About the veterancy system
-
@jip Oh, that's easy - The vast lion's share of the compute time is taken up by function calls across the C++/lua boundary. It's about two orders of magnitude slower than anything else. Potential areas for improvement would be to look for areas where the lua makes repeated, unneccessary calls to engine. I worked with a couple of guys to eliminate all the points in the exe which make stupid calls the other way, so that's already done.
Other than that, you can try using more local variables in hot code - Intel, collision detection, economy events.
-
@icedreamer said in About the veterancy system:
@jip Oh, that's easy - The vast lion's share of the compute time is taken up by function calls across the C++/lua boundary. It's about two orders of magnitude slower than anything else. Potential areas for improvement would be to look for areas where the lua makes repeated, unneccessary calls to engine. I worked with a couple of guys to eliminate all the points in the exe which make stupid calls the other way, so that's already done.
Other than that, you can try using more local variables in hot code - Intel, collision detection, economy events.
Yes - I've been looking into using locals. LOUD applies a similar pattern to optimize functions or to push them as an 'upvalue' which is still better than a global. A wikipedia entry that I've learned from quite a bit: https://springrts.com/wiki/Lua_Performance
Do you happen to know about how the garbage collector works in Lua? I've been trying to find informative examples, but came out short.
-
Oh - and while I have you - how did you found out about the boundary passing being expensive?
-
I profiled the game while it was running and produced a statistical output of how much time was spend in each function, both in engine, in lua, and the time at which the function call was registered.
-
Do you still have that profiler?
-
option 2
-
A few months later, there was something wrong with the implementation of the veterancy system: it could hoard megabytes of worth of tables into memory! Read all about it here:
-
@jip said in About the veterancy system:
o back to the 'on-kill' notion, where the killer takes the mass value of the killee. This prevents
None of the 5 options are optimal.
Option 2 would lead to units vetting while fighting each other, a T1 arty hitting a T2 tank might lead to it vetting, a Sam hitting a Start would vet on 1 volley without killing it. Essentially dependent on the situation you get a restructuring of the balance as some units would by default get more HP during the fight while others wouldn't.
If you implement some kind of time delay to vet up maybe that would prevent this but i imagine it would cause us having to have allocation tables again.Is it possible to spread 50% of the XP gained from killing a unit in an area defined by a certain size around the unit that got killed, i.e. spreading the XP to an army of units that likely were involved in the fight and giving 50% to the unit who got the kill? This would work for arties and nukes across the map if we really want those to vet and at the same time only give them half vet because if it is an arty war, no xp is received in the area where the unit/buildings are killed.
This should also prevent table allocations and is option 6. -
I actually like the idea of units vetting while fighting each other. Sounds chaotic. Imagine a monkeylord fighting a gc. Its gonna vet up way faster than the gc.
-
I see that most people are voting for option 2 here, but there are also downsides of giving veterancy to the unit that has not performed a kill yet. For example, you have tele sacu teleporting in your game ender, let's take mavor. They will get veterancy for every shot that they perform. In case of mavor 1 hp costs 28 mass (225K/8K HP). The damage of sera tele sacu is 400 with 400 dps. Which means tele sacu will get an instant veterancy after performing one shot in something very expensive, mavor, yolo, paragon, etc. In the case of the SMD, for example, if you repair it, you will buff the teleporting unit, cause it is gonna get a veterancy as long as you are repairing it. So the teleport is gonna be way to overpowered with an option 2. The same goes for every ACU teleporting into the game ender and not killing them. With an option 2 you are gonna extremely buff the teleport, because the teleporting unit (SACU or ACU) is gonna get a huge buff for every second damaging the target. Therefore I guess the options 3 is better or 4 is also ok. The veterancy should be given for kill, not for a damage delivered. That's my opinion, thanks!
