IRC logs of #boinc for Tuesday, 2010-04-20

00:52 *** RomW has quit IRC

00:54 *** RomW has joined #boinc

01:58 *** wdsmia has quit IRC

01:58 *** wdsmia has joined #boinc

02:46 *** wdsmia has quit IRC

02:46 *** wdsmia has joined #boinc

03:21 *** MapMan has quit IRC

03:36 * quail just uploaded couple rainbow pics =>

03:48 <efc> Hot. Also the rainbow is nice.

03:49 <efc> I got an almost full semicircle, double rainbow a month or two ago. Have to dig that up sometime

04:17 *** efc has quit IRC

06:24 <wdsmia> &wx 50319

06:24 <Romulus> wdsmia: Weather for Des Moines, IA; Temperature: 46.7°F / 8.2°C | Humidity: 73% | Pressure: 30.04in / 1017.2hPa (Falling) | Conditions: Mostly Cloudy | Wind Direction: South | Wind Speed: 0.0mph / 0.0km/h ; Today - Partly sunny. High in the lower 70s. South wind near 5 mph; Tonight - Partly cloudy with a 20 percent chance of showers. Low in the upper 40s. West wind near 5 mph; Wednesday - Partly (1 more message)

06:24 <wdsmia> &more

06:24 <Romulus> wdsmia: sunny. High in the upper 60s. West wind near 10 mph

06:26 <hawmps> 'lo

06:26 <wdsmia> mornin jay

06:31 <GaelicGrime> morning all

06:47 <hawmps> btw, Doug, finally 'aquired' a JRC NRD-545D in perfect condition.

07:02 <wdsmia> congrats

07:12 *** wetfoot has joined #boinc

07:12 <wetfoot> mornin!

07:14 <hawmps> now to sell my 535-D

07:19 <GaelicGrime> I wish I didnt have other things that take priority, that would be better than most of what I listen on

07:19 <GaelicGrime> never knew there were radio types in the halls LOL

07:20 <wetfoot> uni halls?

07:20 <GaelicGrime> boinc halls

07:20 <wetfoot> ok

07:21 <wetfoot> lol...

07:21 <hawmps> I have more than a few radios..

07:24 <GaelicGrime> I have a small collection, but nothing great these days.  one that I plan on keeping for portability is a degen 1103 but otherwise they are rather unworth mentioning LOL

07:44 <hawmps> The little Degen is a nice portable.

07:54 <hawmps> my only portable in regular use is an Eton E1

07:59 <GaelicGrime> yeah I have loved the degen so far, 2 of them 1 original and 1 later version (memory fails me at present as does my body which has decided sleep is now the plan)

08:14 *** Celelibi_ has quit IRC

08:15 *** Celelibi has joined #boinc

08:16 *** wdsmia_w has joined #boinc

09:37 <hawmps> lol

10:16 <hawmps> Now I really, really need a CHE199 wideband converter...

10:27 *** MapMan has joined #boinc

10:37 <PovAddict> wetfoot: I filled the survey :)

11:23 *** licensed has joined #boinc

12:09 *** zombie67 has joined #boinc

12:23 *** zombie67 has quit IRC

12:49 *** Aeternus has joined #boinc

13:13 *** licensed has quit IRC

13:46 *** yoyo[RKN] has joined #boinc

14:06 <yoyo[RKN]> what could be the reason for this fault

14:06 <Romulus> <> (at

14:06 <yoyo[RKN]> the app and wu runs without problems on most linux systems, but for some useres it crashes

14:07 <PovAddict> does that app use GPU?

14:07 <yoyo[RKN]> no

14:08 <PovAddict> strange since -234 is ERR_MISSING_COPROC

14:10 <yoyo[RKN]> its simple C code

14:15 <yoyo[RKN]> i could link it static

14:16 <PovAddict> the error comes from BOINC

14:41 *** licensed has joined #boinc

15:08 <PovAddict> wetfoot: don't do DCC Chat, it never works properly

15:08 <wetfoot> lol

15:08 <PovAddict> can't you do a normal private chat?

