Simon Slavin-3 wrote > Once you’ve bust the cache things slow down. I do realize that. However, not illinearly. Once I bust the cache, throughput should drop X times and stay there. Here, the speed decreases with the progress. Let me put it this way: say the cache size was 0. I would expect the delete to work very slow, but near-linear with the # of rows being deleted or the progress of deleting them. Here the performance dramatically and constantly decreases with growing of the WAL file. You can literally see how every M of throughput is slower than the previous :) I am trying to find out why. For me it's very important for the migration process.
Simon Slavin-3 wrote > Have you tried these things ? Did the time taken improve or get worse ? Not yet, we will probably implement what you proposed (it's been suggested before). But if I can find out the reason behind the nonlinear DELETE behavior, it would still help greatly. -- Sent from: http://sqlite.1065341.n5.nabble.com/ _______________________________________________ sqlite-users mailing list sqlite-users@mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users