[Survey] How many of you CANNOT run GZDoom's OpenGL renderer

[GooberMan]

it sometimes baffles me how many Doomers 'cannot afford' to even upgrade to a modern entry level graphics card

It's more reasonable if you take in to account the Russian scene. Or the infamous Sgt_Mark from Brazil (and indeed any of the other South American nations where many Doomers come from). And within the EU, Greece and Spain still has greater than a 1 in 4 unemployment rate.

All people "need" in terms of computing is a connection to the internet and a browser these days. Anything else is secondary.

[Gez]

different shader syntax

Would that affect GLDEFS shaders?

[Graf Zahl]

No, they do not access the parts of the spec that changed.

@Gooberman: Ok, I understand that part. But let's be serious: A first generation GL 3.x card can be bought used for approx. $/€ 20 these days and for $/€ 40 something semi-decent can be bought new, for someone with a serious interest in gaming this should be affordable.

We are talking about 8+ year old hardware here, mostly, that can't handle GL 3.3.



And regardless of the situation. anything pre GL 3.3 is a dead end from a developer's perspective. The current engine will serve this hardware well for a long time - but I am absolutely unwilling to maintain two distinctly different implementations of the renderer.

[jpalomo]

I'm using GL 4.4.0.

How old is OpenGL 2.0 anyways?

[Graf Zahl]

Ancient. That's the problem. The first GL 3.x cards were released in 2006, if I remember correctly. Unfortunately Intel was (as usual) late and stalled progress for several years.

Anything with real shader support also supports GL 3.3, that's why I see no point developing a renderer using modern technology but restricting itself to an obsolete OpenGL version.

[GooberMan]

But let's be serious: A first generation GL 3.x card can be bought used for approx. $/€ 20 these days and for $/€ 40 something semi-decent can be bought new, for someone with a serious interest in gaming this should be affordable.

I know that sounds reasonable to us. We're both in IT so we get paid well. €20 to €40 sounds like nothing.

Let me tell a story, something I use to keep perspective myself on these things.

Back in the day, I grew up in a single income family with a mortgage. My dad worked in the public sector, so the job was secure even if the pay was ordinary. But then I got a second brother and the finances meant my mum finally had to get a job. That was all well and good for a few years until some head-office management bullshit eventually saw her get fired from her assistant manager role. Back to a single income then, while the country was still recovering from a recession and the cost of living was worse than what it used to be.

Financial situations kept on changing, but the minimal spending on things that weren't necessary to get by continued. I didn't even have text books in high school because we couldn't afford those fees - yet I still finished top 10% of the state in two subjects in my final year. One of them was a course dedicated to computer programming.

There was always a computer of some kind in the house. My dad would do work at home on the thing when we got a PC. But he would also play lots and lots of games on it. We were all in to gaming. But we couldn't really afford all the games we wanted, so we pirated plenty. I didn't even have a legit copy of the Doom games until I got my own job. The family financial situation was much better by the time I was turning 18, and game purchases were much more common as a result. And I haven't pirated a game since I started programming professionally.

This little story though? It probably seems tame to some people on these forums.

Growing up in Australia means I definitely had it better than some of those countries I mentioned in my earlier post, but there was still a degree of luck to get where I am (mostly of the "right time, right place" kind). I use that same perspective when arguing with colleagues about why we should lower game retail prices and not hate consumers for wanting a cheaper alternative (ie second hand games).

Sometimes that €20 to €40 makes a difference, especially to people earning around or under the average wage. And then there's other countries. Estonia is something like 80km south from where I live and their wages are half or under when compared to Finland.

[Enjay]

I fully accept what you are saying about people in that situation - I see it often enough with some families even in this relatively affluent area, but let's look at it a slightly different way:

If the GZDoom code base moves on to require GL 3.3, what are the consequences for people who can't afford new hardware? Have they lost out? I suggest not really. They will still have the exe that works on their system and they will have access to mods that work with that exe. They will also have access to newer non-renderer features in ZDoom. In other words, they don't lose out on much at all. They can still play the game and do so using an engine targeted at their specs.

Yes, modders will make mods that require the new GZDoom exe. Presumably the new exe will keep pace with ZDoom changes and the old one will not so there may even be mods that do not require new rendering capabilites but which are GZDoom specific and require newer ZDoom features. I can understand that being a bit galling for users with out-dated hardware.

