Posts by Partygott

1) Message boards : Number crunching : Long 2 are a joke (Message 4815)
Posted 16 Dec 2018 by Partygott

I have downloaded and am now running a Long2 work unit. After 8 hours it shows over 14% done, which is a lot quicker than before.

@ PDW,
I am showing all 4 threads on a 4 core computer being recorded at WUProp, 8 hours done and showing 31.6 hours, which seem right to me.

I do have an extra line in my app_config.xml file that may be what is different to what you have.
I have
<cmdline>-t4</cmdline>
<avg_ncpus>4</avg_ncpus>

Not sure if this might be how I am getting all threads picked up or not.

Anyway all seems to be working again,

Thanks
Conan

No, you don't have to integrate an extra line with "avg_ncpus" because you don't want to separate between threads and cpu cores.
You should use the "<project_max_concurrent>1</project_max_concurrent>" command, which tells the project to only use one single wu overall.
At least to me this is usefull in a time where most of the wus are long type.
So the app_config in the FAQ becomes useless at the moment where the project uses different apps (and this happens very often), because in this way every of the 12 possible apps is allowed to use the number of cores you set in the config file.
The result is that you still have 4 threads on a 4 core cpu with the little difference that now the OS switches the cores between the tasks, so you can have a long task with 70% cpu, a long3 task with 20% and 2 short tasks with 5% each, instead of 4x 25%. But in the end, no time is won, because you want to use 100% of you cpu to only one single long task.

So a usefull app_config (with a proper formatting, unlike in the FAQ) where all cpu cores (4 in this case) concentrate to only one wu should look like this:

<app_config>

<app>
<name>srbase</name>
<max_concurrent>1</max_concurrent>
<fraction_done_exact/>
</app>
<app_version>
<app_name>srbase</app_name>
<cmdline>-t4</cmdline>
</app_version>

<app>
<name>srbase2</name>
<max_concurrent>1</max_concurrent>
<fraction_done_exact/>
</app>
<app_version>
<app_name>srbase</app_name>
<cmdline>-t4</cmdline>
</app_version>

<app>
<name>srbase3</name>
<max_concurrent>1</max_concurrent>
<fraction_done_exact/>
</app>
<app_version>
<app_name>srbase</app_name>
<cmdline>-t4</cmdline>
</app_version>

<app>
<name>srbase4</name>
<max_concurrent>1</max_concurrent>
<fraction_done_exact/>
</app>
<app_version>
<app_name>srbase</app_name>
<cmdline>-t4</cmdline>
</app_version>

<app>
<name>srbase5</name>
<max_concurrent>1</max_concurrent>
<fraction_done_exact/>
</app>
<app_version>
<app_name>srbase</app_name>
<cmdline>-t4</cmdline>
</app_version>

<app>
<name>srbase6</name>
<max_concurrent>1</max_concurrent>
<fraction_done_exact/>
</app>
<app_version>
<app_name>srbase</app_name>
<cmdline>-t4</cmdline>
</app_version>

<app>
<name>srbase7</name>
<max_concurrent>1</max_concurrent>
<fraction_done_exact/>
</app>
<app_version>
<app_name>srbase</app_name>
<cmdline>-t4</cmdline>
</app_version>

<app>
<name>srbase8</name>
<max_concurrent>1</max_concurrent>
<fraction_done_exact/>
</app>
<app_version>
<app_name>srbase</app_name>
<cmdline>-t4</cmdline>
</app_version>

<app>
<name>srbase9</name>
<max_concurrent>1</max_concurrent>
<fraction_done_exact/>
</app>
<app_version>
<app_name>srbase</app_name>
<cmdline>-t4</cmdline>
</app_version>

<app>
<name>srbase10</name>
<max_concurrent>1</max_concurrent>
<fraction_done_exact/>
</app>
<app_version>
<app_name>srbase</app_name>
<cmdline>-t4</cmdline>
</app_version>

<app>
<name>srbase11</name>
<max_concurrent>1</max_concurrent>
<fraction_done_exact/>
</app>
<app_version>
<app_name>srbase</app_name>
<cmdline>-t4</cmdline>
</app_version>

<app>
<name>srbase12</name>
<max_concurrent>1</max_concurrent>
<fraction_done_exact/>
</app>
<app_version>
<app_name>srbase</app_name>
<cmdline>-t4</cmdline>
</app_version>

<project_max_concurrent>1</project_max_concurrent>

<report_results_immediately/>

</app_config>

2) Message boards : Number crunching : LLR Version 3.8.20 released (Message 3396)
Posted 6 Apr 2017 by Partygott

Hi.

The mt feature seems to have a problem with scaling and/or hyperthreading on windows.

4c/4t i5-3470 uses all free ressources up to 99.x percent.
2c/4t i5-2520m uses max. 93-97%.

8c/16t ryzen 1700 behaves like this:
1 wu @ 16 vcores: 67%
2 wu @ 8 vcores: 41% per wu
4 wu @ 4 vcores: 22% per wu

Even with restricting to 4 wu at the same time my ryzen sys takes a 5th wu with 4 vcores on a different app, like 4x riesl base + 1x sierpinski/riesel, which results in 5x 19.x percent usage.

btw...
I noticed that on aamd ryzen the crunshing of a wu slows down the more % of it has completed. On intel it has the same speed over 0-100%.