Board index FlightGear Development

ALS vs. Rembrandt (forward vs. deferred rendering)

FlightGear is opensource, so you can be the developer. In the need for help on anything? We are here to help you.
Forum rules
Core development is discussed on the official FlightGear-Devel development mailing list.

Bugs can be reported in the bug tracker.

ALS vs. Rembrandt (forward vs. deferred rendering)

Postby punkepanda » Thu Oct 08, 2015 6:15 pm

Some post was split off from the topic New Canvas GUI

Thorsten wrote in Thu Oct 08, 2015 5:50 am:I think I have said a few times that I would allocate framerate differently from Rembrandt and that is why I keep sticking with a forward renderer - but that's my opinion, not a consensus to discontinue Rembrandt development.

Yeah I see your point. But it is not true when you compare the framerate with detail settings on max with ALS. It is even close to zero more framerate with high end GPU.

I have a Radeon R9 270. In Outerra I get framerates from 50-80 fps with default settings (which is really good looking with superb effects and landscape). In FlightGear looking straight forward into the horizon i usually never get more than 15-25 on the same hardware and OS. And Outerra use some kind of defered rendering system and has stunning optimised effects on default settings.

So with your way of thinking we end up with a simulator good for the computers that are old or low-end. But without the benefits of computer evolution. Why is this a good approach?
Last edited by Johan G on Fri Oct 09, 2015 3:05 am, edited 4 times in total.
Reason: Some post was split off from the topic New Canvas GUI
punkepanda
 
Posts: 237
Joined: Mon Nov 04, 2013 9:11 pm
Callsign: LostFlight
Version: 2.12
OS: Arch Linux

Re: New Canvas GUI

Postby hamzaalloush » Thu Oct 08, 2015 8:00 pm

punkepanda wrote in Thu Oct 08, 2015 6:15 pm:So with your way of thinking we end up with a simulator good for the computers that are old or low-end. But without the benefits of computer evolution. Why is this a good approach?



It's ONE approach, that happens to be compatable with all systems, and a developer that have time for it.

I'll test this zero FPS difference between both rendering solution using my 900 Series.
hamzaalloush
 
Posts: 632
Joined: Sat Oct 26, 2013 9:31 am
OS: Windows 10

Re: New Canvas GUI

Postby punkepanda » Thu Oct 08, 2015 9:17 pm

hamzaalloush wrote in Thu Oct 08, 2015 8:00 pm:
punkepanda wrote in Thu Oct 08, 2015 6:15 pm:So with your way of thinking we end up with a simulator good for the computers that are old or low-end. But without the benefits of computer evolution. Why is this a good approach?



It's ONE approach, that happens to be compatable with all systems, and a developer that have time for it.

I'll test this zero FPS difference between both rendering solution using my 900 Series.


Outerra is also OpenGL which means it can be ported to any OS and planned to support both Windows, Linux and Applestuff, but I guess since its under heavy development they stay for one system untill they are happy about there results.

I get sometimes better framrate on my laptop with NVidia 630M than on my stationary with the powerfull R9 card. Thats when you just have to realize that something is not OK for a modern simulator. Sorry to say but ALS is not the right way. How can realtime lightsources and shadows ever be seen from outside the airplane in ALS mode? Thats essential to make it even look close to realistic in ex. a replay mode or promoting a video on youtube. ALS is like inventing the wheel over again without the rubber.
Last edited by punkepanda on Thu Oct 08, 2015 9:24 pm, edited 1 time in total.
punkepanda
 
Posts: 237
Joined: Mon Nov 04, 2013 9:11 pm
Callsign: LostFlight
Version: 2.12
OS: Arch Linux

Re: New Canvas GUI

Postby hamzaalloush » Thu Oct 08, 2015 9:23 pm

punkepanda wrote in Thu Oct 08, 2015 9:17 pm:Outerra is also OpenGL which means it can be ported to any OS and planned to support both Windows, Linux and Applestuff, but I guess since its under heavy development they stay for one system untill they are happy about there results.


