Author |
Message |
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
Since there isn't an official BOINC client for the PS3, you will need the BOINC client files for the PS3 from a 3rd party source. I've made a file here:
http://www.megaupload.com/?d=NPZZ5TX7
Download, and extract. Instructions are included. Please read, there is important information about not pausing/suspending workunits, and the minor issue with random-cpu benchmarking.
The main difference between this client and the PS3Grid/GPUGrid client is this one is setup to use 6 cpus (SPEs in this case). PS3Grid's client was locked to a lower number when it was compiled.
Note that for some reason, BOINC's cpu benchmark test may fail, don't worry, this will not change anything credit-wise.
Please use this thread for anything PS3 related, or PM me if necessary.
Thanks,
Bryan Little |
|
|
|
I had try this version on my PS3.
But impossible to crunch some WU.
All the WU are going in error.
Here is the message i got in the task
"<core_client_version>6.2.19</core_client_version>
<![CDATA[
<message>
process exited with code 127 (0x7f, -129)
</message>
<stderr_txt>
primegrid_ap26_1.00_powerpc64-ps3-linux-gnu: error while loading shared libraries: libstdc++.so.6: cannot open shared object file: No such file or directory
</stderr_txt>
]]>"
The directory "lib" is present, so i don't understand. I read the readme file, but nothing else to do after installing the client.
I try with my old install, and a new install.
I need some help! |
|
|
|
....
primegrid_ap26_1.00_powerpc64-ps3-linux-gnu: error while loading shared libraries: libstdc++.so.6: cannot open shared object file: No such file or directory
...
same problem here, BOINC client is working (also is working on yoyo's OGR) but AP26 WUs end with the same error
I'm not linux addicted, only few experience |
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
Ok a small note was at the bottom of the readme file about you may need to ldd the binary files. You need the 64bit C++ Library installed (I have this installed by default because the distro was installed as "development install").
In your package manager, add/remove software, search for libstdc++. You should see 3 files, install the first two. (GNU standard C++ library) The main one you need is the ppc64 one for 64 bit support.
Future versions may be linked statically to avoid this issue. |
|
|
|
You need the 64bit C++ Library installed ...
thanks mfl0p, later in the afternoon I will try ;)
NOTE: Since GPUgrid server was not reachable, I downloaded 6 new yoyo OGR WUs. Well, 5 of them started immediately to crunch (total of 6 WU running), then the running GPUgrid WU ended with computational error, then the sixth OGR WU started.
Now I have 6 yoyo's OGR WU crunching... I'm curious about the way they will end, error or not :-D |
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
Let me know how it works out for you.
Also, I would recommend only using this boinc with cc_config.xml saying 6 cpus with PrimeGrid.
This is because yoyo and GPUgrid workunits are made to be split up into pieces and run on the SPEs.
I decided to run a workunit on each spe because they are small enough.
So maybe have a directory for each boinc, and manually switch between the two to run other projects for now. |
|
|
|
Let me know how it works out for you.
libstdc++ for ppc64 installed: it works !
"it works" means: I have 6 PrimeGrid WUs running without errors at the moment (1 hour crunching)
thx, mfl0p
P.S. 6 OGR WUs ended without errors, the last one running together with PrimeGrid WUs. I saw no improvement in OGR performances. |
|
|
|
Still working fine 24h/24
This is my PS3
some strange "error while computing" yesterday 16/07 but I thik they were related to wireless problems (PS3 was ... frozen... for a while), set communication interval= 5 days
RAC = ? - 1000 I suppose |
|
|
|
So far so good, didn't have to install any additional packages either, crunching of 6 AP26s has started, benchmark did not create any errors - looking perfectly fine so far. I'll keep you updated once the first batch has completed.
Really great work, mfl0p, thank you very much!
Edit: Would it make sense to also use the two PPEs for crunching if you're not doing anything else on your PS3? If so, how would I do that, just change the number of CPUs to 8 rather than 6?
____________
|
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
Using a PPE thread to crunch a workunit was the original plan, however until I find a good way to determine all SPEs are busy without special OS privileges given to the app, we're only using the 6 SPEs and part of the PPE at this time. The Cell processor's design intent was to have the PPE dispatch work to the SPEs, not actually do a lot of calculations itself.
Also, the PPE shows as two processors to the OS, but it's similar to Intel's HyperThreading, so maxing two threads on the PPE may be no faster than just one thread. I haven't tested this yet.. I'm already using the PPE for certain calculations the SPE is very slow at. Therefore, if I speed up the SPEs enough, the PPE will be loaded pretty good with no need for a new, dedicated crunching thread.
Increasing the # of cpus in the boinc config file will only cause each workunit to run much slower, the OS will be trying to schedule 8 spu threads on only 6 SPEs. Kinda like running 8 threads on a Q6600 machine, no advantage.
The PS3 app is a work in progress, expect some changes in the future to increase speeds.
Edit: Would it make sense to also use the two PPEs for crunching if you're not doing anything else on your PS3? If so, how would I do that, just change the number of CPUs to 8 rather than 6?
|
|
|
Lumiukko Volunteer tester Send message
Joined: 7 Jul 08 Posts: 164 ID: 25183 Credit: 716,865,909 RAC: 182,289
                          
|
Looking fine here too. First 6 wu's completed and validated.
Tasks for my host are here.
It is PS3 with Ubuntu 9.04 installed to external (USB) disk.
Did take some time before I realised that the libspe2-library that I had was not the same as the ldd claimed "not found". It seems Ubuntu does not deliver the 64bit version at all, only 32bit. So I installed it from
Fedora rpm with "alien", and got everything finally running.
Speed is about 100min "wall time"/"6 wu's".
Thanks mfl0p, great work! Finally something useful to do with my PS3.
--
Lumiukko |
|
|
|
Thanks mfl0p for the explanation.
I can now gladly report that also my WUs have validated and everything is fine. I confirm around 100min/6WUs. Excellent work, thanks again!
____________
|
|
|
|
Any advice for me?
Yellow Dog, using the 6.2.19 manager linked here, GPUGrid runs fine but I'm getting the errors caused by the missing library when AP26 runs. Unfortunately, I'm not missing the library, the package manager shows it installed. I have no idea where they're at, so even if it's something as basic a pathing issue I'd still need some guidance.
http://www.primegrid.com/result.php?resultid=114966934
Appreciate it. :-)
PS: Posted from my PS3 haha. |
|
|
Lexs Volunteer developer
 Send message
Joined: 16 Mar 08 Posts: 61 ID: 20289 Credit: 49,033,000 RAC: 0
               
