[beadman]

I installed the new optimized BOINC 4.19 and started it up. Here's the terminal log:


Last login: Sat Jan 29 11:06:26 on ttyp1
Welcome to Darwin!
[pcp07273623pcs:~] cfreaner% cd boinc
[pcp07273623pcs:~/boinc] cfreaner% chmod a+rx ./boinc_4.19_ppcG4
[pcp07273623pcs:~/boinc] cfreaner% ./boinc*
2005-01-29 11:22:32 [---] Starting BOINC client version 4.19 for powerpc-apple-darwin
2005-01-29 11:22:32 [climateprediction.net] Project prefs: no separate prefs for work; using your defaults
2005-01-29 11:22:32 [SETI@home] Project prefs: no separate prefs for work; using your defaults
2005-01-29 11:22:32 [SETI@home] Found app_info.xml; using anonymous platform
2005-01-29 11:22:32 [climateprediction.net] Host ID is 26432
2005-01-29 11:22:32 [SETI@home] Host ID is 280526
2005-01-29 11:22:32 [---] General prefs: from SETI@home (last modified 2005-01-13 22:59:19)
2005-01-29 11:22:32 [---] General prefs: no separate prefs for work; using your defaults
2005-01-29 11:22:32 [climateprediction.net] Resuming computation for result 1wv6_100110161_0 using hadsm3 version 4.03
2005-01-29 11:22:32 [SETI@home] Started upload of 30ja04aa.13768.3041.361066.107_0_0
Starting model in /Users/cfreaner/boinc/projects/climateprediction.net...
Created shared memory region key = 24515
Env Used=DYLD_LIBRARY_PATH=/Users/cfreaner/boinc/projects/climateprediction.net:../
Starting model ID 1wv6_100110161 Phase 3
Stack size=48.00 MB
Waiting for model startup, this may take a minute...
1wv6_100110161 - PH 3 TS 009505 - 19/06/2051 00:30 - H:M:S=1202:18:09 AVG= 8.20 DLT= 0.00
CPDN Monitor got quit request...
Detaching shared memory...

The last two lines are something new - following this, nothing happens. I can enter a Control-C, then "exit" and it ignores me (also tried "logoff", "quit"). I looked in Activity Monitor and see Boinc still running (process name is "boinc_4.19_ppcG4".

When I launch a new terminal window and attempt to start BOINC, it tells me BOINC is already running, which confirms Activity Monitor. I tried using the "Stop-Boinc.command" from the DomPak, but that didn't do anything either. I've reinstalled the optimized client and tried again, and still get the same problem. Activity monitor shows boinc is running, but nothing seems to be getting done - I've completed no SETI WU or CPDN trickles, even though I had a couple of both that should have completed last night.

Any ideas or suggestions would be appreciated.
Thanks,
beadman

[Shaktai]

I got an error also. (however my integer benchmarks were much higher then before.) This is on Einstein. I'm going to go test the G3 version on the iBook.

[iMac-G5-1-6ghz:~/Desktop/BOINC] davidrob% chmod +x boinc_4.19
[iMac-G5-1-6ghz:~/Desktop/BOINC] davidrob% ./boinc_4.19
2005-01-29 14:41:40 [---] Starting BOINC client version 4.19 for powerpc-apple-darwin
2005-01-29 14:41:40 [Einstein@Home] Project prefs: no separate prefs for home; using your defaults
2005-01-29 14:41:40 [---] Version Change Detected (4.18 -> 4.19); running CPU benchmarks
2005-01-29 14:41:40 [Einstein@Home] Host ID is 2766
2005-01-29 14:41:40 [---] General prefs: from Einstein@Home (last modified 2005-01-28 17:51:51)
2005-01-29 14:41:40 [---] General prefs: no separate prefs for home; using your defaults
2005-01-29 14:41:40 [---] Running CPU benchmarks
2005-01-29 14:42:42 [---] Benchmark results:
2005-01-29 14:42:42 [---] Number of CPUs: 1
2005-01-29 14:42:42 [---] 1463 double precision MIPS (Whetstone) per CPU
2005-01-29 14:42:42 [---] 3276 integer MIPS (Dhrystone) per CPU
2005-01-29 14:42:42 [---] Finished CPU benchmarks
2005-01-29 14:42:43 [Einstein@Home] Resuming computation for result H1_0086.9__0087.3_0.1_T15_Test02_2 using einstein version 4.74
2005-01-29 14:45:05 [---] May run out of work in 0.50 days; requesting more
2005-01-29 14:45:05 [Einstein@Home] Requesting 51105 seconds of work
2005-01-29 14:45:05 [Einstein@Home] Sending request to scheduler: http://einstein.phys.uwm.edu/EinsteinAtHome_cgi/cgi
SIGBUS: bus error
Exiting...

