Collatzs app_config and executable configs

I run solo_collatz only on GPU, if you want to run mini_collatz and collatz, you will need to add more to the app_config below and do additional additions to other exe config files (see below). I use these configs for 7970 class and gtx 670 gpus.
2) app_config.xml - I run one cpu to one gpu wu on AMD bulldozer -
<app_config>
<app>
<name>solo_collatz</name>
<max_concurrent>10</max_concurrent>
<gpu_versions>
<gpu_usage>1.0</gpu_usage>
<cpu_usage>1.0</cpu_usage>
</gpu_versions>
</app>
</app_config>
If you have a fast intel, experiment with 0.5 cpu_usage and compare the times. Our web site strips multiple spaces, so imagine the above neatly indented. ;)

3) Collatz has extra configs in the project directory that match the .exe names that are used to control how the app uses the gpu. For solo colatz you will need to change 2 of them. solo_collatz_4.07_windows_intelx86__opencl_ati_100 .config and solo_collatz_4.07_windows_x86_64__opencl_ati_100.c onfig. The contents I use for both are -
<configuration>
verbose=0
items_per_kernel=22
kernels_per_reduction=9
threads=9
sleep=1
build_options=-Werror
</configuration>
After making the changes, restart the boinc client. You may not see any affect on wus that are already running. When the next one starts your gpu load should jump to 98 to 99%. You don't need to abort your cache of wus, just wait for the next one to start.

Notes on the config file:
1) You have to change both version of the config files, because the boinc client has a habit of switching between the 64-bit version and the 32-bit version. See Slicker's answer here http://www.setiusa.us/showthread.php...86_64-cuda-app

2) The values in the config file affect the responsiveness of your screen refresh especially browsers. If your screen slows down too much switch to
<configuration>
verbose=0
items_per_kernel=20
kernels_per_reduction=9
threads=8
sleep=1
build_options=-Werror
</configuration>
I don't use the PCs that have the 7970s, so I go all out. On my personnel pc, I have a gtx 690. I have 2 boinc clients installed to control the 2 gpus engines separately and to get collatz to run on both NVidia engines (a problem on some projects). I set the gpu that controls my monitors to the these setting and the other to the higher settings. There is only a minor difference in speed, so long as you have cpus free enough to feed the gpu quickly.

3) The seti.usa discussion of optimizing - http://www.setiusa.us/showthread.php...-570-(and-460) That chatter has links to collatz forum chatter. I think the big difference in people results is the lack of cpu to feed the gpu. As I understand it, solo_collatz uses the cpu to check the work done by the gpu and to feed data. Basically it appears the lower the settings, the more frequent the use of the cpu. But, I could be wrong.

Quote:

---

Basically it appears the lower the settings, the more frequent the use of the cpu. But, I could be wrong.

---

All config numbers are powers of 2 so 2^20 vs 2^22 for items_per_kernel means 1048576 vs 4194304 or 4 times as much work per GPU kernel call. After 2^kernels_per_reduction * 2^items_per_kernel numbers have been checked (e.g. 2^22 * 2^9 = 2^31 = 2,147,483,648 numbers using the config above) are calculated prior to the CPU double checking the best results found. So, fewer numbers will make the video more responsive, but the CPU needs to stop and do its validation more often. Also, the CPU is __supposed__ to sit idle while the GPU is crunching but that really depends on the video drivers since the OpenCL spec doesn't guarantee asynchronous (a.k.a. non-blocking) kernel calls.

Thanks! I just figured out why it was taking my 7970 so long to run wu's! Wish id read this a few weeks ago!