Deprecated: Function get_magic_quotes_gpc() is deprecated in /disks/centurion/b/carolyn/b/home/boincadm/projects/beta/html/inc/util.inc on line 663
SET@Home v7 6.99 ati_opencl_sah

SET@Home v7 6.99 ati_opencl_sah

Message boards : SETI@home Enhanced : SET@Home v7 6.99 ati_opencl_sah
Message board moderation

To post messages, you must log in.

AuthorMessage
Yo
Volunteer tester

Send message
Joined: 15 May 12
Posts: 5
Credit: 80,761
RAC: 0
Montenegro
Message 44083 - Posted: 17 Oct 2012, 9:08:04 UTC
Last modified: 17 Oct 2012, 9:10:31 UTC

ATI HD4650
CAT 12.6

[stderr.txt]
WARNING: can't open binary kernel file for oclFFT plan: C:\ProgramData\BOINC/projects/setiweb.ssl.berkeley.edu_beta\clFFTplan_ATIRV730_524288_r1643.bin, continue with recompile...
WARNING: patching required max_kernel_wg_size=32

ATI GPU usage> 0%
WU progress> 0%
ID: 44083 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 29 May 06
Posts: 1037
Credit: 8,440,339
RAC: 0
United Kingdom
Message 44084 - Posted: 17 Oct 2012, 9:33:00 UTC - in response to Message 44083.  

It would be helpful if you unhid your computers,

Claggy
ID: 44084 · Report as offensive
Yo
Volunteer tester

Send message
Joined: 15 May 12
Posts: 5
Credit: 80,761
RAC: 0
Montenegro
Message 44085 - Posted: 17 Oct 2012, 9:36:12 UTC

Done
ID: 44085 · Report as offensive
Profile Mike
Volunteer tester
Avatar

Send message
Joined: 16 Jun 05
Posts: 2530
Credit: 1,074,556
RAC: 0
Germany
Message 44086 - Posted: 17 Oct 2012, 9:39:56 UTC

Not really.

Have you checked with GPUz if OpenCL is active ?


With each crime and every kindness we birth our future.
ID: 44086 · Report as offensive
Yo
Volunteer tester

Send message
Joined: 15 May 12
Posts: 5
Credit: 80,761
RAC: 0
Montenegro
Message 44087 - Posted: 17 Oct 2012, 9:47:02 UTC
Last modified: 17 Oct 2012, 10:26:36 UTC

I just edited S@h beta preferences.
Computer ID: 58720

AstroPulse v6 v6.04 (ati_opencl_100) working fine.

Aborted WU. http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=4156027
ID: 44087 · Report as offensive
Yo
Volunteer tester

Send message
Joined: 15 May 12
Posts: 5
Credit: 80,761
RAC: 0
Montenegro
Message 44088 - Posted: 17 Oct 2012, 10:36:42 UTC

Another WU (suspended at the moment):
[stderr.txt]

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 device 0
Info: BOINC provided device ID used
Used GPU device parameters are:
Number of compute units: 8
Single buffer allocation size: 64MB
max WG size: 128

Build features: SETI7 Non-graphics OpenCL OCL_CHIRP3 ASYNC_SPIKE FFTW AMD specific USE_SSE x86
CPUID: Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz

Cache: L1=64K L2=256K

CPU features: FPU TSC PAE CMPXCHG8B APIC SYSENTER MTRR CMOV/CCMP MMX FXSAVE/FXRSTOR SSE SSE2 HT SSE3
OpenCL-kernels filename : MultiBeam_Kernels_r1643.cl
Info : Building Program (binary, clBuildProgram):main kernels: OK code 0

WARNING: can't open binary kernel file for oclFFT plan: C:\ProgramData\BOINC/projects/setiweb.ssl.berkeley.edu_beta\clFFTplan_ATIRV730_524288_r1643.bin, continue with recompile...
WARNING: patching required max_kernel_wg_size=32
WARNING: can't open binary kernel file for oclFFT plan: C:\ProgramData\BOINC/projects/setiweb.ssl.berkeley.edu_beta\clFFTplan_ATIRV730_8_r1643.bin, continue with recompile...
WARNING: patching required max_kernel_wg_size=32
WARNING: patching required max_kernel_wg_size=32
WARNING: patching required max_kernel_wg_size=32
WARNING: ... repeated hundreds of time ...
ID: 44088 · Report as offensive
Profile Mike
Volunteer tester
Avatar

Send message
Joined: 16 Jun 05
Posts: 2530
Credit: 1,074,556
RAC: 0
Germany
Message 44089 - Posted: 17 Oct 2012, 12:07:07 UTC

The problem is HD 4650 only has max work group size of 128 while 256 is needed for MB apps.