15:08 <wetfoot> ok

15:08 <wetfoot> thanks for filling in my survey

15:08 <wetfoot> really useful stuff

15:08 <wetfoot> I was wondering..... :p

15:09 <wetfoot> could I delete ur survey entry and use it as an interview?

15:09 <PovAddict> yesterday my IRC client had connection problems while the "Accept DCC Chat" message box was left open for hours :P I reported the bug to them

15:09 <wetfoot> would you consent to that?

15:09 <PovAddict> interview?

15:10 <wetfoot> yer, take all that you said and put it in my dissertation as an interview with you over irc

15:10 <wetfoot> putting it as a survey response kind of belittles it a bit

15:11 <wetfoot> and I think it deserves more than that

15:12 <wetfoot> some of your answers were very informed and to say 10 people said this and 1 persoin (you) made this statement kind of puts it the wrong context

15:13 <wetfoot> + it makes me look good if I say I sought out someone who has been about for a long time, started and contributed projects and is exploring a boinc fork

15:14 <wetfoot> contributed to*

15:15 <wetfoot> anyways what do you think?

15:15 <PovAddict> feel free to use it

15:15 <PovAddict> I'm busy right now (and going out in 45 mins) but you could ask more later :P

15:17 <yoyo[RKN]> moin

15:17 <wetfoot> ok man :D

15:17 <yoyo[RKN]> symbol _WinMain@16 not found

15:17 <wetfoot> hello yoyo

15:18 <yoyo[RKN]> in msvcrtd.lib

15:18 <PovAddict> Windows is the problem :)

15:18 <yoyo[RKN]> jup

15:19 <PovAddict> how are you compiling?

15:19 <yoyo[RKN]> created a new project in VisualC++ express and added the cpp file to it

15:20 <yoyo[RKN]> added boinc api+lib directories as include

15:20 <PovAddict> you need to tell Visual C++ that you're writing a console app

15:21 <yoyo[RKN]> oh

15:21 <yoyo[RKN]> I selected the other thing in the wizard

15:25 <wetfoot> will ask you some questions later if it's still ok, thanks ProcAddict, really appreciate you giving your time :D

15:27 <wetfoot> Dissertations are no fun lol, interesting at times but a bit too stressfull!

15:28 <yoyo[RKN]> it compiled and linked

15:31 <wetfoot> how old is yoyo@home?

15:32 <PovAddict> yoyo[RKN]: I should start charging you for tech support already :P

15:32 <yoyo[RKN]> than I must stop to ask

15:33 <yoyo[RKN]> wetfoot : 3 years

15:33 <yoyo[RKN]> nearly

15:33 <PovAddict> already? damn I have no sense of time

15:34 <yoyo[RKN]> according to the oldest news

15:34 <Romulus> Title: News archive (at

16:11 <yoyo[RKN]> oh

16:11 <yoyo[RKN]> CreateProcess() failed - The application has failed to start because its side-by-side configuration is incorrect. Please see the application event log or use the command-line sxstrace.exe tool for more detail. (0x36b1)

16:11 <yoyo[RKN]> strange windows

16:13 *** xcamel has quit IRC

16:14 *** xcamel has joined #boinc

16:14 *** ChanServ sets mode: +o xcamel

16:14 <Romulus> WB, Boss

16:36 *** licensed has quit IRC

16:54 *** wdsmia_w has quit IRC

17:05 *** licensed has joined #boinc

17:08 *** yoyo[RKN] has quit IRC

17:24 *** desti has joined #boinc

17:32 <wdsmia> re-hi

17:42 *** licensed has quit IRC

17:45 <Tank_Master> wb

17:59 *** Aeternus has quit IRC

18:10 <RomW> @yoyo, that generally happens when multiple versions of the c runtime attempt to be loaded at the same time via the application manifests

18:12 <RomW> For instance App A depends on CRT included with VS2008 and another DLL used by the application uses the CRT included with VS2008 SP1

18:13 <RomW> the resulting version clash causes CreateProcess to fail

18:13 <PovAddict> what if app dynamically links to VS2008 CRT and it's not available in the system? does it give that error or a different one?