|
You need to additionally install the 64bit version of the libstdc++ library.
By default only the 32bit libstdc++ is installed. |
|
|
|
I read that above as well, but don't know how to address it. There is no information on the package in the manager, just that it's installed. No options or details. So... needs to be done via command line, but I'm not sure how to go about that on YDL.
I'm not afraid of reading, so if anyone has a link or a doc just send 'em my way. :-) |
|
|
Lexs Volunteer developer
 Send message
Joined: 16 Mar 08 Posts: 61 ID: 20289 Credit: 49,033,000 RAC: 0
               
|
On command line do:
sudo yum install libstdc++.ppc64 |
|
|
|
Thank you muchly. Spent quite a bit of time reading other people complaining about dependencies, the failings of RPM and general crankiness without getting any closer to finding the actual package name to use. ldd now reports a clean bill of health. Thanks, will see about actually making them work now heh. |
|
|
|
Hi all,
A fellow team member has tried running this on his PS3 and got the following error:
<core_client_version>5.10.6</core_client_version>
<![CDATA[
<message>
process exited with code 193 (0xc1, -63)
</message>
<stderr_txt>
Unrecognized XML in parse_init_data_file: fraction_done_update_period
Skipping: 1.000000
Skipping: /fraction_done_update_period
SIGSEGV: segmentation violation
Stack trace (8 frames):
primegrid_ap26_1.00_powerpc64-ps3-linux-gnu[0x1000bb8c]
[0x100418]
/usr/lib64/libspe2.so.2(_base_spe_context_run-0x14804)[0x4000002602c]
/usr/lib64/libspe2.so.2(_event_spe_context_run-0x14464)[0x400000263dc]
/usr/lib64/libspe2.so.2(spe_context_run-0x1c9e4)[0x4000001d5bc]
primegrid_ap26_1.00_powerpc64-ps3-linux-gnu[0x10003cf4]
/lib64/libpthread.so.0[0x80fd37bd08]
/lib64/libc.so.6(clone-0xa825c)[0x80fd19f2ac]
Exiting...
</stderr_txt>
He has tried doing: sudo yum install libstdc++.ppc64
In which the answer is: "Package libstdc++-4.1.2-33.ppc64 already installed and latest version"
Could anyone offer some insight into why this could be happening? Many thanks |
|
|
Lexs Volunteer developer
 Send message
Joined: 16 Mar 08 Posts: 61 ID: 20289 Credit: 49,033,000 RAC: 0
               
|
I guess you're using an old PS3GRID boinc client 5.10.6, try a newer one with SPE support:
http://www.dotsch.de/boinc/boinc6219_10.linux-ps3.tar.gz
don't forget to add cc_config.xml
<cc_config>
<options>
<ncpus>6</ncpus>
</options>
</cc_config>
|
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
you can ignore this:
Unrecognized XML in parse_init_data_file: fraction_done_update_period
Skipping: 1.000000
Skipping: /fraction_done_update_period
But this on the other hand:
SIGSEGV: segmentation violation
Stack trace (8 frames):
primegrid_ap26_1.00_powerpc64-ps3-linux-gnu[0x1000bb8c]
[0x100418]
/usr/lib64/libspe2.so.2(_base_spe_context_run-0x14804)[0x4000002602c]
/usr/lib64/libspe2.so.2(_event_spe_context_run-0x14464)[0x400000263dc]
/usr/lib64/libspe2.so.2(spe_context_run-0x1c9e4)[0x4000001d5bc]
primegrid_ap26_1.00_powerpc64-ps3-linux-gnu[0x10003cf4]
/lib64/libpthread.so.0[0x80fd37bd08]
/lib64/libc.so.6(clone-0xa825c)[0x80fd19f2ac]
That's a pretty nasty error you are getting, what Linux distro? I would recommend using the newer version of BOINC, posted at the top of this thread, and only attaching to PrimeGrid project.
If I was still getting that kind of error i'd probably just format and reinstall the OS. Something very bad has happened along the way. I'd reinstall with an OS like YellowDog, 6.1+ with support for the PS3 already bundled with the distro.
Or, very unlikely, the PS3 may have a hardware problem. |
|
|
|
Next dumb question of the moment! It's up and running but something is very not right. I see 6 WUs marked running, all guestimating 3-5 minutes to completion. Unfortunately the CPU time is incrementing about 1 second a minute. Complete WUs say 180 seconds CPU time in the results here, but took 8 hours!
I checked top and there was a buuuuuunch of orphaned junk still running (dnetc) and the AP26 apps were only getting 2% CPU, so I happily kill -9'd all the dnetc stuff. Unfortunately now top says the system is 80% idle, and all the AP26 apps are still only sitting at 2%. What in the world is wrong with it now?
Should I kill and restart the client/apps?
Here's an example WU, all are similar:
http://www.primegrid.com/result.php?resultid=115528164
Sent 21 Jul 2009 8:48:21 UTC
Received 22 Jul 2009 3:49:24 UTC
CPU time 176.5884 |
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
The AP26 PS3 app uses mostly the SPEs, so using 'top' will only show PPE cpu usage, which is about 2% per thread. To view SPE usage, you can use 'spu-top', and all six SPEs should be busy. However, it should only take about 100 minutes to process 6wu. So make sure something else isnt hogging the SPEs.
Time to completion and final cpu time is wrong in BOINC because it only sees what the PPE thread is doing. BOINC doesn't know what a SPE is :) This is similar to what MilkyWay@home's ATI gpu app was showing, around 8 seconds per workunit when it first came out. |
|
|
|
Thanks to Lexs and mfl0p for your answers :) I have posted this onto 2 of our members who where having problems with it. They are both running Fedora (one is using 5, and the other is using 6).
I highly suspect it was due to the OS that was causing the problem, however it could also be the boinc client as well. now that you mention it. I'll get back to you when i know more.
Thanks again |
|
|
RytisVolunteer moderator Project administrator
 Send message
Joined: 22 Jun 05 Posts: 2649 ID: 1 Credit: 29,840,958 RAC: 165,494
                     