I guess OpenGL is the API for hardware calls to GPU, but dont qoute me on that :)
do they also use OSG, like us?

i too like deffered rendering, but maybe simulated environment is not the same as other AAA games.
hamzaalloush
 
Posts: 632
Joined: Sat Oct 26, 2013 9:31 am
OS: Windows 10

Re: New Canvas GUI

Postby punkepanda » Thu Oct 08, 2015 9:28 pm

hamzaalloush wrote in Thu Oct 08, 2015 9:23 pm:
punkepanda wrote in Thu Oct 08, 2015 9:17 pm:Outerra is also OpenGL which means it can be ported to any OS and planned to support both Windows, Linux and Applestuff, but I guess since its under heavy development they stay for one system untill they are happy about there results.


I guess OpenGL is the API for hardware calls to GPU, but dont qoute me on that :)
do they also use OSG, like us?

i too like deffered rendering, but maybe simulated environment is not the same as other AAA games.

Yes OpenGL is the stargate to the universe if you like :lol:

Please try Outerra and comeback and post your feedback. It is simulating the whole world in a more realistic way than FlightGear at this point. It also has planes, cars, helicopters, boats etc with the same FDM as FlightGear(JSBSim)
punkepanda
 
Posts: 237
Joined: Mon Nov 04, 2013 9:11 pm
Callsign: LostFlight
Version: 2.12
OS: Arch Linux

Re: New Canvas GUI

Postby hamzaalloush » Thu Oct 08, 2015 9:31 pm

i dont see it avaliable on steam, shame, i have some points left.

i thought Outtera had a small team unlike us here, hmmm maybe better to discuss this on another topic, also we dont want to rain on the parade of FG :mrgreen:

It's all we have in the long run.
hamzaalloush
 
Posts: 632
Joined: Sat Oct 26, 2013 9:31 am
OS: Windows 10

Re: New Canvas GUI

Postby punkepanda » Thu Oct 08, 2015 9:41 pm

hamzaalloush wrote in Thu Oct 08, 2015 9:31 pm:i dont see it avaliable on steam, shame, i have some points left.

i thought Outtera had a small team unlike us here, hmmm maybe better to discuss this on another topic, also we dont want to rain on the parade of FG :mrgreen:

It's all we have in the long run.

Is FlightGear on Steam?
http://www.outerra.com Download the demo from there and start test flying :)


If its a small team with those results that make my point even clearer. That we are doing something wrong. But the community is growing big over there. And for a reason!
punkepanda
 
Posts: 237
Joined: Mon Nov 04, 2013 9:11 pm
Callsign: LostFlight
Version: 2.12
OS: Arch Linux

Re: ALS vs. Rembrandt (forward vs. deferred rendering)

Postby wkitty42 » Fri Oct 09, 2015 4:49 am

too bad there's no linux version... i can't even think about trying to look at the tech demo because it is winwhatever only :/
"You get more air close to the ground," said Angalo. "I read that in a book. You get lots of air low down, and not much when you go up."
"Why not?" said Gurder.
"Dunno. It's frightened of heights, I guess."
User avatar
wkitty42
 
Posts: 5552
Joined: Fri Feb 20, 2015 3:46 pm
Location: central NC, USA
Callsign: wk42
Version: git next
OS: Kubuntu 14.04.5

Re: ALS vs. Rembrandt (forward vs. deferred rendering)

Postby Thorsten » Fri Oct 09, 2015 6:55 am

I think I have written a combined twenty pages on how I see the relative merits of ALS vs. Rembrandt over time - I suggest to spend some time reading it up as I won't go through the arguments and explanations every time.
Thorsten
 
Posts: 10819
Joined: Mon Nov 02, 2009 8:33 am

Re: ALS vs. Rembrandt (forward vs. deferred rendering)

Postby Johan G » Fri Oct 09, 2015 4:42 pm

Skimming through the first results from a quick Google search on deferred vs forward rendering makes it clear that both have advantages and disadvantages, and that there are some varying opinions about which one would be best for a certain use (which also makes it possible to cherry-pick which one work best... :roll: ).