-
we can do a combination here :
(2) We keep the system but we provide the veterancy when the damage is applied, instead of when the unit dies. This prevents table allocations.
veterancy is gained on damage, up to the amount needed for next level -1
(3) We go back to the 'on-kill' notion, where the killer takes the mass value of the killee. This prevents table allocations.
xp for the unit that gets the kill will allow it to pass the threshold to vet
could also do a version where the kill could vet units in a certain close radius
this would require some separation of the vet bar from the total mass kill/damaged amount (which we want to keep to see how much value the unit got)
i wonder if a small, visual battlefield indicator of units vetting would be worth adding? (yes ex-dota player here)
-
Note that this is a topic from more than a year ago
, the original debate is no longer of concern. Just found it interesting to post that there was a few things off with the veterancy-based system after all.i wonder if a small, visual battlefield indicator of units vetting would be worth adding? (yes ex-dota player here)
I'm open to this. Do you have a suggestion on an effect to use?
A work of art is never finished, merely abandoned
-
@jip said in About the veterancy system:
Note that this is a topic from more than a year ago
, the original debate is no longer of concern. Just found it interesting to post that there was a few things off with the veterancy-based system after all.good to know, seems forum was starting to re-design it again
i wonder if a small, visual battlefield indicator of units vetting would be worth adding? (yes ex-dota player here)
I'm open to this. Do you have a suggestion on an effect to use?
well originally I was thinking of a brief glisten effect, similar to
https://forum.unity.com/threads/ui-text-shine-effect.467346/
but - this has been done a lot and would require new graphics
a much better faf option would be to use the visuals from units being built, so on vet cybran would display the semi-transparent frame, aeon the metallic form, uef the 3d blueprint, and sera the ghostly shape, for 1 in-game render (should need code only and no new graphics, beyond my skill though!)
-
I'm not so snappy on changing shaders (temporarily). I was more thinking in lines of an emitter / effect
-
@jip said in About the veterancy system:
Note that this is a topic from more than a year ago
, the original debate is no longer of concern. Just found it interesting to post that there was a few things off with the veterancy-based system after all.i wonder if a small, visual battlefield indicator of units vetting would be worth adding? (yes ex-dota player here)
I'm open to this. Do you have a suggestion on an effect to use?
How about having veternacy symbols appear next to units like some other games do? Maybe something similar to
0 vet: no marks
1 vet: 1 gold mark
2 vet: 2 stacked gold marks
3 vet: 3 stacked gold marks
4 vet: 4 stacked gold marks
5 vet: gold starpfp credit to gieb
-
There is another option: Remove veterancy.
-
What we need is less incentive to micro anything in this game, it's too heavy on it as it is
-
@penguin_
These look too contemporary, plus any symbols flashing in a big battle on top of projectiles, explosions, and other visual effects might make it too visually cluttered. Maybe at 5 vet a gold star would be ok since its a rare achievement.
-
@melanol said in About the veterancy system:
There is another option: Remove veterancy.
Remove veterancy bonuses completely.
Remove tracking of damage/veterancy from live games. Track only the number of kills, because that's easy to keep track of.
Keep track of veterancy only in non-live replays.
That way regular games run fast with less memory, and the game becomes less complicated/more streamlined.
But people can still see how much "work" a unit got done when they are watching a replay. The concept of a "5 vet tank" would exist only in replays, not in live games. Nothing of value would be lost. It would be easy enough to rebalance ACU health to make up for the missing veterancy bonuses (e.g. make it easier to repair an ACU or just give them all an extra 20% base health).
Perhaps it makes sense to track "mass killed" on certain units during live games, but only count up the mass value of whatever that unit killed (defined as: delivering the killing blow). that way, you can still see during a match how much damage your nuke launcher did or how much mass your T3 Heavy Arty has killed. But this would only be for a tiny number units (perhaps: nuke launchers, TML structures, T3 heavy artillery, and all T4s except paragon and the novax building)
Do we really need to know that a t1 bomber killed 223 mass worth of units? Isn't it more useful to know that it killed 6 units?
-
No, it is more useful to know how much mass something killed. You compare how much mass a unit kills to how much mass it costs to see how effective it is and if it was used properly. I doubt it is possible to only track that stuff in replays, but I don't know anything.
Rebalancing ACU hp would require work, so I'd just remove veterancy for everything else. Your reward for microing units is that they are more effective and survive to deal more damage. Giving them random HP boosts and regen is just doubly annoying for the person on the receiving end.