|
A new version has been released for public.
____________
|
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
The new version 1.01 is much faster, expect the PS3's RAC to more than double (will be around 2500).
The PS3 will complete 6 workunits in parallel in 45minutes.
A lot of modulo code was changed to spu 8x vector addition/subtraction/compare to achieve this speedup.
The 64bit C++ library is now linked static, so that shouldn't be an issue anymore for some users who don't have it installed by default.
A reminder, suspending boinc/workunit will cause the workunits to error with a signal 11. To avoid this, just exit boinc, then restart boinc when you want to continue. This has to do with the two threads being used for PPE/SPE. I hope to fix this in the future. |
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
A new version 1.02 is coming soon.
Speed is right at 12 workunits an hour!
Almost as fast as my linux Intel quad (about 16wu an hour). And the PS3 only pulls 90watts from the wall while running (65nm cpu version).
RAC should be over 3000, my test machine has been climbing every day, about 2850 right now.
You will notice the PPE thread cpu usage has increased a bit, this is because i've moved the slow 64bit modulo calculations to the PPE, which is better suited to this kind of work.
____________
|
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
Version 1.02 is live.
Anyone interested in parallel computing processors such as cell/gpgpu/clusters should check out this interesting wiki article: http://en.wikipedia.org/wiki/Amdahl%27s_law
It explains Amdahl's Law, which may help you better understand why I'm treating the PS3 as a 6 CPU computer, instead of a single CPU like other BOINC projects. The AP26 workunit has 3 test ranges, which theoretically I could run one range on each of 3 SPEs. This would allow computation of 2 workunits in parallel instead of 6. However, this is where Amdahl's Law comes in, each of the 3 AP26 test ranges take a different amount of time to compute. So, say 2 ranges are pretty fast, but the third range is a bit slower, so the other 2 SPEs are idle while the last range of the workunit completes. There are also portions of the code that cannot be computed in parallel at all.
We can avoid all of this wasted, unnecessary idle time by treating the PS3's CELL as a 6 CPU computer, even though it's only a single CPU. This can be done because of the small workunit size for AP26. Other projects like some LLR workunits that take days to complete on a modern Intel Core CPU would be good candidates to split into pieces and run one workunit at a time on the CELL.
____________
|
|
|
|
I can confirm that the client runs happy on the QS20 Cell Blade.
This platform has 2 full Cell cpus with total of 16 SPU available.
My system runs on Fedora Core 9 , kernel 2.6.30.5.
|
|
|
|
cool tool!
What would be the maximum RAC a PS3 or a Cell Blade could achieve? |
|
|
Lexs Volunteer developer
 Send message
Joined: 16 Mar 08 Posts: 61 ID: 20289 Credit: 49,033,000 RAC: 0
               
|
My current PS3 RAC running AP26 is 2927 (23days uptime running boinc/primegrid)
EDIT: I forgot about the recent hot days: so 3 of those days where limited to 12 hours crunching only)
____________
|
|
|
|
not bad for a 280 Euro computer - should be in the region of a C2D with 2400 MHz. |
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
cool tool!
What would be the maximum RAC a PS3 or a Cell Blade could achieve?
PS3 RAC about 3400 now, has been climbing since 1.02 app release.
root's QS20 Cell Blade would get a bit over 9000 RAC... however the blade alone costs about USD $19000 so one would much rather run fast Cell programs like single precision fpu intensive algorithms on that computer.
____________
|
|
|
|
There is comparably not that much software for Cell platform around, yet...
mfl0p, can You suggest some ?
I run mostly dstributed.net's RC5-72 for now ( no fp just int )
P.S. finally somebody else than my machines called me root ;) |
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
For distributed computing projects there are very few for the Cell that do useful things, which is the main reason I ported AP26 to the platform.
I'm sure you did this already, but being that the cc_config.xml is where you tell BOINC how many SPUs you have, did you change it to 16? I don't know if the boinc client will take a number that high. Would be interesting to know.
Distributed.net's client works very well for RC5 and OGR, they have a talented group of programmers for many exotic platforms. However, being that my ATI Radeon GPU does 435Mkeys/sec on RC5, it makes even the Cell look slow :) And, it seems like it will be a LONG time until the key is found at current rates.
____________
|
|
|
Lexs Volunteer developer
 Send message
Joined: 16 Mar 08 Posts: 61 ID: 20289 Credit: 49,033,000 RAC: 0
               
|
not bad for a 280 Euro computer - should be in the region of a C2D with 2400 MHz.
Remind that the new slim PS3 can't run linux.
So catch the old PS3's as long as they're in stock.
____________
|
|
|
RytisVolunteer moderator Project administrator
 Send message
Joined: 22 Jun 05 Posts: 2649 ID: 1 Credit: 29,840,958 RAC: 165,494
                     
|
However, being that my ATI Radeon GPU does 435Mkeys/sec on RC5, it makes even the Cell look slow :) And, it seems like it will be a LONG time until the key is found at current rates.
Wait, is there finally a public ATI client for RC5? Or are you just stating the rate of that pre-beta client?
____________
|
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
However, being that my ATI Radeon GPU does 435Mkeys/sec on RC5, it makes even the Cell look slow :) And, it seems like it will be a LONG time until the key is found at current rates.
Wait, is there finally a public ATI client for RC5? Or are you just stating the rate of that pre-beta client?
This was public beta 7 on Linux.
____________
|
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
Some info about the new PS3 firmware 3.0 update. There was some confusion about PS3 firmware 3.0 disabling otheros/linux support for all PS3s. This is not true.
Non-slim PS3s can update to 3.0 and still run otheros/linux with no problem. I updated one today.
The new redesigned PS3 (slim) with 45nm cpu CANNOT run otheros/linux at all, regardless of firmware version.
____________
|
|
|
|
The boinc client is happy with 16 clients in cc_config.xml on the QS20 cell blade.
This blade also does the same 435 Mkeys/s with distributed.net client.
and yeah Your are right, the key may never be found... meaning even applying Moore's law of computational power increase, if its found after I R.I.P. does it matter to me ?
It doesn't.
P.S. some say the goal is to chase the rabbit, not to catch it ;-)
|
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
Version 1.03 released today.
The main reason for the new version is to update the memory flow controller code to IBM specifications to ensure proper operation of the application. This is basically a "double check" operation in the memory copy code between application threads.
All MFC code was updated. As a result, PPE thread cpu usage % has dropped a bit due to the more efficient code. The workunits will only be about 21 seconds faster on average than version 1.02 so don't expect much of a change in RAC. The reason for the slight increase in speed is due to less MFC command overhead in this version.
Best of luck finding more APs!
-Bryan Little
____________
|
|
|
|
Hi, I noticed that work for ap26 looked very low. Will this client work well with having two projects, for example this and yoyo. Has anyone tested? I would not want my machines run dry.
GPU-grid (former ps3grid) is closing down for ps3, so I have relocated. Great that this client exists! |
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
Re: AP26 work, the server will generate AP26 workunits very fast when work gets low enough. And, at this time, there is not any shortage of ranges to check with AP26 :) So there is plenty of work.
Going against BOINC's original design intent, I would not use this client with multiple projects, as yoyo's wrapped distributed.net application will try to take control of the PPE and all six SPEs with only one cruncher running. The AP26 application only uses one SPE per cruncher, that's why the client is setup for 6 cpus. I found this the best way to run the program, as none of the SPEs are idle if one workunit finishes before the other.
One thing to remember, since BOINC doesn't natively detect the SPEs, and now the PS3-Slim no longer supports Linux, I do not think many more PS3 or BOINC client applications will be made. BOINC wasn't originally designed for the PS3 and GPU applications, it gets confused since the actual processor (SPE or GPU) isn't being monitored directly like a CPU app.
Hi, I noticed that work for ap26 looked very low. Will this client work well with having two projects, for example this and yoyo. Has anyone tested? I would not want my machines run dry.
GPU-grid (former ps3grid) is closing down for ps3, so I have relocated. Great that this client exists!
____________
|
|
|
|
Thanx for the fast reply.
Ok, I will only run ap26, and hope this project lasts a long time. Yes, I too suppose that sonys decision, to cut out Linux, deals death to any plans for new or improved clients for the ps3. Hopefully this is not true for this client, and that it will get improved and supported as long as possible. There are a lot of numbers after 26. :-)
I guess you must have noted an increase of ps3:s, as a result of gpugrids decision to stop with ps3? I would rather see all the ps3:s do work, rather than just collect dust.
|
|
|
|
I've moved from Ps3grid here with 3 QS20 blades. Which client should I run to use all available SPE's on each blade?
Thank you!
--
Konstantin
|
|
|
|
this one from the first post, you will use 6 spe |
|
|
Lexs Volunteer developer
 Send message
Joined: 16 Mar 08 Posts: 61 ID: 20289 Credit: 49,033,000 RAC: 0
               
