IRC logs of #boinc for Friday, 2012-12-07

16:09 *** Romulus has joined #boinc

16:14 *** PieSpyRom has joined #boinc

16:36 *** Romulus` has joined #boinc

16:36 *** Romulus has quit IRC

16:36 *** Romulus` is now known as Romulus

16:38 *** xcamel has quit IRC

16:38 *** PieSpyRom has quit IRC

16:39 *** PieSpyRom has joined #boinc

16:39 *** xcamel has joined #boinc

16:39 *** ChanServ sets mode: +o xcamel

17:18 *** Caterpillar has quit IRC

17:20 *** Caterpillar has joined #boinc

17:43 *** whynot has joined #boinc

17:44 *** yoyo has quit IRC

17:45 *** yoyo has joined #boinc

17:49 *** Shadow`_ has quit IRC

17:50 *** yoyo has quit IRC

18:17 *** Aeternus has quit IRC

18:36 *** efc has joined #boinc

19:02 *** desti has quit IRC

19:16 *** RomW has joined #boinc

19:16 <RomW> Howdy

19:17 <RomW> @Tank_Master, can you give 7.0.40 a try?  It should attempt to detect the Intel HD 4000.

19:19 *** desti has joined #boinc

19:21 *** Caterpillar has quit IRC

19:26 *** whynot has quit IRC

19:26 *** desti has quit IRC

19:30 *** desti has joined #boinc

20:07 *** efc has quit IRC

20:14 <Tank_Master> ! RomW!

20:14 <Tank_Master> will do that now

20:23 *** Mahmoud has quit IRC

20:57 *** Tank_Master has quit IRC

20:59 *** siamba has quit IRC

20:59 *** Zumu has quit IRC

20:59 <ChristmasCoder> howdy RomW

21:04 *** Tank_Master has joined #boinc

21:04 <Tank_Master> RomW, i dont think its working, and now i see a new error message

21:04 <Tank_Master> 12/7/2012 6:02:15 PM | | Couldn't get Device IDs for platform #2: error -1

21:12 <RomW> Hmmmm

21:12 <RomW> Howdy ChristmasCoder

21:12 <Tank_Master> thoughts? anything i can do to get mroe debug info?

21:12 <RomW> @TM, could you turn on <coproc_debug/>?

21:13 *** siamba has joined #boinc

21:13 *** Zumu has joined #boinc

21:14 <Tank_Master> something tells me this is gonna get spammy :D

21:15 <RomW> :)

21:15 <Tank_Master> odd

21:15 <Tank_Master> it didnt give anything

21:16 <Tank_Master> other than ton the 2 tasks runbning (got ati and nvidia in the system as well)

21:16 <Tank_Master>

21:16 <Romulus> Title: boinc - (at

21:18 <RomW> Is the OpenCL driver installed for the Intel card?

21:18 <RomW> or chip as the case may be

21:18 <Tank_Master> yes, it gets installed with the driver

21:18 <RomW> hmmmm

21:18 <Tank_Master> the gpu hd 4000 driver that is

21:19 <RomW> okay, one sec

21:19 <Tank_Master> i dont have a monitor hooked up to the chip atm, but im not sure if thats needed

21:19 <Tank_Master> its not for the nvidia gpu

21:20 <PovAddict> try it?

21:21 <RomW> Could you run this executable and send me the output in email?

21:22 <Tank_Master> :o

21:22 <Tank_Master> this latest driver now supports dual link DVI!

21:22 <PovAddict> this is weird

21:22 <Tank_Master> i got full qwhd (2560x1440) res

21:23 <Tank_Master> after running it, where does the info go on that util?

21:23 <PovAddict> according to Tank_Master's log, clGetPlatformInfo is returning -1, which maps to CL_DEVICE_NOT_FOUND, which isn't even mentioned in the clGetPlatformInfo documentation as a possible return value

21:25 <RomW> You might have to redirect the output

21:25 <RomW> clinfo > info.txt

21:29 <PovAddict> d'oh, nevermind

21:30 <PovAddict> what's failing is clGetDeviceIDs, of course

21:30 <PovAddict> Tank_Master: did you try connecting a monitor?

