[SCR] Smart Turrets v4.7.2 (AP, 05-04-14), v2.6 (TC, 21-08-13)

Drewgamer · Post by **Drewgamer** » Sun, 4. Sep 11, 15:06

garrry34 wrote:sorry I've been distracted by other matter lately, Shimrod did you remove the turret slot filler in 1.7?

Sorry, coming a little late into the conversation, turret slot filler?

Shimrod · Post by **Shimrod** » Sun, 4. Sep 11, 15:35

There used to be an option called 'Idle Weapons' in the settings menu which would fill the turret with arbitrary lasers before each Wait call. This was designed to mask the visuals of switching lasers, making it look prettier.

However this came at a cost of starving the free laser pool for all the turrets, as well as performance overhead of swapping lasers around before every Wait and clearing them afterwards. So it was never a good thing to enable unless there's a full supply of all lasers for all turrets and mainguns, a pretty rare scenario unless like me you edit all your TLaser volumes to 1

I decided to remove this option in 1.7 to clean up the code around wait calls and cut back on ongoing regression testing.

Drewgamer · Post by **Drewgamer** » Sun, 4. Sep 11, 17:12

Ah, that could explain why I didn't know about it. Only began using the script at version 1.8

Great script by the way, really enjoying it. Of course, I still don't quite get the whole "cache" thing; but as long as it works who cares, right?

I assume you went this method to provide compatibility without having to tailor the mod for each weapon mod out there?

Shimrod · Post by **Shimrod** » Sun, 4. Sep 11, 17:28

Yeah a major goal was was to have something that was going to be compatible with any mod without datafiles.

The laser delay sampling/caching thing is integral to this as the game doesn't make laser delays and DPS data available to scripts.

The other caches are simply for performance:
- to have a shortlist of lasers that are available in cargobay and compatible with a given turret to avoid calculating that every iteration.
- to put the best laser for a given target toward the top of the list.
- to scan targets once per ship and share the data between multiple turrets
- to scan missiles once per sector and share the data between multiple ships

Shimrod · Post by **Shimrod** » Wed, 7. Sep 11, 15:16

I'd be interested to hear whether Smart or 1.8 has any particular deficiency or problem being a barrier to its use, or perhaps any particular usability or performance annoyance that needs addressed.

I can't reproduce the OWP script replacement laser issue with 1.8. I'd appreciate if that could be confirmed as being still a problem or not and I'll spend more time trying to repro it.

Cheers.

garrry34 · Post by **garrry34** » Wed, 7. Sep 11, 15:29

shimrod I've just started testing it as I have working on some ships, I'll give it a roll today, may I make a suggest an addition for a future update, would you add an option in the general settings to reset the options to default...

Shimrod · Post by **Shimrod** » Wed, 7. Sep 11, 16:01

Sure, will do.

Drewgamer · Post by **Drewgamer** » Wed, 7. Sep 11, 16:28

Shimrod wrote:I'd be interested to hear whether Smart or 1.8 has any particular deficiency or problem being a barrier to its use, or perhaps any particular usability or performance annoyance that needs addressed.

I can't reproduce the OWP script replacement laser issue with 1.8. I'd appreciate if that could be confirmed as being still a problem or not and I'll spend more time trying to repro it.

Cheers.

I assume you mean Gazz's OOS rebalance v2 when you say OWP script? I've not tried that script; but when XRM 1.07 is released (gives me an excuse to start a new game), I can go pick up the OWP script and see how it behaves with Smart on my end.

Shimrod · Post by **Shimrod** » Wed, 7. Sep 11, 17:37

Thanks, and yes. A compatibility issue was reported with the OOS rebalance v2 OWP script:
http://forum.egosoft.com/viewtopic.php? ... 08#3665608

I tried to repro that in 1.7 but it worked ok for me:
http://forum.egosoft.com/viewtopic.php? ... 41#3671141

Could be there's some difference in laser availability between my repro environment and the op. I tend to spawn my test ships fully loaded, with laser ware volumes modified to 1 to allow most ships to include every combination of laser.

I don't generally test Smart in laser starvation conditions, though it worked sensibly back in 1.3 or so when I chcked, and works ok on NPC ships in 1.7 which tend to have sparse loads.

Drewgamer · Post by **Drewgamer** » Wed, 7. Sep 11, 19:10

Gotcha! I don't think I'll get time tonight to do any testing, but should definitely get some time as the week winds down (plus, the new version of XRM should be coming out this week so I can test the OWP script at the same time).