|
According to Message 17879:
The boinc client is happy with 16 clients in cc_config.xml on the QS20 cell blade.
So 48 in your case ?
____________
|
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
I've moved from Ps3grid here with 3 QS20 blades. Which client should I run to use all available SPE's on each blade?
Thank you!
--
Konstantin
You should run the client in post #1, but change the cc_config.xml number of cpus to 16 for a QS20 blade with 16 SPEs
____________
|
|
|
|
Hi,
Since GPUGrid closed for PS3, I joined Primegrid. (so about a week now);
I got everything running, as described in this thread, but yesterday my PS3 encountered an error, and the BOINC client stopped running.
I'm not really sure if here's the right place to post it, but maybe someone else saw this kind of behaviour too?
(fyi: after restarting the boinc_client, everything is running smoothly again, ... we'll see for how long...)
Last bit of log from the client:
06-Oct-2009 01:26:53 [PrimeGrid] Finished upload of ap26_15838085_15838088_0_0_0
06-Oct-2009 01:27:02 [---] Contacting account manager at http://bam.boincstats.com/
06-Oct-2009 01:27:04 [---] Account manager: BAM! User-ID: 72641
06-Oct-2009 01:27:04 [---] Account manager: BAM! Host-ID: 187069
06-Oct-2009 01:27:04 [---] Account manager: This host connected 9 times to BAM!
06-Oct-2009 01:27:04 [---] Account manager contact succeeded
06-Oct-2009 01:27:04 [---] Attaching to http://setiathome.berkeley.edu/
SIGSEGV: segmentation violation
Stack trace (8 frames):
./boinc_client[0x10052a5c]
[0x100344]
./boinc_client[0x10006ff8]
./boinc_client[0x10035428]
./boinc_client[0x10014e00]
./boinc_client[0x10040068]
/lib/libc.so.6[0xfe5dc0c]
/lib/libc.so.6(__libc_start_main+0x144)[0xfe5de34]
That segmentation violation (-> attempt to use a memory address that doesn't belong to the process trying to access it) could be caused by:
- a programming error
- mismatching library and/or binaries
So far my troubleshooting... (read: looking up on the net)
Any ideas/hints on how to prevent this?
Thx,
Loe |
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
Hi,
06-Oct-2009 01:27:02 [---] Contacting account manager at http://bam.boincstats.com/
06-Oct-2009 01:27:04 [---] Account manager: BAM! User-ID: 72641
06-Oct-2009 01:27:04 [---] Account manager: BAM! Host-ID: 187069
06-Oct-2009 01:27:04 [---] Account manager: This host connected 9 times to BAM!
06-Oct-2009 01:27:04 [---] Account manager contact succeeded
06-Oct-2009 01:27:04 [---] Attaching to http://setiathome.berkeley.edu/
SIGSEGV: segmentation violation
Stack trace (8 frames):
./boinc_client[0x10052a5c]
[0x100344]
./boinc_client[0x10006ff8]
./boinc_client[0x10035428]
./boinc_client[0x10014e00]
./boinc_client[0x10040068]
/lib/libc.so.6[0xfe5dc0c]
/lib/libc.so.6(__libc_start_main+0x144)[0xfe5de34]
This is an error in the boinc client, not the application binary. I have not run into this error, but it may be related to this:
Contacting account manager at http://bam.boincstats.com/
Any particular reason you're not directly connecting to PrimeGrid?
____________
|
|
|
|
No particular reason no; just the habit of using the account manager to handle all
projects I participate in...
I was wondering if it had to do with trying to get SETI to work in this client...
As right after this message
06-Oct-2009 01:27:04 [---] Attaching to http://setiathome.berkeley.edu/
the client seems to have crashed...
Well, I'll keep an eye on it anyway... client is running ok for 12hours now.
Thx,
Loe |
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
Application version 1.04 released today. Changes:
* Application is about 4 minutes faster per workunit than 1.03. PPE thread CPU usage will be about the same as 1.03. Avearge workunit time is about 27 minutes. Remember the PS3's Cell processor does 6 wu in parallel, so another way to look at the speed is the single PS3 Cell CPU is finishing a workunit in about 4m 30s. This time would be about 1m 42s on Cell Blades with 2 processors (16 SPEs total).
Expect a RAC over 4000 for the PS3 now.
* Application operation has been changed to work like the other x86 apps, you can start a manual search with the command line or with an ".ini" file. The PPE is handling the search range input and sending the info the the SPE via MFC. Previous versions were using a temporary file to transfer the search range data to the SPE.
* And now, the boring technical stuff. I previously thought the AP26 on Cell was about as fast as it was going to get, but I proved myself wrong again with this version. Increases were found by removing two vectors (and about 18 related vector operations) in rarely used areas. Also, by changing the 64bit OKOK arrays to vectors, they are naturally aligned to a quadword boundary for faster loads. But this required more memory on the SPE, since it effectively doubled the size of the OKOK arrays. More LS was found by moving the OK337-OK541 arrays out of the SPE totally. Version 1.03 setup the arrays and sent them to the PPE via MFC, wasting SPE LS. Version 1.04 creates the arrays in the PPE. And, finally, the global OK61 - OK331 char arrays were isolated to a function on the SPEs as temporay arrays to setup the OKOK vectors.
Bryan Little |
|
|
|
Out of curiosity, did credits for 1.04 get fixed back to 12.61/wu instead of 13.61/wu? I'm seeing a lot of 12.61 credited tasts in my completed WU list and want to clarify... |
|
|
John Honorary cruncher
 Send message
Joined: 21 Feb 06 Posts: 2875 ID: 2449 Credit: 2,681,934 RAC: 0
                 
