ATI OpenCL MultiBeam 6.10 problem..

Message boards : Number crunching : ATI OpenCL MultiBeam 6.10 problem..
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · 5 · 6 · 7 · Next

AuthorMessage
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14676
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1086050 - Posted: 11 Mar 2011, 12:38:00 UTC - in response to Message 1086048.  

Aaaaaarrgghh! He's using v6.10.58, and that doesn't do ATI properly yet - no ati_compare.

Upgrade to v6.12.18, or use Claggy's workround - I'm outa here.

hmm...i didn't even know there was a BOINC v6.12.18. even the BOINC website still only has v6.10.58 for download. where would i even get that? a google search doesn't even bring anything up about v6.12.18?

It's still in testing, but close to "release candidate" status - and probably in a better state than the ATI drivers and OpenCL application you're trying to use.

Click the all versions link on the normal download page.
ID: 1086050 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 5 Jul 99
Posts: 4654
Credit: 47,537,079
RAC: 4
United Kingdom
Message 1086062 - Posted: 11 Mar 2011, 13:25:28 UTC - in response to Message 1086048.  
Last modified: 11 Mar 2011, 13:47:53 UTC

...btw Claggy, i did put the alternative cc-config.xml file you suggested above into the BOINC data directory after removing the one with the <use_all_gpus> directive in it. and that actually caused all tasks to error out, including the first one, which i wasn't have a problem with before.


Which device number did you put in instead of 'n' ?, i think it should be 0, i initially said 1,

ATI GPU 0: ATI Radeon HD 2300/2400/3200 (RV610) (CAL version 1.4.900, 341MB, 56 GFLOPS peak)
ATI GPU 1: ATI Radeon HD5800 series (Cypress) (CAL version 1.4.900, 2048MB, 2720 GFLOPS peak)

The confusion crept in because with OpenCL there's only one device, and there the HD5870 is device 0:

Stderr output
<core_client_version>6.10.58</core_client_version>
<![CDATA[
<stderr_txt>
Number of period iterations for PulseFind setted to:2
Number of app instances per device setted to:1
Running on device number: 0
Priority of worker thread raised successfully
Priority of process adjusted successfully, below normal priority class used
OpenCL platform detected: Advanced Micro Devices, Inc.
BOINC assigns 0 device, slots 0 to 0 (including) will be checked
Used slot is 0; Number of period iterations for PulseFind setted to:2
Number of app instances per device setted to:1
Running on device number: 0
Priority of worker thread raised successfully
Priority of process adjusted successfully, below normal priority class used
OpenCL platform detected: Advanced Micro Devices, Inc.
BOINC assigns 0 device, slots 0 to 0 (including) will be checked
Used slot is 0; Info : Building Program (clBuildProgram):main kernels: OK code 0

ar=0.416752 NumCfft=198491 NumGauss= 1134169420 NumPulse= 122038058496 NumTriplet= 17004895469568
Windows optimized S@H Enhanced application by Alex Kan
Version info: SSE3x (AMD/Intel, Core 2-optimized v8-nographics) V5.13 by Alex Kan
SSE3x Win32 Build 177 , Ported by : Jason G, Raistmer, JDWhale

OpenCL version by Raistmer, rev177

Build features: Non-graphics OpenCL USE_OPENCL_HD5xxx IPP AMD specific USE_SSE3 x86
CPUID: AMD Phenom(tm) II X6 1090T Processor
Speed: 6 x 3712 MHz
Cache: L1=64K L2=512K

CPU features: FPU TSC PAE CMPXCHG8B APIC SYSENTER MTRR CMOV/CCMP MMX FXSAVE/FXRSTOR SSE SSE2 HT SSE3
CPU type 0x42
Number of OpenCL platforms: 1


OpenCL Platform Name: ATI Stream
Number of devices: 1
Max compute units: 20
Max work group size: 256
Max clock frequency: 850Mhz
Max memory allocation: 268435456
Cache type: None
Cache line size: 0
Cache size: 0
Global memory size: 1073741824
Constant buffer size: 65536
Max number of constant args: 8
Local memory type: Scratchpad
Local memory size: 32768
Queue properties:
Out-of-Order: No
Name: Cypress
Vendor: Advanced Micro Devices, Inc.
Driver version: CAL 1.4.900
Version: OpenCL 1.1 ATI-Stream-v2.3 (451)
Extensions: cl_amd_fp64 cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_3d_image_writes cl_khr_byte_addressable_store cl_khr_gl_sharing cl_amd_device_attribute_query cl_amd_printf cl_amd_media_ops cl_amd_popcnt


