I have an old K6-2 based computer which I decided to dedicate for the setiathome project.

I have installed FreeBSD 5.4-RELEASE for various reasons (mainly performance but also to get familiar with BSD based UNIX).

However, I am now starting to getting frustrated with setiathome on FreeBSD. After taking almost a week to overcome the 'i386-portbld-freebsd5.4' unkown platform problem (which many users have), I now get segmentation faults at the computation of the work received from setiathome. This is also showed at my account results for the specific computer which report 'Client Error' and 'Computation Error'.

What is going on with FreeBSD? Why is it so difficult to setup? Why these problems appear only in FreeBSD? My other gentoo/linux-based systems work with no problem at all!

The 'boinc-client' and 'boinc-setiathome' have been compiled from ports with CFLAGS=-O3 -pipe -funroll-loops -fomit-frame-pointer

I know plan to remove the -funroll-loops -fomit-frame-pointer flags in case they have something to do with the segmentation fault problem.

Although I have googled a lot, I am not happy with any answers in this forum on anywhere else. They are either incomplete or they do not address setiathome on FreeBSD issues in a guided way.

Given the fact that setiathome is supported on FreeBSD from Berkeley (OR NOT???), I believe it would be very much appreciated if there was a howto or at least a set of instructions to correctly install, setup and run setiathome on FreeBSD.

Any suggestions for the segmentation fault issue would also be appreciated. Any logs that I can generate? Also, kernel reports that a core is dumped but I cannot locate any core file!

Please help!

---

I feel that I also should provide some further information.

Here is the output I get:

root@zod# boinc_client
2006-03-31 09:10:41 [---] Starting BOINC client version 5.2.15 for i386-portbld-freebsd5.4
2006-03-31 09:10:41 [---] libcurl/7.15.3 OpenSSL/0.9.7e zlib/1.2.1
2006-03-31 09:10:41 [---] Data directory: /var/db/boinc
2006-03-31 09:10:41 [SETI@home] Found app_info.xml; using anonymous platform
2006-03-31 09:10:41 [---] Processor: 1 i386 AMD-K6(tm) 3D processor
2006-03-31 09:10:41 [---] Memory: 183.48 MB physical, 0 bytes virtual
2006-03-31 09:10:41 [---] Disk: 1.61 GB total, 720.00 MB free
2006-03-31 09:10:41 [SETI@home] Computer ID: 2294854; location: ; project prefs: default
2006-03-31 09:10:41 [---] General prefs: from SETI@home (last modified 2006-03-29 13:39:08)
2006-03-31 09:10:41 [---] General prefs: using your defaults
2006-03-31 09:10:41 [---] Remote control not allowed; using loopback address
2006-03-31 09:10:41 [SETI@home] Resuming computation for result 25ja03aa.6863.18849.579834.1.183_2 using setiathome version 418
2006-03-31 09:10:41 [SETI@home] Started upload of 23ap03ab.6581.19794.273584.1.191_3_0
2006-03-31 09:10:53 [SETI@home] Finished upload of 23ap03ab.6581.19794.273584.1.191_3_0
2006-03-31 09:10:53 [SETI@home] Throughput 1579 bytes/sec
2006-03-31 09:16:14 [SETI@home] Unrecoverable error for result 25ja03aa.6863.18849.579834.1.183_2 (process got signal 8)
2006-03-31 09:16:14 [SETI@home] Unrecoverable error for result 25ja03aa.6863.18849.579834.1.183_2 (process got signal 8)
2006-03-31 09:16:14 [---] request_reschedule_cpus: process exited
2006-03-31 09:16:14 [SETI@home] Computation for result 25ja03aa.6863.18849.579834.1.183_2 finished
2006-03-31 09:16:16 [SETI@home] Started upload of 25ja03aa.6863.18849.579834.1.183_2_0
2006-03-31 09:16:24 [SETI@home] Finished upload of 25ja03aa.6863.18849.579834.1.183_2_0
2006-03-31 09:16:24 [SETI@home] Throughput 2231 bytes/sec
^Z
[1]+ Stopped boinc_client



Note that the client does not stop executing but stills downloads new work and uploads invalid results!

---

However, I am now starting to getting frustrated with setiathome on FreeBSD. After taking almost a week to overcome the 'i386-portbld-freebsd5.4' unkown platform problem (which many users have), I now get segmentation faults at the computation of the work received from setiathome. This is also showed at my account results for the specific computer which report 'Client Error' and 'Computation Error'.

What is going on with FreeBSD? Why is it so difficult to setup? Why these problems appear only in FreeBSD? My other gentoo/linux-based systems work with no problem at all!

It looks, that the math problems are a *BSD specific problem. I have the simliar problems with NetBSD and OpenBSD on different platforms.

The 'boinc-client' and 'boinc-setiathome' have been compiled from ports with CFLAGS=-O3 -pipe -funroll-loops -fomit-frame-pointer

I know plan to remove the -funroll-loops -fomit-frame-pointer flags in case they have something to do with the segmentation fault problem.

Can you please try to compile it with "-O" only.
I heavly recommend you to let your binary first run against the reference WU. The reference WU is in clients/test_workunits. Copy reference_work_unit.sah to work_unit.sah, copy your client into this directory and start it from the commandline. After your client has computed the reference WU, make a diff between result.sah and reference_result.sah, differences up to 1 % are OK.

