Mark Spiegel wrote:
> [...]
Thanks for the explanation. Yes, I'll have to make sure not to use this
technique when the file is too large. But I think there is definitively
an improvement for us to pull data in cache whenever possible.
> Given that you can't write your own VFS, there is not
Thanks for your quick answer.
RaghavendraK 70574 wrote:
> Try this, use mmap
> (i assume u'r db is readonly)this is much faster and
> better than ifstream read. Also ifstream read can keep the data in
cache as long as no other serious
> i/o occurs.
Our DB is readwrite, though in my example reado
the DB, it only
reduced the timings to 16s, which was still bad for our requirements.
Tests with a larger DB (it can get much bigger than my example) and on
different machines tend to confirm my theory.
Thanks in advance (and a big thank for SQLite which is really nice and
easy to use !),
Ju
3 matches
Mail list logo