21:31 <Tank_Master>

21:32 <Tank_Master> same thing with a monitor pluged in

21:32 <PovAddict> there we go

21:32 <PovAddict> RomW: it's identifying itself as a CPU, not a GPU

21:32 <Tank_Master> only thing left to try is having it pluged in while booting up

21:33 <RomW> Well, the CPU can be used as an OpenCL device

21:34 <RomW> So, we are back to checking to see if the right driver is installed

21:34 <Tank_Master> just installed it off intel's site

21:34 <Tank_Master> the latest version

21:36 <RomW>

21:36 <Romulus> <> (at

21:36 <RomW> Do you have the registry entries list that show up in that forum post?

21:36 <RomW> With the Intel references

21:36 <Tank_Master> looking

21:37 <PovAddict> RomW: is there consensus on an ICD for OpenCL on Windows already?

21:38 <RomW> Consensus?

21:39 <PovAddict> an ICD interface, a way to avoid all vendors to have to provide their own wrapper libraries that can find everybody else's...

21:39 <PovAddict> who ships OpenCL.dll?

21:40 <RomW> They handle it like the OpenGL stuff was handled when MS made the OpenGL.dll library

21:40 <RomW> Registry entries

21:40 <RomW> Khronos

21:41 <RomW> the opencl.dll library just enumerates the DLLs listed in the vendors section and cycles through the APIs dynamically

21:41 <Tank_Master> looks like its all there as far as the openCL dlls

21:41 <PovAddict> RomW: right, but who ships the OpenCL.dll library that enumerates the registry? Microsoft? all individual vendors?

21:42 <RomW> All the vendors

21:42 <Tank_Master> i got it from installing the intel driver

21:42 <Tank_Master> gpu driver that is

21:42 <RomW> But they are supposed to be getting it from Khronos

21:43 <RomW> TM, can you hook a monitor up to the Intel card and then reboot the box?

21:43 <Tank_Master> gpu caps viewer isnt detecting the intel gpu

21:43 <Tank_Master> yeah, ill do that now

21:43 <Tank_Master> brb

21:43 <RomW> I suspect it reacts the same way as the other cards on Win7 and goes to sleep without a monitor or loopback adapter

21:44 <Tank_Master> i got video display atm on the intel port

21:44 *** Tank_Master has quit IRC

21:45 <PovAddict> RomW: see and the three linked forum threads

21:45 <PovAddict> RomW: seems like indeed you need a monitor connected, and there's no way around it

21:46 <PovAddict> "Currently Processor Graphics OCL device in unavailable in the "headless" configuration" from an Intel employee

21:46 *** Tank_Master has joined #boinc

21:47 <Tank_Master> no diff

21:48 <PovAddict> Tank_Master: do you have another GPU connected too?

21:48 <Tank_Master> yes

21:48 <PovAddict> and what cable do you use to connect to the Intel GPU?

21:48 <Tank_Master> amd 7970, and a geforce 680

21:49 <Tank_Master> dvi

21:49 <PovAddict> wow you have all flavors in there

21:49 <PovAddict> Tank_Master: someone reports "For me connecting a monitor via a displayport to dvi adapter did not work. Connecting a monitor with displayport cable worked"

21:49 <Tank_Master> i do testing for a second life viewer

21:49 <Tank_Master> there is no display port on this mobo, only DVI for the intel gpu

21:49 <RomW> Well the third link in the forum post Pov pointed to says the HD 4000 won't work unless it is configured as the primary display

21:49 <PovAddict> someone else reported that he had to disable his nVidia GPU (or not plug a monitor to it)

