Posts by DeleteNull
log in
1) Message boards : News : SSL Certificate Replaced (Message 23158)
Posted 446 days ago by Profile DeleteNull
The certificate was valid until 27th of October 2016. So, now the certificate is expired until it is replaced by a new one.
2) Message boards : Number crunching : No workunits for ATI on Windows? (Message 21894)
Posted 744 days ago by Profile DeleteNull
Your driver doesn't report an OpenCl version (1.02 or 2.0). I had the same issue with my NVidia card. With an updated driver I could get work.
3) Message boards : News : Collatz Sieve v1.21 Released for Linux and OS X (Message 21571)
Posted 821 days ago by Profile DeleteNull
I need new glasses. There are 3 options:
OpenCL for Intel CPUs (opencl_intel_cpu)
OpenCL for AMD/Intel CPUs (opencl_amd_cpu)
and (in a row a few items before)
OpenCL for AMD GPUs (opencl_amd_gpu)

I activated the last.....and now the work is flowing (less than 210 secs for a 270X seems to be o.k.)

Linux (64 bit) with NVidia runs very well.
Linux (64 bit) with AMD/ATI didn't download the application files and therefore no WU. I had to download the files manually and use an app_info.xml to get work.
Is there a configuration missing?

The app_info.xml is this:
<app_info>
<app>
<name>collatz_sieve</name>
<user_friendly_name>Collatz Sieve</user_friendly_name>
<non_cpu_intensive>0</non_cpu_intensive>
</app>
<file>
<name>collatz_sieve_1.21_x86_64-pc-linux-gnu__opencl_amd_gpu</name>
<executable/>
</file>
<file>
<name>collatz_sieve_1.21_x86_64-pc-linux-gnu__opencl_amd_gpu.config</name>
</file>

<app_version>
<app_name>collatz_sieve</app_name>
<version_num>121</version_num>
<platform>x86_64-pc-linux-gnu</platform>
<avg_ncpus>0.5</avg_ncpus>
<max_ncpus>0.5</max_ncpus>
<flops>221633609821706.906250</flops>
<api_version>7.7.0</api_version>
<file_ref>
<file_name>collatz_sieve_1.21_x86_64-pc-linux-gnu__opencl_amd_gpu</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>collatz_sieve_1.21_x86_64-pc-linux-gnu__opencl_amd_gpu.config</file_name>
<open_name>collatz.config</open_name>
<copy_file/>
</file_ref>
<coproc>
<type>ATI</type>
<count>1.000000</count>
</coproc>
<dont_throttle/>
</app_version>
<app_info>


Sure it isn't a setting in your preferences? The collatz preferences allow choosing which type of work to get but I should really remove the CUDA and ATI options because they no longer exist. The only options now are the OpenCL versions and the CPU versions. Make sure that "opencl_amd_gpu" is checked.
4) Message boards : Unix/Linux : LinuxMint and ATI 7770 (Message 21564)
Posted 822 days ago by Profile DeleteNull
My Win7 and ATI 7770 PCs are working fine with collatz.
However, my LM and ATI 7770 are not downloading.

Can someone please advise as to how I can fix this?

If additional info is needed, please advise and will provide.

Highlander


Look here.

The files that are needed you can get from this URL
5) Message boards : News : Collatz Sieve v1.21 Released for Linux and OS X (Message 21383)
Posted 839 days ago by Profile DeleteNull
Linux (64 bit) with NVidia runs very well.
Linux (64 bit) with AMD/ATI didn't download the application files and therefore no WU. I had to download the files manually and use an app_info.xml to get work.
Is there a configuration missing?

