Originally posted by overfuze
View Post
This its a start point to make a better source implementation.
-
holy moly? xD, I need to give to the community what I have learning last years with, and learning with feedback situations give me some ideas what to search for.
This its a start point to make a better source implementation. -
Wow.thank you for your hard works.i cant wait for using your branchOriginally posted by VItorEAFeliciano View PostComment
-
Hey dude ! that would be so cool !!! Thanks a lot ! Sharing your version would be an awsome help and would definitly make my day (any day it shows up of course ! No hurry !Originally posted by VItorEAFeliciano View Post
I'm a newbie with all that github stuffs so i would apreciate a lot a compiled version of all that advanced techs.
I've tested VXGI and Hairworks and it looks very promising but I'd like to use both in UE4 and an "expert" compiled version will be much more easiest to use...
Also, Did anyone know if we can use VXAO in ue4 ? Is there a link to a branch using it ?
And last but not least: Is the "VR left eye only shadows" in VXGI fixed ? It really kills the experience (with HTC vive for me...).
Thanks again a lot VItorEAFeliciano, can't wait for your release !
Have a nice day !Last edited by gonzohot; 04-07-2017, 03:14 AM.Comment
-
[MENTION=183026]gonzohot[/MENTION] VXAO is a part of VXGI branch, you can make VXAO run without GI by typing r.VXGI.AmbientOcclusionMode 1
You can read more about it here: https://developer.nvidia.com/vxao-vo...ient-occlusion (about Unreal Engine section is at the end of the page).Comment
-
Thanks overfuze !
That's true: I now remember that i quickly tested it some times ago but it turns all my scene Black, and I just didn't figured out why...
EDIT: i just tested it again and it seems to work but i don't really see any differences with just HBAO+ enabledComment
-
Yeah I don't really see the difference that much myself either, but you can set the visualization mode to Ambient Occlusion and switch r.VXGI.AmbientOcclusionMode between 0 and 1 and you will see the change.
EDIT: Here is an example https://gyazo.com/b4dac81d1e9c59a1cefc8ccad12815ab (command I entered was r.VXGI.AmbientOcclusionMode 1)Last edited by overfuze; 04-07-2017, 08:07 AM.Comment
-
[MENTION=213912]Andrew.Reidmeyer[/MENTION] So I made a test animation and pushed Flow's couple rate and substepping both to 10, and while the emittance does continue while it's moving quickly there's a very visible stepping between frames. The only solution I was able to find to make it smooth out is to push the grid's simulation rate to 120, which to me sounds like a terrible idea in terms of performance. However, it did result in a 100% completely smooth animation and looks perfect, just what I'm looking for. Are there any other settings that may be able to get that kind of result without taking what sounds like a pretty bruteforce approach? I'm using it on a weapon in an arena fps of sorts, and need to make sure this sort of thing doesn't cause any slowdown.
Also, I've been looking at the Flow VR settings to get ideas on how I could potentially optimize things, but as there isn't any documentation on Flow that covers performance directly (although I have found the gameworks documentation page for it, which helped a lot), what settings would you say to look at first to make it run as fast as possible?Comment
-
There is some latency to make new blocks active. For fast moving emitters, there are two settings. One is 'allocation scale', this can be increased to values greater than 1.0 to allocate a larger active region around the emitter. Very useful for hand held emitters in VR, where it is less trivial to predict the future emitter location. Another is 'allocation predict', this attempts to allocate based on emitter velocity, so is a directional form of prediction.
To visualize the effects of both parameters, use the 'flowvis' command, and look at the wireframe around the emitter.
Lower simulation rates will tend to offer better performance, although 90 Hz is the typical value used with VR, to align with vsync. Performance in general will scale based on the number of active blocks, so flowvis serves as a tool to notice active blocks that might not be contributing enough to justify the allocation.Comment
-
I'll try it out when GitHub isn't completely retarded and they release it somewhere else, or with the engine (like it should be). It's the worse popular service on the entire internet.Comment
-
Originally posted by buzzdx View Post
You right,
Tessellation can't be done with VXGI Voxel's in the same material,
Code:// NVCHANGE_BEGIN: Add VXGI if (RHIIsVoxelizing() && !MaterialRenderProxy->GetVxgiMaterialProperties().bVxgiAllowTesselationDuringVoxelization) { return false; // not allowed } // NVCHANGE_END: Add VXGIComment
-
I did some more testing locally with fast moving emitters, analyzing the substepping interpolation frame by frame.
In the case of unlimited FPS (~200 FPS range in this case), I did see substepping interpolation that could be made smoother. We will work on improved interpolation for this case.
You can verify if this is the issue by setting t.MaxFPS to align with the simulation rate.
This issue did not manifest in previous VR work, where simulation rate was aligned to FPS.Comment
-
This is exactly what I've been seeing on my end in a test scene:Originally posted by Andrew.Reidmeyer View Post
I did the test with setting t.MaxFPS to 60 and can confirm that limiting the FPS did actually fix the sim rate at 60 and made it look as smooth as the 120 fps one. At 120 fps, it almost seemed like substepping wasn't having any effect on the 60 fps sim, where it's clearly working perfectly once it was capped.Comment
-
can someone help me getting a flow emitter to work?
the sample project works fine, but i cant get a new emitter to work.
i add a grid asset, i create a blueprint actor with a sphere and add an emitter component. whats next?Comment
-
Originally posted by VItorEAFeliciano View Post
I vow to stop eating until this is released. We're talking about a man's life here guys!Comment
Comment