2021 (and earlier) DCS Newsletter Discussion Thread

[Worrazen]

7 hours ago, Silver_Dragon said:

DCS World has a international company, not only russian (The CEO, Nick Grey, and the producer, Matt Wagner, has english). The newsletters has always posted on english, and after, translate to others languajes include russian.

Understood, but the raw development notes usually originate from Russian I would assume, at least the patch notes almost always definitely do because there's always a lot of small translation errors there, usually word ordering, not that that's some big deal, usually it's totally fine and I'm used to it. In this case I have no idea what could Indication be, maybe I'll figure it out of I go an play a mission or two.

 

Edited October 17, 2021 by Worrazen

[BIGNEWY]

3 minutes ago, Worrazen said:

Understood, but the raw development notes usually originate from Russian I would assume, at least the patch notes almost always definitely do because there's always a ton of translation errors there, not that that's some big deal, usually it's totally fine and I'm used to it.

 

Newsletter is English, then translated into other languages, if you think there is an error feel free to PM me. But I don't see any issue in this week's newsletter. 

Thanks

[unknown]

20 hours ago, BIGNEWY said:

 

Newsletter is English, then translated into other languages, if you think there is an error feel free to PM me. But I don't see any issue in this week's newsletter. 

Thanks

They are asking what "indication" exactly means in the "multicore" list:

"

At the end of Q3 2021 we accomplished:

• Graphic backend
• EDM models
• Human models
• Atmosphere
• Water and sea
• Terrain engine
• Special effects, particles system
• Night lights for terrains
• Scenes
• Cockpits
• Mirrors
• Indication
• GUI
• Post-effects
• Cascade shadows

"

[Raven434th]

I don't understand what they mean by "accomplished"...have they already incorporated these changes into our installs or just behind the scenes beta testing??

[Worrazen]

On 10/17/2021 at 4:21 PM, draconus said:

I am also interested what Indication means and how there is no AI and FM on the list of multithreading.

 

14 hours ago, unknown said:

They are asking what "indication" exactly means in the "multicore" list:

"

At the end of Q3 2021 we accomplished:

• Graphic backend
• EDM models
• Human models
• Atmosphere
• Water and sea
• Terrain engine
• Special effects, particles system
• Night lights for terrains
• Scenes
• Cockpits
• Mirrors
• Indication
• GUI
• Post-effects
• Cascade shadows

"

I accidentially came across "IndicationTextures" and "IndicationResources" in many Cockpit folders of aircraft module files stored in "Mods" folder, they're exactly that, the GUI indications for HUD, RWR, MFD/MFCD/DDI ... let alone the Mission Editor and F10 map with hundreds if not thousands of icons and does make a lot of sense to me, because you can have all of those indications in the cockpit duplicated in case of a lot of action, and a lot of units on the map need a lot of indications on the HSI for example.

I remember long ago I made a thread about how Mission Editor grid is being rendered and I guess the multi-threading improvements and the new Vulkan API will alleviate the performance hit of some of these indications in ME.

Now, Mission Editor it self is not on the list but we can probably safely assume it'll run a lot lighter as well, no?

[Lurker]

 

10 hours ago, Raven434th said:

I don't understand what they mean by "accomplished"...have they already incorporated these changes into our installs or just behind the scenes beta testing??

 

Yeah the wording is very strange, from the newsletter it appears that many of the so called "multicore" features are already part of the released 2.7.X etc. 

[draconus]

No. The multithreading is not implemented in any of the parts mentioned. It is just in dev/test brand.

[twistking]

On 10/17/2021 at 4:21 PM, draconus said:

I am also interested [...] how there is no AI and FM on the list of multithreading.

I could understand FM not being listed, if it the new architecture still had a master sim thread, which would hold the FM and some other basic functionality. The items on the list would then be the threads that used to run on the main thread but got "outsourced" now.
However i cannot imagine why AI is not on the list. It's obviously a task that can cripple performance and making it a seperate thread should be a priority. I hope it's just an oversight in the newsletter.

2 hours ago, draconus said:

No. The multithreading is not implemented in any of the parts mentioned. It is just in dev/test brand.

Do you know for sure, or is this more of an educated guess? (I would have guessed the same based on the lack of obvious performance gains with recent patches, but wording in the lewsletter was strangely vague about this...)

Edited October 19, 2021 by twistking

[Raven (Elysian Angel)]

1 hour ago, twistking said:

However i cannot imagine why AI is not on the list.

