Scourge of War - Waterloo Diaries Part 3

[Xaver]

Thanks for the reply, is not like see formations by sprites and no companies is critical but maybe you need made a little less chaotic the formation of squares because looks like time to complete formation is a little high with soldiers from one flank in previous formation forming in the oposite flank in the square... or maybe the formations are "completed" with full bonus even when not all soldiers are in position???

Skirmishers... ummm use a numeric limit of 450 is not problematic with certain battalions that were in battle heavy damaged??? and well, in other battles in napoleonic period you can find even detached companies or half battalions to create add-hoc units for a certain mission... you explore the use of a % over a numeric cap??? i refer mantein X% as the minimun battalion size after detach skirmishers and well, use a diferent % by unit type, more trained the unit less troops needs battalion to mantein integrity and infantry type.

Thanks.

[mitra76]

The melee bonus associated to the square is applied only the unit switched from the march state (which is the state used also during the process of formation making) to standing state, but in this first implementation of AI I preferred to be cautious and conservative about the process of cavalry charging and square forming because distances and time are critical, because start a charge a half second before the square is completed means that cavalry going to be destroyed and infantry trying to form square with cavalry already too near is not a good reaction.

Because I cannot control the time of events ( or at least is still not part of my world domination powers ) but I can control space, I introduced a buffer area (variable according to the stance between 70 and 100 yards) under which the infantry cannot form square both because engaged too nearly from other infantry or because cavalry is too near so the battalion commander "evaluate" is too late to form the square (the bad judgement error done from the commanding officer of 69th regiment at Quatre Bras demonstrate that the process was far from linear). This is also the good side effect that action of infantry and cavalry work fine, because cavlary can force the squares to form and leave infantry to destroy the them, or infantry can block the squares to form and leave cavalry to destroy the lines.

At the same time cavalry will stop the attack if at the moment to start the charge (which is more or less equivalent to the buffer area of above) the square making process is already started.

[mitra76]

Skirmishers... ummm use a numeric limit of 450 is not problematic with certain battalions that were in battle heavy damaged??? and well, in other battles in napoleonic period you can find even detached companies or half battalions to create add-hoc units for a certain mission... you explore the use of a % over a numeric cap??? i refer mantein X% as the minimun battalion size after detach skirmishers and well, use a diferent % by unit type, more trained the unit less troops needs battalion to mantein integrity and infantry type.

I preferred have the AI been less intrusive on the skirmishers decision for now; there're players who can appreciate the autonomous use of skirmishers by part of AI, other could prefer hold their battalions strong and could be confused from too many detachments, that is why I used a limit of 450 men and percentage of max 1/6 of organic. In any case players are always free to detach more in autonomy also for AI controlled brigade.

[Gunfreak]

450 sounds very resonable, many miniature wargames wont let smaller units detach skirmishes. And they have to rely on special light units, like jagers ect. To give skirmish support.

[Guest]

Because I cannot control the time of events ( or at least is still not part of my world domination powers ) but I can control space

My home-desktop's CPU is gonna rebel to this once again, Davide! :whistle:

Spatial analysis involving geometrical and trigonometric calculations are very much far more expensive in terms of computational power than, say, 1) posting an AI message into a centralized event queue with delayed dispatch time or 2) updating a timeout variable at every AI tick... Hopefully you're performing SIMD maths at least!

After reading Riccardo's concerns about "low FPS with only a couple of divisions deployed" as well, I'm beginning to suspect this game won't run smoothly either with 60-cores CPUs of super-computers Jim mentioned...

I'll wait another week or so for final judgments. But I can almost bet on the outcomes... Expect a lot of troublesome suggestions this summer. B)

[mitra76]

Nicolò a simple distance check between two points going not kill your CPU; in the fact the game never took more than 45-50% of my CPU time also in the middle of the biggest battles

[Guest]

Nicolò a simple distance check between two points going not kill your CPU; in the fact the game never took more than 45-50% of my CPU time also in the middle of the biggest battles

Sorry for stressing Davide, but even the simplest sqrt is much slower than mov + cmp and, in the long run, can hurt performances at runtime if misused or abused. Figure out if you need to take into consideration the heading of units (sin/cos/atan2)... And also don't forget that overally speaking RAM memory is much cheaper than any kind of instruction executed in a loop. Please guys, turn quickly into more reactive-like AI programming approaches. "Over-dynamicization" is useless for Napoleonic Warfare and comes with this huge price. Playing at 10 fps doesn't mean lag only, but also impossibility for further developments. I believe I wrote it a few times already. :huh:

[Guest]

I forgot to say I'm telling this from a first-hand experience.
I was a very inefficient and unconstrained hobbyist programmer some time ago.
My projects became soon unsustainable from a performance standpoint, thus I had to stop my plan of features adding.
That was the reason I started taking multithreading and optimizations much more seriously then.
Don't follow my footsteps, please. It's a big waste of time after all...

[mitra76]

I forgot to say I'm telling this from a first-hand experience.
I was a very inefficient and unconstrained hobbyist programmer some time ago.
My projects became soon unsustainable from a performance standpoint, thus I had to stop my plan of features adding.
That was the reason I started taking multithreading and optimizations much more seriously then.
Don't follow my footsteps, please. It's a big waste of time after all...

I appreciate the suggestions but it seems me too soon to discuss how good the performances could be done or how the AI logic works before see the exposed code in the SDK.

What I can tell is that respect GB the AI code lines are more than double, the graphic is better and more complex, but the performances are improved.

[Xaver]

Thanks for the reply.

Well, is all a question of play with it and break it hehehe and maybe earlier than i expected :whistle: