Modders Guide to FfH2
- Thread starter Kael
- Start date
I started another thread a while back but got no responses, so I'm going to re-ask my question again here since it is the thread with the most people with programming experience, and in case Kael missed it from his vacation. (Please hit me over the head if this suggestion has already been implemented in the latest release, I won't be able to download things for a while yet)
In all the python event check the code has the gc.getInfoTypeForString() plastered all over the place. Would it not be faster to define all the getInfoTypeForString() variables as global variables earlier, such as on game start?
i.e. (the spacing is messed up for the spoiler, but the idea is still there)
would turn into this:
Or would that just use up more memory resources? Also it would allow the code to be optimized in a few places by changing some of the code to reduce the number of checks in some places by nesting some of the variables in a tuple and doing an if in check to check if it should even start further processing?
found an example
Or if I am wrong, just let me know before I try to mess with my game when I get a copy of the latest version.
In all the python event check the code has the gc.getInfoTypeForString() plastered all over the place. Would it not be faster to define all the getInfoTypeForString() variables as global variables earlier, such as on game start?
i.e. (the spacing is messed up for the spoiler, but the idea is still there)
Spoiler :
would turn into this:
Spoiler :
Or would that just use up more memory resources? Also it would allow the code to be optimized in a few places by changing some of the code to reduce the number of checks in some places by nesting some of the variables in a tuple and doing an if in check to check if it should even start further processing?
found an example
Spoiler :
Or if I am wrong, just let me know before I try to mess with my game when I get a copy of the latest version.
I'm not sure if you can use gc when declaring globals. (But possibly as self.variableName)
I tried defining all my variables in another file (CvHelpers.py) which I import like this:
from CvHelpers import *
But it wouldn't load the variables correctly.
I think you have to declare them before, but set the value in onGameStart (possibly also in onLoadGame)
@Kael: Will the source for patch E be uploaded any time soon?
I tried defining all my variables in another file (CvHelpers.py) which I import like this:
from CvHelpers import *
But it wouldn't load the variables correctly.
I think you have to declare them before, but set the value in onGameStart (possibly also in onLoadGame)
@Kael: Will the source for patch E be uploaded any time soon?
I did it for my modcomp and it seems to work fine (http://forums.civfanatics.com/showthread.php?t=267783). I had to declare the variables in the same class as where they are called. I also had to add some calls to a new function in the class that triggered onGameLoad() and onGameStart(). So I guess the global variables are setup for that particular class.
Actually it was necessary for my Promotion modcomp because the iterations of checking each unit for each promotion on each plot type became too much overhead.
Example of the big variable in my modcomp:
Spoiler :
As you can see quite a beefy variable. After I made that one change in the code, all of a sudden the endgame on a large map was much faster than the endgame on a small map.
Kael
Deity
Patch "e" source is linked in the first post.
@mtagge- I dont know if that would offer increase performance or not. I don't loop on getInfo requests so anytime I run through a loop that uses them I define it outside the loop. As such I think it would be hard to find a place where a global declaration would offer a noticeable improvement.
But thats just my guess, if you can show a change that has any noticable speed improvement I would love to check it out.
@mtagge- I dont know if that would offer increase performance or not. I don't loop on getInfo requests so anytime I run through a loop that uses them I define it outside the loop. As such I think it would be hard to find a place where a global declaration would offer a noticeable improvement.
But thats just my guess, if you can show a change that has any noticable speed improvement I would love to check it out.
Maybe if I get some down time this weekend I will try starting up a new game and getting to the endgame point then timing a few of the turns. Then I'll try pulling out some of the getInfo request out and declaring a few globals then retiming it. If you don't loop it shouldn't be too bad to change a few with the replace command and some mad Ctrl+X, Ctrl+C, Ctrl+V commands. If I get around to it I'll post the results.
Kael
Deity
That would be much appreciated. I can pass you some massive lategame saves if that would make it easier for you?
Actually I am looking forward to a OCC Huge Marathon game for a change of pace. See if I can eek out a altar victory while I am at it. (Besides right now I can't download the same games through my company's firewall).
