The experiment begins.

I have sucessfully attached my PS3 to both PS3GRID and SIMAP and have work from both projects.

This will be interesting to see what happens. I hope things run normally and switching between apps causes no problems.

1 SIMAP just downloaded and started krunchin.

I have my resource shares set at 4096 and 1, so after this first work unit finishes my PS3 will go mostly back to PS3GRID work, unless there is none and SIMAP will only be used to fill in any dry spells, if they have work.

---

Alpha Tester ~~ BOINCin since 10-Apr-2004 (2.28) ~~~ Join team USA

I also have SIMAP attached. So far, I have downloaded two results (both SIMAP so far), and returned one. It took about 99 minutes to crunch. The second result is still crunching as of now. The returned result is still waiting for its wingman to return to see if it validates. You can watch the results here:

http://boinc.bio.wzw.tum.de/boincsimap/results.php?hostid=76483

I have set SIMAP to No New Work until some HMMER results get released, and I can validate that the HMMER application works too. After all, I have other machines that are much faster at crunching SIMAP. My PS3 is primarily for projects that take advantage of its unique qualities. However, it sure would be nice if SIMAP had work all the time, so I could use it to fill in the gaps for when there is no work available here. SETI could do that, if it worked...

---

Reno, NV
Team: SETI.USA

I also have SIMAP attached. ..snip.. , so I could use it to fill in the gaps for when there is no work available here.

Exactly, but I don\'t use NNW, I have both active and set my resource share high for PS3GRID to 4096 and 1 for SIMAP, so 4096:1 gives 99.98% to PS3 and 0.02% to SIMAP. It will only run 1 SIMAP task, which will put that project over budget and will not run another for a long time, unless PS3GRID runs dry of work and my PS3 runs dry of PS3GRID work.

It has run 60 minutes of SIMAP and now that is waiting and probabbly won\'t finish until it gets near deadline, which is 7 days away, because of my small resource share set.

SIMAP has released some HMMER results, and I was able to download one first try. I should have it returned by the end of the day.

---

Reno, NV
Team: SETI.USA

Exactly, but I don\'t use NNW, I have both active and set my resource share high for PS3GRID to 4096 and 1 for SIMAP, so 4096:1 gives 99.98% to PS3 and 0.02% to SIMAP. It will only run 1 SIMAP task, which will put that project over budget and will not run another for a long time, unless PS3GRID runs dry of work and my PS3 runs dry of PS3GRID work.

Unfortunately, that won\'t work for me. My resource share for both are 999. I think you can go only up to 1000 using BAM (which I do). I guess I could use a different venue for PS3, but I tend to forget which machines are using special venues, and then things get messy.

And besides, SIMAP really isn\'t a solution since they rarely have work. I need someone to get the SETI app working, in order to have a real backup plan.

---

Reno, NV
Team: SETI.USA

I need someone to get the SETI app working, in order to have a real backup plan.

I wonder that your did not get any response at the SETI thread. Eventualy a bump with a reminder that it is a persitant problem could help. - Btw. have you verified, that it crashes on more than one WU ? - Sometimes at SETI it could be that some corrupt WUs are getting around, but I think a crash with a exit code 131 is a application failure/problem...
Also, eventualy helps to ask, if you can get the older version from Guarav which was based on the offical SETI sources.

There is also a native PPC Linux SETI application available (http://www.lb.shuttle.de/apastron/boincDown.shtml#gnuppc), which did not use the SPEs, and thought for normal PPC/Power CPUs at Linux. I think they should work (slow) at the PS3. If not please PM me, that I can contact Stefan.

I hope, that the SIMAP example is a starting point for the other projects, and that they see, that a native Linux PPC application also could used at the PS3.

Btw. is there a documentaion at the web available which describes what must be changed in the sources to make usage of the SPEs ?

Maybe the previous results were noisy, or maybe it is the magical reboot, but in any case, I now have a S@H application running for 2 hours and 20% complete. Fingers crossed.

---

Reno, NV
Team: SETI.USA

Maybe the previous results were noisy, or maybe it is the magical reboot, but in any case, I now have a S@H application running for 2 hours and 20% complete. Fingers crossed.

Success!

http://setiathome.berkeley.edu/results.php?hostid=3593325

It took my PS3 31,977 seconds (8.9 hours), and the wingman 13,306 with a C2D E5300 (1.86ghz), using the stock application.

