Replacement For Phil Taylor?

[DadJokeCinema 145]

Phil Taylor's role was lead program manager of the core engine (which powers ESP, FSX and TS2). I expect someone will have to follow his role as surely there needs to be a lead program manager on this. Brett Schnepf just changed roles, so maybe there is some internal restructuring going on. With the X-mas break coming I expect people take their time to find the right person.FSX isn't using software rendering. Phil Taylor in fact wrote a nice blog post explaining the rendering constraints. Since MS is working on a core engine that's built on FSX I would expect evolutionary changes, not revolutionary ones, just like in the past. When TS2 is released next year, we'll get a first insight into FS11 (they'll use the same engine). In fact Phil has been dropping a few hints (like animated people).Christian

To clarify this point from the best of my understanding, FSX uses a software technique for terrain triangulation. At some point, the engine has to take the flat land, made up of a big mesh, and use the terrain data push that mesh into the shape of the terrain. Most games today use vertex shaders on the 3D card to do this. It's referred to sometimes as "hardware transform and lighting". But to maintain the highest level of compatibility, this is done on the CPU with software algorithms. The concept that is being discussed in this thread is that they might take a lot of these algorithms and move them to the GPU where they can be done faster with shaders.

[michal 44]

To clarify this point from the best of my understanding, FSX uses a software technique for terrain triangulation.

That is not correct. The FSX sends to the GPU the physical raw data for the terrain mesh (x,y,z) and then GPU does the triangulation and builds the visual mesh. It actually has been stated many times here that FSX builds the full 3D terrain on the GPU side. Surprisingly you have sometimes on these forums those who believe that FS should "do more" on the GPU side state that the "only" things that FS does on the GPU side is 3D terrain. :(

[MarkRey 16]

So tell me then...why didn't you just buy an old acoustic 6-string guitar with no amp? It would play music, too. Let me guess...you are a "hard core guitar player". Just like someone who can't or won't upgrade their current computer hardware can still fly using FS9.Case closed. Bigger grin.FalconAF

Not to belabor, but I was actually making an honest observation I realized as I wrote - if I balk at spending that kind of money on fs AND music, that probably means I quite literally can't afford both.and I errr... already had an acoustic six string (and now I'm wanting a whole gaming machine's worth of Warwick Thumb bass :( ) so I guess my priorities are clear :( crawling back to the legacy forum now :(

[n4gix 4,947]

That is not correct. The FSX sends to the GPU the physical raw data for the terrain mesh (x,y,z) and then GPU does the triangulation and builds the visual mesh. It actually has been stated many times here that FSX builds the full 3D terrain on the GPU side. Surprisingly you have sometimes on these forums those who believe that FS should "do more" on the GPU side state that the "only" things that FS does on the GPU side is 3D terrain. :(

According to Adam Szofran (the guy who wrote the code!), you are the one who's incorrect:"The latest cutting-edge terrain mesh triangulation and LOD computation algorithms are being designed to run in vertex shaders on 3D hardware [9]. However, since Flight Simulator must run on both older and newer hardware, we run our mesh triangulation algorithms in software using techniques pioneered by Lindstrom [5], Pajarola [7], Duchaineau [11], and others, collectively referred to here as the restricted quad tree triangulation, or RQT. The RQT generates a seamless mesh whose complexity depends upon the roughness of the terrain and is scalable by a simple screen-space error metric. By allowing the end user to set the error metric, Flight Simulator can easily adapt the terrain mesh triangle count to different user preferences and machine configurations.""To avoid monopolizing CPU resources, the terrain engine does not refine the entire mesh every frame. Instead, the mesh is refined and triangulated incrementally and the resulting geometry is added to the scene graph when it becomes available using a double buffering scheme."Further along in Adam's paper he states:"Due to the complexity of the synthesis algorithm, we execute it in software rather than using video hardware. This of course requires us to use a fast, high-quality 2D software rasterizer. We chose Anti-Grain Geometry (AGG), by Maxim Shemanarev [14] because it is fast, free, and comes with full source code making it fully customizable to our needs."No fewer than three times Adam has stated that the rendering is done by the CPU in a memory buffer. Actually there are several buffers in use. While one is being sent the the graphics card, the other is being prepared.What ACES managed to accomplish in FSX SP1 & SP2 was to move some of the final rendering passes to the GPU, where limited use of Shaders allows the unique, "signature features" of FSX to be applied to the final image, but the vast majority of rendering in FSX is still being done by the legacy routines. I strongly recommend reading Adam's paper in it's entirety if you truly wish to understand the process actually used:http://download.microsoft.com/download/5/6..._Terrain_v1.doc

