Feedback Wanted: Scriptable Render Pipelines

I have several questions that I wasn't able to find documentation for, other than writing an entirely new pipeline:

1] How do I inject a command buffer at a specific render point without writing a whole new render pipeline. For example, I want to copy the depth buffer right after it is created and create several downsampled versions of it (half, quarter, etc.). This was easy with a camera command buffer using before depth texture (forward) or before reflections (deferred).

2] I want to inject several other additional command buffers: one right after the opaque objects are rendered and one right before transparent objects are rendered, and a final one right after transparent objects are rendered. How do I do this with the new render pipelines? Is this kind of granularity possible?

3] What is the replacement for light command buffers? For example I want to capture and perhaps modify the cascade and screen space shadow map...?

Will this be different for lightweight vs hd pipeline? What about custom pipelines?

It would be great to hook into rendering points without writing a new pipeline, similar to how camera command buffers work, but maybe that's not how these pipelines are supposed to work...? I am quite new to SRP so please forgive my newness.

 

Hello everyone, i've got some question to you regarding by this article.
https://github.com/Unity-Technologies/ScriptableRenderPipeline/wiki/Upgrading-to-HDRP
You did wrote here: "The HDRP template examples were made to look good but were not realistic, whereas this Scene uses physically correct light values: an afternoon direct sun with no clouds in the sky is much brighter than even the best professional construction spotlight. However, the spotlight is still casting light and shadows on the side of the wall."
SO my question is, how you'll make it realistic further?
Also any news about post process stack 3, with pre-exposure, and rule sunny 16 for physical base lighting setup?
P.S
Can you share evening lux for sun and sky? I can't figure out it

 

That document's intensity values looks wrong.
For the clear sky noon, Sun=100000 and Sky=20000 are typical number for iluminance(lux), not for luminous flux(lumen).
And the surface's emissive intensity must be luminance(nit).

The sun(directional light)'s intensity is ok at 100000lux.
To make the sky 20000lux, the sky's emissive intensity must be 20000/PI nit if the sky is flat white.

The LightBulb is 8500lumen, it's luminous flux(lumen).
So the lightBulb's emissive intensity must be 8500/total surface area nit, if it's light distribution characteristics is isotropic.

 

Ok, but I want to undrerstand what lux settings apply for sky and sun in THE EVENING. 5-6 PM

 

#Feature requset:
Allow for volumetric fog color(single scattering albedo) HDR values here

 

I've noticed the Cameras array in Render( context, cameras[] ) is always length of 1 even when both SceneView and GameView are visible rendering at the same time, is this intended?

*EDIT* Yea it looks like the scene view and game view are called separately.

I was trying to have some per-frame-data render globally across all active cameras and was treating it as if the Cameras array length is the currently active & rendering cameras, so that I know when all of them have rendered, but it doesn't seem to work that way.

What is the intent of having Cameras as an array?

*edit* here is the workaround, you can get an idea of what I am needing to do

Code (CSharp):

1. for (int i = DebugRenders.Count - 1; i > -1; i--)
2. {
3. #if UNITY_EDITOR
4.     int count = Application.isPlaying ? 2 : 1;
5.     if (DebugRenders[i].RenderCount >= count)
6.     //if (DebugRenders[i].RenderCount >= cameras.Length) //TODO cameras.Length always 1 ?
7. #endif
8.     {
9.         DebugRenders[i].Release();
10.         DebugRenders.RemoveAt(i);
11.     }
12. }

 

Very nice!

 

@Tim-C, what's about that?
#Feature requset:
Allow for volumetric fog color(single scattering albedo) HDR values here

 

@Tim-C @phil_lira Is there an API for querying whether or not a given Shader is supported by the active RenderPipeline?

There exists

Shader.globalRenderPipeline

for setting what the active SRP is, but I don't see any way to query if a Shader has the correct

RenderPipeline

Tag. It would be great if either

ShaderData

,

ShaderInfo

, or

ShaderUtil

had some way of querying this for Editor tools.

On a side note, it would be extremely beneficial if the Shader List dropdown in the Material Editor would hide unsupported Shaders by default. Our artists quite frequently complain about there being a lot of "garbage" created when selecting a Shader now, since so many of them are unsupported. It's a Quality of Life feature, for sure, but with an API I could script this type of behavior myself with a custom MaterialEditor.