18:13 <RomW> Should give a file not found error

18:13 <RomW> well DLL not found

18:14 <RomW> Starting with VS 2005 multiple versions of the CRT can be installed in the GAC of the system

18:14 <RomW> at the same time

18:14 <PovAddict> yeah, but BOINC science apps can't rely on the GAC having a certain CRT version...

18:15 <RomW> So VS 2008, VS 2008 SP1, VS 2008 SP1 + hotfix, could all be installed on the same system

18:15 <RomW> true, static linking generally resolves that

18:17 <RomW> Anyway, the easiest way I have found for identifying the bad guy, is to open up all the DLLs that import MSVCRT in VS without a solution and extract the RT_MANIFEST resource

18:18 <RomW> look at it in a text editor

18:18 <RomW> compare with the executable

18:23 <PovAddict> RomW: I have a 256MB GPU

18:24 <RomW> lol

18:24 <RomW> yeah, but you make it a point to not do GPU computing

18:24 <PovAddict> GeForce 8600 GT

18:24 <Tank_Master> PovAddict is behind the times

18:24 <Tank_Master> :P

18:24 <PovAddict> does that card even support CUDA?

18:24 <Tank_Master> yes

18:25 <Tank_Master> its hella slow, but yes, itll run cuda

18:25 <RomW> The better question ais, are there any apps that still support it

18:25 <PovAddict> lol

18:25 <Tank_Master> collatz

18:26 <Tank_Master> i dunno about dnetc, but i think it does

18:26 <Tank_Master> gpugrid will run on it, but you wont finish before the deadline

18:27 <Tank_Master> I believe SETI's cuda app still works on it, but i dont folow SETI much

18:28 <RomW> ah

18:28 <RomW> better than I thought

18:28 *** _Danilo_ has quit IRC

18:29 <Tank_Master> primgrid's cuda app should work for it to, but that subproject is ending

18:29 <Tank_Master> mw wont run on it because they need double resision

18:29 <Tank_Master> presision*

18:30 <Tank_Master> i think latice has a cuda app, but rarly has work

18:30 <RomW> Now that is a surprise

18:30 <Tank_Master> what is?

18:30 <RomW> I would not thought Latice would have had a CUDA app

18:31 <Tank_Master> ah

18:31 <PovAddict> so what do I need to download to enable CUDA?

18:31 <PovAddict> I don't think the Debian-provided driver supports it

18:31 <RomW> Probably the nvidia driver package

18:31 <Tank_Master> just a compatible driver from nvidia and boinc 6.6 or later

18:31 <PovAddict> I have 190.53

18:32 <Tank_Master> thatll work

18:32 <RomW> although i've had bad luck with both the Nvidia and ATI drivers on Linux

18:32 <PovAddict> RomW: I was forced to use ATI drivers recently

18:32 <RomW> both choke on different distros

18:32 <PovAddict> RomW: a capacitor blew up on my nVidia card and I had to use onboard ATI gfx till I got the cap fixed

18:33 <RomW> left my dev box unbootable for a couple of days

18:33 <RomW> ouch

18:36 <PovAddict> actually, a capacitor blew up and I got it fixed soon, but two weeks later another blew up and I had no replacement so I had to use onboard until I got around to go to an electronics shop

18:37 <RomW> ah

18:37 <PovAddict> quite strange... worked for a year or two with no problems, then two caps fail in as many weeks :/

18:38 <RomW> probably heat related

18:38 <Tank_Master> 2 year warenty must have just expiered

18:38 <RomW> lol

18:38 <PovAddict> lol

18:39 <PovAddict> I just configured BOINC, it said *twice* that "Package libnotify was not found in the pkg-config search path."

18:40 <PovAddict> has: LIBNOTIFY_CFLAGS=`$PKG_CONFIG libnotify --cflags`; LIBNOTIFY_LIBS=`$PKG_CONFIG libnotify --libs`;

18:40 <RomW> Easier to just download from

18:40 <PovAddict> any reason why it's not just using the PKG_CHECK_MODULES autoconf macro?

