Dima Dat'ko wrote: > I proceed with playing with the issue. > > sqlite version | flash card type | time from (1) to (2) [milliseconds] > -------------------+----------------------------+------------------------------------------------------ > 3.4.2 | SD SanDisk (fast) | 4335 > 3.4.2 | CF Toshiba (slow) | 4401 > 3.5.6 | SD SanDisk (fast) | 43993 > 3.5.6 | CF Toshiba (slow) | 79568 > > If i comment the UPDATE query (lines from (A) to (B) the results are > the following: > > sqlite version | flash card type | time from (1) to (2) [milliseconds] > -------------------+----------------------------+------------------------------------------------------ > 3.4.2 | SD SanDisk (fast) | 2025 > 3.4.2 | CF Toshiba (slow) | 2099 > 3.5.6 | SD SanDisk (fast) | 2119 > 3.5.6 | CF Toshiba (slow) | 2310 > > > As you can see in this test the version of 3.5.6 can be about 20 times > slower then 3.4.2. I'm wondering is not it a bug? Should I submit a > ticket? >
Yes, this does look like a pretty severe performance regression in SQLite 3.5.6. I think you should create a ticket so the developers can look into it. You have shown quite clearly that it is the update statement that is at fault. I will take a look at the VDBE code generated by the update statement to see if anything jumps out at me. Dennis Cote _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