 

Is there an easy way to revert back to not using a scriptable render pipeline once you've converted to the HDRP? my project is coming along nicely however i found out that some tools i have no longer work in the HDRP and reverting back to not using any scriptable render pipeline will allow my tools to continue working.

 

Remove the pipeline asset from the Graphics Settings, modify all your Materials to no longer be HDRP and remove any special components.

Or, if you're using version control, just revert back to a revision before HDRP was used.

 

I was able to achieve the filtering with some heavy use of Reflection and a blacklist ScriptableObject. It's quite an elaborate and not exactly as futureproof as an official API though.

For this specific case yes, but I can see some additional use cases as well. If a ShaderGUI or Material Editor for an asset needs to change its behavior depending on the active SubShader per RenderPipeline.

 

Just a bit of well intentioned criticism about the HD and LW teams not communicating.

There are semantic differences between hand written shaders in HD and LW which are really not necessary. This is making shader shader conversions from legacy to HD/LW harder than it should be. I guess it is too late to change that now, but in the future, can teams please communicate a bit more in order to keeps things standard? That would help us a lot. Thanks.

 

I've got a question about the possibility of drawing both color and depth output in a single pass and whether or not this is feasible with the scriptable render pipeline. Currently my application renders point cloud data, and it is desirable for the user to load quite high density data. This performs well, however there is a caveat.

In order to improve the visual outcome to the user we also use a screen effect to show the depth differences between each point. This works well, but means we have to draw everything in two passes.

In a shortened pseudo version:

Code (csharp):

1. Pass 1
2. {
3.     vertex shader
4.     geometry shader
5.     pixel shader
6. }
7.  
8. Pass 2
9. {
10.     Tags { "LightMode" = "ShadowCaster" }
11.     ColorMask 0
12.  
13.     vertex shader
14.     geometry shader
15.     pixel shader
16. }

Without the separate ShadowCaster pass the depth simply isn't output and never ends up in the image effect. The problem is this essentially results in drawing e.g. 6 million vertices instead of 3 million. Is this something that can be achieved with the scriptable render pipeline? I've tried to do some research but did not find much on the topic of combining color and depth output into one pass.

 

How to make Light.AddCommandBuffer work in HDRP?

 

The last three weeks HDRP downloaded from GitHub does not work.
I'm using unity 2019.1.0 b3. New project.
I copy the hdsrp files downloaded from github to the "Packages" folder. All except
Tools,
TestProjects,
com.unity.testing.srp.lightweight,
com.unity.testframework.graphics,
com.unity.render-pipelines.lightweight,

How to fix this error?

Packages\com.unity.render-pipelines.high-definition\Runtime\RenderPipeline\HDRenderPipelineAsset.cs(203,32): error CS0115: 'HDRenderPipelineAsset.terrainDetailLitShader': no suitable method found to override
Packages\com.unity.render-pipelines.high-definition\Runtime\RenderPipeline\HDRenderPipelineAsset.cs(211,32): error CS0115: 'HDRenderPipelineAsset.terrainDetailGrassShader': no suitable method found to override
Packages\com.unity.render-pipelines.high-definition\Runtime\RenderPipeline\HDRenderPipelineAsset.cs(219,32): error CS0115: 'HDRenderPipelineAsset.terrainDetailGrassBillboardShader': no suitable method found to override
Packages\com.unity.visualeffectgraph\Editor\PackageInfo.cs(20,80): error CS0117: 'PackageInfo' does not contain a definition for 'GetAll'

 

2019.1 requires 5.x.x HDRP, github master is at 6.x.x and works only in 2019.2. You can grab 5.4.0 from github today and it will work on 2019.1 or if you want to try wip branch, get https://github.com/Unity-Technologies/ScriptableRenderPipeline/tree/release/2019.1 (but be aware this isn't always stable).

 

How to access shadowmaps in HDRP?

 

Is hdrp getting support for raytracing and voxelized shadow maps?
I saw them in the git hub repositories,what can you tell about them?

