Realtime Dynamic GI + Reflections + AO + Emissive plugin - AHR

[RyanTorant]

Dear RyanTorant

For the past 2 months I was too busy to keep on following your thread... and as I see now it became a huge one...!!
Can you please; do me/us a favor...? I think it is time to make an updated small file containing:

AHR's main features, Best advantages, What it covers & limitations as well
Main instructions (How to use) to get best results... something like that...
I may guess that would be very helpful...

Happy Birthday &
Best wishes

Sure, I'll add that to github starting page, and try to keep it up to date.
I haven't quite done that yet because it can be time consuming, and I don't have much of it to begin with D:
In any case, you at right, even if its a small description or tutorial it'll help people get around ahr

[TAN_]

OK OK, I got it. It is my mistake. I did not know about DX12 and tried to activate it on a Windows 8 system. My stupidity, sorry.

[Keiyentai]

Right now DX12 is only a Windows 10 thing. Also unless your running a 9xx Maxwell based nVidia card and the current AMD cards (not sure exactly which ones support what haven't used an AMD/ATi card since the HD3200). Though I tried DX12 with a scene that's particle based lighting and ya >_> it was like "I don't know what to do!?!" so I think I will test DX12 when I get my Titan since it runs "Full DX12" though DX12 is interesting...and it's different Tiers. My 660 though supports the basics so probably won't do anything different then DX11 least right now. I will find a way to make lit particles work >_> I got LPV working when it wasn't suppose to way back when it first got implimented. I like "breaking" UE4 >_>

[DylserX]

Tested AHR out in the Reflections test scene hehe, getting some pretty nice results.

[TAN_]

Tested AHR out in the Reflections test scene hehe, getting some pretty nice results.

Hooooollllyyy.... This is AHR ?! HELL YEAHS !

Can you also give your FPS, AHR settings and system specs ?


EDIT: Ryan do you have a general roadmap ? I just calculated that I will finally have the necessary time to start my game in UE4 in earnest within 3-4 weeks and wondering if it would be possible to use your AHR for it ? Do you plan on a "initial merge" or somethingl like that ? Like merging it main branch in a stage you deem "usable" and then improving on it as the time goes ???

[DylserX]

I get from 25 - 32 fps atm

AMD FX-8120 @ 3.2Ghz
AMD R9 390 8GB GPU
32GB DDR3 1600 Ram

Ignore the "Diffuse samples, usually only have them on 32

[RyanTorant]

That screens made me want to play with that specific scene
Here's a screen comparing reflection types:

EDIT: Ryan do you have a general roadmap ? I just calculated that I will finally have the necessary time to start my game in UE4 in earnest within 3-4 weeks and wondering if it would be possible to use your AHR for it ? Do you plan on a "initial merge" or somethingl like that ? Like merging it main branch in a stage you deem "usable" and then improving on it as the time goes ???

Actually there's just a few things I want to do before asking Epic for integration, and that changes are
-Rewrite part of the tracing algorithm. Want to put it all in one pass, so that you can control the number of rays
-Improve tracing speed by using the binary grid properly, and not how it's being used now
-Multiple bounces

When that things are done, it can be said that AHR is ready for integration with the main branch, as a WIP of course. I'll ask Epic to join it to the main branch. I also want to ask them for an internship there, for next year. I belive I showed that I know what I'm doing, and if they actually want AHR into UE, who better that the guy that created it right ;D
Anyway, still need to do a lot of work. My goal is to have all that done by the end of the year, and also start working on the following "extra" items:
-Particles lighting
-Multiple bounces for reflection
-Improve tracing speed by treating the binary grid as an octree and doing a stackless traverse
-AHR based bake. The idea is to fire a secondary voxelization stage, trace, store that on a voxel grid, blur, and then compact it to a sparse octree. That way you get high quality bakes in seconds, and also you dont need to make lightmap UVs.

To answer your question directly, in 3-4 weeks I hope to have the first two things mostly covered up, so yeah, you can start with AHR.
On the other side, university is taking a toll on my free time ( it's 0 currently tbh ), so that's why development is so slow. I'm quite confident that with 8 hours per day dedication, and working on a team that knows the way around UE (*cough* *cough* aka working at Epic *cough* *cough*. Even if they are not actually working on AHR, having someone to ask a quick question about the engine saves so much time) AHR could be done in 2 - 3 months.

