On 08.02.2012 00:57 CE(S)T, Simon Slavin wrote: > If most of the time really is spent reading from the file, then it > may not be worth parallelising your code. All your data is on the > same disk and you can read only one sector of the disk at a time.
Most of the file should be in the file system cache anyway. And I'm not sure about how efficient my reading code is already. I'm going to look at that, too. But with 2 threads on my Core 2 Duo (no HT) I can see 100% CPU usage instead of just a bit over 50%, and the time it takes is a bit below the middle of the longest single job and all jobs sequentially. So it does seem to help to parallelise it all. (And it's easier for me because reading and plotting is started from the same function which I don't want to split.) > If you can read all your data first in one process, but then > parallelise the graph-drawing elements, then that might take good > advantage of multi-cores and multi-processors. Reading is what takes most of the time for me, plotting is pretty fast. -- Yves Goergen "LonelyPixel" <nospam.l...@unclassified.de> Visit my web laboratory at http://beta.unclassified.de _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