Sorry.

You only can run APs on that card.


With each crime and every kindness we birth our future.
ID: 44089 · Report as offensive
Yo
Volunteer tester

Send message
Joined: 15 May 12
Posts: 5
Credit: 80,761
RAC: 0
Montenegro
Message 44090 - Posted: 17 Oct 2012, 12:54:39 UTC

OK
Thanks
ID: 44090 · Report as offensive
Christoph
Volunteer tester

Send message
Joined: 16 Oct 09
Posts: 58
Credit: 662,990
RAC: 0
Germany
Message 44159 - Posted: 18 Oct 2012, 19:41:20 UTC - in response to Message 44089.  
Last modified: 18 Oct 2012, 19:41:50 UTC

The problem is HD 4650 only has max work group size of 128 while 256 is needed for MB apps.

Sorry.

You only can run APs on that card.


Sorry, again not 100% true.
My HD 5450 also has max 128 and is processing MBs.
One already validated: http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=4158882

So it is the same as with APs, it will work also with 128 at least on some cards.
Christoph
ID: 44159 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 44160 - Posted: 18 Oct 2012, 19:43:52 UTC - in response to Message 44159.  
Last modified: 18 Oct 2012, 19:44:01 UTC

Yes, HD4xxx with WG size of 128 will most probably not work.
ID: 44160 · Report as offensive
Christoph
Volunteer tester

Send message
Joined: 16 Oct 09
Posts: 58
Credit: 662,990
RAC: 0
Germany
Message 44161 - Posted: 18 Oct 2012, 19:56:58 UTC

I understood that max WG size 128 would not work at all never mind the card.
If you and Mike did mean that only HD 4xxx with 128 will not work then I didn't get that correct first.
Christoph
ID: 44161 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 44162 - Posted: 18 Oct 2012, 20:12:25 UTC - in response to Message 44161.  
Last modified: 18 Oct 2012, 20:13:46 UTC

I understood that max WG size 128 would not work at all never mind the card.
If you and Mike did mean that only HD 4xxx with 128 will not work then I didn't get that correct first.


Basicly, it's LHD4k build problem. You could try to find some posts on SETI main and here with this keyword. That build was aimed specifically to HD4xxx with WG128 but never worked stable, unfortunately.
Cause ATi refused for any responsibility for OpenCL on whole HD4xxx family (always "beta OpenCL support", from the start till present time) we can't make them fix their drivers. And for some reason some of oclFFT kernels (most probably) caused driver restart.

HD5xxx with WG128 support was added on some stage already.
ID: 44162 · Report as offensive
Profile Mike
Volunteer tester
Avatar

Send message
Joined: 16 Jun 05
Posts: 2530
Credit: 1,074,556
RAC: 0
Germany
Message 44164 - Posted: 18 Oct 2012, 20:50:14 UTC - in response to Message 44159.  
Last modified: 18 Oct 2012, 20:53:52 UTC

I might add that astropulse is complete different app and has nothing to do with MB.
With each crime and every kindness we birth our future.
ID: 44164 · Report as offensive
Father Ambrose
Volunteer tester

Send message
Joined: 1 May 07
Posts: 556
Credit: 6,470,846
RAC: 0
United Kingdom
Message 44175 - Posted: 19 Oct 2012, 9:44:52 UTC

Hi claggy

have been downloaded several ATI 6.99 WU's tried running

This is a straight stock download.(with nothing added(COULD SAY SET AND FORGET DOWN LOAD)

no progress and about 1% -2% GPU in burst.

After suspending and resuming return to zero time

host 50814 hd4600 GPU GPUz indicates openCL this WU has been left ruuning.05ap10al.27521.4162.140733193388038.14.76_1

Am I missing something.


Michael




ID: 44175 · Report as offensive
Profile Mike
Volunteer tester
Avatar

Send message
Joined: 16 Jun 05
Posts: 2530
Credit: 1,074,556
RAC: 0
Germany
Message 44180 - Posted: 19 Oct 2012, 12:04:39 UTC

If you read my earlier posts you will notice this app won´t work on your GPU i`m afraid.

RV 770 is minimum requirement.