The app_info.xml is this:
<app_info>
<app>
<name>collatz_sieve</name>
<user_friendly_name>Collatz Sieve</user_friendly_name>
<non_cpu_intensive>0</non_cpu_intensive>
</app>
<file>
<name>collatz_sieve_1.21_x86_64-pc-linux-gnu__opencl_amd_gpu</name>
<executable/>
</file>
<file>
<name>collatz_sieve_1.21_x86_64-pc-linux-gnu__opencl_amd_gpu.config</name>
</file>

<app_version>
<app_name>collatz_sieve</app_name>
<version_num>121</version_num>
<platform>x86_64-pc-linux-gnu</platform>
<avg_ncpus>0.5</avg_ncpus>
<max_ncpus>0.5</max_ncpus>
<flops>221633609821706.906250</flops>
<api_version>7.7.0</api_version>
<file_ref>
<file_name>collatz_sieve_1.21_x86_64-pc-linux-gnu__opencl_amd_gpu</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>collatz_sieve_1.21_x86_64-pc-linux-gnu__opencl_amd_gpu.config</file_name>
<open_name>collatz.config</open_name>
<copy_file/>
</file_ref>
<coproc>
<type>ATI</type>
<count>1.000000</count>
</coproc>
<dont_throttle/>
</app_version>
<app_info>
6) Message boards : News : Phasing out Multi-Size (Message 19196)
Posted 1363 days ago by Profile DeleteNull
Thank you for the info.

I am using large_collatz_6.04_windows_x86_64__opencl_amd_gpu.config and solo_collatz_6.04_windows_x86_64__opencl_amd_gpu.config without app_config.xml.
(items_per_kernel=21)

This seems to be 5% slower, so i will switch to items_per_kernel=22
7) Message boards : News : Phasing out Multi-Size (Message 19192)
Posted 1363 days ago by Profile DeleteNull
Large Collatz Conjecture
WIN 7 64, ATI 7970 drivers 14.4
1xWU 18,955.41 sek -----> 237,221.91 pkt
is well


Which config did you use?
8) Message boards : Number crunching : opencl_osx always losing against ati13ati? (Message 17070)
Posted 1655 days ago by Profile DeleteNull
Looking at this WU you can see that your computer is reporting 1630 steps for 2380169206453088055990, but ati3ati is reporting 1692 steps for 2380169206553033385980.

Both values are o.k. (verified with my java prog), but your computer doesn't find the high value. So it seems that the opencl_osx implementation must have a bug :-((

Hi,

looking back at my stats, it turns out that I've never had any WU where an opencl_osx and an ati13ati client would have reached the same results on that same WU. In these cases, a third computer jumps in to verify the results. Now, if this third computer is also an ati13ati, then it verifies always the results of the initial ati13ati computer (if the third computer is not an ati13ati, then it depends). This suggests me that either the ati13ati or the opencl_osx implementation must have a bug. It can't just be a pure coincidence, could it? ;-)

Cheers,
Ádám
9) Message boards : Number crunching : Ati Radeon HD 7750 most WU invalid (Message 13962)
Posted 2082 days ago by Profile DeleteNull
Only 1/3 of the workunits are "valid".

The first wrong message from the software is:
2240 shader units organized in 28 SIMDs with 16 VLIW units (5-issue), wavefront

A 7950 has "only" 1792 shaders because it is using "core next" units (28x16x4) and !_not_! VLIW units with 5 threads.

Therefore the software organizes the calculations incorrect, and only about half the work is done, see this message with "total executed steps":

needed 1873 steps for 2375121519533049873790
224511391507172 total executed steps for 824633720832 numbers

but it should be:
needed 1873 steps for 2375121519519809296446
436376523829883 total executed steps for 824633720832 numbers

So a "valid" WU gets only half of the credits (about 1600 instead of 3200), .....and your wingman too.
10) Message boards : Number crunching : Ati Radeon HD 7750 most WU invalid (Message 13934)
Posted 2085 days ago by Profile DeleteNull
Volunteering?


Yes!
11) Message boards : Number crunching : Ati Radeon HD 7750 most WU invalid (Message 13933)
Posted 2085 days ago by Profile DeleteNull
Volunteering?


Yes!
12) Message boards : Number crunching : Ati Radeon HD 7750 most WU invalid (Message 13925)
Posted 2086 days ago by Profile DeleteNull
I have also a ATI 7950 with driver 12.4 (Win7-64) and Boinc 7.0.26/27.