I then tried it a second time, this time forcing an update of preferences.

[iMac-G5-1-6ghz:~/Desktop/BOINC] davidrob% ./boinc_4.19 -update_prefs http://einstein.phys.uwm.edu
2005-01-29 14:51:16 [---] Starting BOINC client version 4.19 for powerpc-apple-darwin
2005-01-29 14:51:16 [Einstein@Home] Project prefs: no separate prefs for home; using your defaults
2005-01-29 14:51:16 [Einstein@Home] Host ID is 2766
2005-01-29 14:51:16 [---] General prefs: from Einstein@Home (last modified 2005-01-28 17:51:51)
2005-01-29 14:51:16 [---] General prefs: no separate prefs for home; using your defaults
2005-01-29 14:51:16 [Einstein@Home] Resuming computation for result H1_0086.9__0087.3_0.1_T15_Test02_2 using einstein version 4.74
2005-01-29 14:51:16 [Einstein@Home] Sending request to scheduler: http://einstein.phys.uwm.edu/EinsteinAtHome_cgi/cgi
SIGBUS: bus error
Exiting...

It seems to be erroring and then quitting, when it tries to connect.

[drweaser]

I also saw errors when using the 4.19 optimized on my G5. I got the same SIGBUS error when it tried to upload any file to the server. It happened with both Predictor and Einstein. I tried starting and stopping , and reset Predictor nut still had problems. Running the normal 4.19, the files uploaded immediately with no problem. Maybe some error in the general network code from looking at this and the above problem. Just thought I'd leave a note...

[Drash]

Problems here with the optimised 4.19 also when using optimised seti client 4.3 on a G4 eMac - hangs trying to get work from server, CPU at max. running boinc but nothing happens. Old boinc 4.13 optimised client runs fine.

4.19 never gets past this point:
2005-01-30 19:29:25 [---] Insufficient work; requesting more

whereas 4.13 continues on thus:
2005-01-30 19:29:25 [SETI@home] Requesting 152380 seconds of work
2005-01-30 19:29:25 [SETI@home] Sending request to scheduler: http://setiboinc.ssl.berkeley.edu/sah_cgi/cgi
2005-01-30 19:29:36 [SETI@home] Scheduler RPC to http://setiboinc.ssl.berkeley.edu/sah_cgi/cgi succeeded

Regards
Ash.

[Shaktai]

Yes the old 4.13 optimized client worked very well if you were running only one project. Unfortunately it had some problems with multiple projects, and created some errors for projects like Predictor and Einstein which have very complex calculations.

4.13 is no longer recommended, and some projects are stating it will not be accepted after February 1st. The unoptimized 4.19 runs fine as well but something went amiss in the optimizations. Hopefully that can be fixed soon. We'll just have to wait for mikkyo to do his magic. I'm sure he will get to it when he can.

[mikkyo]

I'm not seeing any errors.
Very odd.
Predictor and CP are both working fine for me using the new clients.
I'll see if I can attch to Seti and have problems.

[mikkyo]

So until things are sorted out, The Boinc 4.19 clients have been moved to the beta page.
There are some "test" versions up there as well you can try if you have problems with the other.

[Shaktai]

Okay, I retested both beta versions.

On my iBook G3, both versions work fine.

On my iMac G5 1.6ghz
The optimized version still gets the SIGBUS: bus error. Exiting.. when it trys to connect. However, the un-optimized verion without debug code runs fine, and still provides a substantial increase in performance.

The optimized version provides very little improvement over the unoptimized without debug code.

