AI Duke Out
Moderators: hoijui, Moderators
AI Duke Out
Thought you boys might like to know the results of a few games I ran today whilst doing other stuff - I let the latest versions of NTAI and AAI, and the JCAI that ships with Spring 0.67b3 fight it out on Mars, both as ARM vs ARM vs ARM and same for CORE.... NTAI won hands down both times.
Interestingly though AAI is pretty good, it fails to build anything at all if it starts in the bottom left start position on that map - might need looking into!
I tried to upload a replay so you could watch the action, but FU wasn't playing ball =(
Thanks for taking the time to develop these AIs guys =)
Cheers
Munch
Interestingly though AAI is pretty good, it fails to build anything at all if it starts in the bottom left start position on that map - might need looking into!
I tried to upload a replay so you could watch the action, but FU wasn't playing ball =(
Thanks for taking the time to develop these AIs guys =)
Cheers
Munch
You dont get any errors detailing out of virtual memory at all with NTAI? They're driving me mad, and they're happening in the middle of functions that ahvent changed since 0.25 even tho the problems started with 0.28 (and around the tiem I upgraded versions of spring too hmmm.....).
Currently my games with NTAI v NTAI v AAI v AAI (0.29.05 and 0.32 respectively), go well, NTAI expands quickly but the game runs into the memory problem aroudnt he 11 minute mark every single time, and amsse gets it too, btu his seems mroe around either 11 mintues or 32 min marks.
So far nobody but him and me has complained about this even tho I get these problems in 0.28.10+........
Currently my games with NTAI v NTAI v AAI v AAI (0.29.05 and 0.32 respectively), go well, NTAI expands quickly but the game runs into the memory problem aroudnt he 11 minute mark every single time, and amsse gets it too, btu his seems mroe around either 11 mintues or 32 min marks.
So far nobody but him and me has complained about this even tho I get these problems in 0.28.10+........
Yah but eitherway I am having trouble fidning this bug.
According to the logs the error happens midway through writting to the log, usually I see the gametime, followed byt eh end of the file, or half a tag written out, which makes no sense and points tot eh error being in:
I've tried commenting out the AAI attack code I ported over and edited, the stockpiling routine, I've gotten rid of big masses of attack group handling ode and re-implemented it , I've added lots and lots of checks just to make sure things are working as they should and no problems are occuring, I turned off all the debugging options by default so blobs and the mouse thgins dotn show uelss you type ".mouse" and ".lines", I've commented the D-gunnign and stoickpiling of missiles, and the firing of stockpiled missiles, I've commented out high trajectory and changing mvoe state, and i still get the same results.
The only thign elft it could have been was cains metal class, which I recently tore out and replaced with krogothes mex class , and tho there was idneed an improvement in gameplay, the error still occurs and it is gettign very annoying.
What's mroe annoying si that it only seems to be happening to me ro masse, and I know not of any of the hundred or mroe people who downloaded those versions complaining once about it happening.......
Eitherway I wont eb releasing 0.29 for a while at least till I've overhauled scouting and sorted this mystery out.
According to the logs the error happens midway through writting to the log, usually I see the gametime, followed byt eh end of the file, or half a tag written out, which makes no sense and points tot eh error being in:
Code: Select all
if (logFile){
fputs(gtime.c_str(), logFile);
}
The only thign elft it could have been was cains metal class, which I recently tore out and replaced with krogothes mex class , and tho there was idneed an improvement in gameplay, the error still occurs and it is gettign very annoying.
What's mroe annoying si that it only seems to be happening to me ro masse, and I know not of any of the hundred or mroe people who downloaded those versions complaining once about it happening.......
Eitherway I wont eb releasing 0.29 for a while at least till I've overhauled scouting and sorted this mystery out.
hmmm
Masse has just playtested again and he mentioned that he's been playtesting different situations on AAI 0.32 and has gottent he same virtual memory too low crash I get on NTAI only games, even thot he game was just 4 AAI's in a ffa with no NTAI present.
Thus I have been searching for an AI bug when ti is an engine bug, and a nasty one at that.
AAI 0.32 wont build on comet catcher if it starts in the top east corner, it builds a soalr, waits ages then builds a second then does nought. It also builds too little resources inbetween its first and second factories and ends up stalling towards the end of the second factory.
