This simplify Strata Slab UI, and make it easier to thin-film coat many BSDFs at the same time.
#rb none
#jira none
#preflight 622fa337c51b66df4c248174
#fyi sebastien.hillaire
[CL 19378172 by Charles deRousiers in ue5-main branch]
Reductions *per material*:
SM5
--
FHWRasterizeVS: 832 -> 21
FHWRasterizePS: 104 -> 39
SM6
--
FHWRasterizeVS: 320 -> 9
FHWRasterizeMS: 640 -> 9
FHWRasterizePS: 120 -> 30
Vulkan
--
FHWRasterizeVS: 320 -> 9
FHWRasterizePS: 40 -> 15
Other platforms redacted =)
-- Details
* CLUSTER_PER_PAGE has been fully removed (since we no longer ever run CLUSTER_PER_PAGE=0), which now makes it mutually inclusive with VIRTUAL_TEXTURE_TARGET
* HAS_RASTER_BIN has been replaced with a dynamic branch, since this is just a per cluster index offset based on a simple uniform buffer load
* ADD_CLUSTER_OFFSET has been replaced with a dynamic branch, since this is just a per cluster index offset based on a simple uniform buffer load
* HAS_PREV_DRAW_DATA has been replaced with a dynamic branch, since this is just a per cluster index offset based on a simple uniform buffer load
* NEAR_CLIP (only change to significantly affect codegen) has been turned into a dynamic branch based on FNaniteView - this lets us merge depth clip/clamp rasterizer calls in VSM together instead of relying on HAS_PREV_DRAW_DATA, and a future optimization can now be done to merge local and directional light full Nanite pipeline calls together.
* VISUALIZE permutation removed from VS/MS since it only loaded unform values that passed down per-vertex into fragment stage as nointerpolation parameters. Pixel shader now constructs this uint2 directly under the VISUALIZE permutation
* NANITE_MESH_SHADER_INTERP removed by default but still left in the code, since it is a work in progress potential optimization for DX12 mesh shaders
* Removed explicit Lumen and VSM usage of NANITE_RENDER_FLAG_HAVE_PREV_DRAW_DATA (now the dynamic branch path is only taken if CullRasterizeMultiPass implicitly breaks the rasterization into multiple calls due to NANITE_MAX_VIEWS_PER_CULL_RASTERIZE_PASS overflow)
Performance was tested on a 2080Ti in AncientGame, and the delta is effectively noise (tested cached and uncached VSM). Further testing on other platforms will occur, but important to get this change in for all the benefits and easy to tweak things later if needed.
#rb rune.stubbe
#fyi brian.karis, ola.olsson, andrew.lauritzen, jamie.hayes, daniel.wright, krzysztof.narkowicz
#preflight 622e684c7e2e35638c96a16a
#robomerge FNNC
[CL 19370372 by graham wihlidal in ue5-main branch]
* Moller-Trumbore and Watertight triangle intersections in RayTriangleIntersect.h
* AABB intersection
* Nanite cluster intersection
* Nanite hierarchy intersection using a stack.
#rb Tiago.Costa
#preflight none
[CL 19196324 by aleksander netzel in ue5-main branch]
Scene proxies having the bRayTracingFarField flag set would never be visible in the editor
Removed the VisibleInRaster logic which caused the HLOD1 meshes (having bRayTracingFarField=true) from being hidden.
#rb krzysztof.narkowicz
#preflight 6215a7db0f71e491cce866fd
#ROBOMERGE-OWNER: sebastien.lussier
#ROBOMERGE-AUTHOR: sebastien.lussier
#ROBOMERGE-SOURCE: CL 19091265 via CL 19092640 via CL 19094403 via CL 19095886 via CL 19105082
#ROBOMERGE-BOT: UE5 (Release-Engine-Staging -> Main) (v921-19075845)
#ROBOMERGE-CONFLICT from-shelf
[CL 19146451 by sebastien lussier in ue5-main branch]
This is the start of volumetric support in the path tracer, so a lot of basic infrastructure had to be put into place, making this changelist fairly large. Some shuffling of light parameters had to be done to make room for the volumetric scattering multiplier.