18:40 <RomW> I'm still waiting on Eric to fix that up the right way

18:41 <RomW> instead of my hack

18:48 <PovAddict> yup, manager failed to compile because I don't have gtk-dev, but configure should have noticed and errored

18:50 <RomW> Eric is the build environment guy

18:50 <PovAddict> 20-Apr-2010 19:50:11 [---] No usable GPUs found

18:50 <PovAddict> D:

18:50 <RomW> I only get involved when he breaks my VM builds

18:50 <RomW> lol

18:50 <RomW> cc_config.xml: <gpu_debug>

18:51 <RomW> or whatever the tag is

18:51 <PovAddict> 20-Apr-2010 19:51:21 [---] Unrecognized tag in cc_config.xml: <gpu_debug>

18:51 <PovAddict> guess that was not it

18:51 <PovAddict> coproc_debug it is

18:52 <PovAddict> 20-Apr-2010 19:52:03 [---] No NVIDIA library found

18:52 <RomW> coproc_debug

18:54 <RomW> wasn't found

18:55 <PovAddict> got it, was moved off the nvidia-glx package in debian unstable

18:56 <PovAddict> 20-Apr-2010 19:56:30 [---] NVIDIA GPU 0: GeForce 8600 GT (driver version unknown, CUDA version 2030, compute capability 1.1, 255MB, 76 GFLOPS peak)

18:58 <RomW> Well, now you should be able to do GPU jobs

18:58 <PovAddict> collatz gave me a CPU task

18:59 <RomW> make stage

18:59 <RomW> ops

19:01 <Tank_Master> you need to set your project preferences to use the gpu app

19:01 <Tank_Master> and while your in there, you vcan also tell it to not send you cpu apps

19:02 <PovAddict> I just did that

19:02 <Tank_Master> ok

19:02 <PovAddict> now it's refusing to send work (last req too recent) >_<

19:02 <Tank_Master> LOL

19:02 <Tank_Master> yeah, that is annoying

19:02 <Tank_Master> along with the 24h backoff when theres an error connect

19:02 <PovAddict> I'll try SETI

19:02 <RomW> LOL, the result of not having that feature is pretty annoying too

19:03 <Tank_Master> which one?

19:03 <PovAddict> SETI@home|Message from server: (Project has no jobs available)

19:03 <RomW> Still have nightmares

19:03 <RomW> last request too recent

19:03 <PovAddict> come on, I finally give in to try CUDA and projects conspire against me?

19:03 <Tank_Master> ahh

19:03 <Tank_Master> try dnetc

19:03 <PovAddict> yoyo?

19:03 <Romulus> yoyo is at, PovAddict

19:04 <Tank_Master> yoyo dosnt have gpu

19:04 <Tank_Master>

19:04 <Romulus> Title: DNETC@HOME (at

19:04 <Tank_Master> thats a gpu only project

19:04 <Tank_Master> no cpu apps

19:05 <Tank_Master> ohh, thought they had a linux ap, apparently now

19:05 <RomW> lol, didn't know about that project

19:05 <PovAddict> o_o

19:05 <PovAddict> GPU only?

19:05 <PovAddict> how do they calculate credits then?

19:05 <Tank_Master> its between collatz and MW

19:05 <Tank_Master> thats what he has tuned the credits to

19:05 <PovAddict> so far projects adjusted CPU app to give similar credit to other projects, then let the GPU app give the same credits per WU as the CPU app

19:06 <PovAddict> DNETC@HOME|Message from server: DNETC@Home is not available for your type of computer.

19:06 <PovAddict> Windows-only

19:06 <Tank_Master> he adisus the credit to be between collatz and MW'd GPU for the same card

19:06 <Tank_Master> yeah, noticed that larter, sorry

19:06 <PovAddict> projects are conspiring against me, I tell you

19:06 <Tank_Master> how long is collats backoff?