Standard client
2005-01-31 18:19:05 [---] 885 double precision MIPS (Whetstone) per CPU
2005-01-31 18:19:05 [---] 2841 integer MIPS (Dhrystone) per CPU

No-debug code
2005-01-31 18:21:00 [---] 1454 double precision MIPS (Whetstone) per CPU
2005-01-31 18:21:00 [---] 3355 integer MIPS (Dhrystone) per CPU

Optimized with no debug code (but ends with SIGBUS error)
2005-01-31 18:26:48 [---] 1460 double precision MIPS (Whetstone) per CPU
2005-01-31 18:26:48 [---] 3358 integer MIPS (Dhrystone) per CPU

As you can see the difference is very little in speed. between the no debug code and the optimized. However the no-debug runs correctly, and the optimized errors. Looks like something in the optimizations is causing the SIGBUS error.

EDIT: Still stable on the "less optimized" but no debug code on all my projects. If some other folks try it, then we can know for sure.

[beadman]

I went back to the standard client and it started up just fine. Benchmarks were basically the same as these:

Standard 4.19 client
2005-01-27 21:11:34 [---] 470 double precision MIPS (Whetstone) per CPU
2005-01-27 21:11:34 [---] 1461 integer MIPS (Dhrystone) per CPU

Then, after reading Shaktai's post, I tried installing this one for the G4: Less-optimized safer compiles, no debug code, and it started up with no problems. Benchmarks are

Less-optimized safer compiles, no debug code
2005-01-31 19:52:22 [---] 672 double precision MIPS (Whetstone) per CPU
2005-01-31 19:52:22 [---] 1589 integer MIPS (Dhrystone) per CPU

This can be compared to what I have been getting with the optimized 4.13 for G4

Measured floating point speed 671.79 million ops/sec
Measured integer speed 1588.58 million ops/sec

beadman

[mikkyo]

Just to be clear:
Both versions on the Beta page are optimized clients.
The test versions are the one that seems to be for other projects Im not running.
This is probably due to lack of an optimization that can interfere with shared libraries loading.

So it looks like folks should use the versions with 4.19_ppcGXtest in their names.

After a few more folks reply with success, we can move the test versions back to the main client page.

[beadman]

Thanks, mikkyo! I didn't indicate it in my first post to this thread, but I should have, that I wasn't complaining - I was afraid I was doing something wrong. I appreciate all your efforts to compile and make available to us an optimized version of the BOINC software.

beadman

[Shaktai]

Originally posted by mikkyo:
After a few more folks reply with success, we can move the test versions back to the main client page.

Okay more folks, give the "Less-optimized safer compiles, no debug code" a whirl and make sure they work for you. You can download them
HERE, and if there are no problems we can then move them to here where they have been helpful in getting us a few new members. Once we know they are good to go and stable for all projects, I'll start pimping the new optimized 4.19 versions in the various BOINC forums.

I've tested on Einstein, Climate and Pirates. These clients do make a difference. Worst case scenario, they bring claimed credit on par with the Windows machines. Best case scenario, they actually improve production by a small amount, depending on the project.

[virex]

working good for me, on my eMac, running sah and pah. haven't upgraded the pb yet, but i'm sure it'll be just as flawless

[Drash]

Not sure the G4test client is working for me (SETI opt. client 4.30 eMac 10.3.7) - I'm getting:

2005-02-02 01:38:57 [SETI@home] scheduler init_op_project to http://setiboinc.ssl.berkeley.edu/sah_cgi/cgi failed, error -107
2005-02-02 01:38:57 [SETI@home] init_op_project failed, error -107
2005-02-02 01:38:57 [SETI@home] Deferring communication with project for 1 minutes and 0 seconds
2005-02-02 01:39:57 [SETI@home] Couldn't read master page for SETI@home: error -107
2005-02-02 01:39:57 [SETI@home] Master file fetch failed

then a lot of

2005-02-02 08:11:47 [SETI@home] Couldn't start upload for http://setiboincdata.ssl.berkeley.ed...pload_handler: error -107

now it's deferred right back to 12 hrs (several times), yet my PC is still downloading/uploading every three hours like clockwork. Have reset project twice, but can't get it to stop deferring. Any ideas?

[mikkyo]

Does the default Seti worker function for you?
The optimized Seti workers haven't been rebuilt in a while, so there is a high chance they won't work.
Most folks are using Javalizard's altivec Seti worker instead of the optimized ones.