HD 4 only has OpenCL beta Level support.

regards
Mike

With each crime and every kindness we birth our future.
ID: 44180 · Report as offensive
Stick
Volunteer tester

Send message
Joined: 16 May 06
Posts: 150
Credit: 136,942
RAC: 0
United States
Message 44197 - Posted: 19 Oct 2012, 22:15:02 UTC

My first v7 6.99 ati_opencl_sah task is stuck at the 0.015% mark. The symptoms are very similar to problems I encountered (about 6 months ago) with the Lunatics 6.10 app and which I described in the old Seti Main Catalyst 12.4 and Lunatics 6.10 (ati13ati) thread. That thread generated a lot of discussion and suggestions but (bottom line) nothing helped and I tentatively concluded that my Toshiba Satellite's GPU's 384MB was the limiting factor in the Cat v12.4/6.10 (ati13b) incompatibility. At that time I reverted back to Cat 12.1 and started running the lunatics app successfully again. However, I recently tested Catalyst 12.8. Although the problem w/Lunatics 6.10 immediately returned, I (somewhat unexpectedly) found that with Cat 12.8, the GPU processes AP tasks without any problems. Here, it runs the stock AstroPulse v6 v6.04 (opencl_ati_100) app and on Main it runs Raistmer's Windows x86 rev 1305 app. Hopefully, that's an indication that, with some tinkering, v7 ati_opencl_sah can be made to work, too. And I will be happy to help out in that regard. Just let me know what additional info may be needed.
ID: 44197 · Report as offensive
Profile Mike
Volunteer tester
Avatar

Send message
Joined: 16 Jun 05
Posts: 2530
Credit: 1,074,556
RAC: 0
Germany
Message 44199 - Posted: 19 Oct 2012, 22:27:32 UTC

Maybe you could suspend CPU work until the first unit is finnished.
I will check the result then.
Its just a fact that 12.8 needs more CPU cycles.

With each crime and every kindness we birth our future.
ID: 44199 · Report as offensive
Stick
Volunteer tester

Send message
Joined: 16 May 06
Posts: 150
Credit: 136,942
RAC: 0
United States
Message 44203 - Posted: 20 Oct 2012, 2:56:02 UTC - in response to Message 44199.  

I suspended all other work - but that didn't help. It's still stuck at 0.015%. And, if I suspend/restart it, both time and progress reset to zero and then, after about a minute, it returns to 0.015% and stays there. (This may be the same thing that Father Ambrose reported.)

I also noticed that the WU had two v7 6.99 ati_opencl_sah tasks error out before it was sent to me. So, I would guess that v7 6.99 ati_opencl_sah has more problems going right now than the one I am reporting. Therefore, I think I'll abort this task and try again when the next version comes out.
ID: 44203 · Report as offensive
Richard Haselgrove
Volunteer tester

Send message
Joined: 3 Jan 07
Posts: 1451
Credit: 3,272,268
RAC: 0
United Kingdom
Message 44207 - Posted: 20 Oct 2012, 8:35:18 UTC - in response to Message 44203.  

I suspended all other work - but that didn't help. It's still stuck at 0.015%. And, if I suspend/restart it, both time and progress reset to zero and then, after about a minute, it returns to 0.015% and stays there. (This may be the same thing that Father Ambrose reported.)

I also noticed that the WU had two v7 6.99 ati_opencl_sah tasks error out before it was sent to me. So, I would guess that v7 6.99 ati_opencl_sah has more problems going right now than the one I am reporting. Therefore, I think I'll abort this task and try again when the next version comes out.

I see the two replacements have gone to stock and cuda. I might try downloading the WU and running it on the bench, just for kicks.
ID: 44207 · Report as offensive
Profile Mike
Volunteer tester
Avatar

Send message
Joined: 16 Jun 05
Posts: 2530
Credit: 1,074,556
RAC: 0
Germany
Message 44208 - Posted: 20 Oct 2012, 8:51:11 UTC - in response to Message 44203.  

I suspended all other work - but that didn't help. It's still stuck at 0.015%. And, if I suspend/restart it, both time and progress reset to zero and then, after about a minute, it returns to 0.015% and stays there. (This may be the same thing that Father Ambrose reported.)

I also noticed that the WU had two v7 6.99 ati_opencl_sah tasks error out before it was sent to me. So, I would guess that v7 6.99 ati_opencl_sah has more problems going right now than the one I am reporting. Therefore, I think I'll abort this task and try again when the next version comes out.


Both hosts are using rather old drivers 10.11 and 11.5.
That indicates Rastmers suggestion that at least SDK 2.6 based drivers are required.
10.11 was a very stable driver for example.

With each crime and every kindness we birth our future.
ID: 44208 · Report as offensive

Message boards : SETI@home Enhanced : SET@Home v7 6.99 ati_opencl_sah


 
©2023 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.