 

Looks like it. I suspect they can't say anything about it just yet because that will be one of their big reveals at GDC (which is in 3 weeks).

 

Wink wink, nudge nudge, say no more.

 

Unity 2019.1.0b4

How do I update Reflection Probe via scripting in HDRP?
I'd like to update Reflection Probe on demand because of performance, RealTIme update is too heavy to use.

Here is Reflection Probe Inspector SS in standard render pipeline.
I can see the vis scripting.
RenderProbe() works fine!!


But after install HDRP(5.3.1).
The dropdown list item "via scripting" is gone away form inspector.
And RenderProbe() do NOT works......

 

Hi,
I'm trying to get values from Volume objects at the position of a target object, not at the camera position.
Like “VolumeManager.instance.Update(targetObject, postProcessLayer);”

Are there any recommended way to get the transform of an object in the current scene, in RenderPipeline script?

 

I can't get Volumetric Lighting to work in HDRP (2018.3.6f1). I have this all activated :
https://github.com/Unity-Technologies/ScriptableRenderPipeline/wiki/Volumetric-lighting

And I set up:
Volume (is global)
Experimental Fog / Volumetric Fog
Volumetric Lighting Controller
Volumetric Fog
Point Light, with component "Density Volume" including a Texture3D. Volumetrics is checked to "enable".

I played with all the sliders, but it won't work.

Here are some screenshots on imgur
https://imgur.com/a/CpJ7ERi

And one picture for the forum


Has anyone got it working? Any tips, a tutorial / documentation link?

 

@KYL3R Visual environment > Fog Type > Volumetric fog and enable Volumetric fog component

 

Change Exponential fog to Volumetric Fog under the Visual Environment tab.

 

aaah thanks a lot. Working now. Simple solution, but the whole workflow seems a bit complicated, like everything in SRP. But Once you know how, it's pretty clear. Probably just a side effect of the flexibility.

 

It is not complicated, merely it's not the same as what you have already been doing. In addition, the reason for a lot of this is so you can have moodboxes or volumes that control every aspect of your game from any part automatically...

So you might want exponential in one area or volumetric in another, or adjust shadow quality per area, it makes it trivial and simple!

Problem is this early, there's not many tutorials

 

Could you add for volumetic fog an option for gradient color scattering albedo ?

 

What would it do?

 

It would add gradient color to fog.

 

Yes like in stylistic fog from bit bucket repository

 

Its all about make possible to separately set color for different disrance.
As we now from RL fog aren't just a single color. It's different and that one became with distance. It's called aerial perspective!
Veryfar distance almost white fog
Far have more bluish shade.
Not far light blue.
And etc

 

That's what Atmospheric Scattering does. This seems to be for stylized rendering than anything else.

 

Hey. I got my hands dirty with the LayeredLit Shader in HDRP. It feels weird.

This is what I ended up with after a LOT of fiddling.


It looks okay, you may think. But while painting, it doesn't feel like it fills the crevices first, and when it does - it is very transparent. When I paint more, it already leaks onto the stone's surface too much.

And it only works when I use the height-map of the stones as the Layer Mask Texture. I don't understand why I need that, when I just want to blend the layers using their heightmap and vertex color.

The "Height Transition" slider is like "all or nothing". 0,001 to 0,007 shows small differences, but the rest is like "nope - no changes here". (Same feeling when using the slider for Subsurface Scattering (Index of Refraction by the way)

Here are more images: https://imgur.com/a/ZAryaDz

So, can anyone point out a mistake I am making? Or is this not intended to heightblend stones, grass and sand?

 

Hi, I have 4 problems to report about LWRP, using 4.10.0:
1. There's no checkbox to turn on soft particles wherever (used to be in quality settings). I ended up setting SOFT_PARTICLES_ON macros in the shader manually.
2. Soft Particles doesn't work for LWRP particle shader. It worked after I changed those lines(ShaderLibrary/Particles.hlsl, line 66)

Code (CSharp):

1.  
2. // Fading vertex function
3. #if defined(SOFTPARTICLES_ON) || defined(_FADING_ON)
4. #define vertFading(o, positionWS, positionCS) \
5.     o.projectedPosition = ComputeScreenPos(positionCS);\
6.     o.projectedPosition.z = -TransformWorldToView(positionWS.xyz).z
7. #else
8. #define vertFading(o, positionWS, positionCS)
9. #endif
10.  

