Absolutely.
For what I'm trying to do, and given my experiments thus far, I would love to
replicate the performance of the one-query/thread-per-process concurrency in
the multithreaded case, foregoing the resource optimisations (shared cache
etc.) and just having each query/thread do
"Marcus Grimm" schrieb im
Newsbeitrag news:4b9a01e8.2060...@medcom-online.de...
> Increasing the number of threads does affect the overall
> read performance slightly, example:
> 1 Thread: 11.2 sec
> 16 Threads: 8.2 sec
> Single Process: 11sec.
>
> I still would expect
Just to be clear, from my point of view at least, the big difference is between:
- Multiple threads performing concurrent queries (with or without the shared
cache)
vs
- Multiple processes performing concurrent queries
In the latter case (for a number of queries chosen to lie within the
Hi again,
I have played again with some variations of the speed test:
I'm not able to reproduce the reported behavior that a single process
queries faster than a single thread. During my tests I don't get a
big difference, even when I turn off some of the locking
via SQLITE_OPEN_NOMUTEX.
I have
Hi Luke,
Luke Evans wrote:
> Hi Marcus,
>
> Well, I'd certainly be interested in looking at your code. Can you mail a
> zip, or post to a web or file hosting site? Thanks.
ok, please try this:
http://www.exomio.de/SqliteSpeedTest.c
I havent tried yet to compare the numbers when using one
Hi Marcus,
Well, I'd certainly be interested in looking at your code. Can you mail a zip,
or post to a web or file hosting site? Thanks.
Your results seem to broadly agree with mine: multithreaded querying can save a
percentage of time (10-30%?) over the same queries issued serially with no
I have followed the discussion about this issue with interest since
my usage of sqlite involves threads and sharedcache mode as well.
I have also generated a little speed test program that uses
multible threads and shared cache to read *some* rows out of the
sqlite DB. It might not help here but
Well, I guess this is looking more an more like a bug then.
I just went to the sqlite.org site to find out how I log a bug and it referred
me back to here. Perhaps this means I just need to *DECLARE BUG* here ;-)
Actually, I'll probably have some time soon to try out a profiler to see if I
can
8 matches
Mail list logo