It actually wasn't that bad to change all the globals, the downside is that I am not up to date with the release version, but the results should be the same regardless of the version. I'll post my results here on Monday.
PS Now that I have spent oodles of time looking through the python code, you don't put actual loops in, you have a tendency to write all the code out longhand. A good example is when you define each hero for each civ, you just list each of them on a seperate set of lines.
Sto
Should i code today ...
I've tried to found where the scrub were added , but I've found nothing . I've made a search in all python files , SDK and XML with 'FEATURE_SCRUB' , nothing . Since iAppearance = 0 , the feature generator should not add some . So how and when the scrub have been added please ?
Tcho !
Tcho !
Kael
Deity
iAppearance is the chance that feature will spread (as forests do). The game places any features on terrain as long as the terrain is valid. So int he scrubs place because we set desert as an acceptable terrain the map generator will put them there. If we take that terrain out it wont autoamtically place them and then could place them via our own script or whatever.
Sto
Should i code today ...
I'm sorry but i still do not understand why there is some scrub . iAppearance is the probability (/10000) that the feature generator place some feature on a plot . In fact , if the default SDK function is launched there should not be any forest forest ,jungle or ice . The generic function to place feature is addFeatures in CvMapGeneratorUtil.py :
Spoiler :
As you can see , if you don't tell the process to place a feature , it will not be placed . The chance that a feature spread is based with <iGrowth> like for forest or jungle (in doFeature) . The features like flood plains (with iappearance at 10000 will always be place , oasis with 500 will be placed with a chance of 500/10000) then if there is still no feature , it try to put a forest, jungle or ice .
15 min later ...
I've understood now where the scrub are placed ... i've generated a huge map with only 2 players and lots of deserts . The only scrub were near starting locations . This should be normalizeAddExtra that add them thinking this is a good add , but in fact replace some flood plains . This should be great to put <bNoRiver>1</bNoRiver> with scrubs . The simple way to place some would be to put iAppearance to 300-500 like oasis perhaps .
Some other things i've noticed :
_ Sometimes a barbarian city spawn on an unique feature :i've fixed that with CvGamesUtils.cannotFoundCity:
Code:
iImprovement = pPlot.getImprovementType()
if iImprovement != -1 :
if gc.getImprovementInfo(iImprovement).isUnique() :
return True_ with normalization , a bonus can be added to an unique feature (those that don't spread a bonus) , that would be great to check if the plot don't have an unique improvement in CvPlot::canHaveBonus . This will prevent the game to add some bonus into an unique improvement plot on normalize starting plot functions. But also when you add some bonus during the game like with rite of omega (... not sure of the name) . I've still not wrote the code ,tell me if want me to do it .
Tcho !
Kael
Deity
Ahh, you are absolutly right. I was confusing iApperance and iGrowth. I'll block them from founding by rivers and the block for unique improvements and barb cities as you suggested.
Here are the results from my testing over the weekend:
All actual tests using lategame save data did not produce reliable data. It seemed that the code was "remembering" bits of data from the code that was used when the new game started (i.e. errors put in to trigger refered to places and lines that did not exist any longer). Also on my computer I was getting 17 seconds maximum between turns playing marathon, huge, 3 extra civs, living world, no AI building requirements on turns when all civs were at war with at least one other civ. On the upside the tests with the changed FFH code did allow me to find out that the game creates the initial units (and onUnitCreation()) before it runs the onGameStart() code so I got errors for the checks to see if the initial units were elementals or undead. It wasn't a problem onGameLoad()
So, what I did was go a more scientific route and added the following code to the original game:
When I turn cycled I took lag time readings (spinning globe time) and got a range of 15.10 seconds to 15.56 seconds (although I am by no means a great stopwatch user)
Then I changed the code to this:
This time I took readings and got a range between 1.35 seconds and 1.9 seconds (again my reflexes are far from perfect, which is why I play CIV rather than any FPS)
From these results it is hard to determine how much overhead is caused by gc calls in the code. Although 10000000 is a large number of times to cycle through the small block of code my laptop doesn't seem to have much of a problem running through all the FFH code on a huge/marathon game running Vista and streaming media to a 360 at the same time. For most machines changing to global variables could have significant gains (or for all I know it could actually slow down the game by having to store too many variables in memory) but it will be hard to get definitive numbers. If someone wants to test the code with a slightly older comp I would be interested in the numbers. Also for some of the most frequently used code (onUnitMove() onUnitKilled() code as well as many of the every turn plot functions maybe although the changelog shows that you moved some of that to the SDK) I don't think it would hurt to make some global variables.
