I'm working on a VR version - early alpha

[duzenko 557]

I guess I'll need to double-check that the vertexcache changes don't break shadow mapping.

Don't worry about shadow maps, I will look into that myself after trunk merge

...

Trunk shows no change in fps with glgeterror skipped on AMD. Will clone your git and give it a try.

...

First impressions: buttons in the main menu flickering when hovering mouse over them?

...

Rightful Property won't load - stuck somewhere in image load.

[cabalistic 377]

So, I've got a puzzle for you. I experimented some more with parallelization and made some experimental changes to call AddActiveInteraction in R_AddModelSurfaces in parallel or serially depending on a cvar switch. In each case I timed how long it takes. I can clearly see the times go down with the parallel version, yet the time the frontend takes in total hardly changes at all. Goes down slightly in some maps, actually seems to rise in others. Really don't get it.

[cabalistic 377]

Actually, never mind. The scenes I was just testing that were strongly frontend-limited yesterday or suddenly backend-limited, so that's why I see no improvements. Something strange is going on with those backend timings, I don't get it.

 

Edit: So it seems Afterburner is another thing that somehow eats time on the backend. Very weird...

Edited December 11, 2017 by cabalistic

[cabalistic 377]

Some more careful timing analysis shows that the frontend time does indeed go down in the multithreading scenario. However, the time between two frames (i.e. after backend and frontend finish and until they start again) suddenly increases and eats up any time savings.

 

I investigated more closely, and what suddenly takes a lot more time is unmapping the vertex caches. If I move that inside the backend, then some other operations still seem to block. My current hypothesis is that even though the backend seemingly finishes early, the GL backend is actually still working. And the next (blocking) GL call thus has to wait. So my seemingly amazing savings after commenting out glGetError might just have been hiding the true backend costs, because that way the backend was waiting for frontend instead of GL...

 

At this point it's still an open question whether I can turn the multithreading savings into something tangible.

[cabalistic 377]

Well, here's my attempt at parallelizing R_AddModelSurfaces. It works fine, as far as I can tell. Just doesn't help, because the backend (or GL) will still block.

 

https://github.com/fholger/thedarkmod/pull/2/commits/87880984c894fd2dd68ba87e8bf34785ff5eaf1c

[duzenko 557]

Intel test (github repo)

Map loads - whew!

Fps is lower than trunk - 36 vs 46.

Second load aborts with error: Not all interactions removed.

This was with the last commit before TBB addition.

Ignore gl errors shows no impact on fps to me.

Console quit hangs in idSessionLocal::WaitForFrontendCompletion

[cabalistic 377]

Interesting - and disappointing. Thanks for testing. Was this on Rightful Property? What settings did you use for shadows etc.? There's also a commented line in RenderWorld.cpp:1498 that you might try to enable. It pregenerates some of the static interactions and might take some load of the system, but it causes problems with savegames I haven't yet tracked down.

 

As for the error, I'll have to look into that. Might be some missing cleanups, or just an oversight with the new interactionTable.

[cabalistic 377]

So I just tested this with the two devices I have with Intel graphics.One is my work notebook (it's technically an Optimus hybrid of onboard Intel 530 and a GTX960m), the other is my GPD Win (Intel Atom with onboard graphics).

 

First of all, I cannot reproduce any framerate drops in Rightful Property. The GPD Win goes from 19-20 fps to 21-23 fps. The notebook stays just under 50 fps in both cases. Looking at the logs (r_logSmpTimings), I can see a fairly significant reduction in frontend drawing duration on the notebook, which does not increase the frame rate because the backend stays the same and limits it.

 

Unfortunately, the notebook also shows some serious graphical artifacts with the vertexcache version. I tried different versions of the Intel driver to no avail. The GPD Win is fine. The Nvidia card on the notebook is also fine.

 

You did not see any visual artifacts, right? They were fairly noticable, so you'd probably have mentioned them. As for the performance, can you give me more information about the graphical settings you are using? Resolution, AA etc. as well as shadow maps or stencil, ... And what build are you using, release x86 or x64?

 

I'm not sure where to go from here. If the changes cause performance drops or even visual artifacts on some machines, then I'm not sure we want to apply them. But I don't really have a clue how to fix either.

[duzenko 557]

Maybe leave this for now and return later with a fresh head

[cabalistic 377]

In the meantime, I finally took the time to port my actual VR code from the 2.05 codebase to 2.06. Performance is about the same, as far as I can tell. Not great, because it will cause reprojection at one point or another in most maps, but at least some lighter maps might be very playable.

 

I also experimented with changing the vertexcache buffers to persistent buffers, i.e. buffers you don't have to unmap to draw from, which eliminates driver syncing overhead. It doesn't improve performance per se, but with the syncing overhead of glUnmapBuffer gone, I can now actually reap some benefits from parallelizing the frontend. Nothing spectacular, but for VR, every bit counts.

