Turn time

[Sparth]

Ok guys in free time I tested C2C and one thing makes me mad
Turn time.

I tested some other big mods (RI, RoM, RoM AND, PAE, AND) and measure turn times in the first turns. Every other mod has similar turn time (less then 1s, maybe little more in PAE) but in C2C turn time from the start is 3-4 longer. I tested v28 and v34 of C2C.

Ofcoure all setting was the same in testing (standard map, 6AIs)

Do someone make some serious analysis of this problem in the past?

 

[JosEPh_II]

Ok guys in free time I tested C2C and one thing makes me mad
Turn time.

I tested some other big mods (RI, RoM, RoM AND, PAE, AND) and measure turn times in the first turns. Every other mod has similar turn time (less then 1s, maybe little more in PAE) but in C2C turn time from the start is 3-4 longer. I tested v28 and v34 of C2C.

Ofcoure all setting was the same in testing (standard map, 6AIs)

Do someone make some serious analysis of this problem in the past?

Yes koshling did.

JosEPh

 

[Sparth]

Yes koshling did.

JosEPh

I only found his threads about memory usage.

BTW do someone have archival versions of the C2C and can check in which version turn time start grow up? In RoM AND (which is C2C father) turn time is "normal" (~1s).

 

[Hydromancerx]

If anything turn times have gone down from what they use to be before koshling helped speed them up.

In addition all those other mods are tiny compared to C2C.

 

[Thunderbrd]

Aside from a problem we're trying to find in the Assault AI (naval) that's causing THAT to take far too long, the only real explanation for C2C having a far longer turn time is the deep evaluations the AI makes when making decisions that really does account for a better core AI mechanism (though it can still use a lot of tinkering to make much better in some places.) It's more adaptive to the changing mod the way it is than most AI structures but on the down side it takes a long time to evaluate some decisions - much like a more advanced player in a computerized chess game will cause turn times to take much longer. Caching throughout the AI evaluation coding helps to minimize the impact a bit but it can't completely eliminate it.

And AND must've made some big improvements there because Hydro's right... when we picked up from where AND left off at the time our turn times were just as horrendous if not worse.

 

[strategyonly]

If anything turn times have gone down from what they use to be before koshling helped speed them up.

In addition all those other mods are tiny compared to C2C.

Hydro is correct, you CANNOT compare any mod that even comes close to C2C, so no comparison to look at, sorry.

RI - 1.58 GB
RoM - 965 MB
RoM Dawn - 812 MB 1.62 GB
Total War - 1.79 GB
History - 989 MB
FfH2 - 662 MB

C2C - 3.38 GB

So if you look at these and go by the total then C2C should be 2 times as long (for turn times) as the biggest one marked here and 3 times longer than RoM.

 

[JosEPh_II]

SO you're a bit behind on RoM-AND currently it's 1.62GB SVN version 700. So it's half C2C's size. And it's called AND2.1 now. But you are right turn times for C2C are twice as long as AND2.1.

@Sparth,
C2C takes longer because it has More to load in for that 1st turn. AND2.1 is sparse compared to C2C for that 1st turn and for the 1st 2 eras to C2C's 1st 3 eras. Kind of hard to compare.

And poor RoM lies neglected.

JosEPh

 

[Sparth]

Ok I write PM to Koshling. Maybe he remember which aspect is so problematic.
And yes I know that other mods are small compared to C2C but I dont think turn time is not linked with weight of the mod (in MBs) but rather with some new mechanisms which C2C is using (like properites). Anyway Ill wait for Koshling answer.

 

[Stormwind]

And bare SVN C2C without usersettings and autolog is 'only' about 2.85 GBs.

 

[Sparth]

And bare SVN C2C without usersettings and autolog is 'only' about 2.85 GBs.

And I have a plan to lower the weight for another 200-300 Mbs or even more

 

[alberts8]

Ok I write PM to Koshling. Maybe he remember which aspect is so problematic.
And yes I know that other mods are small compared to C2C but I dont think turn time is not linked with weight of the mod (in MBs) but rather with some new mechanisms which C2C is using (like properites). Anyway Ill wait for Koshling answer.

It's not the weight in MB but the weight in number of Techs, Buildings, Units, things like the combat mod, and so on................

ROM AND2 now has all of Koshling improvements to the dll but with less stuff it has to be faster.

 

[JosEPh_II]

It's not the weight in MB but the weight in number of Techs, Buildings, Units, things like the combat mod, and so on................

ROM AND2 now has all of Koshling improvements to the dll but with less stuff it has to be faster.

+1

JosEPh

 

[strategyonly]

ROM AND2 now has all of Koshling improvements to the dll but with less stuff it has to be faster.

All of then, even the graphic paging

 

[Hydromancerx]

Yeah graphics are a big load on any mod. I am sure the flavor units we have are not helping, but they sure do look nice.

I wonder how big those other mods would be if they did not have animated leaderheads like we do. Or I should say how HUGE C2C would be WITH animated leadeheads. We have cut many corners in C2C to try to make things go faster when it comes to graphics. Heck most buildings now don't even have graphics anymore. And imagine how big C2C could get if we ever finished adding all the audio quotes and wonder movies.

 

[strategyonly]

Yeah graphics are a big load on any mod. I am sure the flavor units we have are not helping, but they sure do look nice.

I wonder how big those other mods would be if they did not have animated leaderheads like we do. Or I should say how HUGE C2C would be WITH animated leaderheads. We have cut many corners in C2C to try to make things go faster when it comes to graphics. Heck most buildings now don't even have graphics anymore. And imagine how big C2C could get if we ever finished adding all the audio quotes and wonder movies.

I just looked at it and its not even a (no offense to them) but a contender. Its so easy and really doesnt have much of the thinking C2C has or should i say the Human Player needs to make. Its just, well not much content?? I didnt want to stick my foot in my mouth here.
Still cant even be compared to C2C, not even in the slightest. . .

 

[masaykh]

my observations tell what turn times CPU-dependable and so it can be optimized.
I already suggested to try and compile dll using clang or gcc. But it need to have code changed so it will be compatible with such compilers.
Produced DLL wil be compatible and will work.

 

[alberts8]

my observations tell what turn times CPU-dependable and so it can be optimized.
I already suggested to try and compile dll using clang or gcc. But it need to have code changed so it will be compatible with such compilers.
Produced DLL wil be compatible and will work.

I have tried using gcc a while ago but in the end it wasn't working because i don't have much experience using gcc, mingw....
You say it could be possible but how much work is it to get it 100% working? We already don't have many programmers in the team i don't think we have time for this experiment. If you have experience with those compilers under windows and want to try this feel free to do so and if you have questions about i try to help you if i can.


You can compile the dll with the intel c++ compiler 11.1 this works 100% turn times are reduced by at least ~15-25% maybe more. But the problem is no one has that compiler because it's not free.

 

[masaykh]

my friend work with Intel compiler so he can produce dll for us....

i posted errors what produced by gcc then i give him code..

 

[Dancing Hoskuld]

my friend work with Intel compiler so he can produce dll for us....

i posted errors what produced by gcc then i give him code..

Then when you and your friend are not available we get stuck. . It is not a viable long term solution to rely on software that everyone does not have access to.

 

[masaykh]

but you can periodicaly get optimized dll so not bad..