Given the fact that setiathome is supported on FreeBSD from Berkeley (OR NOT???), I believe it would be very much appreciated if there was a howto or at least a set of instructions to correctly install, setup and run setiathome on FreeBSD.[quote]
Seti will run on FreeBSD. But Berkeley do not make any binaries. But I think, it is better, if a volunteer releasing well running and tested binaries for FreeBSD. It will help the users to install it quickly, and have not to compile it via the ports.
I will look what I can do....

[quote]Any suggestions for the segmentation fault issue would also be appreciated. Any logs that I can generate? Also, kernel reports that a core is dumped but I cannot locate any core file!

You can try to look at with ktruss, what happens. But your seti app crashs with signal 8 (SIGFPE) it is a floating point error. So it looks, that it is better to recompile the seti app first, or wait till a volunteer release tested binaries for FreeBSD.

---

---

Hello again, thanks for your response.

I have made several attempts to produce a working binary - but with no luck! I have updated my source tree, compiled setiathome from ports and also individually (both boinc-client and setiathome). In addition, I tried it with no optimization at all (only with the -march=k6-2 and -pipe options).

I am still getting floating point exceptions! What could be the problem?

I would rather build the binary myself than use a precompiled binary. There are several reasons for that including the fact that I do not want X support and also want to pass cpu-specific and/or optimization options.

Any suggestion at this point will be much appreciated!

Thanks!

---

I have made several attempts to produce a working binary - but with no luck! I have updated my source tree, compiled setiathome from ports and also individually (both boinc-client and setiathome). In addition, I tried it with no optimization at all (only with the -march=k6-2 and -pipe options).

I am still getting floating point exceptions! What could be the problem?

I would rather build the binary myself than use a precompiled binary. There are several reasons for that including the fact that I do not want X support and also want to pass cpu-specific and/or optimization options.

Any suggestion at this point will be much appreciated!

Can you please collect some debuging informations for me :
- Copy a WU and to a seperate directory
- Copy your setiathome binary to this seperate directory
- Start ktruss : "ktruss ./setiathome > out 2>&1 "
- Let it run for about 5 minutes, then abort.
Can you please compress the file "out" and mail it me (seti_boinc at dotsch dot de)

Btw. the including X support and optimistation for the a special CPU will not be heavy faster. There is a known performance problem with FreeBSD. - The FreeBSD binaries are really slow.
But I am not shure, if we should invest to much work in the seti app, because the seti enhanced will be released soon...

---

I have made several attempts to produce a working binary - but with no luck! I have updated my source tree, compiled setiathome from ports and also individually (both boinc-client and setiathome). In addition, I tried it with no optimization at all (only with the -march=k6-2 and -pipe options).

I am still getting floating point exceptions! What could be the problem?

I would rather build the binary myself than use a precompiled binary. There are several reasons for that including the fact that I do not want X support and also want to pass cpu-specific and/or optimization options.

Any suggestion at this point will be much appreciated!

Can you please collect some debuging informations for me :
- Copy a WU and to a seperate directory
- Copy your setiathome binary to this seperate directory
- Start ktruss : "ktruss ./setiathome > out 2>&1 "
- Let it run for about 5 minutes, then abort.
Can you please compress the file "out" and mail it me (seti_boinc at dotsch dot de)

Btw. the including X support and optimistation for the a special CPU will not be heavy faster. There is a known performance problem with FreeBSD. - The FreeBSD binaries are really slow.
But I am not shure, if we should invest to much work in the seti app, because the seti enhanced will be released soon...

Hello,

------------------------------------------------------------------------------------------------
I am having problems sending you the traces to the address seti_boinc@dotsch.de
(Remote host said: 554 Sorry, no mailbox here by that name. (#5.1.1))
------------------------------------------------------------------------------------------------

I decided to setup FreeBSD 6.0-RELEASE on a virtual PC to check if the same
problems occur as with 5.4-RELEASE. I have installed boinc-setiathome from
ports and I have success with it!

I do not get any floating point segmentation faults on the virtual PC
(running on a Linux 64-bit host (32-bit FreeBSD 6.0 client)). However,
there are differences between the reference_result.sah and the result.sah
produced from setiathome on FreeBSD 6.0. I attach these differences as
well. Could you please take a look at them and tell me if these differences
are acceptable?

I know plan to install FreeBSD 6.0 on the old PC. If the same binary does
not work I will give it another go by recompiling it on that PC. If that
does not work... then I think there is a hardware problem and the PC will
go to the garbage!

I will let you know how it goes...

---

I am having problems sending you the traces to the address seti_boinc...
(Remote host said: 554 Sorry, no mailbox here by that name. (#5.1.1))

Can you please try it again.

I decided to setup FreeBSD 6.0-RELEASE on a virtual PC to check if the same
problems occur as with 5.4-RELEASE. I have installed boinc-setiathome from
ports and I have success with it!

I do not get any floating point segmentation faults on the virtual PC
(running on a Linux 64-bit host (32-bit FreeBSD 6.0 client)). However,
there are differences between the reference_result.sah and the result.sah
produced from setiathome on FreeBSD 6.0. I attach these differences as
well. Could you please take a look at them and tell me if these differences
are acceptable?

I will verify it.

I know plan to install FreeBSD 6.0 on the old PC. If the same binary does
not work I will give it another go by recompiling it on that PC. If that
does not work... then I think there is a hardware problem and the PC will
go to the garbage!

I think, that your PC are OK, and the math problems are a *BSD problem.

If you try out, if your PC computes right, you can also install the seti enhanced binary (http://www.lb.shuttle.de/apastron/boincDown.shtml#freebsd) and let a reference WU running against the binary. - If you wan't to test, I can mail you the reference WU...

---