Complete.Org: Mailing Lists: Archives: freeciv-dev: January 2004:
[Freeciv-Dev] Re: (PR#7278) Profile of current cvs server
Home

[Freeciv-Dev] Re: (PR#7278) Profile of current cvs server

[Top] [All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index] [Thread Index]
To: undisclosed-recipients: ;
Subject: [Freeciv-Dev] Re: (PR#7278) Profile of current cvs server
From: "Arnstein Lindgard" <a-l@xxxxxxx>
Date: Wed, 21 Jan 2004 10:26:20 -0800
Reply-to: rt@xxxxxxxxxxx

<URL: http://rt.freeciv.org/Ticket/Display.html?id=7278 >

On Wed, 21 Jan 2004 08:37:57 -0800 Per I. Mathisen wrote:

> Don't look to hard on this table. As the rest of the profile no doubt will
> show you, it is certain AI functions which consumes CPU power in the
> server. A very large number of very small improvements is much less useful
> than a few well-placed optimizations of key algorithms.

No doubt you are right. I can believe the best way of optimizing is
something like what Ross just described in general terms, caching
tilepointers etc.

The raw number of functions calls overhead is what can be eliminated
by macro/inline. I got a ~12% speedup using about 14 macros. This is
relevant because it seems that the primary anti-argument for not
allowing inlining at all, at the decision IRC meeting, was that we
may or may not gain this relatively small performance improvement on
some compilers, and that's deemed unimportant.

Most of Ross' verbosity on the not-allowing-inline subject, in the
end, concentrate on the uncertainity of whether or not the function
will be inlined. It seems he has not convinced anyone yet that there
are further arguments.


Arnstein




[Prev in Thread] Current Thread [Next in Thread]