I'm successfully running both boinc-client and boinc-setiathome on a FreeBSD-6.0 athlon-XP box. I saw that the sources spec'ed in the Makefile are from 2005-06-01. Is there anyone here that has used more recent versions of the sources to compile the binaries? And did doing so give any appreciable benefits?

thz

T F

I'm successfully running both boinc-client and boinc-setiathome on a FreeBSD-6.0 athlon-XP box. I saw that the sources spec'ed in the Makefile are from 2005-06-01. Is there anyone here that has used more recent versions of the sources to compile the binaries? And did doing so give any appreciable benefits?

I think, that some people has compiled it newer sources on FreeBSD.
The boinc in the ports is version 4.68. There was a lot of changes and bugfixes in the boinc client.
The seti app is 4.07. There is no need to compile the newer version (4.18). In short the new enhanced app will be released....

---

I spent this week by updating the ports, they are now available for testing at http://people.freebsd.org/~pav/boinc.html

Now I'm checking out various optimized linux seti clients to see if any of them will work on FreeBSD under Linux emulation, because the performance of seti client built from sources is abysmal.

---

I spent this week by updating the ports, they are now available for testing at http://people.freebsd.org/~pav/boinc.html

Now I'm checking out various optimized linux seti clients to see if any of them will work on FreeBSD under Linux emulation, because the performance of seti client built from sources is abysmal.

I think the lower performance is a *BSD problem. Some people has reported this on different forums here.
My OpenBSD and NetBSD port are slighty slower (10 to 15 %) than the windows version.
I am not shure, if the linux build on *BSD gave better performance.

Which which compiler flags do you have compiled the seti app ? - How many is it slower than the linux build on the same or comparable system ?

---

Now I'm checking out various optimized linux seti clients to see if any of them will work on FreeBSD under Linux emulation, because the performance of seti client built from sources is abysmal.

I think the lower performance is a *BSD problem. Some people has reported this on different forums here.
My OpenBSD and NetBSD port are slighty slower (10 to 15 %) than the windows version.
I am not shure, if the linux build on *BSD gave better performance.

Which which compiler flags do you have compiled the seti app ?

-O3 -march=athlon64

How many is it slower than the linux build on the same or comparable system ?

I can't compare with Linux, I don't have dualboot. But I'm working on amd64 (x86_64) system, and native 64-bit clients are known to be very slow. I'm still waiting for completion of first work unit, I have 74% done with 10.5 hours of CPU time.

Harold Naparst's SSE client for Linux can do the reference work unit slightly under an hour on that very box.

---

I can't compare with Linux, I don't have dualboot. But I'm working on amd64 (x86_64) system, and native 64-bit clients are known to be very slow. I'm still waiting for completion of first work unit, I have 74% done with 10.5 hours of CPU time.

Harold Naparst's SSE client for Linux can do the reference work unit slightly under an hour on that very box.

Strange, it is to much out of range. Is a 32 bit app faster on the 64 bit systems ?
But Harolds Clients have a modified sources...

The seti enhanced is announced, it looks like it will released soon. So it is the question, is it worth to spend a lot of work for the old seti...
The new release can use the libfft, which increases performance.

---

Strange, it is to much out of range. Is a 32 bit app faster on the 64 bit systems ?

Obviously, it is.

But Harolds Clients have a modified sources...

As long as it returns valid results, which it does, there is no problem.

The seti enhanced is announced, it looks like it will released soon. So it is the question, is it worth to spend a lot of work for the old seti...

Not up to me to judge...

The new release can use the libfft, which increases performance.

You mean FFTW, as of www.fftw.org fame? We already use that in the port.

---

Strange, it is to much out of range. Is a 32 bit app faster on the 64 bit systems ?

Obviously, it is.

I recommend you, to post you problems to the boinc_opt mailinglists.

But Harolds Clients have a modified sources...

As long as it returns valid results, which it does, there is no problem.

Depends on the sight of view...

The new release can use the libfft, which increases performance.

You mean FFTW, as of www.fftw.org fame? We already use that in the port.

Yes. But, so far I know, are the libfftw not used/implemented in the standard seti source.

---