if
> it's worth compiling a whole bunch of these for different processor
> architectures and making them available for download.
>
> So far we've only tested on AMD processors and are seeing about 13%
> improvements for AthlonXP and slightly more (15-20% ??) for 64-bit systems
> (x86_64). Given these improvements I wondered what sort of interest there
> would be.
>
> If you would be interested, please post your processor architecture (Athlon
> XP, PIII, P4 etc) and I might make some optimized clients available for public
> download :)
>
> And thanks again to Chris Bosshard :)
>
> Regards,
>
> Ned
>
>

Hey Ned...is Chris using the FFTW3 version?

> Hey Ned...is Chris using the FFTW3 version?
>
>
As usual, you've lost me ben!

We've been using the public seti sources from 1st Dec 2004 (V4.70).

Ned

---

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

> If you would be interested, please post your processor architecture (Athlon
> XP, PIII, P4 etc) and I might make some optimized clients available for public
> download :)
>

Hi Ned!

I'm already using your optimized 4.19 clients and would be interested in clients for this processor architectures:

athlon-xp
athlon
k6-3
k6-2
pentium3
I guess for the Pentium 2 i've to use i686, right?

So if you don't mind doing all this work, it would be lovely.
But anyway, thank you *very* much for all the good work you've done so far.

rattelschneck

---

> As usual, you've lost me ben!
>
> We've been using the public seti sources from 1st Dec 2004 (V4.70).
>

One of the submitted optmizations to the sourceforge project was a replacement of the FFT code that seti uses with the public FFTW3 library FFT.
- Fastest Fourier Transform in the West.

The reason I guessed at this was their shown improvement was around 13-15%.

---

Is it all compiler flags, or is some of it code changes?

Hi Benher

We recompiled original sources from Berkeley archives.
Only played with different GCC flag combinations and tested them against the original Seti clients in use.

As Ned mentioned the results on 32bit AMD systems are in the range of 13 percent.

The tests on my X86_64 system indicate an improvment of 18 percent when comparing with the original seti client.

Is anybody using a 64bit Linux version (x86_64) and is interested in testing the client for this architecture?

Cheers
Chris

---

Chris Bosshard
Visit my homepage
astroinfo SETI page

Yes ! Please compile some clients and make them availible !

>
> Is anybody using a 64bit Linux version (x86_64) and is interested in testing
> the client for this architecture?
>
> Cheers
> Chris
>
>
>

Yes definitely, I could never get the SETI client itself to compile, so one that's already compiled and optimised would be very welcome.

OK, we need a couple more beta testers this week to further test the clients before releasing them publically. If you'd like to volunteer, you can contact me by e-mail (address is on my website linked in signiture) or at my team's seti forum here:

http://forums.pcper.com/forumdisplay.php?f=17

Chris has primarily been working on the x86_64 client and I've been doing the 32-bit clients. We've been using berkeley's original v4.70 sources unmodified. The client I've been testing has now returned and validated over 3,000 credits without error, but we now need a few more testers to further test some static builds for speed and compatability.

Regards,

Ned

---

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

> If you would be interested, please post your processor architecture (Athlon
> XP, PIII, P4 etc) and I might make some optimized clients available for public
> download :)
>
> And thanks again to Chris Bosshard :)
>
> Regards,
>
> Ned


ned,I would be interested in a optimized client for my p4 and would be willing to help in any way. However I'm just a landscaper and any skills that I have regarding computers is truly minimal.

---

Please all be aware...that the optimized clients are good for Seti's science, but may not increase your credit scores.

Your machine will be claiming xxxx seconds of work, with the benchmarks you already have (faster benchmarks with Ned's optimized boinc)...and so even though you will really be doing 13-18% more work your claimed credits will be for xxxx seconds for your host.

Now given that the 2 or 3 other hosts doing the same WUs may claim higher scores, it is likely yours will end up being the lowest...tossed..claim and you probably will get more than you claim, but be aware.

Benher is absolutely right.

The optimized seti client reduces the calculation time per work unit.
It will let you do more science per machine, which I think is our goal.
Automatically this will reduce the ammount of credit you claim.

On the other hand, you still have the chance to receive more credit than your own client claims... :-)

The real effect on the average credits per machine will be seen once we use the new clients for some weeks.