The integration strategy is to use null tracking (with Spectral MIS) for choosing a random scatter point along the ray. This point is chosen similarly to transparent hits, so surface and volume shading are unified. However, these volume hits are chosen proportionally to transmittance times scattering which is not optimal for lights embedded in the volume. To handle the latter, we also allow the main trace call to return a volume segment over which we can compute the direct lighting from local light sources. To simplify the handling of overlapping media and inter-mixed transparent hits, we stochastically select a ray segment. From this point on, we can evaluate direct lighting using equi-angular sampling. The MIS combination of equi-angular sampling and null tracked spectral density sampling was prototyped but found not to bring any improvement for the currently implemented volume types. This will likely have to be revisited in the future.
To improve quality and reduce the amount of ray-marching required, the volume API allows for analytic implementations of transmittance for cases where this is possible to do more efficiently than by ray-marching.
Implementation details for Atmosphere: This volume type is a planet sized, spherically symetric model. Because the default units in UE are centimeters, objects like the planet that are kilometer sized will run into all sorts of numerical precision artifacts. To solve this, an implementation of double-word arithmetic was added which allows enough decimal digits to robustly intersect the planet, as well as cary out the lookups required. The Transmittance is cached in a lookup table indexed by height above the ground and viewing angle cosine. This is similar to the LUT used by the realtime version but with a different parameterization which covers the full range of angles/heights with high precision. This lookup table is automatically baked on demand when atmosphere parameters change. The volumetric sky model is only used when "reference atmosphere" is enabled in the post process volume. This is because the existing approached (cached into a skylight) is generally a bit faster and supports clouds. This toggle may be removed as the support for volumes matures.
Implementation details for ExponentialHeightFog: This volume is represented as a finite slab centered around the camera. Transmittance is easily computed analytically for this volume. We only add this volume when the volumetric fog checkbox is enabled, as the default parameters are not fully physically based. We limit the fog to be present only within a certain radius of the camera, to prevent rays from scattering forever.
#preflight 620ad32d583261b0a66af216
#rb Sebastien.Hillaire,Patrick.Kelly
#preflight 620dd2270931bfd925e5936b
[CL 19031069 by chris kulla in ue5-main branch]
- Completely changed the way we evaluate and represent the material topology.
- A topology is now represented as a tree to allow any horizontal/vertical operations (almost like a kdtree).
- This allows for correct translucent material coverage/transmittance evaluation, as well as correct BSDFs luminance weight (finally).
- All the operations to be able to do that are expended in HLSLTranslator instead of letting the compiler handle it (otherwise was too slow, leading to bugs and could not even debug shaders). This means no conditional BSDF packing.
- There is no more Material/layer/BSDFs array: only a Tree or operators, some of them being BSDF at leaves.
- Parameter blending has also been updated: when enabled on a node, it will be forced on all the oprator child / in the sub tree.
- Parameter blending is applied on an inlinedbsdf on the StrataData. When it is the root of the parameter blending subtree, it is promoted to a fully fledged BSDF operator node to allow other regular operators.
- The inlinedbsdf is not longer in premultiplied mode. Horizontal/Vertical/Add parameter blending operators have been updated to reflect that. This allows for simpler/unified/clearer code.
- Vertical blending now uses the new simple "layering of two slabs with uncorelated coverage" math I have RnD'd. See GetVerticalLayeringInfo in ShadingCommon.ush.
Follow up to this CL:
- CompileStrataBlendFunction should use parameter blending.
- Add more compilation debug output
- Update material vizualisation
- Fix Rough refraction
- Fix decals
- Fix debug probes
- Pack FStrataBSDF
- Move all StrataCompilationInfoCreateSingleBSDFMaterial into into the strata tree and rework cost evaluation and material LOD
- Expand graph visit such as UpdateBSDFWeightAfterOperatorVisit to be specific to operators for the compiler to have to do less search to inline what we know already
- Fix STRATA_TODO: operation using parameter blending are not actually discarded so they still occupy a pot in the opeartion arrays in the compiler and in the shader also.
#rb charles.derousiers
#preflight https://horde.devtools.epicgames.com/job/620a5f5d803d9066e67de938
#fyi charles.derousiers
[CL 18993400 by Sebastien Hillaire in ue5-main branch]
This changes allow to bind a larger amount of rect. light texture, and allows future support for rect light in forward & cluster passes.
#jira none
#rb sebastien.hillaire
#preflight 61fcebd0b5092d45ad110db4
[CL 18861192 by Charles deRousiers in ue5-main branch]
Move visualize enum/defines to shared header file.