As usual though, the only way to really have an influence is to roll up ones sleeves, acquire the knowledge and skills needed, start working, and discuss the ideas on the mailing list when underway (in essence not when done, people would want to review and get familiar with the code while you are working on it in order to feel comfortable with committing it). It may very well take years, but it actually is the way things usually get done (both in FlightGear and other open source projects).
Low-level flying — It's all fun and games till someone looses an engine. (Paraphrased from a YouTube video)
Improving the Dassault Mirage F1 (Wiki, Forum, GitLab. Work in slow progress)
Johan G
Moderator
 
Posts: 5480
Joined: Fri Aug 06, 2010 5:33 pm
Location: Sweden
Callsign: SE-JG
IRC name: Johan_G
Version: 3.0.0
OS: Windows 7, 32 bit

Re: ALS vs. Rembrandt (forward vs. deferred rendering)

Postby hamzaalloush » Fri Oct 09, 2015 11:42 pm

ok, i might turn up here are basicaly go for all out data, people who can review such code are people familiar with shader work and rendering engines, we dont have much of those, one of those used to be on this forums(we lost a very informative member), we also have by name if you dont mind, wlbragg(sp), and abbasign which is also working on effects.

anyway, expect to see benchmark shortly, backed up by video.
hamzaalloush
 
Posts: 632
Joined: Sat Oct 26, 2013 9:31 am
OS: Windows 10

Re: ALS vs. Rembrandt (forward vs. deferred rendering)

Postby Thorsten » Sat Oct 10, 2015 5:52 am

The benchmark is basically meaningless, because

a) what framework runs best depends strongly on your hardware - I never managed to get more than 15 fps out of Rembrandt with proper shadows even with my gaming machine, others with nominally much weaker hardware report 25+ fps, so there's more here than meets the eye

b) ALS at high quality and Rembrandt *roughly* give the same order of framerates because Rembrandt does a very simple light. fog and texturing scheme and dumps all performance to do the deferred technique whereas ALS at high quality executes roughly 10x more instructions per pixel to get detailed lighting, fogging and procedural texturing. So that's my point - you can use the performance to either do deferred stuff or do realistic atmosphere stuff, but you can't use it twice. And if we had extra, I'd still not spend it on deferred techniques but on other visually more striking things.
Thorsten
 
Posts: 10819
Joined: Mon Nov 02, 2009 8:33 am

Re: ALS vs. Rembrandt (forward vs. deferred rendering)

Postby Hooray » Sat Oct 10, 2015 7:08 am

I would not say that a benchmark comparing Rembrandt with non-Rembrandt is necessarily meaningless, as long as the majority of features in each mode is disabled, so that conclusions can be drawn about the impact of certain features - IIRC, that is basically how Richard came up with the hypothesis that Rembrand't performance issues may be in some way due to OSG-level issues involving camera/vsync-limits: http://wiki.flightgear.org/FlightGear_N ... _and_Vsync

Now, I don't know if you, or anybody else familiar with the effects/GLSL and OSG side of things, ever got around to looking at those findings, but it would definitely make sense to revisit this at some point, simply because Rembrandt may currently be the main component in FG using slaved cameras to this degree, but as you know, we do have other cameras/components making fairly extensive use of OSG cameras that should not necessarily be vsync limited, e.g. Canvas coming to mind - and the other thing being CompositeViewer support: http://wiki.flightgear.org/CompositeViewer_Support

Admittedly, I belong to the camp of those people (like Thorsten) with fairly powerful hardware, who are still not able to get very impressive frame rates - a few months ago, F-JJTH was politely suggesting that we simply would not understand how to configure our systems properly. From my standpoint, I never got sufficiently motivated to look at these issues beyond exchanging osg-stats with Thorsten.
But like Richard's tests have shown, it is indeed possible to come up with tests/benchmarks and come up with useful data, or even a working hypothesis.

So personally I would hope that we don't simply disregard such information or shut down such efforts too prematurely, especially given the very low number of active developers familiar with effects, shaders and the OSG/C++ side of all this (largely unmaintained code).

Looking back in time, some fairly significant bugs were discovered, and subsequently fixed during the last 1-2 release cycles, many of which could have been identified/fixed earlier, had we paid more attention to end user feedback, and not just disregarded things.

