Re^7: Threads and DBD::SQLite?

[BrowserUk]

Thanks again erix. Those are some intriguing numbers.

ISTM that it doesn't retrieve enough.

Have you seen/disabled this from the posted code?:

    $Q->nq( $_ ) for (shuffle 1 .. $N)[ 0 .. $R ];
#.....................................^^^^^^^^^^^

[erix]

OK, so my numbers for "Read all 100000 records using 4 threads" were really

" Read ", $R * 100 / $N, " % of all records using 4 threads"

that means the test was doing 0.1 % of what the output-line was saying.

Here is a run without that mistake:

#  The offending code line changed to:
#  $Q->nq( $_ ) for (shuffle 1 .. $N);

SQLite -   5.90120077 s  - Populate db with 100000 rows
SQLite -   0.15266800 s  - Create primary index
SQLite -   0.53022408 s  - Read all 100000 rows, unthreaded
SQLite -   7.23427510 s  - Read all 100000 rows, 4 threads

    Pg -  23.40106297 s  - Populate db with 100000 rows
    Pg -   0.53076911 s  - Create primary index
    Pg -  13.71553111 s  - Read all 100000 rows, unthreaded
    Pg -   7.40499115 s  - Read all 100000 rows, 4 threads

[BrowserUk]

so my numbers

Not only your numbers, mine also. Another forgotten artifact from when I was testing this a couple of months ago.

Your latest output makes for intriguing reading. The Pg 1 thread versus 4 threads is sort of what you'd expect.

The sqlite numbers make no sense at all. It looks like it has to re-build the cache from scratch for the new connections?

Those kind of numbers combined with my inability to get a memorydb combined with shared db via multiple connections tell me that this stuff isn't ready for prime time yet.

Thanks for your help.