OMFG HAVOK WILL BE FREE IN MAY!

[el_matarife]

Yeah but binaries could be used by commercial devs, GPL is too scary for most of them.

Yeah, this will be great for commercial or indie developers. Unfortunately, if it isn't GPL or BSD or some other GPL compatible license we can't use it in Spring.

[SwiftSpear]

I don't really think havok would even be useful for spring... this is much bigger news to independent shooters, and platformers. I'm not sure havok could be even made to work in any sensical way with our animation system. Traditionally it's been applied to quite advanced skeletal animation frames.

Also, havok is popular, and havok was an innovator, but honestly, it's not all that impressive a physics engine.

[smoth]

it is pretty good for a FREE one.

I posted this here because I presume other people who develop projects that are not spring may be here.

[trats20050]

Currently whats the physics engine Spring is using ?

[aegis]

spring is an engine

[Argh]

...and Spring's physics is mainly OK, other than the FALL event, certain checks against angles (Units can do some very odd things on map slopes), and the lack of another animation / model format other than BOS / S3O.

[hoijui]

i guess the question was, if spring uses a dedicated physics engine (which eg, there could be hardware acceleration available with dedicated physics cards).
spring has does all physics calculations internal -> no dedicated/specification based physics engine like the ones supported by cards.
i dont really know how these cards work though.

[el_matarife]

Dedicated physics cards are dead and everyone will be using OpenCL running against your videocard for physics real soon now. Spring won't, because OpenCL requires DirectX 10 level cards like the 8xxx series and the 3xxx series from ATI and most of our players are on stuff way older than that. (DirectX 10 isn't a requirement for OpenCL exactly, but I don't think ATI and nVidia are going to provide support for their older cards and unified shaders may be necessary.) http://www.nvidia.com/object/cuda_learn_products.html I can't find a list of ATI cards but I assume the list is similarly limited. I assume we may require OpenCL for Spring somewhere around 2015 at the current rate we've been requiring new hardware. (We just required SSE which has been on every mainstream processor for ~8 years and the Via chips for ~6 years.)

[imbaczek]

(We just required SSE which has been on every mainstream processor for ~8 years and the Via chips for ~6 years.)

...and we got complaints about it.

[zwzsg]

yes but would it be worthwhile? I really think so. if i can figure out how to get input from the bos files to a C program so it can do some math on it, i'll redo all the units with true walk scripts and a coordinate smoother.

Most walkscripts are horrible broken copypasta of broken Cavedog scripts. Trying to automatically convert them to beautiful won't work. Better off rewriting them from scratch.

Wai, may... 2008? Looks like trats20050 cast a necromant spell.

[el_matarife]

Wai, may... 2008? Looks like trats20050 cast a necromant spell.

Yeah, I knew the thread was really old when I replied but I had to make sure people knew that the future of physics lies in OpenCL or DirectX compute shaders. I suggest locking it now.

...and we got complaints about it.

People are always going to complain, but if open source projects don't keep up with (reasonably) modern hardware what's the point? Firefox and a few other projects dropped support for Windows 9X / ME over loud complaints a few years ago, FFMpeg was requiring SSE2 a few years ago, and any one who suggested maintaining application support for Linux kernel 2.0 or 2.2 would be laughed out of most serious mailing lists.

At some point in time we have to say "Sorry, we can either have these really cool features here or we can keep Spring working on your really old PC. Please go buy a cheap desktop or laptop with a good videocard so we can keep innovating and adding new features."

Desktops and laptops with nVidia's ION platform should make this a lot easier. The Acer Aspire Revo desktop is $300 with a dual core processor and a DirectX 10 compliant videocard that will do OpenCL. Imagine what will happen over the next two or three years, especially if Intel cleans up their act on integrated video?

I think the 5 year mark is a good rule of thumb. Once a feature has been available on every processor or videocard available for 5 years, it is safe to require it. I would however ignore Intel's integrated video platform since it has been slow in deploying new features and the performance is near universally awful.

[Jazcash]

Isn't PhysX free already?

[el_matarife]

nVidia only and proprietary. We're going to need to build our own physics engine in OpenCL or find someone else who's built a GPL one.

[YokoZar]

(We just required SSE which has been on every mainstream processor for ~8 years and the Via chips for ~6 years.)

...and we got complaints about it.

And we also get complaints for Spring being slow. Requiring newer hardware is always a tradeoff, and at some point that tradeoff becomes worth it.

Maybe it's 5 years as El Matarife suggests, maybe it's less if hardware gets cheap enough. Either way it's something we'll want to do eventually.

[el_matarife]

And we also get complaints for Spring being slow. Requiring newer hardware is always a tradeoff, and at some point that tradeoff becomes worth it.

