[OpenCL] app v1.20/v1.21 feedback thread

HV Margus*
HV Margus*
Joined: 10 Jan 12
Posts: 17
Credit: 45000
RAC: 0

Oliver i tested debs again

Message 79050 in response to message 79049

Oliver i tested debs again somehow it not work but i got this i dont know if that help. Supplied catalyst is older then now.

08-Feb-2012 17:37:56 [---] OS: Linux: 3.0.0-15-generic
08-Feb-2012 17:37:56 [---] Memory: 3.87 GB physical, 3.99 GB virtual
08-Feb-2012 17:37:56 [---] Disk: 472.75 GB total, 443.57 GB free
08-Feb-2012 17:37:56 [---] Local time is UTC +2 hours
08-Feb-2012 17:37:56 [---] ATI GPU 0: AMD Radeon HD 6900 series (Cayman) (CAL version 1.4.1664, 2048MB, 1932MB available, 6758 GFLOPS peak)
08-Feb-2012 17:37:56 [---] OpenCL: ATI GPU 0: Cayman (driver version CAL 1.4.1664, device version OpenCL 1.1 AMD-APP (851.4), 1024MB, 1932MB available)
08-Feb-2012 17:37:56 [---] ATI GPU 0 is OpenCL-capable

Stderr output

7.0.14

process exited with code 255 (0xff, -1)

[17:49:25][2413][INFO ] Application startup - thank you for supporting Einstein@Home!
[17:49:25][2413][INFO ] Starting data processing...
[debug] BOINC: Trying to find requested ATI OpenCL device #0
BOINC: No OpenCL platform found (-1001)!
[17:49:25][2413][ERROR] Failed to get OpenCL platform/device info from BOINC (error: -1)!
[17:49:25][2413][ERROR] Demodulation failed (error: -1)!
17:49:25 (2413): called boinc_finish

]]>

Alex
Alex
Joined: 1 Mar 05
Posts: 88
Credit: 398734
RAC: 0

I've updated my main system

I've updated my main system to be able to run albert wu's
first finished (HD6950) is wuid=37677
resultid=110257
on this pc hostid=1767

just if someone wants to compare performance

Alex
Alex
Joined: 1 Mar 05
Posts: 88
Credit: 398734
RAC: 0

next wu

next wu finished.
HD5850
resultid=116606
wuid=35510
hostid=1767

choks
choks
Joined: 24 Feb 05
Posts: 5
Credit: 1110845
RAC: 0

Hi I had this error with

Hi

I had this error with the 1.2 version:

[debug] BOINC: Trying to find requested ATI OpenCL device #0
[07:09:27][29377][ERROR] Application caught signal 11.
......
Frame 2:
Binary file: /usr/lib/libamdocl32.so (0xf5fcad78)
Offset info: +0xead78
------> End of backtrace

It turned out on debian64, the AMD driver is installed on /usr/lib32/libamdocl32.so and not /usr/lib/libamdocl32.so.
I don't know why but I had an old version in /usr/lib.
Just copied the new lib over, and it's now working.

Christophe

HV Margus*
HV Margus*
Joined: 10 Jan 12
Posts: 17
Credit: 45000
RAC: 0

RE: Hi I had this error

Message 79054 in response to message 79053

Quote:

Hi

I had this error with the 1.2 version:

[debug] BOINC: Trying to find requested ATI OpenCL device #0
[07:09:27][29377][ERROR] Application caught signal 11.
......
Frame 2:
Binary file: /usr/lib/libamdocl32.so (0xf5fcad78)
Offset info: +0xead78
------> End of backtrace

It turned out on debian64, the AMD driver is installed on /usr/lib32/libamdocl32.so and not /usr/lib/libamdocl32.so.
I don't know why but I had an old version in /usr/lib.
Just copied the new lib over, and it's now working.

Christophe


Thanks for hint,

Oliver i tested not debs or supplied catalyst but its 12.1 installer and found that opencl missing on system nothing helped even installing sdk after 12.1 install. Only what is work is supplied in ubuntu and installing sdk after but this combination is terrible old.