|
Out of curiosity, did credits for 1.04 get fixed back to 12.61/wu instead of 13.61/wu? I'm seeing a lot of 12.61 credited tasts in my completed WU list and want to clarify...
Yes, that is the case. With the recent speed improvement of the AP26 apps, we needed to make an adjustment. We'll soon be testing "doubling" the WU. And yes, credit will be doubled as well. :)
____________
|
|
|
|
i saw that all wu reported today (25th) get 24 granted credit, so it almost 100% more, is this will be normal now or its only test or something?
You wrote that we cen see rac over 4000 but in this case it can be over 6 or even 7k per day. |
|
|
Scott Brown Volunteer moderator Project administrator Volunteer tester Project scientist
 Send message
Joined: 17 Oct 05 Posts: 2181 ID: 1178 Credit: 9,078,854,925 RAC: 13,113,590
                                      
|
Looks like there has been an error somewhere. Doubled credit, but work times are unchanged.
____________
141941*2^4299438-1 is prime!
|
|
|
John Honorary cruncher
 Send message
Joined: 21 Feb 06 Posts: 2875 ID: 2449 Credit: 2,681,934 RAC: 0
                 
|
Looks like there has been an error somewhere. Doubled credit, but work times are unchanged.
No "real" error...just an overlap in transition. We doubled the credit a bit early. If you haven't already, you'll soon notice that WU's are now doubled.
____________
|
|
|
|
I noticed it now, i swapped over this morning while being frustrated not having found one lousy prime since the last challenge in ppsLLR and was pleasantly surprised by the credit throughput... ;)
But it is still very high.
____________
|
|
|
|
I hope somebody will use some lines to explain me how to get this working. As I am sure you will find out shortly Linux is not my way of life yet :-)
I have two PS3 with YDL 6.2 which I have used for dcnet and would like to try on this project, but my Linux knowledge stops me.
I downloaded BOINC client for the PS3 from here: http://www.megaupload.com/?d=NPZZ5TX7 and put on the desktop. I right click on boinc6219_10.linux-ps3.tar.gz and choose Exctract here.
I got a folder called opt and a subfolder called boinc
Inside boinc I have folders: doc, lib and projects
Inside boinc I have files: boinc_clinet and boinc_cmd
I opened a Terminal and typed gedit .bash_profile and edit file to be:
# .bash_profile
# Get the aliases and functions
if [ -f ~/.bashrc ]; then
. ~/.bashrc
fi
# User specific environment and startup programs
PATH=$PATH:$HOME/bin
export PATH
LD_LIBRARY_PATH=/opt/boinc/lib ; export LD_LIBRARY_PATH
Save and close, then double clicked on boinc_client
I get 4 new files in folder but nothing more seems to happend:
client_state.xml
gui_rpc_auth.cfg
lockfile
time_stats_log
I expected some graphical user interface as I am used to from Windows...
So I added cc_config.xml as suggested somewhere in this forum, but no luck.
Then I deleted the 4 files made, rebooted and played a bit more with the .bash_profile file as put in doc\install i.e. my understanding of how to do it:
# .bash_profile
# Get the aliases and functions
if [ -f ~/.bashrc ]; then
. ~/.bashrc
fi
# User specific environment and startup programs
PATH=$PATH:$HOME/bin
LD_LIBRARY_PATH=/opt/boinc/lib:$LD_LIBRARY_PATH
export LD_LIBRARY_PATH
export PATH
but still not luck :-(
I tried to install the 64 bit library mentioned:
sudo yum install libstdc++.ppc64
but got an error saying: Roald is not in the sudoers file
So I ran it with su
I got the library installed.
But still the same happends with 4 files generated and nothing more
So obviously I miss something.
I wonder if I need to to the rest in doc\install - it seems to be all about SETI...
Thanks in advance!
|
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
First, download the client in post #1.
Use the command line in a terminal to start the boinc client. You don't have to do anything with libraries or worry about libstdc++ with YDL 6.2. So really it's easier than you are thinking :)
The most easy way would be to extract the .tar to a directory, then open a terminal and go to the boinc directory. Then run:
./boinc_client
Then open another terminal, and go to the boinc directory and run:
./boincmgr
This will first start the main boinc client program, then the manager will attach to it so you can configure with a GUI. After you are done configuring, I suggest closing the boincmgr GUI because it uses quite a bit of CPU time on the PS3. Keep the other termial running (./boinc_client). When you want to stop boinc, just ctrl-c in the terminal window running the ./boinc_client.
If you really want to keep the GUI open, start it with command line:
nice ./boincmgr
this will keep it from hogging the CPU too much.
There are other ways to run the client, such as under a private user account, via "screen", as a daemon, and setting up via boinc_cmd but for the average PS3 user the "easy" way is the best I can think of. |
|
|
|
Thanks for you reply
Re:
Then run:
./boinc_client
--> Works ok :-)
Then open another terminal, and go to the boinc directory and run:
./boincmgr
--> Do not work, I only have
boinc_client
and
boinc_cmd
in that folder.
Should I dl some more or is my extraction faulty somehow? |
|
|
|
Some time ago I tested crunching yoyo and GPUgrid.
I still had the extracted files on desktop.
I copied the boincmgr file I had in that folder. Seems to work as I got 25 tasks.
Only running two in parallell. |
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
--> Do not work, I only have
boinc_client
and
boinc_cmd
in that folder.
Should I dl some more or is my extraction faulty somehow?
you said in a prior post you downloaded "boinc6219_10.linux-ps3.tar.gz"
This is not the file in post #1 at http://www.megaupload.com/?d=NPZZ5TX7 it is named "ps3-boinc.tar.gz"
Try downloading that version, it has all the files.
____________
|
|
|
|
I have to blame my non existing Linux skills for that, maybe in combination with use of Firefox - as I'm Norwegian, I'm more used to Opera :-).
I thought the dl file ended up on my desktop. Apparently not. I now have dl the linked file on a Windows PC and unzipped it, and I see it's a different version then the one I played with yesterday night (Norwegian time).
I'll sort this out when I'm home from work.
Thanks for pointing my mistakes out for me! |
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
I have identified a minor bug in the PS3 application.
Rarely, a workunit may hang after restarting boinc-client and resuming computation. This is due to the SPE handling the checkpointing. If the SPE was in the middle of a checkpoint write when the boinc-client was killed, then the state file may be blank upon resume. This will cause the task to appear to be running in boinc, but only 5 threads will actually be running.
I will fix this by moving the checkpoint code to the PPE thread. This way checkpointing will be safe inside a boinc_critical_section, and this will not be an issue.
Until I do this, if you find a hung workunit just abort it, and things will continue as usual.
____________
|
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
The checkpointing hang bug has been fixed in version 1.05 released now.
Also, the application will now use your boinc config settings on checkpoint/result write to disk time intervals. Previous apps wrote to disk whenever a result or checkpoint was reached while the program was running. One can imagine disk operation was very frequent.
The results are now buffered between checkpoints, so there will not be any duplicate results reported if boinc is stopped and resumed later (this could have happened in previous versions due to result file being updated before next checkpoint).
Checkpoint interval is now 7 times per K, or about every 1m 17s. This was changed from default 85 times per K because on PS3 every CPU is running at the same speed, so such a frequent checkpoint is not needed.
The only remaining issue I am aware of is the signal 11 error on all running workunits during a cpu-benchmark or if you suspend computation. This has been a "bug" since the first release. So avoid suspending or running cpu-benchmarks, if possible. I have tried a few things to avoid this problem, but was not successful.
____________
|
|
|
Lumiukko Volunteer tester Send message
Joined: 7 Jul 08 Posts: 164 ID: 25183 Credit: 716,865,909 RAC: 182,289
                          