Chris

---

Chris Bosshard
Visit my homepage
astroinfo SETI page

An update:

Unfortunately, we're having a few problems getting statically compiled clients to run on other systems. Static clients are needed to ensure all the libraries are present on the users system. The client runs fine on the system it was compiled on, but atm we can't get it to work on other systems, meaning others can't test or use it just yet.

We're working on it, and if/when we get this issue resolved then we'll start sending out the client for more testing.

I'll post back here when we've made some progress!

If anyone can offer any help with regards to static builds of the seti app, we have a thread going here:

http://forums.pcper.com/showthread.php?t=381586

Thanks to everyone that's contacted me by e-mail offereing to help :)

Ned

---

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

> Your machine will be claiming xxxx seconds of work, with the benchmarks you
> already have (faster benchmarks with Ned's optimized boinc)...and so even
> though you will really be doing 13-18% more work your claimed credits will be
> for xxxx seconds for your host.

I don't mean this to be critical, but while optimizing BOINC is generally good, isn't an "optimized benchmark" a contradiction?

I mean, isn't the whole idea of a benchmark to accurately predict the performance of the computer when it gets work?

---

> I mean, isn't the whole idea of a benchmark to accurately predict the
> performance of the computer when it gets work?

Ned (Ludd),

The "optimized boinc" clients are for *nix only.
They bring the benchmark results for *nix much closer to the benchmarks produced by the windows boinc client (on identical machines).

All projects currently overestimate how long WUs will take on any given machine...and so either the benchmark numbers are too low, or the estimated work for each WU is too large.

Anything that bring the estimated completion time closer to actual does the science of each project a favor.

Besides which, the optimizations probably reduce the overhead of the *nix boinc manger allowing more CPU time for the science application.

> > I mean, isn't the whole idea of a benchmark to accurately predict the
> > performance of the computer when it gets work?
>
> Ned (Ludd),
>
> The "optimized boinc" clients are for *nix only.
> They bring the benchmark results for *nix much closer to the benchmarks
> produced by the windows boinc client (on identical machines).
>
> All projects currently overestimate how long WUs will take on any given
> machine...and so either the benchmark numbers are too low, or the estimated
> work for each WU is too large.
>
> Anything that bring the estimated completion time closer to actual does the
> science of each project a favor.
>
> Besides which, the optimizations probably reduce the overhead of the *nix
> boinc manger allowing more CPU time for the science application.

If the optimizations bring *nix closer to the Win 2.2x clients, then that is a good thing, but "faster" benchmarks are not necessarily good unless the results are valid.

---

Yes, that's correct. When I first started the optimized boinc clients it was because the Windows version gave benchmark score about double that compared to the linux version running on the same hardware. This disparity caused two issues - first it affected the prediction of the amount of work required (which is based on benchmark performance) and secondly, windows users were claiming almost double the credit compared to linux users. Optimizing the linux boinc client reduced the disparity between the two OS platforms.

Our current work is aimed at optimizing the seti client - something that's intended to assist in the science as it will allow users to process work units faster, thus allowing more "science" to be performed in a given timeframe. Again, the type of increases we are seeing makes the linux seti app more comparable to the windows version, maybe even fractionally faster (the linux version has always been about 10-20% slower right back to the seti classic days).

Ned

---

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

> Yes, that's correct. When I first started the optimized boinc clients it was
> because the Windows version gave benchmark score about double that compared to
> the linux version running on the same hardware. This disparity caused two
> issues - first it affected the prediction of the amount of work required
> (which is based on benchmark performance) and secondly, windows users were
> claiming almost double the credit compared to linux users. Optimizing the
> linux boinc client reduced the disparity between the two OS platforms.

... and if I'm not mistaken, BOINC 4.2x does a better job on Windows hosts because the benchmark doesn't get "optimized" by Visual C++.

---

>
> ... and if I'm not mistaken, BOINC 4.2x does a better job on Windows hosts
> because the benchmark doesn't get "optimized" by Visual C++.
>

Hmmm. I guess throwing away useless bits of code is a vital feature for a
compiler that is used by Microsoft itself :o)

Regards Hans

---