[Freeciv-Dev] Re: Corecleanup patch update with alternate map topologies

["Ross W. Wetmore" <rwetmore@xxxxxxxxxxxx>]

At 12:38 PM 01/08/13 +0200, Thue wrote:
>On Monday 13 August 2001 06:46, Ross W. Wetmore wrote:
>> Not a direct comparison as the map and game evolution are different.
[...]
>> Donut world, generic topology fixes
>>
>> real    15m40.701s
>> user    2m22.100s
>> sys 0m0.570s
>
>In my experience freeciv is highly chaotic, ie the smallest change to 
>any code (like changing the sequence of iterating) will make the result 
>completely unrecognizable. So don't believe too much in comparisons 
>like these.
>
>-Thue

Not only that, but the system seems to have shifted a lot of partial
timeslices to the idle process when the game spends a fair bit of its 
life sleeping in select in sniff_packets. 

I finally figured out where the hooks needed to go, so I can do reasonable
autogame timings. First, it now skips out when you set game.timeout to -1 
in sniff_packets and resets the GAME_OVER_STATE so it will actually quit. 
Finally it does a final auto savegame() in svr_main.c on the way out. When 
you build a release version the GAME_MIN_TIMEOUT value will be set back to 
a positive value so it won't be easy to trigger this code. Will do a mini
patch of this next time I snarf a CVS snapshot.

I ran a few tests with the USE_MAP_FUNCTIONS flipped. It turned up a couple 
more cleanup bugs, but with these fixed the behaviour in either mode is now
identical except for the timing stats.

There seems to be about a 25% reduction when you avoid all the stack
frame operations :-).

Flat Earth, generic topology fixes, server timeout patch, macro

real    9m0.859s
user    8m44.880s
sys     0m2.070s

3765516784 86826 civgame1998.sav.gz

Flat Earth, generic topology fixes, server timeout patch, function

real    11m56.264s
user    11m41.370s
sys     0m1.480s

3765516784 86826 civgame1998.sav.gz

Cheers,
RossW