|
Version 1.05 erroring all workunits in my PS3: exceeded disk limits (I have 100GB reserved for Boinc in preferences):
Sun 01 Nov 2009 09:09:30 PM EET|PrimeGrid|Starting ap26_17795886_17795896_0_0
Sun 01 Nov 2009 09:09:30 PM EET|PrimeGrid|Starting task ap26_17795886_17795896_0_0 using ap26 version 105
Sun 01 Nov 2009 09:09:30 PM EET|PrimeGrid|Starting ap26_17795189_17795195_0_0
Sun 01 Nov 2009 09:09:30 PM EET|PrimeGrid|Starting task ap26_17795189_17795195_0_0 using ap26 version 105
Sun 01 Nov 2009 09:09:30 PM EET|PrimeGrid|Starting ap26_17795554_17795562_0_0
Sun 01 Nov 2009 09:09:30 PM EET|PrimeGrid|Starting task ap26_17795554_17795562_0_0 using ap26 version 105
Sun 01 Nov 2009 09:09:30 PM EET|PrimeGrid|Starting ap26_17796012_17796018_0_0
Sun 01 Nov 2009 09:09:30 PM EET|PrimeGrid|Starting task ap26_17796012_17796018_0_0 using ap26 version 105
Sun 01 Nov 2009 09:09:30 PM EET|PrimeGrid|Starting ap26_16134903_16134910_0_1
Sun 01 Nov 2009 09:09:30 PM EET|PrimeGrid|Starting task ap26_16134903_16134910_0_1 using ap26 version 105
Sun 01 Nov 2009 09:09:30 PM EET|PrimeGrid|Starting ap26_17789932_17789940_0_1
Sun 01 Nov 2009 09:09:30 PM EET|PrimeGrid|Starting task ap26_17789932_17789940_0_1 using ap26 version 105
Sun 01 Nov 2009 09:09:56 PM EET|PrimeGrid|Aborting task ap26_17795886_17795896_0_0: exceeded disk limit: 16.94MB > 9.54MB
Sun 01 Nov 2009 09:09:56 PM EET|PrimeGrid|Aborting task ap26_17795189_17795195_0_0: exceeded disk limit: 14.10MB > 9.54MB
Sun 01 Nov 2009 09:09:56 PM EET|PrimeGrid|Aborting task ap26_17795554_17795562_0_0: exceeded disk limit: 11.31MB > 9.54MB
Sun 01 Nov 2009 09:09:56 PM EET|PrimeGrid|Aborting task ap26_17796012_17796018_0_0: exceeded disk limit: 16.61MB > 9.54MB
Sun 01 Nov 2009 09:09:56 PM EET|PrimeGrid|Aborting task ap26_16134903_16134910_0_1: exceeded disk limit: 14.95MB > 9.54MB
Sun 01 Nov 2009 09:09:56 PM EET|PrimeGrid|Aborting task ap26_17789932_17789940_0_1: exceeded disk limit: 16.31MB > 9.54MB
Sun 01 Nov 2009 09:10:01 PM EET|PrimeGrid|Computation for task ap26_17795886_17795896_0_0 finished
Sun 01 Nov 2009 09:10:01 PM EET|PrimeGrid|Computation for task ap26_17795189_17795195_0_0 finished
Sun 01 Nov 2009 09:10:01 PM EET|PrimeGrid|Computation for task ap26_17795554_17795562_0_0 finished
Sun 01 Nov 2009 09:10:01 PM EET|PrimeGrid|Computation for task ap26_17796012_17796018_0_0 finished
Sun 01 Nov 2009 09:10:01 PM EET|PrimeGrid|Computation for task ap26_16134903_16134910_0_1 finished
Sun 01 Nov 2009 09:10:01 PM EET|PrimeGrid|Computation for task ap26_17789932_17789940_0_1 finished
Sun 01 Nov 2009 09:10:01 PM EET|PrimeGrid|Starting ap26_17795429_17795434_0_0
Sun 01 Nov 2009 09:10:01 PM EET|PrimeGrid|Starting task ap26_17795429_17795434_0_0 using ap26 version 105
Sun 01 Nov 2009 09:10:01 PM EET|PrimeGrid|Starting ap26_17795413_17795419_0_0
Sun 01 Nov 2009 09:10:01 PM EET|PrimeGrid|Starting task ap26_17795413_17795419_0_0 using ap26 version 105
Sun 01 Nov 2009 09:10:01 PM EET|PrimeGrid|Starting ap26_17795276_17795283_0_0
Sun 01 Nov 2009 09:10:01 PM EET|PrimeGrid|Starting task ap26_17795276_17795283_0_0 using ap26 version 105
Sun 01 Nov 2009 09:10:01 PM EET|PrimeGrid|Starting ap26_17795510_17795515_0_0
Sun 01 Nov 2009 09:10:01 PM EET|PrimeGrid|Starting task ap26_17795510_17795515_0_0 using ap26 version 105
Sun 01 Nov 2009 09:10:01 PM EET|PrimeGrid|Starting ap26_17795540_17795546_0_0
Sun 01 Nov 2009 09:10:01 PM EET|PrimeGrid|Starting task ap26_17795540_17795546_0_0 using ap26 version 105
Sun 01 Nov 2009 09:10:01 PM EET|PrimeGrid|Starting ap26_17795420_17795428_0_0
Sun 01 Nov 2009 09:10:01 PM EET|PrimeGrid|Starting task ap26_17795420_17795428_0_0 using ap26 version 105
Sun 01 Nov 2009 09:14:20 PM EET|PrimeGrid|Sending scheduler request: To report completed tasks. Requesting 0 seconds of work, reporting 6 completed tasks
Sun 01 Nov 2009 09:14:24 PM EET|PrimeGrid|Scheduler request succeeded: got 0 new tasks
Sun 01 Nov 2009 09:14:56 PM EET|PrimeGrid|Aborting task ap26_17795429_17795434_0_0: exceeded disk limit: 180.74MB > 9.54MB
Sun 01 Nov 2009 09:14:56 PM EET|PrimeGrid|Aborting task ap26_17795413_17795419_0_0: exceeded disk limit: 183.09MB > 9.54MB
Sun 01 Nov 2009 09:14:56 PM EET|PrimeGrid|Aborting task ap26_17795276_17795283_0_0: exceeded disk limit: 172.59MB > 9.54MB
Sun 01 Nov 2009 09:14:56 PM EET|PrimeGrid|Aborting task ap26_17795510_17795515_0_0: exceeded disk limit: 169.05MB > 9.54MB
Sun 01 Nov 2009 09:14:56 PM EET|PrimeGrid|Aborting task ap26_17795540_17795546_0_0: exceeded disk limit: 171.52MB > 9.54MB
Sun 01 Nov 2009 09:14:56 PM EET|PrimeGrid|Aborting task ap26_17795420_17795428_0_0: exceeded disk limit: 192.20MB > 9.54MB
Sun 01 Nov 2009 09:15:02 PM EET|PrimeGrid|Computation for task ap26_17795429_17795434_0_0 finished
Sun 01 Nov 2009 09:15:02 PM EET|PrimeGrid|Computation for task ap26_17795413_17795419_0_0 finished
Sun 01 Nov 2009 09:15:02 PM EET|PrimeGrid|Computation for task ap26_17795276_17795283_0_0 finished
Sun 01 Nov 2009 09:15:02 PM EET|PrimeGrid|Computation for task ap26_17795510_17795515_0_0 finished
Sun 01 Nov 2009 09:15:02 PM EET|PrimeGrid|Computation for task ap26_17795540_17795546_0_0 finished
Sun 01 Nov 2009 09:15:02 PM EET|PrimeGrid|Computation for task ap26_17795420_17795428_0_0 finished
--
Lumiukko
|
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
It's hard to tell at the moment if it's an app problem since most of the PS3s haven't switched over to 1.05 yet. The few PS3s I do see running 1.05 already seem to be running fine. However, they are all running a different version of Linux than your PS3. I have run this version for a few days on my test PS3 before submitting it.
I did some searching on that error, and it can be due to many things. Low disk space or allocation with BOINC settings, or possibly an error log (boinc logging or application stderr.txt) is getting huge for some reason.
If you can, check the slot folder while boinc is running to see if the stderr.txt is rapidly increasing in size, and if so, post what it contains. |
|
|
Lumiukko Volunteer tester Send message
Joined: 7 Jul 08 Posts: 164 ID: 25183 Credit: 716,865,909 RAC: 182,289
                          