21:49 <PovAddict> and there's the primary display thing too, yeah

21:50 <PovAddict> but this is all anecdotal

21:50 <Tank_Master> atm i have 3 monitors pluged into the ati card, one into the intel

21:50 <PovAddict> the only official information I have seen is that you need a monitor connected, but not anything more specific

21:50 <Tank_Master> intel is the default adaptor in the bios, but wasnt the main screen in windows

21:52 <Tank_Master> humm

21:52 <Tank_Master> now that i set intel as the main display in windows, gpu cap sees it

21:52 <PovAddict> chances are BOINC will too

21:53 <RomW> Well, a better chance anyway

21:53 <RomW> :)

21:53 <RomW> This is a first try at debugging the new code

21:54 <Tank_Master> LOLOLOL

21:54 <Tank_Master> 12/7/2012 6:53:54 PM | | OpenCL: ATI GPU 0: Intel(R) HD Graphics 4000 (driver version, device version OpenCL 1.1, 1624MB, 1624MB available)

21:54 <PovAddict> RomW: hey I said BOINC would see it

21:54 <PovAddict> RomW: it's still possible it'll crash and burn when it does

21:54 <PovAddict> :)

21:54 <RomW> LOL

21:54 <PovAddict> or, whaddayaknow, classify it as ATI

21:54 <Tank_Master> see above, wats wrong with that?? :D

21:54 <RomW> Hmmm

21:55 <RomW> Lets see if it is fit and finish work or something more serious

21:55 <PovAddict> it's fit and finish

21:55 <Tank_Master> nothing is runing on it atm

21:56 <PovAddict> RomW: client/gpu_detect.cpp

21:56 <PovAddict> -        intel_gpu_opencls[i].description(buf, GPU_TYPE_ATI);

21:56 <PovAddict> +        intel_gpu_opencls[i].description(buf, GPU_TYPE_INTEL);

21:56 <RomW> No peek flops calculation though

21:56 <RomW> ah

21:56 <RomW> good catch

21:57 <Tank_Master>

21:57 <PovAddict> that *only* affects the description though

21:57 <Romulus> Title: boinc2 - (at

21:58 <Tank_Master> thats now how it shows at startup

21:59 <Tank_Master> oh hey, i got work from astroids now

22:01 <PovAddict> RomW: erm, I don't see the flops peak for *any* OpenCL line in his log

22:01 <PovAddict> RomW: only for CUDA and CAL

22:01 <Tank_Master> in the past, when the openCL SDK was installed, i would see "openCL device detected, not using"

22:02 <Tank_Master> for both the ati and nvidia cards

22:03 <PovAddict> RomW: can you get Tank_Master a Windows build with this patch?

22:03 <PovAddict> so we know if the flops are being calculated properly

22:04 <Tank_Master> is there a howto on compiling boinc for windows?

22:06 *** smj has quit IRC

22:09 <RomW> TM, try this drop:

22:10 <RomW> Well, there is code to do it, but I don't think it is being called at the moment

22:11 <Tank_Master> 12/7/2012 7:11:27 PM | | Suspending computation - CPU benchmarks in progress

22:11 <PovAddict> nice timing

22:11 <Tank_Master> yeah, that looks better

22:11 <Tank_Master> 12/7/2012 7:11:26 PM | | OpenCL: intel_gpu GPU 0: Intel(R) HD Graphics 4000 (driver version, device version OpenCL 1.1, 1624MB, 1624MB available)

22:12 <Tank_Master> though shouldnt it be 'Intel GPU', not intel_gpu?

22:13 <PovAddict> huh

22:13 <Tank_Master> (symantics, i know)

22:13 <PovAddict> RomW: did you use my patch?