Indeed: it would open the door for AFM/PFM/EFM to be used on aircraft you’re actively engaged with at close range (obviously there’s no need to have an advanced AI FM for distant enemies), making dogfights in SP a lot more interesting…

[Lurker]

All the listed items in the newsletter are parts of the rendering engine\pipeline. This is why AI and FM etc. haven't been mentioned. I wouldn't worry too much about it. The graphical engine is the most demanding part of the equation and is obviously the most difficult to code or rather convert into a proper multicore setup. 

Edited October 19, 2021 by Lurker

[Worrazen]

In my ears it sounds the other way around, accomplished for development branch, I think the changes would make quite a noticable difference without any patch notes.

[Worrazen]

9 hours ago, twistking said:

I could understand FM not being listed, if it the new architecture still had a master sim thread, which would hold the FM and some other basic functionality. The items on the list would then be the threads that used to run on the main thread but got "outsourced" now.
However i cannot imagine why AI is not on the list. It's obviously a task that can cripple performance and making it a seperate thread should be a priority. I hope it's just an oversight in the newsletter.

Do you know for sure, or is this more of an educated guess? (I would have guessed the same based on the lack of obvious performance gains with recent patches, but wording in the lewsletter was strangely vague about this...)

 

From listening and hanging out in various dev circles, there will always be a master thread that hold everything together, it just won't necessairly have the highest CPU usage all the time or at any time, and if it does have it, in case of calculations that simply can never be fully parallelized, it wouldn't be as high than the other threads compared to without multi-core enhancements.

However there is some special programming in Doom Eternal where it keeps workload balanced across threads in a dynamic way is what I'm guessing is happening, litterally shifting work to another thread and the most active ones end up taking very similar amount of cycles, almost like a neck to neck horse race, but that game is not really comparable to DCS in many other aspects.

 

On 10/15/2021 at 5:54 PM, Willdass said:

Very excited about the engine updates, curious when it will go live.

I would say at least 6 months more to go, some of those items still on the list are big ones.

 

On 10/16/2021 at 9:19 PM, schroedi said:

There are 21 points under the topic of multithread, sound not mentioned it would be 22.

Does this mean DCS is supporting up to 22 threads in the future?

Naoooo! It doesn't work like that, it's not a literal list of a number of threads, just a list of the component that are redesigned to be part of the multi-core enhancements upgrade, whatever that might be is up to the implementation only the developers best know. A proper dynamic system spawns as many threads as your system can optimally handle and group workload into them evenly. Technically each AI unit could be it's own thread, but that won't make sense on most computers, not only because each AI unit won't nearly take up enough processing power to max out a single CPU core of most modern processors, but because if you have 100 threads, your OS Thread Scheduler would just put those 90 into only a few CPU cores if you have a 6 or 8-core CPU, even tho it would feel like that there's so many threads doing the work depending on what kind of monitoring software you're looking at, in that case so many software threads can actually be detrimental, you're not going to lose much performance if you put 100 AIs onto one thread, as long as this thread doesn't completely max out a CPU core, however delays in AI processing aren't that big of a deal in long-.term, they shouldn't interrupt the core DCS engine and you wouldn't feel any slowdown at all, the AI would just take 0,xxx seconds longer to respond to some command or whatever, there's only a few areas where this would count, like missile avoidance, but that should run in a separate AI thread with very high priority, so there's many types of AI's, or components to an AI that can work independently, and thus can be parallelized to separate threads, and they should also be separated by priority indeed, the top-level coalition AI that controls strategic moves for the Dynamic Campaign would have the lowest CPU priority because delays even up to 10 seconds would not really be a big of a deal and wouldn't sway the outcome of the battles.