|
If you can, check the slot folder while boinc is running to see if the stderr.txt is rapidly increasing in size, and if so, post what it contains.
Yes, the stderr.ttxt is rapidly increasing in size. It begins:
spu_create(): Function not implemented
Debug: 4095 , unexpected SPE->PPE interrupt mbox value
Debug: 4095 , unexpected SPE->PPE interrupt mbox value
Debug: 4095 , unexpected SPE->PPE interrupt mbox value
Debug: 4095 , unexpected SPE->PPE interrupt mbox value
Debug: 4095 , unexpected SPE->PPE interrupt mbox value
Debug: 4095 , unexpected SPE->PPE interrupt mbox value
...
...
Debug: 4095 , unexpected SPE->PPE interrupt mbox value
Debug: 4095 , unexpected SPE->PPE interrupt mbox value
So it is full of this "Debug: 4095"-message.
--
Lumiukko |
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
Good, we found the problem. Something appears to be 'broken' in your linux spufs support. As I have no idea what distro, or how it was installed, this is about all the information I can give you.
____________
|
|
|
Lumiukko Volunteer tester Send message
Joined: 7 Jul 08 Posts: 164 ID: 25183 Credit: 716,865,909 RAC: 182,289
                          
|
... As I have no idea what distro, or how it was installed....
It's Ubuntu. Just upgraded from 9.04 "Jaunty" to 9.10 "Karmic", so the upgrade probably broke it. I think I'll replace it with YDL.
--
Lumiukko
|
|
|
|
Hi gang
A real newb question here.
Thanks to mfl0p I have Yellow Dog running on a PS3 and have downloaded his file. I'm even posting via the PS#So far all so good. I am running at very low res, 480i. The only problem at the moment is that when I come to try to attach to a project the nice graphical interface shows me a lovely picture of a cup of coffee but the unresizable box isn't big enough to show the data fields.
Can some one explain how I jump what I hope is the last hurdle? I am sure there must be a fairly simple way of doing this from a terminal but I don't know it. Bear in mind it would be difficult to under estimate my computer skills so really easy steps please!
Cheers!
T
ps Isn't swine flu a boar? |
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
You can use the boinccmd program to attach to your account instead of the GUI, it's just a bit more complicated. ./boinccmd --help should get you started.
Or, turn off the PS3's firewall (or open BOINC's RPC port) and remotely attach to the PS3 with another computer's BOINC manager. You will need the GUI RPC key off the PS3.
Both aren't really "easy" if you are new to linux, but without having at least a 720p output monitor then the GUI is probably not going to look like it should, so you are left with these "harder" options.
____________
|
|
|
|
Thanks again for the help mfl0p, much appreciated.
I scrounged the use of an HD tv and reinstalled Yellow Dog. The user interface is certainly different! I managed to get BOINC running but is seems not to have up loaded results. I will look into it if I ever recover from this swine flu. I tried ringing the swine flu helpline but all I got was crackling.
Cheers!
T
|
|
|
|
Thanks for this app! I was using the old PS3grid one and this one has been a huge improvement in speed and efficiency. Wish I could have found it months ago... |
|
|
kto Send message
Joined: 13 Jan 09 Posts: 27 ID: 34164 Credit: 41,469,365 RAC: 0
                
