On Mon, 18 Sep 2006 09:41:35 +0200
"A.J. Venter" <[EMAIL PROTECTED]> wrote:
> I ran wole through gprof recently to see if I can speed it up a
> bit... imagine my surprize to learn that 47% of wole's total CPU time
> is spent - not on drawing, not on calculating, not on launching
> children - not even the very complex recursive function used to
> compare version numbers but purely on uppercase(ansistring).
gprof has a unit of 10ms. Maybe your program runs in total less than 1
second and the rest are external libs? Then the gprof numbers are
useless.
> Now indeed I do use a few (about 5) such calls, mostly for the
> purpose of doing case independent string checks such as
> if pos('<ITEM>',uppercase(RSS)) = 0 then
>
> Now some things to considder:
> RSS here is the single ansistring format of an entire RSS source - so
> yes, it's quite big, probably several kilobytes (but not megabytes),
> so that could be part of why uppercase takes so much time - also
> there are at least 3 such sources in a typical run - so this
> particular one runs three times.
>
> Nonetheless - should uppercase take up THAT much CPU time ?
>
> The RSS parser code I use makes quite a lot of use of the abilities
> that come from reading tstringlist as a single string, could be VERY
> hard to rewrite, but even so the rest of it doesn't seem to be using
> much CPU at all anyway.
>
> Should I rewrite this test ? If so... how would you suggest I do it ?
Make the test longer.
Mattias
_________________________________________________________________
To unsubscribe: mail [EMAIL PROTECTED] with
"unsubscribe" as the Subject
archives at http://www.lazarus.freepascal.org/mailarchives