19:07 <PovAddict> 1 second left

19:07 <Tank_Master> not to bad then

19:07 <PovAddict> Collatz Conjecture|Message from server: No work sent

19:07 <PovAddict> no reason given

19:07 <Tank_Master> o0

19:07 <RomW> ugh

19:08 <PovAddict> no extra message in the sched_reply*.xml file so it's not that the client failed to show it, the server didn't send any more info

19:09 <Tank_Master> odd

19:09 <PovAddict> and DNETC is giving a ridiculously low backoff

19:09 <RomW> I wonder if I should bug John about an SSL account again

19:09 <PovAddict> 7-second backoff for "project doesn't support Linux"?

19:10 <PovAddict> it's not like they're going to add Linux support while we're talking

19:10 <Tank_Master> lol

19:10 <RomW> IIRC the min time is a different config.xml setting

19:11 <PovAddict> SETI keeps saying no jobs

19:11 <Tank_Master> its tuesday, are they dont messing around with the db?

19:11 <RomW> otherwise it just falls back to exp backoff

19:12 <PovAddict> I think unsupported platform should give a higher default backoff...

19:12 <Tank_Master> looks like aqua is out of fokker WUs again

19:12 <RomW> I think we left it that way for places like St. Jude hospital

19:12 <RomW> they bring apps up and down

19:14 <PovAddict>

19:14 <Romulus> Title: pastebin - nicolas - post number 1870593 (at

19:14 <PovAddict> the task queue is totally empty for both CPU and GPU, why does it sometimes ask CPU only and sometimes GPU only?

19:15 <PovAddict> and never both?

19:16 <Tank_Master> its BOINC, you expecting things to make sence?

19:17 <PovAddict> okay, I might have luck with gpugrid this time

19:17 <RomW> turn on work_fetch_debug

19:17 <PovAddict> I got a task

19:17 <RomW> or is it workfetch_debug

19:17 <Tank_Master> you wont finish it in time

19:17 <PovAddict> Tank_Master: I don't care, I want to see if it starts at all

19:17 <Tank_Master> ahh, ok :P

19:17 <PovAddict> oh awesome

19:17 <PovAddict> "Ready to start"

19:18 <PovAddict> *there are no other tasks*

19:18 <RomW> ah

19:18 <RomW> the new code might be working

19:18 <RomW> lol

19:18 <Tank_Master> do you have "run gpu alwys"?

19:18 <RomW> need to add some more logging though

19:18 <RomW> ooo good one Tank_Master

19:18 <PovAddict> GPUGRID|[coproc_debug] Insufficient CUDA for p44-IBUCH_A011_pYEEI_long_100413-2-4-RND6677_1; 0 < 1

19:19 <PovAddict> chance of a user seeing that: 0.01% (it's under a debug flag)

19:19 <RomW> that shouldn't have even been assigned to you to begin with

19:20 *** desti has quit IRC

19:21 *** desti has joined #boinc

19:22 <RomW> did you build BOINC? or did I?

19:22 <PovAddict> I built BOINC from latest trunk

19:24 <PovAddict> building again with optimizations so I can use gdb and keep my sanity

19:25 <xcamel> 'lo Rom

19:25 <PovAddict> cp->available_ram[0] == 70004224

19:25 <PovAddict> rp->avp->gpu_ram = 249561088

19:26 <RomW> 'lo

19:26 <PovAddict> in get_integer_assignment()

19:27 <RomW> cpu_sched.cpp:1117

19:28 <PovAddict> yup

19:28 <PovAddict> in line 1107, memory needed > memory available, so defer_sched was set to true

19:28 <PovAddict> and it left the loop with nfree==0

19:29 <PovAddict> kind of impressive I have only 66MB of free GPU RAM...

19:30 <RomW> How much is X using?

19:30 <PovAddict> I don't know how to tell

19:33 <PovAddict> I just disabled KDE desktop compositing

19:33 <PovAddict> available RAM went up to 91.41MB

19:33 <PovAddict> I added a debug message to the client to show available and required RAM

19:33 <PovAddict> 20-Apr-2010 20:33:07 [GPUGRID] [coproc_debug] GPU has 91.41 MB RAM, task needs 238.00 MB

19:34 <PovAddict> quitting web browser, 105.23MB free

19:35 <RomW> Is your GPU card one of the funky ones where half of the GPU ram is actually system memory?