|
Hi. I'm crunching with my PS3 running the latest Yellow Dog Linux. Holy cow I'm averaging more credits with my PS3 than my 2.3 GHz AMD quad Phenom. Well done on porting this to the PS3, Bryan (mfl0p). You've got decent tuning skills.
I, too, am a C/C++ developer (20 years specializing in high performance programming, multithreading, etc.). I'd like to ask:
Would it be possible to port of the other prime search algorithms to the PS3 other than AP26? Could I help? I understand Linux on the PS3 is a dying platform, but the Cell is a compelling chip - and I can see other users are doing some seriously major crunching with Cell workstations.
Point me at some code and lets get crackin! :)
thanks again,
kev (kto). |
|
|
Lexs Volunteer developer
 Send message
Joined: 16 Mar 08 Posts: 61 ID: 20289 Credit: 49,033,000 RAC: 0
               
|
For the project staging area, there's already a version of prpclient including phrot available for the PS3. Although they're not using the cell SPU.
I have genefer-2.2 working on the cell SPU's but the speed is as slow as on the PPC, so there is no advantage here.
(It might scale better on Cell-Blades with much more than 6 SPUs.)
At the moment I'm trying to port genefer to Cuda and Stream, but I have not much hope they will perform better. (because of slow double precision and memory transfers on GPU)
____________
|
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
It was just a matter of time until Sony did this:
http://blog.us.playstation.com/2010/03/28/ps3-firmware-v3-21-update/
Firmware 3.21 will disable OtherOS on the old (fat) PS3. They claim it's for security reasons. If you have a PS3, you'll have to update to keep playing your games/sign on to PSN. However, if you have a dedicated PS3 for number crunching only, you don't HAVE to update and remove a feature of your PS3.
____________
|
|
|
Lexs Volunteer developer
 Send message
Joined: 16 Mar 08 Posts: 61 ID: 20289 Credit: 49,033,000 RAC: 0
               
|
It was just a matter of time until Sony did this:
http://blog.us.playstation.com/2010/03/28/ps3-firmware-v3-21-update/
Firmware 3.21 will disable OtherOS on the old (fat) PS3. They claim it's for security reasons. If you have a PS3, you'll have to update to keep playing your games/sign on to PSN. However, if you have a dedicated PS3 for number crunching only, you don't HAVE to update and remove a feature of your PS3.
Let's hope that this is just an april fool, as the current
PS3 firmware version is 3.15 and the update shall be 3.21,
sounds like an unusal jump.
On the other side, now there really is a reason to hack
the box's hypervisor and unfold PS3's full potential for
homebrew software.
____________
|
|
|
mackerel Volunteer tester
 Send message
Joined: 2 Oct 08 Posts: 2468 ID: 29980 Credit: 449,690,393 RAC: 298,401
                           
|
The first line of that page lists:
The next system software update for the PlayStation 3 (PS3) system will be released on April 1, 2010 (JST), and will disable the “Install Other OS” feature that was available on the PS3 systems prior to the current slimmer models, launched in September 2009.
That certain date is on it, although if it is one, technically shouldn't it have been posted then?
Having said that, while I appreciate the AP26 throughput of the PS3, Linux on it is cripplingly slow from lack of ram, so I certainly wont miss that. |
|
|
mfl0p Project administrator Volunteer developer Send message
Joined: 5 Apr 09 Posts: 224 ID: 38042 Credit: 878,946,796 RAC: 824,785
                       
|
It is official. At least the Cell Blades can still work on the AP26 project. IMO, Sony just killed the Cell processor by doing this. Most all websites teaching Cell programming were teaching people on the PS3 running linux. I'll bet the PS4 will have a processor a lot more similar to desktop PCs.
From Sony's PS3 website:
PlayStation®3 System Software Version 3.21
SettingsInstall Other OS / Default System
The [Install Other OS] and [Default System] features have been deleted. If you update your system software to version 3.21 or later, you will not be able to use [Install Other OS] or [Default System] under (Settings) > (System Settings).
Notice for customers who are using another operating system
Before performing the system software update, follow the steps below:
Back up data from the other operating system
If you install the system software update, you will not be able to use the hard disk partitions used by the other operating system. The data saved in these partitions cannot be restored after the system software has been updated. Back up all data to storage media before updating the system software.
Using the hard disk partition for the PS3™ system software
If you want to allocate all of the hard disk for use by the PS3™ system software, you must format the hard disk. In this case, follow the steps below:
1.Back up data
Back up the necessary data by selecting (Settings) > (System Settings) > [Backup Utility]. For details about the backup utility feature, select here .
2.Format the hard disk
Format the hard disk by selecting (Settings) > (System Settings) > [Format Utility]. During the format operation, you will be asked to select a hard disk partition. Select [Use All for the PS3™ System].
For details about the format utility feature, select here .
3.Update the system software
Perform the system software update.
4.Restore the data
Restore the data that was backed up in step 1 by selecting (Settings) > (System Settings) > [Backup Utility].
Notice for customers who want to continue using another operating system
Do not update the system software if you want to keep using the [Install Other OS] and [Default System] features. If you choose to continue using another operating system, the following features will no longer be available:
The ability to sign in to PlayStation®Network
The ability to use online features that require you to sign in to PlayStation®Network, such as chat.
The ability to use the online features of PlayStation®3 format software
Playback of PlayStation®3 format software or Blu-ray Disc videos that require PS3™ system software version 3.21 or later
Playback of copyright-protected videos that are stored on a media server (when DTCP-IP is enabled under (Settings) > (System Settings) > [Enable DTCP-IP])
Use of new features and improvements that are available on PS3™ system software 3.21 or later
.PlayStation®3PSP®(Playstation®Portable)
____________
|
|
|
|
Sony hasn't killed anything. Just don't upgrade to the new firmware. =)
____________
FUCK THIS PROJECT AND THEIR COMMIE BULLSHIT POLITICS |
|
|