Claggy
ID: 1086062 · Report as offensive
Profile Sunny129
Avatar

Send message
Joined: 7 Nov 00
Posts: 190
Credit: 3,163,755
RAC: 0
United States
Message 1086066 - Posted: 11 Mar 2011, 14:22:16 UTC

yes, i used device "0" in the cc_config.xml you suggested to me. it read exactly like this:

<cc_config>
<options>
<ignore_ati_dev>0</ignore_ati_dev>
</options>
</cc_config>


anyways, let me try BOINC v6.12.18 when i get home and see if it works in conjunction with the cc_config.xml fix. by the way, i have lots of tasks in the que - am i going to lose all of those when i update BOINC? can i just install v6.12.18 over v6.10.58, or do i have to treat it like i would drivers (uninstall the old version first, then install the new one)?
ID: 1086066 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14676
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1086070 - Posted: 11 Mar 2011, 14:51:31 UTC - in response to Message 1086066.  

anyways, let me try BOINC v6.12.18 when i get home and see if it works in conjunction with the cc_config.xml fix. by the way, i have lots of tasks in the que - am i going to lose all of those when i update BOINC? can i just install v6.12.18 over v6.10.58, or do i have to treat it like i would drivers (uninstall the old version first, then install the new one)?

I suggest you use either v6.12.18, or a cc_config.xml file - not both at once.

I've lost count of the number of times I've installed test versions of BOINC v6.12 - probably over 50, given the number in my download archive and the number of machines I test on. Not lost any tasks yet. But, having said that, take precautions if you feel the need. Disable network activity, shut down BOINC, and backup the entire "D:\Documents and Settings\All Users\Application Data\BOINC" folder structure, including all sub-folders.

As you'll have gathered from your experiences to date, care and attention to detail are important when you venture into the realms of test and non-standard applications. Make sure you download the correct version (32-bit or 64-bit) of BOINC - the normal page detects the OS for you, on the test page you have to do it yourself. By "correct", I mean the same as you had before (32-bit it looks like - but check yourself) - that makes the upgrade easier. And read the installer screens - check in particular that that D:\... data drive has been picked up correctly.
ID: 1086070 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 5 Jul 99
Posts: 4654
Credit: 47,537,079
RAC: 4
United Kingdom
Message 1086072 - Posted: 11 Mar 2011, 14:54:23 UTC - in response to Message 1086066.  
Last modified: 11 Mar 2011, 14:58:43 UTC

You can install Boinc 6.12.18 straight over the top of 6.10.58, you shouldn't loose any Wu's,

But to be safe, suspend all the Wu's first, then you can see which GPU is detected, before unsuspending a few to see if they run correctly,
(only suspend the running tasks, after you're suspended the unstarted tasks, Boinc 6.10.58 has a bug where if you suspend everything,
as Boinc ripples through the tasks, some of them get marked as started, but the slot for that Wu doesn't get filled, so errors out next time that Wu is attempted)

Remember 6.12.18 is an Alpha build, it may still have unreported bugs in it, the only Bugs i've reported is it doesn't close Collatz Cuda tasks down on exit,
and errors out the next task if i suspend a running Cuda task, both issues are probably due to my 128Mb 8400M GS being very slow, and lacking in memory,

Claggy

Edit: or do as Richard suggests,
ID: 1086072 · Report as offensive
Profile Sunny129
Avatar

Send message
Joined: 7 Nov 00
Posts: 190
Credit: 3,163,755
RAC: 0
United States
Message 1086077 - Posted: 11 Mar 2011, 15:10:38 UTC - in response to Message 1086070.  