All actual tests using lategame save data did not produce reliable data. It seemed that the code was "remembering" bits of data from the code that was used when the new game started (i.e. errors put in to trigger refered to places and lines that did not exist any longer). Also on my computer I was getting 17 seconds maximum between turns playing marathon, huge, 3 extra civs, living world, no AI building requirements on turns when all civs were at war with at least one other civ. On the upside the tests with the changed FFH code did allow me to find out that the game creates the initial units (and onUnitCreation()) before it runs the onGameStart() code so I got errors for the checks to see if the initial units were elementals or undead. It wasn't a problem onGameLoad()
So, what I did was go a more scientific route and added the following code to the original game:
Spoiler :
Then I changed the code to this:
Spoiler :
From these results it is hard to determine how much overhead is caused by gc calls in the code. Although 10000000 is a large number of times to cycle through the small block of code my laptop doesn't seem to have much of a problem running through all the FFH code on a huge/marathon game running Vista and streaming media to a 360 at the same time. For most machines changing to global variables could have significant gains (or for all I know it could actually slow down the game by having to store too many variables in memory) but it will be hard to get definitive numbers. If someone wants to test the code with a slightly older comp I would be interested in the numbers. Also for some of the most frequently used code (onUnitMove() onUnitKilled() code as well as many of the every turn plot functions maybe although the changelog shows that you moved some of that to the SDK) I don't think it would hurt to make some global variables.
xienwolf
Deity
The Worker Promotions Thread has a great idea which I want to implement. So any advice on how to add Workrate into the PromotionInfos XML? I know I'll have to change the Schema, and I am confortable with that, but where else will I have to make adjustments?
Vehem
Modmod Monkey
- Joined
- Nov 22, 2005
- Messages
- 3,219
PromotionInfos (in CvInfos.cpp) will need a new variable to store the tag's value and you'll need to have it read/write it in all the relevant places (easiest way is to find another int value to copy - the screenshot shows where those are for iFirstStrikesChange - you'd just have to add iWorkRateChange in exactly the same places - CodeBlocks doesn't allow you to Copy/Paste the search results - hence the clumsy screenshot).
Then you need to build that into this function in CvUnit.
Code:
int CvUnit::workRate(bool bMax) const
{
int iRate;
if (!bMax)
{
if (!canMove())
{
return 0;
}
}
iRate = m_pUnitInfo->getWorkRate();
iRate *= std::max(0, (GET_PLAYER(getOwnerINLINE()).getWorkerSpeedModifier() + 100));
iRate /= 100;
if (!isHuman() && !isBarbarian())
{
iRate *= std::max(0, (GC.getHandicapInfo(GC.getGameINLINE().getHandicapType()).getAIWorkRateModifier() + 100));
iRate /= 100;
}
return iRate;
}You can either check all promotions and work out the work-rate bonus here, or add a variable to CvUnit to store the workrate bonus which is updated when the promotion is added.
Attachments
xienwolf
Deity
Thanks much for the help Vehem. I'm making my first attempt at it now, hopefully it all goes well.
Another question I had was about the Schema Files: Why aren't all the values just set to MinOccurs = 0? Then the individual XML files could be considerably smaller, and much easier to browse through and change.
EDIT: Having one final problem with my formula. Is this the correct form?
iRate *= (pUnit->getWorkRateChange() + 100);
Using
iRate *= (m_pUnitInfo->getWorkRateChange() + 100);
Gave me an error because getWorkRateChange isn't declared as a part of CvUnitInfo.