[Drash]

Apologies, my earlier post wasn't very clear. I'm using Javalizard's Altivec Setiworker (4.30) and having problems with the less optimised safer compile for G4 (I'm using an eMac with 10.3.7). I run seti 100% using the excellent SETI Control by Bob Delaney. Problem is that it has tried and failed so many times that it is deferring for such long periods of time (currently still 6 more hours 'til it tries again) that it's making testing a pain. I've reverted back to the std 4.19 to see if it connects ok when these 6 hours are up. That would probably confirm that there are problems with the less optimised boinc client.

Anyone know of a way to get boinc to reset the deferral time back to near zero so I can take less than two days to change between a couple of test clients?

Thanks
Ash

[mikkyo]

-detach and then attach.
Remember if you remove the altivec worker you need to remove the app_info.xml file as well.
Also remember that though there may be work for the default mac worker, there may not be work for the custom altivec worker as they are treated separately.
So you might just be seeing no work for the custom workers, when there is work for the default worker available.

Also, javalizard's altivec client may need to be rebuilt as well as it links with the boinc library, and that may have changed significantly enough in boinc_4.19 to require a worker rebuild.


BTW, to do testing you can just make a new directory called something like Test and put a copy of the client and your account_setiathome.berkeley.edu file in it and run the client in that directory.

[Drash]

Thanks mikkyo, now I'm learning something!

I created a sub-folder "/Applications/BOINC/TEST/" to my main boinc folder and put "account_setiathome.berkeley.edu.xml" and "boinc_4.19_ppcG4test" in it. I changed SETI Control to point to "/Applications/BOINC/TEST/boinc_4.19_ppcG4test" and let it do its thing. Their seti client "setiathome_4.02_powerpc-apple-darwin" was dowloaded along with a work unit and off it crunched.

I moved the whole of "/Applications/BOINC/TEST/", created a new TEST in its place, and put the account_seti... and the more optimised boinc4.19_ppcG4 in, altered SETI Control accordingly and it broke - just sat eating all CPU cycles thus:

2005-02-03 10:47:53 [---] 675 double precision MIPS (Whetstone) per CPU
2005-02-03 10:47:53 [---] 1596 integer MIPS (Dhrystone) per CPU
2005-02-03 10:47:53 [---] Finished CPU benchmarks
2005-02-03 10:47:54 [---] Insufficient work; requesting more

I had to Force Quit SETI Control and kill boinc using Activity Monitor to get it to die.

So it looks like javalizard's optimised setiathome_4.3_ppcG4.7450.altivec client may be broken, at least with the new boinc 4.19 code, and the more optimised boinc_4.19_ppcG4 doesn't appear to work, for me at least.

Hope that helps.
Ash

[beadman]

As I mentioned earlier, I'm currently running G4: Less-optimized safer compiles, no debug code version of 4.19 on a 1 GHz PowerBook G4. What I forgot to mention is that I'm running both SETI (Javalizard's Altivec Setiworker (4.30) and CPDN. It was the CPDN that hung on me when I tried the most-optimized version of the 4.19. Both are working fine on the no-debug version.

beadman

[mikkyo]

The optimized "test" clients have been renamed and moved to the Team's client page.
A Dominator Pack with clients for all cpus and setup and execution scripts for those with large farms is there as well.
There are no combo archives with the optimized client and Seti worker since the custom compiled Seti workers appear to need a rebuild to work with boinc_4.19.

Clients can be found here, Team MacNN Boinc Clients

[zac4mac]

guess I missed this one - "new" 4.19 client. Just saw CPDN tell me as it was uploading...

Thanks for the optimize, mikkkyo, compared to old benchmarks(optimized 4.13), floating point stayed about the same but integer jumped by about 700. Dunno if this will have my G5 doing cobblestones faster yet tho.

Tried running Einstein but got no units, so only running CPDN on G5, DPG4 and G3 iBook, old S@H on two OS9 Macs; 7500/604e-233 and 8500/G3-454. Will get new clients over to the other OS X Macs now... At some point soon, I expect to shift from CPDN to Einstein on the OS X Macs.

Z

[beadman]

When you do start Einstein, be aware of <a href="http://forums.macnn.com/showthread.php?s=&threadid=245727">this.</a> It may not affect your machines, but it certainly affects my PowerBook G4.

beadman

[LarryB56]

Whenever I attempt to download, both Safari and IE open the link as a web page of gobbledegook!!!
Now what???
Thanks;
Larry

[reader50]

To force a download of a link, hold down 'option' while clicking on the link.