19:35 <PovAddict> I bet it's not possible for X to use 18MB of RAM even without a desktop environment

19:35 <PovAddict> no, it's 256MB real video memory

19:35 <RomW> hmmmmmmm

19:36 <PovAddict> don't underestimate KDE4's hit on video memory :)

19:37 <RomW> 1117 talks about usabe instances of CUDA devices

19:37 <RomW> ah

19:37 <RomW> okay

19:37 <RomW> I see the code ahead

19:37 <RomW> of that debug block

19:37 <PovAddict> nfree has the number of usable instances (0)

19:38 <PovAddict> nfree was not incremented to 1 because the one and only CUDA device didn't have enough memory

19:39 <PovAddict> hmm I killed Plasma and it went up to 110.16 MB... way too little change

19:39 <PovAddict> I wish I could see what is actually using GPU memory

19:42 <RomW> No GPU has a way to do that yet

19:42 <RomW> About the only thing to do is drop to the command line in your case

19:44 *** PovAdct_cli has joined #boinc

19:44 *** PovAddict has quit IRC

19:45 <PovAdct_cli> just logging out of my KDE session got it to this:

19:45 <PovAdct_cli> [coproc_debug] GPU has 228.09 MB RAM, task needs

19:45 <PovAdct_cli> 238.00 MB

19:46 <PovAdct_cli> X is still running, showing login screen

19:46 <RomW> lol

19:47 <PovAdct_cli> killing X altogether...

19:47 <PovAdct_cli> [coproc_debug] GPU has 227.50 MB RAM, task needs

19:47 <PovAdct_cli> what the hell?

19:47 <Romulus> what the hell is that, PovAdct_cli

19:47 <PovAdct_cli> I kill the login manager and the free memory goes *down*?

19:48 <RomW> PCIX? Or AGP?

19:49 <PovAdct_cli> is there any hardware-visible difference between PCI and PCIe?

19:49 <RomW> Do you know what the ISA bus looked like?

19:49 <RomW> and EISA?

19:49 <PovAdct_cli> very wide compared to PCI?

19:51 <PovAdct_cli> afaik ISA is ancient, which probably means that I haven't seen it in real life

19:51 <RomW> Okay...

19:51 <RomW> Ummmm

19:51 <RomW> I'm sure there are a couple of pictures online

19:51 <PovAdct_cli> that means restarting X :)

19:51 <RomW> lol

19:51 <RomW> true

19:52 <RomW> But, my roomate just got home, and we are supposed to go out to eat dinner, so i'll be back later

19:52 <RomW> ttyl

19:52 <PovAdct_cli> anyway, it's not AGP, I don't know the visible differences between PCI and PCIe, but if it was "normal" PCI it would be way too slow... so that leaves PCIe

20:06 *** zombie67 has joined #boinc

20:09 *** PovAdct_cli has quit IRC

20:11 *** efc has joined #boinc

20:14 <Tank_Master> zombie67, you mised a historic event...

20:14 <Tank_Master> pov has gone cuda!

20:14 <zombie67> oh?

20:14 <zombie67> I didn't realize he was an ATI fanboi

20:14 <zombie67> or do you mean that he is GPU crunching

20:14 <zombie67> ?

20:15 <Tank_Master> no, he was no gpu at all till today

20:15 <zombie67> I seem to recall he said he would never GPU crunch

20:15 <Tank_Master> for crunching anyway

20:15 <zombie67> brb

20:15 <Tank_Master> k

20:16 *** PovAddict has joined #boinc

20:55 <RomW> i'm back

20:56 *** quail has quit IRC

20:58 <RomW> well, gone again...

20:58 <RomW> i'll be back later tonight

21:01 <zombie67> pov, i hear you went to the dark side...

21:26 <PovAddict> I what?

21:26 <PovAddict> zombie67: ^

21:27 <zombie67> you are GPU crunching?!