I am familiar with hamzaalloush's benchmarking efforts, and I really hope that we can grow a library of tests and extend fgfs to enable end-users to make better bug reports, even unrelated to ALS/Rembrandt in particular.
Please don't send support requests by PM, instead post your questions on the forum so that all users can contribute and benefit
Thanks & all the best,
Hooray
Help write next month's newsletter !
pui2canvas | MapStructure | Canvas Development | Programming resources
Hooray
 
Posts: 11321
Joined: Tue Mar 25, 2008 8:40 am

Re: ALS vs. Rembrandt (forward vs. deferred rendering)

Postby hamzaalloush » Sat Oct 10, 2015 8:46 am

Thorsten wrote in Sat Oct 10, 2015 5:52 am:The benchmark is basically meaningless


i disagree, look at the original argument, he says we cannot utilize powerful systems to make a scene pretty, because we(i use this word generously) optimized Flight Gear too much to the point of lowest common denominator graphics, well, I’m willing to prove this wrong, if that means doing the same benchmark over and over again using different systems: i have Intel, ATI, and Multiple Nvidia GPU's(all the way from Intel 3000 to GTX 980), you cannot regard this as non informative to people, "useless" is a strong word to use here.

we can come up with a table of features to enable to meet set performance target. not just merely drawing a line in the sand and flying randomly to produce these results/finding either.

Thorsten wrote in Sat Oct 10, 2015 5:52 am:a) what framework runs best depends strongly on your hardware - I never managed to get more than 15 fps out of Rembrandt with proper shadows even with my gaming machine, others with nominally much weaker hardware report 25+ fps, so there's more here than meets the eye


It also depends on data/content developers and how well their 3d elements work with the rendering engine that we have i.e to the OSG scene. (refer to my latest comments about HW vertex limits and osg drawables), my opinion is to use a scene optimized enough not to kick in hardware limits for extra drawables, but produce results compare the two renedering schemes with a minimal subset of features.

Thorsten wrote in Sat Oct 10, 2015 5:52 am:b) ALS at high quality and Rembrandt *roughly* give the same order of framerates because Rembrandt does a very simple light. fog and texturing scheme and dumps all performance to do the deferred technique whereas ALS at high quality executes roughly 10x more instructions per pixel to get detailed lighting, fogging and procedural texturing. So that's my point - you can use the performance to either do deferred stuff or do realistic atmosphere stuff, but you can't use it twice. And if we had extra, I'd still not spend it on deferred techniques but on other visually more striking things.


now this is useful info.
hamzaalloush
 
Posts: 632
Joined: Sat Oct 26, 2013 9:31 am
OS: Windows 10

Re: ALS vs. Rembrandt (forward vs. deferred rendering)

Postby hamzaalloush » Sat Oct 10, 2015 9:02 am

@Hooray, thanks for informative comment, yes we always need to reach out and have second opinions,

one person Emilian Huminiuc "i4dnf" does indeed look into this finding, and indeed understand how drawables work and the how OSG does meshes, i wish he was here to discuss this with us.

Richard finding is already helpful if we want to extend the deferred rending scheme :) but let's do this one step at a time.

initially, i will make a "pretty" scene video upload for punkepanda with ALS against Rembrandt on full features, just to compare with Outerra, looking into extending the advanced weather code to give me full cloud coverage and cloud shadows, instead of clouds popping on range(i have 6GB VRAM).

another technichal issue i'm sorting out now is to find an OpenGL screen recorded on Windows that captures an OpenGL application with uncapped performance, because currently Nvidia's Shadowplay does not capture OpenGL, but rather just throws out the full desktop display(that includes the Windows compositor affecting the results). It works good with DirectX application though!
Last edited by hamzaalloush on Sat Oct 10, 2015 9:07 am, edited 1 time in total.
hamzaalloush
 
Posts: 632
Joined: Sat Oct 26, 2013 9:31 am
OS: Windows 10

Next

Return to Development

Who is online

Users browsing this forum: No registered users and 1 guest