-
Hello, In the documentation, it is just described as useful for virus scanners, so my main question is if there is any good use in having a high number set in the settings for performance. If ime not completely mistaken, the limit just shouldent be lower then the global active torrent limit, right? |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment
-
The reason there is a pool is to avoid high costs to
The size of the file pool is not really related to the number of torrents you have. It needs to be set below the per-process file descriptor limit (with margin to allow for peer connections, since those are also file descriptors). As for a lower limit, the more different files are actively read from or written to, the higher you want the pool size to be in order to get the most effect. |
Beta Was this translation helpful? Give feedback.
The reason there is a pool is to avoid high costs to
open
andclose
calls. Both of those can become very expensive with virus scanners, but also on MacOS with HFS+,close()
can be extremely slow. In more recent versions of libtorrent, it's not justopen
andclose
, but alsommap
andmunmap
calls that are saved by caching them in the pool. I'm less familiar with how kernels normally handle frequent mapping and unmapping, but there's a risk each call have to alter the page table and could be somewhat expensive.