21:28 <PovAddict> I compiled BOINC 6.10.something, installed libcuda, and tried attaching to projects

21:28 <PovAddict> SETI says there are no jobs

21:29 <PovAddict> dnetc has no Linux app

21:29 <PovAddict> Collatz only gave me CPU tasks

21:29 <PovAddict> finally GPUGrid gave me a WU and it refused to start due to low GPU memory

21:30 <PovAddict> so you could say I never actually got to GPU-crunch

21:32 <zombie67> wait, you have to use the right settings

21:32 <zombie67> collats works just fine with small CUDA cards

21:32 <PovAddict> right settings for what?

21:32 <zombie67> in your proj prefs

21:32 <zombie67> what GPU?

21:32 <zombie67> how much memory?

21:32 <Romulus> It has been said that how much memory is in the machine, zombie67

21:32 <PovAddict> 8600 GT, 256MB RAM

21:33 <zombie67> gimme the url to your machine on that project

21:33 <PovAddict>

21:33 <Romulus> <> (at

21:34 <zombie67> it shows only 255, but even so, it should work with that much

21:34 <PovAddict> I get "no work sent" without a reason

21:35 <zombie67> did you set your proj prefs to send only CUDA?

21:35 <PovAddict> after getting a CPU task and aborting it, I set the prefs to send CUDA and not CPU

21:36 <zombie67> and then what?

21:36 <PovAddict> Collatz Conjecture|Requesting new tasks for GPU

21:36 <PovAddict> Collatz Conjecture|[sched_op_debug] CPU work request: 0.00 seconds; 0.00 CPUs

21:36 <PovAddict> Collatz Conjecture|[sched_op_debug] NVIDIA GPU work request: 8640.00 seconds; 1.00 GPUs

21:36 <PovAddict> Collatz Conjecture|Message from server: No work sent

21:37 <zombie67> you mean, the mighty BOINC gives you no reason/

21:37 <zombie67> ?

21:37 <PovAddict> yup

21:38 <PovAddict> I thought maybe the server was returning extra info that the client wasn't showing me

21:38 <PovAddict> but sched_reply_boinc.thesonntags.com_collatz.xml doesn't have anything else

21:38 <zombie67> did you post about it?  Slicker is very active and helpful

21:39 <PovAddict> nope, </experiment>

21:39 <zombie67> experiment with a tolerance of a single failure?

21:41 <zombie67> [18:40]  <zombie67> ah!

21:41 <zombie67> [18:41]  <zombie67> cuda 2.2?

21:41 <zombie67> [18:41]  <zombie67>

21:41 <Romulus> <> (at

21:43 <PovAddict> how do I even know? :P

21:43 <PovAddict> "CUDA version 2030, compute capability 1.1"

21:43 <zombie67> ah, even more

21:43 <zombie67>

21:43 <Romulus> Title: Applications (at

21:43 <zombie67> there is no linux cuda default app

21:44 <zombie67> you have to use anon platform:

21:44 <Romulus> <> (at

21:44 <zombie67> and now we know.

21:44 <zombie67> so install it and try again.

21:46 <PovAddict> 'tis crunchin'

21:47 <PovAddict> and I'm sure it's using the GPU

21:49 <PovAddict> because it lags the GUI

21:49 <PovAddict> if I suspend the WU, it still takes some seconds to get the GUI framerate back to normal

21:50 <zombie67> heh

21:50 <zombie67> yeah, slower GPUs can lag

21:50 <PovAddict> I'm trying glxgears

21:50 <PovAddict> it runs at 2800fps

21:50 <zombie67> notice all the platforms supported by default or anon-platform?

21:50 <zombie67> that is why i was so surprised you didn't get work

21:51 <PovAddict> when I resume the WU it drops to 40-60fps

21:51 <zombie67> he supports EVERYTHING

21:51 <zombie67> you know you can set it so that it suspends when GPU is in use?

21:51 <PovAddict> I'm doing this just because RomW asked for help testing low-RAM GPUs

21:52 <zombie67> oh?