However, what we are talking about is a relatively small percentage of the potential audience versus the significant inconvenience of Graf having to keep the old code and also hold back GZDoom as a result. In other words, the majority of users, the developer and the engine itself (I suppose, kind of) suffer as a result of catering for a fairly small minority that are only disadvantaged to the extent of not being able to play some of the, as yet unreleased (not even planned), projects that would require the newer GZDoom.

Over time, there would, of course, be more and more mods requiring the new exe but, equally, there would be fewer and fewer people unable to play them. By the time those mods are available, the number of people with unsuitable hardware will be less because there is only ever going to be a diminishing percentage of users in that situation.

I know it sounds harsh but I don't believe that GZDoom should be held back because a small proportion of the potential users of a newer renderer are unable to upgrade to a standard that was superseded almost a decade ago. Keep the old exe available, people can play suitable mods on it and when they have the hardware, they can access the newer stuff, just like they will be be able to with any other game or piece of software that requires specs above what they currently have.

I don't think this is an unreasonable standpoint. I'm not arguing that Graf should abandon half of the current user base or force them to upgrade (that would be unreasonable) but, rather, I'm saying that I feel that it is reasonable to move on to a better, more flexible standard that the vast majority of players can already access and which will benefit almost everyone involved at the minimal "cost" of a small proportion of users not being able to play some as yet unreleased mod immediately it is released.

[GooberMan]

I don't think this is an unreasonable standpoint.

Quite honestly, neither do I. Old versions for old hardware is completely reasonable. Moving to a newer version of GL is feasible now that OSX finally properly supports something other than 2.x. My own ideas for improving the renderer include ignoring older versions of GL (especially ditching fixed function support).

Multi-platform support is quite ingrained in to me though. G/ZDoom does have a small number of maintainers though, hence why I still see 2.x as quite worthwhile - especially considering where computing is going (ie off the desktop and on to mobile/tablet).

[Graf Zahl]

(especially ditching fixed function support).

Precisely my point. That stuff gives me serious headaches because it makes the code extremely hard to handle.
And also getting rid of immediate mode and all means that I could add some multithreaded processing. As it stands there's little point because it all gets stored in some intermediate structure and during rendering even more processing is done so only a relatively small amount of processing could be spread across cores. That'd be different if data is written to a buffer instead of being handled by immediate mode functions.


And besides, nobody has said that the old engine can't be kept up to date with ZDoom's new additions - even if I wouldn't do it myself.


Regarding mobile: GLES 2.0 is closer to 3.3 than to regular GL 2.0 if you ask me, it got less features for sure but like GL 3.x core it requires use of shaders, vertex buffers and user space matrix handling.
Common GL2 hardware does not - and actually is far too weak for a shader based engine so that point is moot anyway. GL 2.0 desktop wouldn't even work without fixed function features on hardware that's not GL3 compatible.
And GLES 3.0 is definitely coming so with the insane replacement cycle in mobile it shouldn't take too long until support for it is widespread enough. Even now GLES 1.1 is obsolete and I don't expect 2.0 to outlive its successor for long.

[Blue Shadow]

I agree with Enjay's points. But since I've already written this, I might as well post it:

Personally, I'd say go for it. Once you get the time and the opportunity - just go for it; it's something that's long overdue. I'm not saying that out of selfishness because I have support. Heck, my current machine doesn't run GZDoom as smooth as I'd hoped. I can run light-to-medium mods fine, I suppose. But anything heavier, and I start to suffer framerate drops.

I shy away from GZDoom-centric mods, in general, because of that. That's why ZDoom is my main engine of choice. Besides, I pretty much prefer the software renderer over the OpenGL one (if only we had truecolor support in the software renderer... *sigh*).

So yeah, as you can see, I'm still alive and well. GZDoom and GZDoom-based mods aren't a necessity of life, and you don't need to have them to enjoy playing the game you like, as ZDoom can still fulfill that (to me it does, anyway).

[Enjay]

(especially ditching fixed function support).

And besides, nobody has said that the old engine can't be kept up to date with ZDoom's new additions - even if I wouldn't do it myself.

True. I had actually meant to mention that someone might feel able and willing to keep the older renderer in step with ZDoom changes as long as there was a perceived need to do so.