Rename a few cvars for consistency.
#rb graham.wihlidal
[FYI] ola.olsson
#preflight 61f9d8921d7ca8ed2d628ccd
#ROBOMERGE-AUTHOR: andrew.lauritzen
#ROBOMERGE-SOURCE: CL 18819821 in //UE5/Release-5.0/... via CL 18819825 via CL 18822904
#ROBOMERGE-BOT: UE5 (Release-Engine-Test -> Main) (v910-18824042)
[CL 18825054 by andrew lauritzen in ue5-main branch]
- Store Position and Bounds in translated world space
- Not using TilePosition+RelativeWorldPosition to avoid increasing struct size.
- FPathTracingLight arrays seem to be calculated every frame so shouldn't be a problem.
- LightGrid fully built in translated world space.
- TraceLight/SampleLight/EstimateLight/InitLightPickingCdf now expect rays and positions in translated world space.
#preflight 61f98c744b0bc1c4176461df
#rb chris.kulla
#ROBOMERGE-AUTHOR: tiago.costa
#ROBOMERGE-SOURCE: CL 18813277 in //UE5/Release-5.0/... via CL 18813296 via CL 18822754
#ROBOMERGE-BOT: UE5 (Release-Engine-Test -> Main) (v910-18824042)
[CL 18824289 by tiago costa in ue5-main branch]
Ran Strata and non Strata code path.
QAGame cooked on Swi.
ShoorterGame cooked and ran on PC.
#rb charles.derousiers
#preflight https://horde.devtools.epicgames.com/job/61d47489db0309127dfae2a2
#ROBOMERGE-AUTHOR: sebastien.hillaire
#ROBOMERGE-SOURCE: CL 18509133 in //UE5/Release-5.0/... via CL 18509155
#ROBOMERGE-BOT: STARSHIP (Release-Engine-Staging -> Release-Engine-Test) (v899-18417669)
[CL 18509183 by sebastien hillaire in ue5-release-engine-test branch]
We always compile the tile vertex shader and this includes strata.ush. We will fix compilation issue there when we get to it.
#jira https://jira.it.epicgames.com/browse/UE-137866
#rb none
#preflight https://horde.devtools.epicgames.com/job/61d42e9f932a02483cb8d000
[FYI] charles.derousiers
#ROBOMERGE-AUTHOR: sebastien.hillaire
#ROBOMERGE-SOURCE: CL 18506891 in //UE5/Release-5.0/... via CL 18506899
#ROBOMERGE-BOT: STARSHIP (Release-Engine-Staging -> Release-Engine-Test) (v899-18417669)
[CL 18506901 by sebastien hillaire in ue5-release-engine-test branch]
Implement tiled dispatch in the path tracer to reduce the likelyhood of GPU timeouts when rendering at high resolution. This also reduces the memory requirements for path state when running with path compaction enabled.
Change from a uint buffer to a structured buffer for storing path states which gives a small speedup.
Add indirect dispatch support to launch less work for compacted bounces (off by default as it does not seem to provide a speedup so far)
#jira TM-6595
#rb Juan.Canada
#preflight 61b27c6a2b48d03df526ce85
#preflight 61b28773ee0de9822e0f02de
#ROBOMERGE-AUTHOR: chris.kulla
#ROBOMERGE-SOURCE: CL 18426885 in //UE5/Main/...
#ROBOMERGE-BOT: STARSHIP (Main -> Release-Engine-Test) (v897-18405271)
#ROBOMERGE[STARSHIP]: UE5-Release-Engine-Staging Release-5.0
[CL 18426911 by chris kulla in ue5-release-engine-test branch]
To maintain consistency with the rasterizer implementation, keep track of background visibility (1.0 where there is background, 0.0 where there are solid objects). MRQ will flip this quantity before outputing to disk.
#rb Juan.Canada
#preflight 6149fccbe594c90001d9c558
#ROBOMERGE-AUTHOR: chris.kulla
#ROBOMERGE-SOURCE: CL 17584574 in //UE5/Main/...
#ROBOMERGE-BOT: STARSHIP (Main -> Release-Engine-Test) (v871-17566257)
[CL 17584582 by chris kulla in ue5-release-engine-test branch]
