Interim Vulkan CM2x build?

[Howler]

I can now run Quake 3 on my Raspberry Pi Model 4 (RPI4) using Vulkan. Frame rates have gone from 30 (OpenGL) to 700+ (Vulkan).

Any chance at all that we can get CM2x titles migrated to Vulkan as we wait for CM3x?

An awful lot of platforms are offering solid support for Vulkan. It would be pretty nifty to run CM2x on a RPI4...

[Schrullenhaft]

Unfortunately converting from OpenGL to Vulkan is NOT a straightforward, simple process. Vulkan consists of lower-level graphics calls that require very extensive rewrites of the graphics code especially if you want to see significant performance gains. I'm not sure what accounts for the 700+ FPS on Quake 3 that you see on a a RPi4, but even with a highly optimized Vulkan rewrite I doubt you would see more than double the current frame rate in CM2 games. Admittedly such a performance boost would be very welcome, but a highly optimized Vulkan rewrite would come at the cost of potentially delaying CM3 for another 2-3 years.

[Howler]

It's basically the FPS difference between the two technologies at a stand still. OpenGL is not Vulkan and would require an extensive retrofit.

I was fishing for some confirmation that CM3 is: a) being worked on; b) the GFX layer will be Vulkan; and c) multithreaded.

I'll keep fishing.

[Sgt Joch]

The problem is always trying to predict what will be the optimum engine 5-10 years down the road.

When CMx1 was being worked on, DirectX seemed to be the better bet and BFC coded the engine assuming NVIDIA and ATI would follow DirectX protocol in their drivers, but both card makers tweked their drivers in unusual ways which caused problems with the game.

So when CMx2 was being designed, BFC chose OpenGL as the underpinning since it resolved the CMx1 graphic issues, was more flexible than DirectX and seemed to be what most games would be using in the near future, but then most games switched to DirectX and both NVIDIA and AMD let their OpenGL drivers languish.

So Vulkan is now the new kid on the block, assuming BFC makes the immediate decision to switch to Vulkan, will it still be the gold standard in 2025 when the 1st CMx3 game would come out?

Edited October 28, 2020 by Sgt Joch

[AlexUK]

Would be great to know where BFC is heading. 

 

Any chance of CM2 on the Mac silicon (not sure what the best term is)? 

[Schrullenhaft]

I don't think that CM2 in its current form may work with the new ARM-based Macs that are coming out in the future. There are libraries now (that require recompiling the games, to my knowledge) that will convert OpenGL calls to Metal API, but they cost money (MoltenGL). So any patch that adds this feature will probably need to come out as a 'paid-for upgrade'. This library would have to be ported to ARM and then it may be possible to run a patched version of CM2 on the new Macs, but there may be other issues that require recompiling and emulation of the current code to work in an ARM environment. Rosetta 2 would almost certainly be necessary and that may have potential bottlenecks and incompatibilities with CM and other games. 

I highly doubt that Apple is going to provide any native support for OpenGL on the ARM-based Macs. They will likely support the Metal API and nothing else. The situation similar to this in the past that comes to mind is when Apple came out with OS X (on PowerPC CPUs initially) and OS9's 3D RAVE API was initially supposed to work in some sort of emulation mode to work with OS X's OpenGL. This ended up not happening and the CM1 series had no way of working, even though there was a 'Rosetta' emulation layer for this Apple transition.

[Howler]

12 hours ago, Sgt Joch said:

The problem is always trying to predict what will be the optimum engine 5-10 years down the road.

...

So Vulkan is now the new kid on the block, assuming BFC makes the immediate decision to switch to Vulkan, will it still be the gold standard in 2025 when the 1st CMx3 game would come out?

OpenGL is cross platform and and still being used 20 years later. Vulkan is positioned to be it's successor.

I can't speak to anything else as I have no insight into the deliberations BFC may be having on target platform(s). But, better performing OpenGL games have been released since CMSF. Not to mention games with multi-threaded support for 64bit environments being not uncommon today.

You seem to be indicating that we are five years away from being five years away from CM3x...

As I said, I was fishing for info as their target platform would need to be defined now if they were expecting a release by 2025. If it's still under consideration then a 2030 window seems more likely.

[Erwin]

11 hours ago, Howler said:

if they were expecting a release by 2025. If it's still under consideration then a 2030 window seems more likely.

OMG.  2030 for a new engine release?  After working on essentially the same game concept for over 20 years (since the late 1990's) would either be bored and looking to move on, or thinking about retirement.   

[Howler]

 

2 hours ago, Erwin said:

OMG.  2030 for a new engine release?  After working on essentially the same game concept for over 20 years (since the late 1990's) would either be bored and looking to move on, or thinking about retirement.   

I've studiously avoided any mention of raytracing.

If it's not uncommon now to wait 3-5 years for a module to an existing title, what makes you think a complete re-implementation using *new* technology is going to be any faster?

Don't get me wrong, I'm comfortable with "it is what it is".

Again, I'm only fishing for bones ahead of the usual year end feeding frenzy...

[Bulletpoint]

4 hours ago, Howler said:

If it's not uncommon now to wait 3-5 years for a module to an existing title, what makes you think a complete re-implementation using *new* technology is going to be any faster?

Don't get me wrong, I'm comfortable with "it is what it is".

I'm not. I would like to see this game progress. It has such potential.

[PIATpunk]

Steve dismissed Vulkan in a thread a year or two ago.