21:52 <zombie67> I have a low ram iMac w/ cuda

21:54 <PovAddict> my system just hanged pretty badly for a minute

21:55 <PovAddict> also, I think I can actually hear the fan changing speed when the WU runs

21:57 <wdsmia> low ram iMac w/ cuda... thats sad

22:49 *** infinisoft has joined #boinc

22:54 <RomW> The help in testing was for the case where the amount of available GPU ram dips below the required ram for the task and BOINC shuts the task down.  It shouldn't attempt to start the next GPU task until at least the required amount of ram is available for either the next GPU task or the currently selected GPU task

22:54 <RomW> Pov helped test the changes without even knowing he did

22:54 <RomW> :)

22:56 <RomW> Although with his feedback we'll be tweaking the status of the task so that it says something to the effects of 'ready to start when more GPU Ram is available' instead of just ready to start'

22:56 <PovAddict> yet another WU state?

22:56 <PovAddict> there is a "waiting for memory" state already...

22:56 <RomW> yeah

22:57 <RomW> waiting for memory just refers to system memory

22:57 <RomW> i'm not sure what David is going to call it though

22:59 <PovAddict> ah looks like it's not really a state in the GUI RPC protocol

22:59 <PovAddict> it's an independent boolean (too_large)

23:06 <zombie67> I have seen a common problem

23:07 <zombie67> where the GPU memory is shared with system memory

23:07 <zombie67> and even though it shows 256mb

23:07 <zombie67> it reports something less to BOINC

23:07 <zombie67> and a task that needs 256, and will really get it, won't ever start.

23:08 <RomW> IIRC, the Nvidia API returns just what is on-board ram, doesn't add in the system memory allocated for that purpose

23:09 <RomW> Nvidia hasn't given any intruction for determining that

23:09 <zombie67> odd

23:09 <RomW> since that value can be changed via the BIOS

23:10 <zombie67> for example, my iMac shows I have 256mb GPU memory

23:10 <zombie67> but the projects show my machines as 255mb

23:10 <RomW> Then I must be thinking about a different problem

23:11 <zombie67> so it's not that boinc is not adding system memory, it is that boinc is counting less than all the gpu memeory

23:11 <zombie67> or there is a rounding problem

23:11 <RomW> well, some of that ram gets used by the OS shell

23:11 <zombie67> could be

23:12 <RomW> in the case of Pov and KDE4 100+MB was used

23:12 <zombie67> but since people use round numbers, when setting requirements...

23:12 <zombie67> there needs to be a fudge factor, or "close enough"

23:12 <RomW> he couldn't start crunching until after shutting down X Windows

23:13 <RomW> well, that needs to be a server-side configuration option

23:13 <RomW> or server-side fudge factor

23:13 *** quail has joined #boinc

23:13 <zombie67> but the server side does not see my GPU as it really is

23:14 <PovAddict> RomW: even after shutting down X, there wasn't enough VRAM

23:14 <RomW> @Pov, really?

23:14 <PovAddict> yeah, short by a few MBs

23:14 <RomW> ouch

23:14 <PovAddict> like I said as PovAdct_cli

23:15 <RomW> @Zombie, the client reports what the Nvidia APIs return as the card configuration

23:15 <PovAddict> now I'm running Collatz along with KDE4 with desktop compositing; there's some GUI lag but doesn't make the desktop unusable

23:16 <RomW> Didn't Collatz say it requires 256mb?

23:16 <zombie67> collatz runs fine on my "255" gpu

23:16 <PovAddict> RomW: GPUGRID app required 238MB; when I logged out of KDE, free RAM went to 228MB; when I killed X (taking the KDE login manager down), free RAM went *down* to 227.5MB

23:17 <PovAddict> no idea what the hell happened there... close X, slightly *more* VRAM in use?

23:17 <RomW> oh, I was thinking about GPUGRID then...

23:17 <RomW> <Shrug>

23:17 <RomW> It is just a number returned to us from Nvidia

23:18 <zombie67> I read somewhere that it runs with something like 100mb

23:18 <zombie67> collatz does, I mean

23:21 <zombie67> this thread shows the memory problems:

23:21 <Romulus> <> (at

23:31 <RomW> I've forwarded on to David, maybe we can presure Nvidia/ATI in proving better tools to track memory usage

23:32 <RomW> i.e. provide breakouts by process for how much VRAM is being used

Generated by 2.4 by Marius Gedminas - find it at!