Maybe it's 5 years as El Matarife suggests, maybe it's less if hardware gets cheap enough. Either way it's something we'll want to do eventually.

What we need is a really transparent easy to follow guideline for when we're going to require features. That can either be a calendar date like X number of years from this feature being put on all videocards or CPUs, or it could be something like "Valve's hardware survey has this feature hitting X%". Maybe we should run our own hardware survey?

Either way, as long as we have a guideline everyone can follow the response to complaints should be "You knew that we were going to require this feature at around this time for over a year now and you didn't get ready. How is this our problem?"

[zwzsg]

Spring used to work in 2004. I still have my pre-2004 computer I first ran Spring with, and I'd be pretty irritated if Spring stopped working on it. Old games shoud not start requiring new hardware when patched. Support, fine, require, no!

You're all a bunch of nerdy bearded linux geeks, so of course you're so much into computer things you're craving for every new hardware pieces, but not every player is like you. For instance, some are kids, with no money, and harsh parents. Others have wife and several children to feed. Some live in the backwater areas of the world that are not America and where standard of living don't make upgrading computers that affordable.

[el_matarife]

Spring used to work in 2004. I still have my pre-2004 computer I first ran Spring with, and I'd be pretty irritated if Spring stopped working on it. Old games shoud not start requiring new hardware when patched. Support, fine, require, no!

Spring from 2004 will still work on your 2004 era hardware. There's no way we can take advantage of the new advancements in videocards or CPUs if we don't start requiring those features at some point. Meanwhile, at some point supporting older computers becomes more troublesome than it is worth. Do you really want to be supporting fixed function DirectX 7 type rendering well into the next decade? Or would you rather support a nice clean OpenGL 2.1 or 3 rendering pipeline?

I'm not necessarily opposed to splitting off an older codebase called "SpringBasic" or something similar that would be targeted at older, slower PCs without modern features. I just don't think we should be actively adding features to it. Five or more years is plenty of time to say that we should be retiring older PCs. In fact, five or six years is roughly the lifespan of console generations and they force people to upgrade if they want newer games.

[zwzsg]

If you made Spring 2 and it looked like Crysis, ok I would accept it'd require new hardware.

But if Spring 0.79.1.3 suddendly required new hardware just to play with the same old ugly 3do ripped from a 1998 game, then I would not be happy.

You can't make an old PC version for game that is almost exclusively multiplayer, since the old versions would not sync with the new ones.

In fact, five or six years is roughly the lifespan of console generations and they force people to upgrade if they want newer games.

If they want newer games. Not if they want to keep playing the same ole stuff with friends.

[el_matarife]

You can't make an old PC version for game that is almost exclusively multiplayer, since the old versions would not sync with the new ones.

No, people with old PCs would be forced to play with each other and anyone with newer PCs who would install and play with them.

I don't think Spring is going to become Crysis overnight, but I think we could at least shoot for looking as good as some 2004 era RTS like Dawn of War, Rome: Total War, and other games that came out around that time. Sure, the models will still suck but we could shoot for snazzier effects and there's several Spring mods building better models. Meanwhile, the CPU changes are about adding performance for newer systems at the cost of dropping support for older models. Would you trade a 10-20% performance boost at the cost of dropping support for the Athlon XP and Pentium 3? I know I would.

Also, I did the math. 5 years is 3 cycles of Moore's law which states that transistor count will double every 18 months to 2 years. Assuming that we can say transistor count is roughly equivalent to performance, a 2004 era system is 1/8th as powerful as current systems or buying equivalent performance should be 1/8th the cost as we can see with netbooks or refurbished PCs. In fact, my math is a little conservative since I compared the Geforce 6600 which is a midrange card from 2004 with an MSRP of $199 and the Geforce GTS 250 which is nVidia's current card in the $150 price bracket. The GTS250 is 10x to 20x more powerful if you look at the texture and pixel fill rates. With new hardware this cheap, why should we be crippling what kind of features we write because it might break older PCs?

[Tobi]

If you made Spring 2 and it looked like Crysis, ok I would accept it'd require new hardware.

But if Spring 0.79.1.3 suddendly required new hardware just to play with the same old ugly 3do ripped from a 1998 game, then I would not be happy.

You can't make an old PC version for game that is almost exclusively multiplayer, since the old versions would not sync with the new ones.

In fact, five or six years is roughly the lifespan of console generations and they force people to upgrade if they want newer games.

If they want newer games. Not if they want to keep playing the same ole stuff with friends.

Kernel Panic contains only 3dos ripped from a 1998 game? S44 contains only 3dos ripped from a 1998 game? I didn't know that.

Your reactions lately make it sound like you really should fork Spring and forbid any improvements so you can be sure no new bugs or requirements are introduced ...