HV Margus*
HV Margus*
Joined: 10 Jan 12
Posts: 17
Credit: 45000
RAC: 0

Now i got improvement but its

Message 79055 in response to message 79054

Now i got improvement but its far from working
Stderr output

7.0.14

process exited with code 11 (0xb, -245)

[08:07:21][2740][INFO ] Application startup - thank you for supporting Einstein@Home!
[08:07:21][2740][INFO ] Starting data processing...
[debug] BOINC: Trying to find requested ATI OpenCL device #0

[08:07:22][2740][ERROR] Application caught signal 11.

------> Obtained 5 stack frames for this thread.
------> Backtrace:
Frame 5:
Binary file: ../../projects/albert.phys.uwm.edu/einsteinbinary_BRP4_1.21_i686-pc-linux-gnu__atiOpenCL (0x80b6d32)
Offset info: pthread_mutex_lock+0x672
Source file: erp_boinc_wrapper.cpp (Function: sighandler / Line: 166)
Frame 4:
Binary file: ../../projects/albert.phys.uwm.edu/einsteinbinary_BRP4_1.21_i686-pc-linux-gnu__atiOpenCL (0x842eff6)
Offset info: _ZNSsD1Ev+0x26
Frame 3:
Binary file: ../../projects/albert.phys.uwm.edu/einsteinbinary_BRP4_1.21_i686-pc-linux-gnu__atiOpenCL (0x842eff6)
Offset info: _ZNSsD1Ev+0x26
Frame 2:
Binary file: /usr/lib32/libamdocl32.so (0xf600ad78)
Offset info: +0xead78
------> End of backtrace

08:07:22 (2740): called boinc_finish
Frame 1:
Binary file: [ ((nil))
Offset info: nil
'[': No such file

]]>

choks, can you upload the silly catalyst opencl files somehow i not have right ones.

choks
choks
Joined: 24 Feb 05
Posts: 5
Credit: 1110845
RAC: 0

Hi, I updated Catalyst to

Message 79056 in response to message 79055

Hi,

I updated Catalyst to version 11.12.
The file I have is:

-rw-r--r-- 1 root root 25369464 4 févr. 07:10 /usr/lib32/libamdocl32.so

I think it is written when you run amd-driver-installer-12-1-x86.x86_64.run.

Christophe

HV Margus*
HV Margus*
Joined: 10 Jan 12
Posts: 17
Credit: 45000
RAC: 0

You guys use linux for years

Message 79057 in response to message 79056

You guys use linux for years i use maybe in total one month i not guru yet.
Is needed install catalyst in root or not, whay it is broke always something or not work.

terencewee*
terencewee*
Joined: 2 Feb 12
Posts: 5
Credit: 4500
RAC: 0

http://albert.phys.uwm.edu/re


http://albert.phys.uwm.edu/results.php?userid=333612

All WUs validated.
v1.19 and v1.20 - running on 5850 (Cypress) and M7820 (Juniper).
Will continue test run again when finished with POEM++ OpenCL WUs.

Good job!

-- terencewee* Sicituradastra.

Oliver Behnke
Oliver Behnke
Moderator
Administrator
Joined: 4 Sep 07
Posts: 320
Credit: 8545955
RAC: 0

RE: It turned out on

Message 79059 in response to message 79053

Quote:
It turned out on debian64, the AMD driver is installed on /usr/lib32/libamdocl32.so and not /usr/lib/libamdocl32.so.
I don't know why but I had an old version in /usr/lib.
Just copied the new lib over, and it's now working.

I experienced the same issue on my machine. I turns out that this is an inconsistency between the APP SDK and Catalyst. Both install this 32-bit compatibility library, but in different locations, hence the mix-up that leads to the segfault. I already filed a bug report with AMD.

I recommend to not install the APP SDK, you don't need it to run OpenCL GPU applications as long as you installed Catalyst >= 11.3 (Windows, 11.11 for Linux). The SDK is for developers only.

Cheers,
Oliver

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.