It could be something like this:

• DCS Main Thread
• DCS Graphic Thread 1
• DCS Graphics Thread 2
• DCS Sound Thread 1
• DCS GUI Thread 1
• DCS Asset Loading Thread 1
• DCS Asset Loading Thread 2
• DCS Asset Loading Thread 3
• DCS Asset Loading Thread 4
• DCS Asset Loading Thread 5
• DCS Asset Loading Thread 6
• DCS Asset Loading Thread 7
• DCS Asset Loading Thread 8
• DCS ... etc
• DCS Dynamic-Campaign AI Thread 1:
  • Coalition 1 Master Command AI
  • Coalition 1 Ground Commander AI
  • Coalition 1 Naval Commander AI
  • Coalition 1 Airborne Commander AI
  • Coalition 1 Base Operations and Logistics AI
• DCS Dynamic-Campaign AI Thread 2:
  • Coalition 2 Master Command AI
  • Coalition 2 Ground Commander AI
  • Coalition 2 Naval Commander AI
  • Coalition 2 Airborne Commander AI
  • Coalition 2 Base Operations and Logistics AI
• DCS Unit AI Thread 1:
  • Coalition 1 Flight 1 Commander AI (DCS Unit Group AI - Unit #1)
  • Coalition 1 Flight 2 Commander AI
  • Coalition 1 Flight ... Commander AI
  • Coalition 1 Flight 20 Commander AI
• DCS Unit AI Thread 2:
  • Coalition 1 Unit Group 1 Unit #2 AI
  • Coalition 1 Unit Group 1 Unit #3 AI
  • Coalition 1 Unit Group 1 Unit #.. AI
  • Coalition 1 Unit Group 1 Unit #99 AI
  • etc
• DCS Thread Misc ... etc

 

However, the implementation of the AI might be completely different, as mentioned it has something to do with neural-processing, it might not resemble a structure that we see in practice at all and it may be a lot more integrated in it's thinking but also a lot more parallelizable under the hood because AI stuff can run on GPUs which are the kinds of parallel processing, so I might be totally off, maybe DCS will surpass this by far, but it could be like this 10-15 years ago if one wanted to do it, you culd do it like this and it would be better than , while NOT comparing it to how it is in DCS right now, because AI thinking really is not a bottleneck and I've never done AI testing.

We also need to be careful what the heck do we mean with AI, because what I mean with AI generally is the pure AI unit/airplane/tank deciding what to do and to respond, pathing, etc, I usually don't count weapon/radar guidance/detection as an AI, even tho it probably does classify under the larger AI umbrella, there's a lot of physics involved and sometimes it's not that obvious which is AI, to a programmer it may go by the type of calculation rather than what we humans percieve it as "thinking" or not, but for sake of practical communication I think that a bomb falling through air even if it's laser guided, most of it isn't the AI doing it, the weapon AI is just correcting, most of that are still physics claculations, no? Then again I might be wrong, underestimating the cost of real-time corrections.

This is a bit of a tangent that I wasn't really that deep into lately so I think you should not conclude absolutely anything with what I said.

WHOA ... wait a minute, the benefit of this kind of AI command struture, as a practical implementation of the segments, not about AI multi-threading, would be if these battlefield commanders actually are physical units somewhere on the map and losing them you actually lose that specific AI ability and the thinking that specific AI field was responsible for, and the AI units relying on that would no longer receive these instruction (until it's replaced, if allowed by game rules). Huge gameplay depth opportunity.

So if you lose Naval Commander on the map, you would lose Coalition Naval Commander AI functionality, that's kinda interesting if DCS goes to that length to simulate what a headless army would do, I think that's so interesting for gameplay, the lower levels would finish all the pending tasks from above but keep their own thinking and keep doing what they do by default without, stuff like self-defense and area patroling (if that's what they were taked to do in perpetuity) isn't something they should need the higher levels to order about every time an enemy unit is sighted, it's a bit of a process to determine what these default behaviors should be (tho in RL even perhaps self-defense use of weapons might need to be authorized so it's not so simple), definitely the lowest level unit keeps their AI thinking going, feeding information back to higher levels, their buddies in the squadron might hear his warnings of an enemy in the bushes but the nonexistend higher level wouldn't be able to forward that info to other commanders/squadrons/units, with a loss of a big commander, that force may kinda lose the big picture and probably lose effectiveness, you would have basically lack of strategic orders to move there or here but it shouldn't all crumble down completely, over a long period it would have quite a big impact ... it's actually a big topic so I don't want to get too deep into it, but boy it was worth brainstorming with the AI stuff so I got to this.

Then there's even a more complicated aspect of the Top level bypassing the incapacitated middle command and giving order down to the lowest level units, that should work, so if Master Commander AI makes those big strategic moves to move forces to a completely new area, that should get through and units would listen .... but here's a catch, how would they execute that move may be crippled, they might not use the best path to get there possibly taking the riskiest approach because there's no middle command thinking about how do it in the best and safest but also effectivest ... this is some complex AI stuff there, you'd have to account for all possible combinations of KIA commanders, etc, I'd be surprised if DCS would think about this at all, but if Dynamic Campaign is such a big deal and the returns are worth it I'd say go ahead!