Shimrod · Post by **Shimrod** » Wed, 7. Sep 11, 19:12

Great, thanks. I've got itchy feet with these sparse XRM turret loadouts, think I'll cycle back to vanilla for foreseeable future. Been some time since I checked Smart on vanilla.

garrry34 · Post by **garrry34** » Wed, 7. Sep 11, 23:16

If you want I can look to the XRM side of things if you wish, , I don't think the missile defense is as effective in 1.8 as it was before, grant its still early in the testing, but from what I've seen so far it only fire 1 single laser instead of the multiple single lasers in previous editions...

Shimrod · Post by **Shimrod** » Thu, 8. Sep 11, 01:09

Is this a drop in antimissile performed by the playership or an AI driven one?

1.7 Antimissile target cap
I added a performance optimization in 1.7 where the missile array produced by the central ship task for use by turrets is ordered by increasing range and cropped to something like 20 items by default. This cap can be adjusted in the performance menu and also disabled - toggle it to where it would hit zero and this becomes disabled.

This was to reduce lag when large missile spams would pass by clusters of smart enabled ships, and each of the turrets on each of the ships check to see if the missiles are in the firing arc. The cap was an effective tradeoff allowing missile interception to be enabled without significant performance impact on NPCs.

1.8 - cap lifted for the playership
In 1.8 I excluded the player ship from the cap as it had decreased antimissile effectiveness where experience counted the most, particularly in my usual drone spawning tests where some turrets just wouldn't shoot as its targets were further away than the other turrets.

Since 1.7 antimissile traverses missiles by ascending range
Also note that prior to 1.7 missile targets weren't sorted in ascending range order but in descending threat, which might have been anything if the targets were all the same thing and not far apart. Sorting by range was both required for the cropping, but also it seemed most effective to begin antimissile by shooting the closest.

Ordering by range versus a barrage would show reduced spread in the individual gun fires as typically the warheads in a swarm are very close to each other, so it might appear that the turret is firing at the same target.

In 1.6 it might fire 2 guns per target
If there were less missile targets than guns in a turret I made it so the turret might potentially fire 2 guns at an antimissile target, to get more dps out within the same delay framework (and partly to address your interest in firing multiple shots at drones!). However when there are more targets than guns it should only fire 1 gun.

This could easily be made to use up all the gun slots instead of capping 2 shots per target however I didn't want to be firing full volleys at single drone targets.

I've found and fixed a code discrepancy in this piece this evening but its unclear if this would have any impact.

My observations of this evening
In 1.8 I've enabled only a single turret on an ocelot, spawned a xenon M7M and ran a script causing it to barrage me with flails. I've also spawned drones around the ship.

The turret is definitely engaging multiple targets with individual guns. There's certainly a noticeable visual difference over the old days. I have a strong suspicion that the iteration of targets by range rather than arbitrarily by calculated threat, is flattening things out when intercepting barrages.

Ocelots are still horribly effective when pointing their ass at the barrage when they use the up/down/back PALC, instantly destroying missiles.

Anubitus · Post by **Anubitus** » Thu, 8. Sep 11, 02:38

hi,

i been using this script a week, and tbh, im somewhat confused..

what is the dps cash delay for..?

i have noticed that a very fast fireing laser (lets say 200 rounds/min) suddenly fires extremly slow (lets say something like 30 rounds/min), why is this..? and if i check out the delay it has something like 2569.. if i set it to 1 it fires a never ending stream of bullets wich appears to be faster then the 200 rounds it can actually fire..?

im kinda lost here.. pls enlighten me..

sry if this is a stupid question..

EDIT: sometimes they dont even fire at all for a while..

AnU

Drewgamer · Post by **Drewgamer** » Thu, 8. Sep 11, 04:06

Here is my (preliminary) findings. Using a Teladi (XRM) Pteranodon with "default" loadout (plus a full compliment of Energy Bolt Chainguns) in sector (XRM) Saiinen II. This was against randomly spawned M3's, M6's and a single M7M.

+ SMART seems to do very well with limited loadout (just enough guns to fit turrets, or with a few extra). It switches rapidly enough to keep full coverage and will (unintentional?) even keep a turret a few lasers short so that it can keep optimal coverage on other turrets.

+ SMART (once its got a good sampling) uses laser data very effectively (almost too effectively

). It abused the (XRM) Gauss Cannon so much that it was obliterating fighters with them, instead of using Energy Bolt Chainguns (FAA was still preferred at close range).

+ SMART doesn't like large swarms of missiles. With a small volley incoming, the turrets quickly dispatch the missiles (a few might get through). However with huge volleys (such as a barrage from the M7M) the turrets seemed to get overwhelmed. At one point (during the SWARM barrage) the turrets seemed to actually "give up" and stop firing. You would think that with Energy Bolt Chainguns (and their ridiculous fire-rate) very few (if any) missiles should be getting through.