[Guest christian]

Ok, fair bit of confusion here in rgds to terrain rendering, I just add a few comments but don't have the time to discuss this heavily.The CLOD algorithm doesn't have anything to do with transform and lighting per se. Yes, it's done in software, but the only thing it does is recalculate a surface level of detail algorithm on the CPU. It doesn't do any transformations nor lighting, just recalculates the vertices. This stuff is really old and CLOD algorithms were a problem a few years ago. I would think that you would save much CPU by transferring CLOD onto the GPU as it's running pretty fast on modern computers. CLOD is really a very minor part in rendering, all the surface objects and aircraft fully use the GPU, otherwise you wouldn't get bumpmapping and other shader effects. As Phil explained, one of the big problems is the heavy load of the autogen. There is a document from MS somewhere which explains how to limit DX drawcalls which put a heavy burden onto systems.I won't argue that the FSX engine is behind current technology, FarCry 2 seems to achieve a whole lot more (although I heard that the paging of new terrain is fairly slow which isn't a workable solution for flight sims. Brian - love your blog ;)

[maxter 0]

Now i'm really confused :( I guess if the new sim is going to appear in 18 months or so (rumour???) many/most of the tech directions would already have been made already. I do wonder however what the core direction of the new version is going to be. To clarify, I think the current version development is/was heavily skewed in favour of the GA pilot and not so much for the tubliner jockey. Maybe it was just the drain on system resources that make this apparent and the new systems midrange being released more recently will alter that balance.Interesting debate though, I just wish some of the MS guys would chip in some time.Cheers,

[Adam Szofran 0]

Ok, I'd like to clarify a few things. > we run our mesh triangulation algorithms in software True. At least for FSX anyway. We are currently experimenting with terrain mesh rendering techniques that rely much more heavily on vertex shaders. Hopefully that will allow us to implement neat things like LOD morphing to prevent "popping" and real-time mesh deformation for explosions or water waves. I don't know if any of this will ever see the light of day, but at least we're experimenting with it. > Due to the complexity of the [texture] synthesis algorithm, we execute it in software rather than using video hardware. > No fewer than three times Adam has stated that the rendering is done by the CPU in a memory buffer. This applies to the synthesis of terrain textures only. Once the texture for a particular patch of terrain is done being synthesized, it is attached to the corresponding patch of terrain mesh and rendered by the GPU. > What ACES managed to accomplish in FSX SP1 & SP2 was to move some of the final rendering passes to the GPU... Sort of. The main thing we did for terrain in SP1 was to parallelize the terrain texture synthesis routines by allowing them to run on multiple CPU cores simultaneously. Thanks for reading my paper, by the way. -Adam

[n4gix 4,947]

Thanks for reading my paper, by the way. -Adam

Thank you for writing it! I read it the day the link was first published, and have referred to it many times since... :( Thanks also for confirming that I wasn't completely misunderstanding the concept(s). :(

[michal 44]

We are currently experimenting with terrain mesh rendering techniques that rely much more heavily on vertex shaders. Hopefully that will allow us to implement neat things like LOD morphing to prevent "popping" and real-time mesh deformation for explosions or water waves. I don't know if any of this will ever see the light of day, but at least we're experimenting with it.

Oh, thanks Adam for chiming in.Anyway the way I read you post (correct me if I am wrong) this heavier reliance on vortex shaders, if it ever "sees the light of day", is not something that will have miraculous effect on frame rates OR that you are held back by compatibility considerations.

