Since the message below I have found the below config_aux entry. I have asked 
the project admin what values he has set. Other users have suggested its taking 
1024 tasks before trying a different app. Other than this is there anything 
else the project admin can do so it tries the various apps more quickly?

MarkJ

<version_select_random_factor>X</version_select_random_factor>
In predicting which app version will be faster for a given host, multiply the 
projected FLOPS by a uniform random variable with mean 1 and this standard 
deviation (default 0.1).


Begin forwarded message:

> From: TarotApprentice <tarotapprent...@yahoo.com>
> Date: 5 November 2017 at 5:08:27 pm AEDT
> To: BOINC Alpha <boinc_al...@ssl.berkeley.edu>
> Subject: Project app selection reset
> 
> I have an issue with Asteroids where the project app selection doesn’t work 
> properly. They have 3 period search apps for the CPU, being sse2, sse3 and 
> avx. All my x64 machines support avx as the highest instruction set available 
> from the project. I have 4 identical Ryzen 7 Machine, two of which have 
> received and are running the avx app. One got the sse2 and another got the 
> sse3 app. None of them have tried the other apps according to the application 
> details for each host.
> 
> Its not limited to AMD, I also have 8 i7’s-6700’s again identical builds and 
> its a similar story. Six got the avx app, one the sse2 and the other one got 
> the sse3 app. Again no history of trying the other apps to determine the one 
> with the highest APR. For these two “slow app” machines I have detached them 
> from the project and am waiting on their 90 days of no contact before I can 
> delete the host before reattaching it in the hope a new host record will 
> start with the avx app.
> 
> Is there any way to get it to try the other apps like its supposed to, or at 
> least to reset the host app details? I have contacted the admin, Kyong but he 
> isn’t aware of anything he can do. I have also seen other people on their 
> message boards complaining of the same problem.
> 
> MarkJ
_______________________________________________
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
https://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Reply via email to