Comments on: ATC Part I – Real Physics https:/2011/01/atc-part-i-real-physics/ Developer resources for the X-Plane flight simulator Tue, 01 Feb 2011 19:02:43 +0000 hourly 1 https://wordpress.org/?v=6.6.1 By: Benjamin Supnik https:/2011/01/atc-part-i-real-physics/#comment-20 Tue, 25 Jan 2011 09:06:41 +0000 http://www.x-plane.com/dev_blog/?p=8#comment-20 "Because on a quick test on 9.65, frame rate went from just below 60 without AI to about 25 with 19 AI planes."

Quick follow-ups:
– In v9, the AI planes don't ever taxi, so you can't measure the cost and/or optimization of ground contact in v9. (In fact, ground contact costs are very low with lots of AI planes because they don't taxi.)
– In v9, foil-airstream calculations dominate the FM when a plane is in the air – 17% of CPU with 20 planes on my machine. (In this configuration, total FM takes 50% of CPU, so you can see how multicore is a big win here.) I believe Austin has already restructured the code to be faster in v10.
– Scenery checks are always there – 5.5% of CPU time spent checking for ground height in the autopilot AI as it tries not to crash into things. (This relatively high cost for a simple collision check is partly because the spatial coherency is bad, which causes collision cache thrash.)

]]>
By: Benjamin Supnik https:/2011/01/atc-part-i-real-physics/#comment-21 Tue, 25 Jan 2011 07:35:21 +0000 http://www.x-plane.com/dev_blog/?p=8#comment-21 Judith, I believe X-Plane 9 does reduce the number of ground checks when a plane is not in ground effect, but I a not 100% sure.

MatthewS, we are not going to make a version of simconnect – we'd be asking to be sued. A third party could make a simconnect -> XPLM bridge just like they have made an FSUIPC bridge.

MatthewS & Judith, a forum may be a better place to discuss simconnect.

]]>
By: HDOnlive https:/2011/01/atc-part-i-real-physics/#comment-22 Tue, 25 Jan 2011 06:20:24 +0000 http://www.x-plane.com/dev_blog/?p=8#comment-22 MatthewS: FSX 3rd party ATC modules won't be anywhere near the XP-10's ATC. I don't want to start an FSX vs XP10 discussion here but I guess you are one of those whoe just want to see the eyecandy and not the realism. AI-Aircrafts will be handled by the real flight model and you can ACTUALLY see them suffer when there is some turbulence or crosswinds or gusts. I think BEN can explains this in more detail.

]]>
By: Arista https:/2011/01/atc-part-i-real-physics/#comment-23 Tue, 25 Jan 2011 06:17:37 +0000 http://www.x-plane.com/dev_blog/?p=8#comment-23 Ben, does the version 9 AI turn off ground testing when "clearly in the air" as well (which they always are in version 9, I guess), or is this one of the possible future optimizations? Because on a quick test on 9.65, frame rate went from just below 60 without AI to about 25 with 19 AI planes.

Of course, if the flight model will indeed become completely independent of the main thread, I couldn't care less, but in version 9, the AI certainly is a big performance hit (on my first-generation Mac Pro anyway).

Off topic (please feel free to delete the following if you feel it's inappropriate here):

MatthewS, how many FSX add-ons are actually out there that only use SimConnect, without also using FSUIPC or some other IPC mechanism, like the panel SDK? I'm only aware of a very few. And there's already a reimplementation of the FSUIPC interface for X-Plane, in the form of XPUIPC (which is far from complete, but quite usable with a lot of FS9/X add-ons).

Judith

]]>
By: Anonymous https:/2011/01/atc-part-i-real-physics/#comment-24 Mon, 24 Jan 2011 21:21:35 +0000 http://www.x-plane.com/dev_blog/?p=8#comment-24 Looking forward to the blog post.

Im sure the XP10 ATC is much better than the FSX ATC, but FSX has 3rd party ATC modules (eg Radar Contact and Proflight) that completely replace the default FSX ATC. I very much doubt the default XP10 ATC comes close to these 3rd party products.

It would be nice if whatever API primitives the default XP10 ATC uses to direct aircraft can be exposed via the plug-in SDK so that 3rd parties can develop their own replacement ATC plugins and turn off the default XP10 ATC.