Closing notes: I really only got a good solid 30 minutes or so of actual testing so more needs to be done on my end. Also, the missile "problem" may be due, in part, to my inability to adjust the settings correctly. Also, I was playing around with the Valhalla and noticed (not really SMART's fault) that the Terran Capital ships lack any good missile defense (the poor Starburst Shockwave Cannon fires too slow for large volleys).

Cheers,
Drew

Requiemfang · Post by **Requiemfang** » Thu, 8. Sep 11, 05:01

XRM Terrans have the Electro-Magnetic Repeater, that gun is more rapid fire then the Electro-Magentic Plasma gun which is the vanilla mainstay fighter weapon.

The EMR should alone be able to cover missile coverage defense so long as you have enough of them on board a ship, that is of course if Smart isn't aware of this or hasn't been programed to use them. Although they don't have the same range has the EBC or PAC or other rapid fire weapons with over 2 kms range it still has decent range for 2 kms which should be enough to take out heavy hitting missiles before they come into contact range to nuke you with their area of effect blast radius.

Shimrod · Post by **Shimrod** » Thu, 8. Sep 11, 12:13

Anubitus wrote:what is the dps cash delay for..?

It's a creative workaround for the game not making laser DPS or refire delays available to scripts. Smart estimates these values and stores the result in the DPS cache. Actually the only meaningful value is the delay: DPS is calculated from delay, factoring in dmg per shot which is available to scripts.

The sampling code basically performs millisecond waits then tries to shoot the laser. If can't shoot it waits again. Once a shot is fired the sample is taken as the sum of the waits.

This is generally pretty accurate except when collecting samples when SETA is involved or slideshow framerate lag. There's a disparity between how long the wait calls sleep for and how long the game will allow before shooting a laser, under these conditions. This seems to be a game engine quirk and I can't do anything about it.

Anubitus wrote:i have noticed that a very fast fireing laser (lets say 200 rounds/min) suddenly fires extremly slow (lets say something like 30 rounds/min), why is this..? and if i check out the delay it has something like 2569.. if i set it to 1 it fires a never ending stream of bullets wich appears to be faster then the 200 rounds it can actually fire..?

The refire delays in the cache are in milliseconds. 200 rounds per min should be a 300 ms refire delay. 2569ms is definitely not a good delay for a 200 rounds per min laser. I've never seen the delay sampling produce generate such a wildly inaccurate number for a fast laser.

Under SETA or heavy lag it can produce samples that are way too low (e.g. 300ms for a 1200ms laser), but values don't tend to go too high above the real one - at least from my observations.

If you poke a value of 1 in the cache it'll fire the laser every frame. I'd not recommend setting delays below 50ms.

Try pushing the reset option at the bottom of the DPS cache menu to clear all the samples, then invoke the training option on your ship to have it train up again, and be sure not to engage SETA.

Anubitus wrote:sometimes they dont even fire at all for a while..

Ships perform a scan for enemies within a certain range and if there aren't any, Smart will sleep for a longer time. Spawning enemies right on top of the ship will show a few seconds delay before the turrets start shooting.

The only other case I know where the guns don't appear to shoot, is where beam lasers fire upon and destroy proximity detonation missiles. It appears that the detonation destroys the beam "bullet" before it gets displayed.

Hektos wrote:+ SMART seems to do very well with limited loadout (just enough guns to fit turrets, or with a few extra). It switches rapidly enough to keep full coverage and will (unintentional?) even keep a turret a few lasers short so that it can keep optimal coverage on other turrets.

Smart has a throttling mechanism where less laser slots are filled as laser energy decreases below 50%, which would make more lasers available in the pool. There's no other special code though. I'll have a look at a limited loadout pteranodon this evening and see if there's anything worrying about it.

Hektos wrote:+ SMART (once its got a good sampling) uses laser data very effectively (almost too effectively ). It abused the (XRM) Gauss Cannon so much that it was obliterating fighters with them, instead of using Energy Bolt Chainguns (FAA was still preferred at close range).

The XRM gauss cannon is a Smart's historical nemesis. It's super fast firing with long range and high DPS, making it the perfect choice for just about everything. Only in practice it can often miss a lot on fighters. It has better DPS and speed than the EBC, and better shield DPS and speed than the FAA, making it the weapon of choice.

In 1.8 I made changes to detect beams and make them less desirable for fighter targets. The gauss cannon gets categorized as a beam, which is why FAA are even getting a chance to fire. FAA is preferred over EBC as it has better DPS all round at more or less the same range with faster speed. You might see EBC used more on the Pteranodon if FAA is put in the excluded laser list. If its not being used there may be some error involving ammo based guns I need to look into. I'll look into it.