 

On 10/16/2021 at 9:59 PM, Raven (Elysian Angel) said:

Hopefully there will be enough potential for scalability built-in, so that over time when more and more people get CPUs with large core counts (12+), DCS will be able to take advantage of that as more and more effects will be added in over the years.

Indeed it most likely would, but the number of threads it might spawn may be dependent on your system so it may not be the same with all the players, people shouldn't be worried if they don't see 100 threads for so many components that can technically be parallelized, the system should probably spawn threads in mid-game if it needs to do so under heavy load. Purely more threads on a CPU that can't handle all of them simulateneously isn't benefitial at all, if you have 100 boxes, you'll still have to put them on 8 trucks with an 8-core CPU, so it's easier to just use 32 or 16 bigger boxes instead. But I don't really know what kind of load balancing DCS will have, if we have huge battles with thousands of units, there is no need to have a 1000 threads for each of the unit's AI's and other components, so much of that would be grouped together, because so many threads would actually stress the OS and internal CPU scheduling and introduce a performance penalty, right the opposite effect, AFAIK. Keeping those cores that you have well fed is the key thing rather than the pure amount of them. I'm generally speaking btw, not as if the quoted doesn't know this already.

Edited October 19, 2021 by Worrazen

[Worrazen]

Also, let's pay attention, the "MultiCore" list sounds to be more like a reference to the overall progress of the rewrite to the newer engine with render dependency graph technology, so it's not directly about threading at all. Parts or whole of some of those points may not be parallelizable so they'll run serially anyway, even if they would appear in separate threads, those would need to wait for their turn.

 

 

Edited October 20, 2021 by Worrazen

[Eldur]

On 10/15/2021 at 6:17 PM, sthompson said:

I get notifications for this thread, but not for the Official Updates forum or for the "Official News 2021" thread even though I'm "following" each and have the option selected to receive notifications. In general the notification system has seemed flaky ever since the new forum software was rolled out. Wish they would fix it. Sorry to bitch here but it seems at least as relevant as some of the other comments.

BTW, I also do not seem to receive the newsletter reliably. I'm subscribed and sometimes it shows up, but often does not.

The twist is that ED staff write the news posts in an internal area, probably for cross-reading and then move the posts to the threads we can see. The board software doesn't register them as new posts, so no notifications are sent and sometimes it even doesn't get that yellow "new post here" thing. The only solution I have in mind to that is creating a new post in the news threads and C&Ping the content from the internal ones (from the edit function WYSIWYG box) right into it so it actually counts as a completely new post.

I'm kinda excited for the multicore thing, which probably will bring more benefit that just going Vulkan. I mean it's a big "let's do all that stuff" package after all.

BTW as it has been mentioned as well in one of the last newsletters, I wonder what the VR improvements actually are going to be. There's a lot more than just increasing performance by any means (Single Pass Stereo, adding a fixed and eye-tracked Foveated Rendering feature, adding DLSS support and such) as there's also a good amount of controller issues (WMR especially with positioning of viewports and hands including the wrist angles) and general convenience/usabilty things that could be addressed like VTOL VR-like grab the thing and flick it with actual moment control, hands casting a shadow, the still bugged and mostly invisible cursor, option to disable the laser pointer style of control for those who don't like it and have the space, general possibility to remap the controllers, virtual hand fading out after a settable time, mechanisms to keep the controllers alive and prevent them from going to deep sleep, direct fingertip click interaction model with intelligent boundary detection to prevent accidental triggering of nearby switches... just to name a few things. Would be great to know what's actually planned or even just considered from all those suggestions in the VR section.

[Lurker]

36 minutes ago, Eldur said:

VTOL VR-like grab the thing and flick it with actual moment control,

I would kill for this to be implemented. Seriously, this would be the best thing ever. I mean fumbling aroind for the mouse? Instead of just naturally using one of the controllers (which I can conveniently park next to my throttle) to hit switches? Man that would be the best thing ever. That laser pointer thing needs to die.

[Pizzicato]

3 hours ago, Lurker said:

I mean fumbling around for the mouse? Instead of just naturally using one of the controllers (which I can conveniently park next to my throttle) to hit switches?. 