 

Of course, persistent buffers are an OpenGL 4.4 feature, so this probably won't find its way back into trunk.

[cabalistic 377]

Haven't posted here in a while, because I haven't had much time, and progress has been slow as a consequence. But I'm still working on it and still working on getting performance to another level.

 

So far, I have been copying the BFG style approach to stereo rendering by simply doing each eye frame individually, meaning that both back- and frontend basically have to carry double the load. And that causes problem in many places where the CPU load just is too high to do double the load at 90 fps. So instead I'm now aiming to do things smarter. The frontend will only assemble a single view from the head position, which is then rendered twice with slight offsets for each eye. That already saves the double load on the frontend and is already working, aside from a few minor quirks.

 

But to truly profit from that, I also need to lower the CPU load on the backend side. To do that, I have been working on an all new OpenGL4 renderer, which is now in a state where it can render all basic elements of the game. It doesn't do any advanced stuff, yet (no soft shadows or advanced interactions etc.), and it won't until I think I might actually have some room in stereo rendering for it. But what it does, it does with less CPU overhead than the standard renderers, which is good and boosts the VR framerate in tight situations. Also, I can now make use of fancy new OpenGL features to basically render both eye views with a single pass, which will further reduce CPU overhead. So I'm cautiously optimistic that after I've done that, performance will be up at a level where I deem roomscale viable, even if it may not be perfect

[Bienie 277]

So glad to head that! I would think minimizing the CPU bottleneck would be the key to TDM's performance in VR, since that seems to be where it already is in standard TDM. Best of luck in your endeavours and I can't wait to one day jump into TDM-VR!

[BuckleBean 36]

That's great news! I still play missions using your previous implementation. Looking forward to where this goes!

[cabalistic 377]

Great to hear the initial alpha is finding some use

Just to manage expectations, though: due to current time constraints, it's unlikely that any new release will be ready before May, and that will still be a seated version. So please be patient with me

[Bikerdude 3736]

Ive been trying to get my best friend to play TDM for years, seeing as he now has an Oculus rift, this might be the push he needs to try it out.

[dracflamloc 0]

I am really looking forward to this continuing. Even if progress is slow due to other commitments, thanks for working on this!

[Dram 123]

Just thought to chime in, thank you for working on this, I have a rift my end too and look forward to giving this a whirl!

[BuckleBean 36]

Hi, is this compatible with 2.06? In other words, if i update to 2.06, is there anything I need to do to keep things working?

[cabalistic 377]

I'm afraid the currently released version is not compatible with 2.06. At the very least, the executable is still on 2.05 code and probably won't play well with 2.06 assets.

 

I have ported my modifications to 2.06 and continue to work on it, but it's currently not in a state ready to be released, sorry.

[BuckleBean 36]

I'm afraid the currently released version is not compatible with 2.06. At the very least, the executable is still on 2.05 code and probably won't play well with 2.06 assets.

 

I have ported my modifications to 2.06 and continue to work on it, but it's currently not in a state ready to be released, sorry.

 

No worries at all! I'll just keep from updating for the time being. Thanks for the quick reply.

[Diego 169]

Decided to give this another shot and it still doesn't launch in VR. SteamVR just sits there showing me the room and tdm runs on the screen. I even tried creating a steam shortcut for TDM with "run in VR" option.

 

Get to work, cabalistic! work for my entertainment! *cracks whip*

 

In all seriousness I'm always baffled by my incompetence when it comes to all technical things in TDM.

[cabalistic 377]

First things first sorry if these are too primitive:

 

1) Do you have a 2.05 installation of TDM? It may not run on a 2.06 installation.

2) Are you starting the right executable? (TheDarkMod.exe from the 0.1.3 release)

3) Did you put 'vr_enable 1' in your autoexec.cfg?

[Diego 169]

Yup. And if I pull the console during gameplay and check for the vr_enable it says it's 1.

 

This executable is the normal one from installation, right?

 

I have my suspicions that steamvr might be the one not playing along. It's known to be difficult.

[Samson- 5]

Hi guys - if there's anything I can do to help with the VR implementation let me know. ( I'm the D3BFG Fully Possessed VR mod dev )

 

I know you're working on engine optimizations now, and most of the work I've done is on the BFG engine so isn't directly compatible, but I'm happy to lend a hand if I can.

[cabalistic 377]

Hey, thanks for the offer! I studied your code when I attempted my initial VR modifications. Even though it wasn't directly applicable, it certainly helped to guide me in the right direction

 

I'm currently working on a small FBO overhaul for the renderer. After that, I need to port the VR mod to the latest sources. Unfortunately, that may still take a while, but after that's done, there's more than enough work to share The most pressing would be to make the UI and HUD VR-friendly. And then, of course, going for actual room-scale and controller input.