GZDoom 3.5.1 Released

[Graf Zahl]

No, there won't. The Vintage version is just meant to support old systems with OpenGL 2. This is a mere 2.5% of all our users at the moment.
The vintage build is not meant to be used on more modern systems. It contains an old version of the renderer that performs worse on modern hardware and will fall even further behind in the future because it no longer gets developed. The only reason it exists is because more recent changes to the renderer had a negative effect on OpenGL 2 performance.

[SPZ1]

The speed of this release is back to normal for me!

[Zen3001]

wow it's really a lot faster, sitching between windowed and fullscreen mode, resolution changing done instantly without a short frozen screen like it used to

[SuaveSteve]

Will Vulkan improve performance for scenarios where there are many lights in a scene?

[Graf Zahl]

No. That's one of the things where only the hardware performance matters. In terms of driver overhead, rendering a wall with or without lights is identical.
The only thing that may help you with that is a better graphics card and/or a faster CPU.

[Master O]

No. That's one of the things where only the hardware performance matters. In terms of driver overhead, rendering a wall with or without lights is identical.
The only thing that may help you with that is a better graphics card and/or a faster CPU.

Speaking of hardware performance, what hardware would be enough to run nuts.wad in GZdoom at full speed? Currently, once the enemies start attacking in that level, the frame rate turns into a clip show and it's like moving in slow motion.

[wildweasel]

No. That's one of the things where only the hardware performance matters. In terms of driver overhead, rendering a wall with or without lights is identical.
The only thing that may help you with that is a better graphics card and/or a faster CPU.

Speaking of hardware performance, what hardware would be enough to run nuts.wad in GZdoom at full speed? Currently, once the enemies start attacking in that level, the frame rate turns into a clip show and it's like moving in slow motion.

Try it on your computer. If it's not full speed, you need a faster one.

[Rachael]

GZDoom is not made to run nuts.wad. Period.

If that is the metric whereby you are judging the source port, then move along because GZDoom is not the correct port for you. PRBoom+, or any other port that does not have a ton of overhead with enemy processing in every frame, would be better alternatives for that. But with GZDoom's scripting backends, such a map is cost-prohibitive, even on more modern systems, for GZDoom to run decently, and that is not going to change too quickly in the near future.

People use GZDoom for the OTHER things it can do, which is NOT nuts.wad! (To be clear, GZDoom can technically run it, and doesn't really have a lot of bugs with it, but it would take a system that can achieve a very high CPU clock speed in order to do so smoothly, and as far as I know, no such system exists today outside of supercomputers)

[Graf Zahl]

Actually, the biggest single hits on Nuts's performance are that sprites are drawn translucently and the calculation of dynamic lights on all the sprites. If all sprites were drawn in opaque mode, front to back, it'd perform a lot better. The only issue here is that this can cause some real problems on realistic maps if sprites are in the same place.

So, to sum it up: To play Nuts at decent speeds you'd need a computer with a CPU approximately 1.5 times as fast as the fastest thing currently available, plus a high end GPU to better deal with the large number of sprites - and you need to disable dynamic lights.

All that said, I really see no point trying to make this map run well. It's a total edge case with minimal geometry and total overkill in areas where the engine does not scale well.

[Enjay]

I've never really understood why people consider making nuts.wad playable to be some sort of gold standard to aim for. The map is a joke ( I mean literally ).

When it was released, nothing could run it at a decent speed. The only reasonable way to "complete" it is to run to the exit leaving most of the bad guys alive. Texture-wise, it looks awful, there's no gameplay and it's not architecturally interesting either. Even if you consider it a good map for stressing ports or computers, there are other maps that can do the same or even more and many that provide a far more realistic stress test. It's also a map that any mapper could emulate in about 10 minutes with GZDoombuilder.

So if a port or computer can run nuts.wad I say "so what". It's not a relevant standard to aim for. It's just a bragging rights edge case that was intentionally beyond runnable by design at its time of release. An actual playable map can run smoothly long before nuts.wad can, and that's what I want to play.

[Graf Zahl]

Anyone claiming to have managed to run nuts.wad well must have stripped down the game engine to a degree where it only can do the most basic stuff. In other words: This level and a good feature set are mutually exclusive.

PrBoom is at an advantage here, after all it got no portals, no 3D floors, no Hexen, Heretic or Strife, no means to run user provided code which all boils down to the core collision detection code running approximately twice as fast as GZDoom's (because it's only doing half the work!) and that makes all the difference in a map like this, combined with a far more basic renderer.

[Rachael]

Sorry if I came off a little harsh the first time I replied to the post. I took some bite off of it after thinking better of it, but the basic point still remained: GZDoom is not for nuts.wad. It's for running about 95% of the stuff that appears in this set of forums: viewforum.php?f=41 (outside of the occasional 3DGE and ECWolf project that appears there, but both those ports are pretty well advertised here anyhow)

[drfrag]

There's an interesting discussion going on @doomworld about increasing sprite drawing performance and they are mentioning nuts.wad.
https://www.doomworld.com/forum/topic/1 ... s-sorting/

[Master O]

Such hostility!

Anyway, Crispy Doom runs it well enough, but I was merely curious if I could run it at full speed in GZdoom.

That was all I wanted to know.

[Graf Zahl]

There's an interesting discussion going on @doomworld about increasing sprite drawing performance and they are mentioning nuts.wad.
https://www.doomworld.com/forum/topic/1 ... s-sorting/

The interesting thing with that discussion is that people got totally hung up on Nuts, even though that is probably the worst test case imaginable. Whatever info you get out of profiling that particular map won't have much use in more common scenarios because dpJudas is correct when pointing out that it won't trigger any of the real time wasters. Even the profiling of the sort algorithms yields a mere 1ms difference for one of the most hostile special cases imaginable for QuickSort (i.e. the data is mostly sorted already, which normally isn't the case in conventional editing scenarios.)