But also I guess if a XP10 version of SimConnect.DLL could be written then a lot of FSX utilities might work "out of the box" with XP10. Imagine if things such as Radar Contact (http://www.jdtllc.com/) could work with XP10 without changes, just a new XP10 specific version of SimConnect.dll

Regards, MatthewS

]]>
By: Benjamin Supnik https:/2011/01/atc-part-i-real-physics/#comment-25 Mon, 24 Jan 2011 20:58:48 +0000 http://www.x-plane.com/dev_blog/?p=8#comment-25 MatthewS:

I don't know how many we'll support. It might be 20, it might be more. They might be virtualized, they might not be. I will describe how the feature set is planned in more detail tomorrow.

If you set up a check list of FS X vs. X-plane 10 ATC, I can almost guarantee that you will be disappointed; what we will ship in 10.0 will basically be brand new; it will (like the scenery system) take a few patches to mature and develop some of the deeper features.

Using a real flight model is _not_ an "excuse" for 20 planes – it's not an either or trade-off. We didn't go "do we want 20 planes with real physics or 100 planes with some fake physics."

Rather, like all development, the system is built incrementally.

]]>
By: Anonymous https:/2011/01/atc-part-i-real-physics/#comment-26 Mon, 24 Jan 2011 20:45:12 +0000 http://www.x-plane.com/dev_blog/?p=8#comment-26 How many AI aircraft will XP10 support?

If its only 20 then its going to look very poor in comparison to FSX that can handle over 100 on the ground and in the air at one time.

Sure the FSX AI physics is poor compared to what your suggesting for XP10 but who cares when you can land at EGLL with 50 planes sitting at gates and other 30 or so in the air, including some on final and some having just departed.

I think far more people would be impressed with busy airports and realistic traffic levels than with how 'super accurate' the AI flight model is.

I feel that you're using the accurate AI flight model as an excuse for not having realistic AI traffic levels.

XP10 will look a very lonely place if big airports have just a handful (20) of AI aircraft.

Regards, MatthewS

]]>
By: Anonymous https:/2011/01/atc-part-i-real-physics/#comment-27 Mon, 24 Jan 2011 14:45:55 +0000 http://www.x-plane.com/dev_blog/?p=8#comment-27 Refreshing to hear all of this! One important aspect here is to remember computers evolve very quickly, where a new generation of CPUs/GPUs by far exeeds the performance of their earlier counterparts. Quadcore CPUs are already standard or close to standard in new, medium class, computers. With the latest Sandy Bridge processor there really -is- a bucket load of performance to play with. And one really should. The same goes with GPUs. Even an old nVidia 260 gfx card produces very good framerates in XP.

I find that a part of the XP community are feeling afraid that their 3 year old computer wont run XP. If it was a medium end computer: It sure will! But technology advances and with refined and faster technology we also expect higher quality content to reflect the hardware found in our computers.

There really is not much to add as I am feeling confident the XP team already are implementing sliders with broad intervals in options for graphics, effectively making a huge spectrum of options available to make the simulator look just right for everyone's computer.

Just having bought a new Sandy Bridge computer for XP10 running at 4.5GHz with a GTX580 card and 8GB of RAM all I can say is that on my part I am ready to try out next version of XP. Kudos to ben for keeping us updated with all these details!

]]>
By: Elizaveta https:/2011/01/atc-part-i-real-physics/#comment-28 Mon, 24 Jan 2011 14:33:32 +0000 http://www.x-plane.com/dev_blog/?p=8#comment-28 Hi ben. I completely agree with you. I would love to see some improvements in the FM regarding the Wheel-ground collision system. I think at present it needs some refinement so that taxi, takeoff and touchdowns look more real. The wheel-ground collision should be improved so that tires ACTUALLY sit on the ground properly!. Secondly real FM should handle the AI-Aircrafts. I would love to see how those AI-planes handle some serious crosswind.

]]>
By: Anonymous https:/2011/01/atc-part-i-real-physics/#comment-29 Mon, 24 Jan 2011 14:17:04 +0000 http://www.x-plane.com/dev_blog/?p=8#comment-29 I am definitely looking forward to the improved AI. Any chance that we will be able to run the ATC as a stand-alone server on a second computer?

-Carrotroot

]]>