Masse has just playtested again and he mentioned that he's been playtesting different situations on AAI 0.32 and has gottent he same virtual memory too low crash I get on NTAI only games, even thot he game was just 4 AAI's in a ffa with no NTAI present.
Thus I have been searching for an AI bug when ti is an engine bug, and a nasty one at that.
AAI 0.32 wont build on comet catcher if it starts in the top east corner, it builds a soalr, waits ages then builds a second then does nought. It also builds too little resources inbetween its first and second factories and ends up stalling towards the end of the second factory.
new releases are good if your current release is a pain in the ass. But yea, the same thing happened for me on Comet Catchers with AAI 0.32. Though AAI haven't sucked up my memory during play yet. I've complained about it, or least I think I have, but either case, I'm complaining now :) cept for mine sometimes just locks up, need to switch my OS back onto the fast harddrive.
And you're not sending em descriptions or AILogs at all?!!?!?!?!?!?!?!? Are there hudnreds fo suffering people out there who refuse to give data that could fix the problem?
I take it the majority fot hese crashes are VM is low emssages and their associated freezes.
Suffice to say the new copy of NTAI now runs on my machien 40 mins+, and I ahve a new set of bugs that i exchanged for the VM low bug, but these are fixable!
I take it the majority fot hese crashes are VM is low emssages and their associated freezes.
Suffice to say the new copy of NTAI now runs on my machien 40 mins+, and I ahve a new set of bugs that i exchanged for the VM low bug, but these are fixable!
OTAI
Sort of. It was surprisingly hard to get the other three AIs to play against each other, even though I'd succesfully played games against each of them (albiet a lot of the games I started crashed before the end). OTAI on the otherhand wouldn't even start a game with me, so I didn't get as far as trying to incorporate that in the test - sorry. I might give it a go at some other point.CrowJuice wrote:You forgot OTAI :wink:
Whilst I'm here, does anybody know how to start a game without any human players? The way I got the game working was to start with me as player/team 0 and have the AIs as teams 1-3. I then self-D as soon as the game started to become a spec. Any change in that configuration seems to cause problems. This is a shame as it means that if for example I did want to add OTAI, I couldn't set it up with one AI in each corner of the map (which is the obvious thing to do).
Munch
It isnt the logging, loggign was in NTAI before the error cropped up, NTAI's log files ahve reduced in size drastically and it logs in plaintext now, and thatd dint fix the bug, I looked at the cosntruction code and changed everythign so ti didnt use pointers to unti objects btu refrences. This made the error go away, ro become none existant in most cases, btu ti broke the con code at the same time by generating exceptions, I'm just fixing those new bugs now.
Code: Select all
[GAME]
{
Mapname=CometCatcher.smf;
StartMetal=1000;
StartEnergy=1000;
MaxUnits=1000;
StartPosType=0;
GameMode=0;
GameType=xta_se_v066.sdz;
hostip=localhost;
hostport=8452;
MyPlayerNum=0;
NumPlayers=1;
NumTeams=4;
NumAllyTeams=4;
[PLAYER0]
{
name=yourname;
team=0;
}
[TEAM0]
{
TeamLeader=0;
AllyTeam=0;
Color=0;
Side=Core;
Handicap=0;
AIDLL=aidll\globalai\NTAI.dll;
}
[TEAM1]
{
TeamLeader=0;
AllyTeam=3;
Color=3;
Side=Core;
Handicap=0;
AIDLL=aidll\globalai\AAI.dll;
}
[TEAM2]
{
TeamLeader=0;
AllyTeam=2;
Color=1;
Side=Core;
Handicap=0;
AIDLL=aidll\globalai\NTAI.dll;
}
[TEAM3]
{
TeamLeader=0;
AllyTeam=1;
Color=2;
Side=Core;
Handicap=0;
AIDLL=aidll\globalai\AAI.dll;
}
[ALLYTEAM0]
{
NumAllies=0;
}
[ALLYTEAM1]
{
NumAllies=0;
}
[ALLYTEAM2]
{
NumAllies=0;
}
[ALLYTEAM3]
{
NumAllies=0;
}
NumRestrictions=0;
[RESTRICT]
{
}
You start out in the bottom left corner, but you're sharing your commander with one fo the NTAI's, just do .cheat .spectator and watch what happens.