[GooberMan]

That'd be different if data is written to a buffer instead of being handled by immediate mode functions.

Quite honestly, you should be able to ditch immeiate mode with the current code base. Vertex arrays are supported by just about everything, and the difference between vertex arrays and vertex buffers are the values you put in to the bind functions (arrays work off in-memory locations, but buffers require offsets from zero).

GLES 2.0 is closer to 3.3 than to regular GL 2.0 if you ask me, it got less features for sure but like GL 3.x core it requires use of shaders, vertex buffers and user space matrix handling.

This is all doable in the 2.0 API. Optional, sure, but there's nothing stopping you enforcing a 3.3 feature set while sticking to 2.x APIs. It's pretty similar to how I handle my iOS/PC cross compatible code. Direct3D 11 even does something comparable with its 9.x hardware profiles.

And GLES 3.0 is definitely coming so with the insane replacement cycle in mobile it shouldn't take too long until support for it is widespread enough.

That rampant replacement is really only applicable for high end phones. I think the Galaxy S5 is the only phone out on the market with 3.0 support at the moment. That low-to-mid range segment will stay on the cheaper 2.x hardware for quite a few years - long enough for a delevopment roadmap to slowly be over to 3.0 only features at the very least.

[Graf Zahl]

That'd be different if data is written to a buffer instead of being handled by immediate mode functions.

Quite honestly, you should be able to ditch immeiate mode with the current code base. Vertex arrays are supported by just about everything, and the difference between vertex arrays and vertex buffers are the values you put in to the bind functions (arrays work off in-memory locations, but buffers require offsets from zero).

The current code base is entirely unsuitable for such a rewrite.

This is all doable in the 2.0 API. Optional, sure, but there's nothing stopping you enforcing a 3.3 feature set while sticking to 2.x APIs. It's pretty similar to how I handle my iOS/PC cross compatible code. Direct3D 11 even does something comparable with its 9.x hardware profiles.

It may be doable with the API but not with the hardware that cannot support GL 3.x. All the hardware that can do decent shader stuff with GL 2.x also supports 3.3 so it'd be a totally counterproductive move to do a shader-only engine while sticking to an outdated API that's incompatible with current development.

As for using a 3.3 feature set with a 2.x API, it means that I'm still squarely stuck in the past, depending on extensions to do stuff. Not so much a problem on Windows where there's still the compatibility profile, but certainly a complete no-go on MacOSX. It's either 2.1 or 3.3 core. Apple is quite strict about not supporting compatibility profiles. And if there's one thing I do not want is a code base that doesn't work on Macs. It's not about supporting most of today's hardware but also having a forward compatible code base. And if there's something GL 2.x is not, it's being forward compatible - quite the opposite!

And GLES 3.0 is definitely coming so with the insane replacement cycle in mobile it shouldn't take too long until support for it is widespread enough.

That rampant replacement is really only applicable for high end phones. I think the Galaxy S5 is the only phone out on the market with 3.0 support at the moment. That low-to-mid range segment will stay on the cheaper 2.x hardware for quite a few years - long enough for a delevopment roadmap to slowly be over to 3.0 only features at the very least.[/quote]


Let's see. GLES 1.1 is already gone completely and it hasn't been that long that GLES 2.0 went mainstream.
I'm also waiting for GLES to be ditched completely eventually. Now, with GL's core profile I wouöd expect that regular GL and GLES will eventually merge.

[GooberMan]

GLES 1.1 is already gone completely and it hasn't been that long that GLES 2.0 went mainstream.

Now, that's a rhetorical argument at best. The number of users still on Android 2.3.x indicates both people holding on to hardware for longer and the low-to-mid range market purchasers buying 2.3.x phones. The rush to get a smart phone is over. Expect newer hardware features to take longer to propagate.

[Graf Zahl]

... and the low-to-mid range market purchasers buying 2.3.x phones.

They'd be very hard to find these days - and certainly not in the mid-range segment anymore. 2.3 may still exist as bargain-bin offers, but that's the ultra-cheap crap, the lowest of low end.
As a mobile game developer I get some statistics and from the looks of it, 2.3's share among game downloaders is significantly lower than among all app users - at least with my company's products.

Ultimately these users will be as relevant to game development in mobile as low end desktop PC users are to Windows based game development, i.e. they'd be mostly ignored eventually.