[Odaat 0]

and the fact that FS11 probably won't be seen for at least another 1 and 1/2 years, anybody who honestly expects to be able to "max everything out" on ANY computer available when FS11 is released is dreaming. It will be the same as always...if you want to be able to max everything out with FS11, you probably won't be able to do it for at LEAST a year after it is released...if not longer.FalconAF

and the craziness continues... Why oh why can't Microsoft release something that actually runs well on current machines at the time?! This business of 'buy a computer 2 years after FS release that will finally run the software only to have a new release the next year, repeat, repeat' is just a silly business model to me. Oh, I know they will not change it and people who have plenty of money to spend think it's great. But us working folk with families lose out. I've used FSX since it was released and even with core 2 duo, I cannot get it to run anywhere near where I would like it to be. I will continue to use FSX as computers improve, but I seriously doubt I will update to FS11 when it is released.Paul

[mgh 89]

and the craziness continues... Why oh why can't Microsoft release something that actually runs well on current machines at the time?!

If anyone really thinks that FS with all its latest features could run on current machines with all sliders maxed out then I suggest they not in the real world. Microsoft's approach enable users to tune FS to meet their own individual requirements. Microsoft offers users a choice by providing sliders. Users have different priorities. Some put priority on terrain, others on AI, others on AutoGen etc etc. This way most users can set their own priority. The alternative would be for Mocrosoft to remove the sliders and determine once-and-for-all the users' priorities. This would almost certainly result in a rather mediocre middle of the road performance that would satisfy very few.

[michal 44]

and even with core 2 duo, I cannot get it to run anywhere near where I would like it to be.

And I run FSX "well" on my 3+ year old PC but I define "runs well" a bit different than you. There is really no objective "runs well" to speak of.I am glad that FSX comes with whole array of sliders/options and I can tailor this simulation to where I want it to be. It would be a rather poor software if it 1. refused to run outright because it determined your computer was insufficient OR 2. run "well" with many options enabled but offered very little improvement over prior version.

[meshman 109]

...and even with core 2 duo, I cannot get it to run anywhere near where I would like it to be.

I run a C2D in my two year old system and FSX won't run anywhere near where other people think it should be. But it works just fine for me and my expectations. I am not one to buy into the "my sliders are further right than your's are" syndrome. I set a level that my system can handle and accept what it gives me, then I go fly. My DENSE scenery and NORMAL autogen settings gives me sufficient eye-candy. I've yet to stop the plane and count the number of trees that are displaying on a texture tile.I did three upgrades to my prior system to gain better performance with FS2004. My choice, as nothing was forced upon me. I'll probably get a newer video card by Christmas, as the old ATi is lacking. But it's not a choice forced upon me by any desire for bragging rights or a feeling that somehow the sim is now deficient in some manner.Perspective and how it's implemented can go a long way towards getting enjoyment while being a simulated pilot, sitting at a desk and grasping a joystick.

[Gate99 246]

Hi mgh, I agree. What happened to Phil was disgusting! and frankly, the forum moderators bare some of the blame here. Here we had a high level FSX developer joining our community, sharing his valuable insight, answering questions, and offering valuable advice. Constructive criticism and dicussions is one thing, and we should all feel free to offer our opinion respectfully, but some members here decided to treated him as their personal punching bag. All the while the system moderators sat quitely on the sidelines allowing things to escalate before finally stepping in and locking threads. Prehaps if the moderators took a more proactive approach by warning/banning certain users earlier Phil would still be with us. To those members who took this approach I hope your happy with yourself. I hope Phil finds a more civil bunch of people in his future endevours. What happened here was shameful.Mark

It's my impression that Phil Taylor wasn't officially appointed, but that he chose to take on the role. Given his concerns about the way he was treated by some people on thse forums I don't expect there will be many volunteers. Re-read his own comments about the way he felt!

[Geofa 108]

Hi mgh, I agree. What happened to Phil was disgusting! and frankly, the forum moderators bare some of the blame here. Here we had a high level FSX developer joining our community, sharing his valuable insight, answering questions, and offering valuable advice. Constructive criticism and dicussions is one thing, and we should all feel free to offer our opinion respectfully, but some members here decided to treated him as their personal punching bag. All the while the system moderators sat quitely on the sidelines allowing things to escalate before finally stepping in and locking threads. Prehaps if the moderators took a more proactive approach by warning/banning certain users earlier Phil would still be with us. To those members who took this approach I hope your happy with yourself. I hope Phil finds a more civil bunch of people in his future endevours. What happened here was shameful.Mark

" All the while the system moderators sat quitely on the sidelines allowing things to escalate before finally stepping in and locking threads."Prehaps if the moderators took a more proactive approach by warning/banning certain users earlier Phil would still be with us."Really?