Well, it really helps knowing what's going on to scale the results to other
situations. Depending on what it's doing, there are SO
MANY factors - does it depend on the record number, the mode of access, the location
of the database? Do multiple accesses to the
same record behave the same as accesses to multiple records, debug rom vs release rom,
etc, etc.
True, I can do extensive test cases and begine to infer all this, but no answer is as
definite as knowing how the access works.
But while we're on the topic of timing, I must ask a question about timing in the
debugger:
(1) All the documentation and headers suggest that under windows emulation, the
emulator provides 1,000 ticks per second, but
SysTicksPerSecond still returns 100. Has anyone seen millisecond timing under windows?
(2) I know the Emulator is doing a hardware level simulation of a real device, but I
get timing differences between emulator runs
and the real thing as high as a factor of two. How cycle accurate is the emulator?
(E.g., the ARM emulator has several different
modes of realism...)
Thanks for the advice!
- Jeff
P.S. -> simple tests not addressing all the issues in my first paragraph showed record
access to be pretty quick in human terms -
about 6,600 GetResource+RelaseResource per second, but taking about 3,000 cycles to
execute, which makes me again suspect there's
more than simple code execution going on here.
Richard Burmeister wrote:
> > From: DIAMOND JEFF
> >
> > I'm trying to decide in designing a project whether to use feature
> > memory or a database for a commonly accessed buffer.
> >
>
> If you are concerned about speed, why not create two test cases - one using
> features and the other using a database - and see which takes longer to do
> 1000 or maybe 10,000 accesses?
--
For information on using the Palm Developer Forums, or to unsubscribe, please see
http://www.palmos.com/dev/tech/support/forums/