22:13 <PovAddict> RomW: the coproc.cpp change doesn't seem to be there

22:14 <RomW> I only saw the change to gpu_detect.cpp

22:14 <RomW> in IRC

22:14 <PovAddict> see

22:15 <Tank_Master>

22:15 <Romulus> Title: boinc3 - (at

22:15 <RomW> ah

22:15 <PovAddict> Tank_Master: ahh the benchmark re-run is because of the version number change, not just bad timing

22:15 <Tank_Master> ya

22:17 <RomW> The OpenCL properties are just an add-on to the existing property blocks for the other card types

22:18 <RomW> general assumtion was there was always going to be a native property block

22:18 <RomW> Intel kind of threw a monkey wrench into that

22:21 <RomW> TM, could you send me your request file for a project like Collatz?

22:21 <RomW> I would like to see how the Intel GPU is send to the scheduler

22:22 <RomW> Be careful about losting it to pastebin or some such, it'll contain your authenticator

22:22 <Zumu> PovAddict: I used the same email address

22:23 <Zumu> valer: I didn't type it, used autocompletition

22:23 <Tank_Master> 12/7/2012 7:14:42 PM | Server for testing Bolpex | [sched_op] intel_gpu work request: 86580.00 seconds; 0.07 devices

22:23 <PovAddict> RomW: I once had to contact SETI admins to get an authenticator changed because someone posted his in this channel...

22:23 <Tank_Master> is that what your looking for?

22:24 *** Zumu has quit IRC

22:24 *** Zumu has joined #boinc

22:24 <PovAddict> Tank_Master: no, the sched_request_*.xml file

22:24 <PovAddict> for that project

22:24 <PovAddict> (send it *privately* to Rom)

22:25 <Tank_Master>

22:25 <Romulus> <> (at

22:26 <PovAddict> pri-va-te-ly

22:26 <PovAddict> your account key is in there!

22:26 <Tank_Master> did you get it RomW?

22:26 <RomW> Ugh, delete it asap

22:26 <Tank_Master> done

22:26 <PovAddict> you should've censored it at least

22:26 <Tank_Master> meh

22:27 <Tank_Master> or pmed it to him

22:27 <Tank_Master> but i was lazy

22:27 <PovAddict> Tank_Master: now I can login to the gpugrid forums with your account and insult the admin \o/

22:27 <RomW> lol

22:28 <Tank_Master> i rarly log into forums

22:28 <PovAddict> come to think of it, I twice got the account key *of* an admin

22:46 <RomW> be back in a bit

22:46 <RomW> hooking up the XML

22:46 <RomW> So servers can see it

22:47 <Tank_Master> k

22:48 <Tank_Master> are you going to make the intel string similer to the others? (not intel_gpu)

22:49 <PovAddict> Tank_Master: I think something is exactly backwards wrt that

22:49 <Tank_Master> ?

22:50 <PovAddict> Tank_Master:

22:50 <Romulus> Title: #622004 KDE Pastebin (at

22:50 <PovAddict> Tank_Master: I think the log message should use the second function rather than the first

22:51 <Tank_Master> is what i see

22:51 <Romulus> Title: request - (at

22:51 <PovAddict> yeah, same problem in the init messages

22:51 <Tank_Master> yeah

22:52 <Tank_Master> but.. its a good start

22:53 <Tank_Master> at least thigns are rollin

23:35 <RomW> TM, are you still up?

23:36 <Tank_Master> nope!

23:36 <Tank_Master> :D

23:36 <RomW> LOL

23:36 <Tank_Master> im pacific TZ

23:37 <RomW> Can you try this one...

23:38 <RomW> I'm interested in both the message log and the sched_request

23:40 <Tank_Master>

23:40 <Romulus> Title: boinc 4 - (at

23:41 <PovAddict> tsk tsk, still not printing the flops

23:42 <RomW> doah, I was actually trying for that too, but as part of COPROC_INTEL instead of messing with the OpenCL structure

23:42 <Tank_Master>

23:42 <Romulus> Title: request2 - (at

23:43 <Tank_Master> RomW, get my pm?

23:57 <Tank_Master> ill bb in a hr

23:57 <Tank_Master> thx for your time, RomW

Generated by 2.4 by Marius Gedminas - find it at!