I suggest you use either v6.12.18, or a cc_config.xml file - not both at once.

well as i stated previously, i had problems using the suggested cc_config.xml file w/ BOINC v6.10.58. so i suppose the next logical step is to give v6.12.18 a try without the help of a cc_config.xml file. if i still have problems, i'll add the following cc_config.xml file to the BOINC data directory and see it that fixes things:

<cc_config>
<options>
<ignore_ati_dev>n</ignore_ati_dev>
</options>
</cc_config>


thanks again guys for the guidance.
ID: 1086077 · Report as offensive
Profile BilBg
Volunteer tester
Avatar

Send message
Joined: 27 May 07
Posts: 3720
Credit: 9,385,827
RAC: 0
Bulgaria
Message 1086122 - Posted: 11 Mar 2011, 18:05:40 UTC - in response to Message 1086077.  

I suggest you use either v6.12.18, or a cc_config.xml file - not both at once.

well as i stated previously, i had problems using the suggested cc_config.xml file w/ BOINC v6.10.58. so i suppose the next logical step is to give v6.12.18 a try without the help of a cc_config.xml file.
if i still have problems, i'll add the following cc_config.xml file to the BOINC data directory and see it that fixes things:

<cc_config>
<options>
<ignore_ati_dev>n</ignore_ati_dev>
</options>
</cc_config>


thanks again guys for the guidance.


How can line like this:
<ignore_ati_dev>n</ignore_ati_dev>
"fix things"??

"n" have to be substituted by 0 or 1 or 2 ..., e.g.:
<ignore_ati_dev>0</ignore_ati_dev>
or
<ignore_ati_dev>1</ignore_ati_dev>


 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 
ID: 1086122 · Report as offensive
Profile Sunny129
Avatar

Send message
Joined: 7 Nov 00
Posts: 190
Credit: 3,163,755
RAC: 0
United States
Message 1086130 - Posted: 11 Mar 2011, 18:37:18 UTC - in response to Message 1086122.  


How can line like this:
<ignore_ati_dev>n</ignore_ati_dev>
"fix things"??

"n" have to be substituted by 0 or 1 or 2 ..., e.g.:
<ignore_ati_dev>0</ignore_ati_dev>
or
<ignore_ati_dev>1</ignore_ati_dev>


yes i know...that "n" should be a "0" since GPU_0 is my HD 3300 integrated video and GPU_1 is my HD 5870. i tried to edit my above post to reflect that, but by the time i realized it, it was too late...i can no longer edit that post. if you look up 5 or 6 posts, you'll see that i did in fact try that cc_config.xml file under BOINC v6.10.58 to no avail (and yes i remembered to replace n with 0 when i tried it).
ID: 1086130 · Report as offensive
Profile Gundolf Jahn

Send message
Joined: 19 Sep 00
Posts: 3184
Credit: 446,358
RAC: 0
Germany
Message 1086149 - Posted: 11 Mar 2011, 19:25:50 UTC - in response to Message 1086130.  

(and yes i remembered to replace n with 0 when i tried it)

Did you also try to replace n with 1, in case the numbering in message 1085649 was the right one?

Gruß,
Gundolf
ID: 1086149 · Report as offensive
Profile Sunny129
Avatar

Send message
Joined: 7 Nov 00
Posts: 190
Credit: 3,163,755
RAC: 0
United States
Message 1086151 - Posted: 11 Mar 2011, 19:33:26 UTC - in response to Message 1086149.  
Last modified: 11 Mar 2011, 19:33:48 UTC

(and yes i remembered to replace n with 0 when i tried it)

Did you also try to replace n with 1, in case the numbering in message 1085649 was the right one?

Gruß,
Gundolf

no i haven't, b/c the BOINC start-up dialogue told me that GPU_0 was in fact the HD 3300 onboard video (which is of course the one i would like to disable in BOINC). nevertheless, i suppose i should try the cc_config.xml file with the number n=1 before i update BOINC to v6.12.18 just to be sure...i'll do that first and let you know if anything positive comes of it.
ID: 1086151 · Report as offensive
Profile Raistmer
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 16 Jun 01
Posts: 6325
Credit: 106,370,077
RAC: 121
Russia
Message 1086190 - Posted: 11 Mar 2011, 22:40:44 UTC - in response to Message 1086050.  

It's still in testing, but close to "release candidate" status - and probably in a better state than the ATI drivers and OpenCL application you're trying to use.


