Hello Oskar.
The PPL is not using multithreading in its own computations.
I have worked on a (conditionally) thread-safe version of the library, which is available in the git branch "devel". The support for thread-safety needs to be enabled at configuration time (--enable-thread-safe). Beware that the standard allocation library may become a bottleneck, so that you may want to switch to TCMalloc.
See the following messages for a few hints about using PPL services from multiple threads:
http://www.cs.unipr.it/pipermail/ppl-devel/2016-September/019137.html
http://www.cs.unipr.it/pipermail/ppl-devel/2018-January/019187.html
Feel free to ask questions if you face problems.
Cheers, Enea.
On 09/12/19 23:15, Schneider, Oskar wrote:
Hello,
i want to know if ppl is using more than one core for convex hull calculations? If not is it possible to run 20 instances on a 20 core machine and compute 20 hulls in parallel or is the performance not good if too many instances run on the same machine?
Thank you very much.
Greetings,
Oskar _______________________________________________ PPL-devel mailing list PPL-devel@cs.unipr.it http://www.cs.unipr.it/mailman/listinfo/ppl-devel