3. Depth textures doesn't work if I only turn "Depth texture" on in LWRP asset and turn off "Opaque texture".
4. All shaders have compile errors if I upgrade LWRP(from 4.9.0 to 4.10.0), I have to delete the whole Library folder and reopen Unity to make them work again.

I reported 1,2,3 from bug reporter inside Unity too.

There're screenshots of depth texture problem:

https://imgur.com/a/fv8YviN

I tried to create a depth prepass myself to work around this. The best injection point would be IBeforeRenderPass, but that it's run before "SetupForwardRenderingPass", so variables are not set up for rendering. There's no way to fix this without changing the source code of LWRP.

If I turn both "Depth Texture" and "Opaque Texture" options on, the frame rate drops from 60 to 40 instantly. There's too much overhead for those 2 textures. I captured API calls with RenderDoc and found out that Unity was rendering to a FBO(let's call it A) with a renderbuffer set as depth buffer. If I enable "Depth Texture" and "Opaque Texture", Unity will create another FBO(let's call it B) with a texture2d as depth buffer. After opaque pass, the color texture and depth texture will be copied to another 2 textures. After all renderings were done, B will be blit to A and A will be blit to backbuffer. That's just too many render target switches and copying, it cost 7ms for an empty scene on a Snapdragon 636 device.

Why can't we just set a texture2d as depth buffer for A, and copy it to a separate texture when opaque renderings are done? I didn't find an easy way to do this with LWRP

 

Another problem is that: why expose IRenderSetup interface but leaving no way to inherit from DefaultRenderSetup?

Also, what's the best way to change bits and pieces from LWRP and still get updates when you guys update the code base on github?

How can I setup IDE integration when I want to change LWRP's source code? It's very hard to write without code completion.

 

I submitted a couple of bugs regarding Cubemap rendering in the LWRP. However, on some of them regarding Stereo Cubemap Rendering being broken, they just get closed with the message "this feature is not supported, thus this is not a bug".
I can't find any information about that - neither the LWRP nor the HDRP documentation nor the github repo contain any information about which features are supposed to work, which are deliberately omitted, and which are known to currently being broken.

With LWRP out of preview, I find this very disturbing as it means that people (like us with the Cubemap rendering) might stumble across those issues halfway throughout development.

Am I missing something? Is there a clear list of supported/unsupported features somewhere?

Also, @Tim-C the documentation link in your first post is broken.

 

If the system will not use, please open to use the rest float3 of the unity_RenderingLayer for general purpose.

That UnityPerDraw buffer is valuable for custom SRP with SRPbatcher (without ECS?), so if I can set that float3 value on mesh component from script, it can use for many purpose.
Make variation per object, Pass the object’s volume for scale detail map repeatation and other actual size values baked on asset, and so on.

 

How to get emissive material in SRP?

 

It's on the shader?

 

What do you mean on shader?

If I use material emisive light, and put it on object, it still have shadows. But emissive materials shouldn't have shadows.

Also, I don't know if I should use albedo color with emissive color or put it as white or black?

 

You need to turn shadow casting off on the mesh. Emissive objects do not receive shadow.

 

What does this mean ?

Code (CSharp):

1. Recursive rendering is not supported in SRP (are you calling Camera.Render from within a render pipeline?).
2. The targets array should not be used inside OnSceneGUI or OnPreviewGUI. Use the single target property instead.
3. UnityEditor.Experimental.Rendering.HDPipeline.HDLightEditor:OnEnable()
4. The serializedObject should not be used inside OnSceneGUI or OnPreviewGUI. Use the target property directly instead.
5. UnityEditor.TransformInspector:OnEnable()

 

I am having this issue using LWRP



It should look like the image below



The details for the device which the issue occurred is on the image below

 

I'm reading through this: srp-batcher-speed-up-your-rendering and I see the "UnityPerDraw" CBuffer but whenever I put any of the items in the CBuffer (say, unity_WorldToObject) I get an error telling me that it's already been declared. What am I doing wrong?