I would not put all together. App status is not "release candidate", it's released :P
ID: 1086190 · Report as offensive
Profile Raistmer
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 16 Jun 01
Posts: 6325
Credit: 106,370,077
RAC: 121
Russia
Message 1086191 - Posted: 11 Mar 2011, 22:44:01 UTC

And speaking more seriously, "release candidate" could not mix so different cards and provide single APP plan for them, knowing that HD3xxx only brook/cal capable. Do we have OpenCL plan class already? W/o it I use ati13ati as bandage...
ID: 1086191 · Report as offensive
Profile BilBg
Volunteer tester
Avatar

Send message
Joined: 27 May 07
Posts: 3720
Credit: 9,385,827
RAC: 0
Bulgaria
Message 1086205 - Posted: 11 Mar 2011, 23:19:14 UTC - in response to Message 1086130.  


How can line like this:
<ignore_ati_dev>n</ignore_ati_dev>
"fix things"??

"n" have to be substituted by 0 or 1 or 2 ..., e.g.:
<ignore_ati_dev>0</ignore_ati_dev>
or
<ignore_ati_dev>1</ignore_ati_dev>


yes i know...that "n" should be a "0" since GPU_0 is my HD 3300 integrated video and GPU_1 is my HD 5870. i tried to edit my above post to reflect that, but by the time i realized it, it was too late...i can no longer edit that post. if you look up 5 or 6 posts, you'll see that i did in fact try that cc_config.xml file under BOINC v6.10.58 to no avail (and yes i remembered to replace n with 0 when i tried it).


I read all posts before posting mine.

But the last of yours sounded like you want to say:
"because I tried <ignore_ati_dev>0</ignore_ati_dev> (5 posts above) and it has no effect
I will try <ignore_ati_dev>n</ignore_ati_dev> in hope it will (magically) fix things
(maybe BOINC knows what to disable and will auto-choose the "n" device)"


 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 
ID: 1086205 · Report as offensive
Profile Sunny129
Avatar

Send message
Joined: 7 Nov 00
Posts: 190
Credit: 3,163,755
RAC: 0
United States
Message 1086209 - Posted: 11 Mar 2011, 23:25:02 UTC
Last modified: 11 Mar 2011, 23:27:02 UTC

I read all posts before posting mine.

But the last of yours sounded like you want to say:
"because I tried <ignore_ati_dev>0</ignore_ati_dev> (5 posts above) and it has no effect
I will try <ignore_ati_dev>n</ignore_ati_dev> in hope it will (magically) fix things
(maybe BOINC knows what to disable and will auto-choose the "n" device)"


lol sorry, i didn't mean to give you the impression that i literally thought the letter "n" would fix my problem.


at any rate, i think the problem is solved, but i'm going to let a few WU's run to completion just to be sure. i went ahead and replaced n=0 with n=1 despite the fact that BOINC's start-up dialogue is telling me the GPU_0 is my HD 3300 onboard video. with the <ignore_ati_dev>1</ignore_ati_dev> directive in the cc_config.xml file, S@H MB GPU tasks are no longer erroring out. granted, my 5870 is still only crunching one task at a time, but the important thing is that when i resume a 2nd task while one is already crunching, it goes to "ready to start" status, rather than erroring out the instant i resume it. BillBG pointed out earlier in the thread how to run 2 tasks at once, but i'm going to hold on that until i know everything is stable and without errors.
ID: 1086209 · Report as offensive
Profile BilBg
Volunteer tester
Avatar

Send message
Joined: 27 May 07
Posts: 3720
Credit: 9,385,827
RAC: 0
Bulgaria
Message 1086213 - Posted: 11 Mar 2011, 23:36:21 UTC - in response to Message 1086209.  
Last modified: 11 Mar 2011, 23:43:50 UTC

BilBG pointed out earlier in the thread how to run 2 tasks at once, but i'm going to hold on that until i know everything is stable and without errors.


Raistmer (as the main/only programmer of the ATI apps) can tell you exactly what to change.

<count> is one of the things to change but "there is another" (Yoda) :)


Do you use GPU-Z to monitor temperature and load of the GPUs?
http://www.techpowerup.com/gpuz/


 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 
ID: 1086213 · Report as offensive
Profile Sunny129
Avatar

Send message
Joined: 7 Nov 00
Posts: 190
Credit: 3,163,755
RAC: 0
United States
Message 1086218 - Posted: 11 Mar 2011, 23:53:05 UTC - in response to Message 1086213.  
Last modified: 11 Mar 2011, 23:57:08 UTC

Raistmer (as the main/only programmer of the ATI apps) can tell you exactly what to change.