Half of the WU's are invalid because the app version 2.09 does not work with 79xx, .....because the hardware changed (no longer "VLIW5-Unified-Shaders"). So the 79xx does not calculate all steps, as you can see here:
http://boinc.thesonntags.com/collatz/workunit.php?wuid=50175393

Computer 99633 (ATI 7950) reports:
needed 1661 steps for 2375109180790110377473
151680369242353 total executed steps for 824633720832 numbers

Computer 77981 (and also 93176) reports:
needed 1842 steps for 2375109180842987849598
415354125465393 total executed steps for 824633720832 numbers

If anyone has the source code for app version 2.09 ..... perhaps we can make the modifications so that the sun shines for Tahiti (79xx)?!
13) Message boards : Number crunching : ATI 9.12 / Linux64 / BOINC problem... (Message 6383)
Posted 2881 days ago by Profile DeleteNull
Thanks Crunch3r,

after hard work (a few problems with the ATI-drivers) i got my box running with 2 GPU's, and both GPU's were detected by BOINC! Now your version 1.5 is able to use GPU 0 and 1.

It was a stab in the dark since i can't test the code myself (got no box here running an ati on linux).


Yes, that's not easy. The kernel must have a version below 2.6.31 !!! I had to downgrade my box to Suse 11.1 (kernel 2.6.27).
14) Message boards : Number crunching : ATI 9.12 / Linux64 / BOINC problem... (Message 6357)
Posted 2883 days ago by Profile DeleteNull
Hi,

i have testet the new version.
2 GPU's detected but all WU's running on GPU 0.

The difference to Windows is, that the second screen was black (Windows uses it with colors ;-).

I have explicit activated the second screen in the ATI-config tool. After the reboot the "init 5" fails (with Xinerama and without) :-((

So i think, in about 5 years (when ATI has proper linux drivers), i try again.....
15) Message boards : Number crunching : ATI 9.12 / Linux64 / BOINC problem... (Message 6347)
Posted 2883 days ago by Profile DeleteNull
Hi Guiseppe,

look here (one of your WU's) http://boinc.thesonntags.com/collatz/result.php?resultid=20552437

Running Collatz Conjecture (3x+1) ATI GPU application version 2.01 by Gipsel (Linux64, CAL 1.4 R1.2)
instructed by BOINC client to use device 1
.....
Initializing lookup table (16384 kB) ... done
Starting WU on GPU 0
.....

I testet it also in one of my boxes with two 4890, same result...

2 GPU's, but all WU's were assigned to GPU 0. It seems that the application assigns all WU's to the default (0).

16) Message boards : Number crunching : Linux 32-bit ATI App (Message 6323)
Posted 2885 days ago by Profile DeleteNull
Hi,

i am testing collatz_2.09_i686-pc-linux-gnu__ati14.tgz

1. untar
2. copy libbrook.so to /usr/lib/
3. starting collatz_2.09_i686-pc-linux-gnu__ati14
Result:
Shared Object NOT Found! libbrook_cal.so

This file is missing in collatz_2.09_i686-pc-linux-gnu__ati14.tgz
17) Message boards : Cafe : Last one to post here win! (Message 6228)
Posted 2892 days ago by Profile DeleteNull
Just wait 'til Andre Lange is winning the bobsleigh contests (I hope so)! ;-P


Now winning after the PrimeGrid contest!
18) Message boards : Cafe : Last one to post here win! (Message 5611)
Posted 2923 days ago by Profile DeleteNull
Time for a WIN!
19) Message boards : Number crunching : Interesting Statistics (Message 5610)
Posted 2923 days ago by Profile DeleteNull
Opps. I was thinking 4350 = 710. The RV740/RV770 ranges from 4730 to a 4870. There is a big difference in FLOPs between a 4730 and a 4870 and no good way to tell which is which from the data available on the server.


The RV740/RV770 ranges from 4730 to 4890 (!), because the code used in the coproc_detect.cpp assings all these chips to this class (RV740/770).
You (or the BOINC developers) have to edit this file to get a more granular detection like this: http://boinc.thesonntags.com/collatz/show_host_detail.php?hostid=12881
20) Message boards : Cafe : Last one to post here win! (Message 5524)
Posted 2927 days ago by Profile DeleteNull
7 hours of winning is enough for cats!


Next 20

Main page · Your account · Message boards


Copyright © 2018 Jon Sonntag; All rights reserved.