Now for the problem. I have my setting to switch tasks every 2 hours. BOINC paused the PS3GRID result, crunched the SETI result for 2 hours, then wend back to crunching the PS3GRID result. A day later, the result was up to 30 hours and still going. So I rebooted the PS3 to see if that would make any difference. The clock started over at 0 for the PS3GRID result. When I woke up this AM, the SETI result had been completed, and the PS3GRID result showed about 4 hours of crunch time. I am getting the impression that the PS3GRID application, or the BOINC client doesn\'t play well with others, when it has to switch back and forth.

---

Reno, NV
Team: SETI.USA

There is a fundamental difference between the Cell processor and a normal processor: The spes are not sharable nor are accounted in the boinc CPU time.

If you start a PPE+SPE program and then the PS3GRID application (also PPE+SPE), it would look that it is running but instead it is just polling to see when the 6 SPEs are free. When the first ends then PS3GRID app really starts the calculation, but it may be to late to get credits because the BOINC client saw the PPE at work and summed up CPU time.

This may be the cause of some very long run that you have experienced in the past.

There is a fundamental difference between the Cell processor and a normal processor: The spes are not sharable nor are accounted in the boinc CPU time.

If you start a PPE+SPE program and then the PS3GRID application (also PPE+SPE), it would look that it is running but instead it is just polling to see when the 6 SPEs are free. When the first ends then PS3GRID app really starts the calculation, but it may be to late to get credits because the BOINC client saw the PPE at work and summed up CPU time.

This may be the cause of some very long run that you have experienced in the past.

I\'m not really sure I understand all of that. But I have the BOINC client to \"Run Always\", and the PS3 is doing nothing else except running BOINC inside YDL.

FWIW, the task is back up to over 11 hours now. I am guessing it will take the full 22 hours to run, as if it started over from the beginning. 30 hours down the tube.

If this is always going to happen every time BOINC switches to another task, the concept of \"backup\" projects will not work.

---

Reno, NV
Team: SETI.USA

Yep, it took the full 22 hours, so it started all over again. I guess the checkpointing didn\'t work. That makes the run time over 52 hours. Ugh.

---

Reno, NV
Team: SETI.USA

OK, My SIMAP work unit finished. With a low resource share heres what happened. Resumed SIMAP and got one task from SIMAP, The PS3 task paused and the SIMAP task ran 59:49, paused, went back to running PS3GRID task which finished ok in the 22 hours. That was July 31st. I got several more PS3GRID tasks over the next few days. Then it got near deadline so the SIMAP finished, another 45 minutes. It completed OK and validated. Then my PS3 went back to finishing the PS3GRID task, which also finished in 22 hours and validated. So at least with the SIMAP app which is PowerPC only it does not interrupt PS3GRID work. All mine over the last 10 have completed in about the same time in the range of 79,700 to 80,200 minutes. I do not get any taking 30 hours.

Of course SIMAP only had work for 1 day and is now out until September 1st, so it does not make a good backup project for when PS3GRID runs out.

This weekend I will try the seti app again.

As a further note I saw ResielSieve has a PowerPC app, but I could not figure out how to manually download it and the references in the forum are outdated with non working links. I did not have time to fully pursue or ask anyone on the forum.

Docking at home has a PowerPC app, of course they have no work until they move.

I haven\'t checked many other projects.

I haven\'t seen anything on any other projects about anyone developing specific PS3 apps (PPC+SPE) except for the one on the Seti forum (outside of PS3GRID that is).

@zombie,
Which seti app were you running, regular, enhanced or the one optimized to use SPE\'s mentioned somewhere in the forum ?
Were did you download it from ?
I would like to try the same one and see what happens for me.

I also did not have any problems with the 3 SIMAP results I crunched.

Docking at home has a PowerPC app, of course they have no work until they move.

That PPC Docking app is for Mac/PPC, not Linux/PPC (assuming we are looking at the same thing).

@zombie,
Which seti app were you running, regular, enhanced or the one optimized to use SPE\'s mentioned somewhere in the forum ?
Were did you download it from ?
I would like to try the same one and see what happens for me.

I used the one referenced in the thread over on S@H:

http://setiathome.berkeley.edu/forum_thread.php?id=39910

This line added to get past the SPAM filter. Ignore. This line added to get past the SPAM filter. Ignore. This line added to get past the SPAM filter. Ignore. This line added to get past the SPAM filter. Ignore. This line added to get past the SPAM filter. Ignore. This line added to get past the SPAM filter. Ignore.

---

Reno, NV
Team: SETI.USA

I also did not have any problems with the 3 SIMAP results I crunched.

Thats good.

My point of posting here is no so much that these other applications work, is that they work in co-existance with PS3GRID app without causing problems to PS3GRIS or vice-versa.