Some may ask why use AHR instead of VXGI for the main branch, and that's a fair question. VXGI is more stable, and normally looks better. There are two main differences (currently, the particles and bake stuff would make it 4) one being the performance gap (AHR is still faster than VXGI, by a large margin depending on the settings. Specially specular tracing, that's crazy slow on VXGI) and the other, and to my opinon the main one when it comes to integration, is that AHR is running directly in the engine, all the code is there and doesn't rely on external libraries like gameworks. That makes it quite more flexible, as Epic can actually see what AHR is doing and improve from there.

[RyanTorant]

I get from 25 - 32 fps atm

AMD FX-8120 @ 3.2Ghz
AMD R9 390 8GB GPU
32GB DDR3 1600 Ram

Could you make a profilegpu and post a screen? I want to see how the frame time is distributed.

[DylserX]

Such a stark difference haha :P

[Keiyentai]

Ooooh shiney. That is awesome looking with AHR. Still working on a forest scene can't wait to see how well it works with that scene. *grin*

[Bladerskb]

i have a 970 and would love to test things out. but a question i have is, whats going on with the metal surfaces. they look whitish cloudy.

[John Alcatraz]

Tim Sweeney has retweeted the tweet from DylserX about AHR



So Epic knows how Epic this is

[Keiyentai]

Whoa Tim retweeting it? That is freaking awesome. Hopefully it's a good sign for the future of UE4 and AHR. I know I love AHR.

[TAN_]

That screens made me want to play with that specific scene
Here's a screen comparing reflection types:


Actually there's just a few things I want to do before asking Epic for integration, and that changes are
-Rewrite part of the tracing algorithm. Want to put it all in one pass, so that you can control the number of rays
-Improve tracing speed by using the binary grid properly, and not how it's being used now
-Multiple bounces

When that things are done, it can be said that AHR is ready for integration with the main branch, as a WIP of course. I'll ask Epic to join it to the main branch. I also want to ask them for an internship there, for next year. I belive I showed that I know what I'm doing, and if they actually want AHR into UE, who better that the guy that created it right ;D
Anyway, still need to do a lot of work. My goal is to have all that done by the end of the year, and also start working on the following "extra" items:
-Particles lighting
-Multiple bounces for reflection
-Improve tracing speed by treating the binary grid as an octree and doing a stackless traverse
-AHR based bake. The idea is to fire a secondary voxelization stage, trace, store that on a voxel grid, blur, and then compact it to a sparse octree. That way you get high quality bakes in seconds, and also you dont need to make lightmap UVs.

To answer your question directly, in 3-4 weeks I hope to have the first two things mostly covered up, so yeah, you can start with AHR.
On the other side, university is taking a toll on my free time ( it's 0 currently tbh ), so that's why development is so slow. I'm quite confident that with 8 hours per day dedication, and working on a team that knows the way around UE (*cough* *cough* aka working at Epic *cough* *cough*. Even if they are not actually working on AHR, having someone to ask a quick question about the engine saves so much time) AHR could be done in 2 - 3 months.

Some may ask why use AHR instead of VXGI for the main branch, and that's a fair question. VXGI is more stable, and normally looks better. There are two main differences (currently, the particles and bake stuff would make it 4) one being the performance gap (AHR is still faster than VXGI, by a large margin depending on the settings. Specially specular tracing, that's crazy slow on VXGI) and the other, and to my opinon the main one when it comes to integration, is that AHR is running directly in the engine, all the code is there and doesn't rely on external libraries like gameworks. That makes it quite more flexible, as Epic can actually see what AHR is doing and improve from there.

Tim Sweeney has retweeted the tweet from DylserX about AHR

[ATTACH=CONFIG]57549[/ATTACH]

So Epic knows how Epic this is

HOOOLY ****

**** YEAHS !!!

Way to go Ryan ! As an old saying goes "One step closer to victory ! "

I would really like to see you at Epic beating the **** out of this tech. Also great to hear that it would be ready to use within 3-4 weeks. Hopefully I'll make a good showcase out of your tech as a game.

Also please don't forget my little wish; Realtime Sunshafts

@ Keiyentai Will be looking forward to that forest scene

[DylserX]

Could you make a profilegpu and post a screen? I want to see how the frame time is distributed.

Yeh sure mate, almost home now will post it soon.