The way the fighter laser selection works is it finds the best DPS with comfortable speed, where this is target max speed plus a multiple of the speed further modified by range. If nothing has sufficient speed it'll pick the fastest it finds - which is generally going to be the gauss cannon sitting at the end of the list. A fast target at longer range can easily exceed the 1200m/s speed of the EBC. If you spawn something slow like a falcon you might see gauss cannon used less often.

Smart used to always pick the fastest bullet speed laser for fighters. The heuristic selection algorithm is fairly new and an attempt to use better DPS when at closer ranges. It might need further optimization though, like maybe introducing a falloff as clearly a 1200 m/s bullet is sufficiently fast to hit most fighters at reasonable ranges. I'll look into that...

Hektos wrote:+ SMART doesn't like large swarms of missiles. With a small volley incoming, the turrets quickly dispatch the missiles (a few might get through). However with huge volleys (such as a barrage from the M7M) the turrets seemed to get overwhelmed. At one point (during the SWARM barrage) the turrets seemed to actually "give up" and stop firing. You would think that with Energy Bolt Chainguns (and their ridiculous fire-rate) very few (if any) missiles should be getting through.

I'll look at antimissile further this evening. This might link with Garry's observation about antimissile being less effective in 1.8.

Hektos wrote:Closing notes: I really only got a good solid 30 minutes or so of actual testing so more needs to be done on my end. Also, the missile "problem" may be due, in part, to my inability to adjust the settings correctly. Also, I was playing around with the Valhalla and noticed (not really SMART's fault) that the Terran Capital ships lack any good missile defense (the poor Starburst Shockwave Cannon fires too slow for large volleys).

I know certainly in vanilla terrans lack a good antimissile laser, and I've perhaps set the bar too loosely on qualifying antimissile lasers in an attempt to accommodate terrans. With the XRM Valhalla Smart is selecting M/AM as an antimissile laser, which is far too slow and I think I'd better review what qualifies as an anitmissile laser to rule out high delay lasers.

I appreciate your time on this, this is useful feedback that gives much food for thought.

Thanks!

Requiemfang wrote:XRM Terrans have the Electro-Magnetic Repeater, that gun is more rapid fire then the Electro-Magentic Plasma gun which is the vanilla mainstay fighter weapon.

I'll take a look this evening at why this mightn't be getting utilized in antimissile for terrans.

Drewgamer · Post by **Drewgamer** » Thu, 8. Sep 11, 15:56

The Electromagnetic Repeater does seem like a good choice for Terran anti-missile. Not sure why I didn't think of that. I won't get any time tonight (I'm positive this time

) But I will have some free this weekend, I can put SMART through it's paces.

Drew

Shimrod · Post by **Shimrod** » Thu, 8. Sep 11, 21:47

I've made a series of antimissile optimizations, the ocelot is very much cleaning up on incoming missiles right now. I'll continue testing.

- The ship missile scan will prune dead missiles from the sector missile list
- The turret missile scan will prune dead missiles from the ship missile list
- There is no longer a 50ms sleep call on entry to antimissile
- At the end of antimissile rather than waiting for the longest laser delay in that round, find the highest delay gun from this or previous round that we haven't wait for, subtract time waited already in this round and wait for that time.
- Reduce qualifying laser refire delay for antimissile lasers from 1000 to 675 (disqualifying XRM M/AM launches while allowing EMP cannon depending on accuracy of sampled delays)

I stood toe to toe with a cobra and Smart mopped up its whole missile barrage at short range, while taking time out to kill the cobra. I'm not sure if the pruning of dead entries, removed entry wait or generally lower exit wait is having the effect but its certainly a dramatic change over what I was seeing yesterday. Mind you it does have PBEs and PALC available, which makes it one of the best antimissile boats.

Valhalla is making good use of electromagnetic repeater from its top turret.

Some nice intercept efforts from a pack of 10 hcp's, I'm quite happy with its performance now. I'll gear up for releasing 1.9.

Shimrod · Post by **Shimrod** » Thu, 8. Sep 11, 23:02

Okay 1.9 is up there with the improved antimissile and reset buttons in the menus. Don't worry about the upgrade, just replace existing files and it'll upgrade itself on game load.

Pteranodon observations:
1. Versus Xenon M there's a definite preference for gauss cannons.
2. Ejecting the FAA it will use EBC, noticeably less effective (actually less than Gauss).
3. Spawning xenon falcons the turrets aren't picking gauss, they gauge other lasers sufficiently fast.
4. Low laser conditions : working happily with less lasers. When all the good ones are in use it inevitably fires big slow guns at fighters, until something better available.

The falcon in XRM is 180 m/s. In vanilla that's pretty fast for a fighter, M 300 m/s. I fear that if I lower the bar too greatly, in vanilla turrets will shoot PPC at nearby M3. I'm not unhappy with the Pteranodon performance versus 30 M's and falcons, even with flak ejected.