<count> is one of the things to change but "there is another" (Yoda) :)


thanks for the tip Jedi Master ;)


on that note, none of my S@H GPU tasks are currently "suspended" (i.e. they are all "ready to start" in the que) while waiting for the 1 active task to finish before the next one begins, and i'm getting zero errors! yay!

the only other issue i've encountered since adding the cc_config.xml is a Milkyway@Home issue. i figured it would be a good idea to test out the one other GPU app i use, and upon resuming my MW@H tasks, i noticed that my 5870 only crunches 1 MW@H task at a time now, when it used to crunch 2 at a time. it really doesn't bother me, as it used to take ~3:00 min. to crunch 2 MW@H tasks, and now it takes ~1:30 min. to crunch 1 MW@H task. so it seems that despite going from crunching 2 MW@H tasks at a time to crunching only 1 at a time, my 5870's MW@H crunchng efficiency is still essentially the same as before. so like i said, it really doesn't bother me since the "average time per WU" hasn't changed, but i am still curious as to why that cc_config.xml file made a difference in the number of active simultaneous MW@H tasks.


Do you use GPU-Z to monitor temperature and load of the GPUs?
http://www.techpowerup.com/gpuz/


actually i'm using MSI Afterburner b/c it allows me to both monitor vitals AND manipulate them, whereas GPU-Z only allows me to monitor. sometimes i run them both at simultaneously out of paranoia. but rest assured that i'm keeping temps under control.
ID: 1086218 · Report as offensive
Profile dnolan
Avatar

Send message
Joined: 30 Aug 01
Posts: 1228
Credit: 47,779,411
RAC: 32
United States
Message 1086219 - Posted: 11 Mar 2011, 23:54:23 UTC
Last modified: 11 Mar 2011, 23:55:12 UTC

To run 2 at a time, you would want to change [edit] in app_info.xml [\edit]
<count>1</count>
to
<count>0.5</count>
and change
<cmdline></cmdline>
to
<cmdline>-instances_per_device 2</cmdline>
or add the -instances_per_device 2 into whatever other commands you already have.

-Dave
ID: 1086219 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14676
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1086226 - Posted: 12 Mar 2011, 0:11:03 UTC

May I, yet again, plead that you - everyone - try to understand what is going on, and propose changes that are appropriate both to the hardware being used, and the skill level of the person using it? We all might learn something then.

If setting <ignore_ati_dev>1</ignore_ati_dev> results in the device which BOINC itself describes as device 0 being ignored, and device 1 being used without error, then that's good to know. I'm not quite sure how we're going to write it up in the FAQ, though.

But, what do the startup Event Log (now you're using v6.12) messages say? Which device is flagged for use, and which is being ignored? What does the stderr_txt of the successful tasks say about the device they ran on?
ID: 1086226 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 5 Jul 99
Posts: 4654
Credit: 47,537,079
RAC: 4
United Kingdom
Message 1086233 - Posted: 12 Mar 2011, 0:21:04 UTC - in response to Message 1086226.  
Last modified: 12 Mar 2011, 0:29:48 UTC

Richard, it's even better than that, Sunny129 is still using Boinc 6.10.58,
the site is showing he's got a CAL ATI Radeon HD 2300/2400/3200 (RV610) (341MB) driver: 1.4.900,
and he's crunching with OpenCL on his disabled HD5870 via app_info,

What i want to know is why Wu's errored out when he disabled the HD3300,

Claggy
ID: 1086233 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 14676
Credit: 200,643,578
RAC: 874
United Kingdom
Message 1086235 - Posted: 12 Mar 2011, 0:27:23 UTC - in response to Message 1086233.  

Richard, it's even better than that...

Oh, I give up on ATI. I'm going to bed ;-)
ID: 1086235 · Report as offensive
Previous · 1 · 2 · 3 · 4 · 5 · 6 · 7 · Next

Message boards : Number crunching : ATI OpenCL MultiBeam 6.10 problem..


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.