EDIT 2: I think I got it now, I changed the initial Definition of the iRate to:
iRate = std::max (0, m_pUnitInfo->getWorkRate() + getWorkRateChange());
It is compiling now at least, so we'll see what other errors I have in there, and if none then attempt to run the dang thing
EDIT 3: All compiled now. But is this warning normal? Sure isn't related to what I changed, but might indicate I set things up wrong.
Linking dynamic library: Final Release\CvGameCoreDLL.dll
Creating library Final Release\CvGameCoreDLL.lib and object Final Release\CvGameCoreDLL.exp
LINK : warning LNK4089: all references to 'KERNEL32.dll' discarded by /OPT:REF
Process terminated with status 0 (19 minutes, 11 seconds)
0 errors, 1 warnings
EDIT 4: Whatever the warning was it didn't matter Mod works, posting it now.
Another question I had was about the Schema Files: Why aren't all the values just set to MinOccurs = 0? Then the individual XML files could be considerably smaller, and much easier to browse through and change.
EDIT: Having one final problem with my formula. Is this the correct form?
iRate *= (pUnit->getWorkRateChange() + 100);
Using
iRate *= (m_pUnitInfo->getWorkRateChange() + 100);
Gave me an error because getWorkRateChange isn't declared as a part of CvUnitInfo.
EDIT 2: I think I got it now, I changed the initial Definition of the iRate to:
iRate = std::max (0, m_pUnitInfo->getWorkRate() + getWorkRateChange());
It is compiling now at least, so we'll see what other errors I have in there, and if none then attempt to run the dang thing
EDIT 3: All compiled now. But is this warning normal? Sure isn't related to what I changed, but might indicate I set things up wrong.
Linking dynamic library: Final Release\CvGameCoreDLL.dll
Creating library Final Release\CvGameCoreDLL.lib and object Final Release\CvGameCoreDLL.exp
LINK : warning LNK4089: all references to 'KERNEL32.dll' discarded by /OPT:REF
Process terminated with status 0 (19 minutes, 11 seconds)
0 errors, 1 warnings
EDIT 4: Whatever the warning was it didn't matter
Mod works, posting it now.
Vehem
Modmod Monkey
- Joined
- Nov 22, 2005
- Messages
- 3,219
I've wondered that myself - though a lot of them couldn't be set to that because they need to have some sort of value. Of course - a lot don't...
xienwolf said:
That one is normal - there's no need for Kernel32 to have an entrypoint into CvGameCore as only the Civ4/BtS exe's ever need to use it (Kernel32 is kinda core to Windows and can be used to run code from inside DLLs). I did at one point have it compiling without that warning (in Visual Studio) but the only result was that the DLL was significantly larger.
TheCowSaysMoooo
Warlord
Where is the function that controls the reversion of mana nodes to raw mana when crossing cultures? And how difficult would it be to make that function also remove the improvement on the node that's being reverted?
I poked around (quickly) through the python, but didn't see it anywhere... so I'm guessing it's in the dll?
I poked around (quickly) through the python, but didn't see it anywhere... so I'm guessing it's in the dll?
Vehem
Modmod Monkey
- Joined
- Nov 22, 2005
- Messages
- 3,219
CvPlot.cpp, in setOwner
Code:
//FfH: Added by Kael 09/01/2007
if (getBonusType(NO_TEAM) != NO_BONUS)
{
if (GC.getBonusInfo((BonusTypes)getBonusType((TeamTypes)NO_TEAM)).getBonusClassType() == GC.getDefineINT("BONUSCLASS_MANA"))
{
if (getImprovementType() == NO_IMPROVEMENT || GC.getImprovementInfo((ImprovementTypes)getImprovementType()).isUnique() == false)
{
setBonusType((BonusTypes)GC.getDefineINT("BONUS_MANA"));
}
}
}
//FfH: End AddShort version - if there is (a) no improvment or (b) a non-unique improvement on a tile with a Bonus of BONUSCLASS_MANA (any mana-type has this class), reset the tile to have the Bonus described as "BONUS_MANA" in the global-defines.
You'd want a setImprovementType here as well, to remove the improvement (possibly checking that the improvment is a mana-related one first)