That PPC Docking app is for Mac/PPC, not Linux/PPC (assuming we are looking at the same thing).

Oh, You are correct, I had no looked there in a long time and that was just from memory, I remember people mentioning PPC on the forum.

I used the one referenced in the thread over on S@H:

http://setiathome.berkeley.edu/forum_thread.php?id=39910

OK, that is the one I tried before and only got errors from the tasks.

I\'ll try again, maybe something has been fixed.

Stay tuned....

As a further note I saw ResielSieve has a PowerPC app, but I could not figure out how to manually download it and the references in the forum are outdated with non working links. I did not have time to fully pursue or ask anyone on the forum.

I think, it would be worth to put a posting at the forums...

I haven\'t checked many other projects.

There is a table with the different projects, there supported platforms and a lot of more things at the inoffical BOINC wiki :
http://www.boinc-wiki.info/Catalog_of_BOINC_Powered_Projects
http://www.boinc-wiki.info/The_Current_List_of_Alpha_Test/Beta_Test_Projects

I haven\'t seen anything on any other projects about anyone developing specific PS3 apps (PPC+SPE) except for the one on the Seti forum (outside of PS3GRID that is).

I seen a posting from Gaurav at the einstein@home forums, that he would port the einstein@home application for the PS3. - I think, that he would work one PPE+SPE version...
Also, I not knew any other projects which would support the PS3 or PPC/Power on Linux.
But, I am involed in a alpha project, eventualy I can port the application to PPC Linux. But I have the problem, that I have no PS3 and only access to PPC/Power Linux systems and that the PPE+SPE needs heavy source code changes, which is a lot of work...

There is a fundamental difference between the Cell processor and a normal processor: The spes are not sharable nor are accounted in the boinc CPU time.

If you start a PPE+SPE program and then the PS3GRID application (also PPE+SPE), it would look that it is running but instead it is just polling to see when the 6 SPEs are free. When the first ends then PS3GRID app really starts the calculation, but it may be to late to get credits because the BOINC client saw the PPE at work and summed up CPU time.

This may be the cause of some very long run that you have experienced in the past.

Have you forwarded this problem to David Anderson ? - I think he would be very interested at the PS3 at BOINC and would help you to sort this out.

Continued from another thread where we got off topic...

I\'m actually running at PS3GRID = 136.61 and Yoyo = 24 and SIMAP = 1.
I need the 24 for my other computers, but on the PS3 this gave me about 6 tasks a week for PS3GRID (until they got a little longer) and then 19 hours a week for Yoyo. I have not had any problems in the past 2 or 3 weeks running this way. Generally it will run PS3GRID and Yoyo and only fall back to some SIMAP if there is not work available from the others.

I will watch closely, but since my last restart I have 2 completed PS3GRID and the 3rd task running with a 4th ready to go, so it should only be a couple of more days before some yoyo has to run.

I\'ve been running these several months in combination, without any problem.

Do you have, in your general preferences for the venue the PS3 runs under,
Leave applications in memory while suspended?
(suspended applications will consume swap space if \'yes\')
set to NO or unchecked

if NO, this should remove the app from memory when switching occurs and that would have to release the SPE\'s.

Several of the users had this problem until changing above setting, then it went away.

---
Todays findings:

With my current preferences I should get 19 hours of Yoyo per week and the rest PS3GRID. SIMAP is so low that the ones that recently have run should have put its debt over and they should not run again for a long time, unless the other two run out of work.

Since my last restart heres what has happened so far:

*1st PS3GRID task begins
*2nd PS3GRID on Ready to Run
*1st PS3GRID Completed, 2nd PS3GRID Began
*3rd PS3GRID on Ready to Run
*2nd PS3GRID Completed, 3rd PS3GRID Began
*4th PS3GRID on Ready to Run and 1st Yoyo on Ready to Run
*3rd PS3GRID Completed, 4th PS3GRID Began
*After 1:03:03 hours The 4th PS3GRID task suspended and the 1st Yoyo began, this would be correct for a switch applications setting of 60 minutes as I have, as it is the next checkpoint after 60 minutes.
*1st Yoyo completed after 3:11:56 hours, it does not checkpoint as it is a wrapper application, so there would not be a switch until it finished, no percent completed shows. Checking yoyo shows completed status and credit granted.
*4th PS3GRID restarted.
*5th PS3GRID on Ready to Run

Now to see if it finishes, 10:39 hours gone shown on screen, checking log shows it started at 1:10:00 AM and it is now 10:39 AM, so that is correctish 1:03 hour s previously plus 9:39 hours now. The last three took 24:5X hours.