If you can conveniently park your controller next to your throttle, surely you could just as easily conveniently park the mouse next to the throttle to reduce said fumbling?

(I only mention it because that's exactly the approach I take.)

[Rmnsvn]

Ok now. What happened to F16 g onset tuning for Cat 1? Doesn't appear in the changelog. Has it been moved back?

[Lurker]

16 hours ago, Pizzicato said:

If you can conveniently park your controller next to your throttle, surely you could just as easily conveniently park the mouse next to the throttle to reduce said fumbling?

(I only mention it because that's exactly the approach I take.)

Unfortunately not an option, I suck at leftie mousing. Seriously though, I'm looking for a trackball which I could sort of ducktape to my throttle, unfortunately I've yet to find a small and suitable one. 

[ShaunX]

45 minutes ago, Lurker said:

Unfortunately not an option, I suck at leftie mousing. Seriously though, I'm looking for a trackball which I could sort of ducktape to my throttle, unfortunately I've yet to find a small and suitable one. 

I can recommend this solution for VR if you're a rightie. The trackball is hotglued to the joystick. It looks fragile but hotglue is amazing at creating an instant rock solid bond over a small surface area but can also be removed again with a sharp twist and doesn't leave any residue (ask any woodworker who makes temporary jigs). The trackball is excellent and is an "ELECOM Relacon Wireless trackball mouse" which has 9 assignable buttons and track speed setting. Not widely available but I can see there are a few on ebay at the moment. It's a short, reliable movement from my joystick grip to trackball, thumb on the ball and scrollwheel, and there are 2 triggers on the front of it (not in the photo) for left and right buttons, so my grip and fingering is essentially the same as on the joystick. Equally, you could hot glue to your throttle I guess (forget duct tape !)- of course you'll need a hot glue gun but they're pretty cheap.

[Lurker]

Thanks ShaunX, I will see if I can order one. 

[SMH]

47 minutes ago, ShaunX said:

I can recommend this solution for VR if you're a rightie. The trackball is hotglued to the joystick. It looks fragile but hotglue is amazing at creating an instant rock solid bond over a small surface area but can also be removed again with a sharp twist and doesn't leave any residue (ask any woodworker who makes temporary jigs). The trackball is excellent and is an "ELECOM Relacon Wireless trackball mouse" which has 9 assignable buttons and track speed setting. Not widely available but I can see there are a few on ebay at the moment. It's a short, reliable movement from my joystick grip to trackball, thumb on the ball and scrollwheel, and there are 2 triggers on the front of it (not in the photo) for left and right buttons, so my grip and fingering is essentially the same as on the joystick. Equally, you could hot glue to your throttle I guess (forget duct tape !)- of course you'll need a hot glue gun but they're pretty cheap.

THAT is why I'm completely uninterested in VR! LOL!

[ShaunX]

25 minutes ago, SMH said:

THAT is why I'm completely uninterested in VR! LOL!

Why ? Because I had to get creative and do a spot of DIY ?

[Eldur]

This might possibly be both the ugliest and the most practical solution for many DCS pilots at the same time

[Lace]

1 minute ago, Eldur said:

This might possibly be both the ugliest and the most practical solution for many DCS pilots at the same time

Practical? You mean completely unnecessary.  I have been flying exclusively in VR for a couple of years now and never needed to touch a mouse, trackball or VR controller.  Map a stick hat to LMB/RMB/Mouse scroll up/down and simply look at the switches to operate.  If you don't have enough buttons a modifier on your throttle lets you double up.  I can't understand why so many people don't do this.  Don't fight the HOTAS, use it!

[ShaunX]

8 minutes ago, Eldur said:

This might possibly be both the ugliest and the most practical solution for many DCS pilots at the same time

Thank you I'll take that.  Can't see the thing in VR anyway 

3 minutes ago, Lace said:

Practical? You mean completely unnecessary.  I have been flying exclusively in VR for a couple of years now and never needed to touch a mouse, trackball or VR controller.  Map a stick hat to LMB/RMB/Mouse scroll up/down and simply look at the switches to operate.  If you don't have enough buttons a modifier on your throttle lets you double up.  I can't understand why so many people don't do this.  Don't fight the HOTAS, use it!

Each to their own buddy. It works perfectly for me thank you.

I'm sure this is not the thread to discuss this anyway. I only posted to assist someone looking for a solution in the spirit of this forum. Wasn't expecting a critique.