This leaves me 15:49 hours this period, 7 days since restart, of yoyo to run and after the current PS3GRID finishes about 2 more of those unless the long term debt has changed those, I should have reset all those to zero.

Now I\'m confused, that percentage comes out wrong. I can not figure out where I came up with these numbers of 136.61/24/1 gives me 142.01/24.94/1.03 hours per project per week which is not what I wanted as that will not give me 6 PS3GRID tasks a week. I changed something somewhere, or forgot to. Oh it should have been 186 not 136, Can\'t read my own writing I guess or it was late one night.

Recomputing now for current length gives me this combination 205.38/24/1 which gives me 149.76/17.5/0.72 hours per project per week. I\'m trying to complete 6 PS3GRID tasks a week for PS3GRID but need to keep the 24 and 1 shares for the other projects as those are shared with other computers running different combinations of projects, that is why the weird setting for the resource share on PS3GRID.

Check back tomorrow for an update.

Continued...

*4th PS3GRID Resumed.
*5th PS3GRID Ready to run.
*4th PS3GRID Completed. Note 1.
*5th PS3GRID started.
*2nd Yoyo downloaded and Ready to Run.
*6th PS3GRID downloaded and Ready to Run.
*5th PS3GRID stopped, ran from 12:55:51 AM to 1:57:03 AM, again just over 1 hour.
*2nd Yoyo started and ran to completion of 1:16:41 hours.
*5th PS3GRID resumed and still running, now at 06:38:00 hours CPU time.

Note 1: So the 4th PS3GRID completed successfully. This was run 1 hour, then removed from memory to run the 1st yoyo, then resumed where it ran in full to completion. The completion time was in the normal range 1 day 00:50:46 hours which is within a dozen seconds of the previous ones time completed.

Checking results logs on yoyo, shows no errors, but they just cleared November off and I missed recording it. When I viewed it a few days ago, I saw no errors. I have about 70 yoyo recorded for October, all success.

In this time of two months, there has been no hung work, every task switch is normal, with one exception but that was when I was moving from Harddrive to USB drive to flash card and back to USB drive. I attributed that to my fooling around.

Total run so far PS3GRID 4 completed at 99:30:50 hours(out of 6 for a week) and 4:28:37 hours of yoyo (out of 17.5 hours estimated for a week). This equates to 103:59:27 hours total so far. Percentage PS3GRID 95.69%, Yoyo 4.31%. BOINC shows my current shares at PS3GRID 89.14% and Yoyo 10.42%. So PS3GRID is a little high and yoyo needs to catch up. There are 64:00:33 hours left in the week (since restart). It looks correct to me, the long term debt is coming into effect some, hence the higher percent on PS3GRID. I also changed my share during this, so that may have thrown it off a little, but that only took 7 hours off of yoyo\'s share and gave it to PS3GRID.

I have not had any problems with the applications of either project \'hanging\' (that is running and showing no progress) as others have reported. Mine seems to be running and switching between the two without problem and at the correct interval per scheduler and checkpointing rules.

Continued...

*5th PS3GRID Completed 1day 00:50:25 hours, again this was a split run.
*6th PS3GRID started, ran 01:01:07 hours then suspended.
*3rd Yoyo stated and ran to completeion at 2:46:20 hours.
*6th PS3GRID restarted and now running.
*7th PS3GRID downloaded and Ready to Run.

Still switching and running smoothly.

Total for period:
PS3GRID 5 completed at 124:21:15 hours (out of 6 for a week) 95.94%
and
Yoyo is at 7:14:57 hours (out of 17.5 hours estimated for a week) 5.59%
---
Total 129:36:12 out of 168 for a week.

PS3GRID is still over percentage, but Yoyo has caught up some from 4.31% to 5.59% (should be at 10.42%).

So by my estimate after this 6th PS3GRID finishes, Yoyo should run for about 10 hours of work. Now I\'m guessing that since I already have 1 PS3GRID ready to go, the same thing that happened on the last two will happen, and that it will start, I will get some work from yoyo that will be ready to run, and then after 1 hour the PS3GRID will suspend and yoyo will run. The question will be will it only run 1 yoyo or will it run more to catch up ?

Now since I didn\'t really reset the LTD to zero when I started, according to BOINC it may already be within range.

Stay tuned, PS3GRID task #6 ends in 12 hours.

(This is presented as a sample to new users on how BOINC works with multiple projects and resource shares).

Thanks KK. All these notes are with \"leave in memory\" set to no? Can you please confirm?

---

Reno, NV
Team: SETI.USA