- Some static variables made it tricky, so added some support to make it simpler on users of the static variables
#rb David.Harvey
[CL 31431760 by josh adams in 5.4 branch]
- D3D12 PC Bindless needs descriptor heaps managed on the CPU; we cannot update them on the GPU timeline.
- Each context now has a FD3D12ContextBindlessState that contains the per-context GPU descriptor heap as well as descriptor rollbacks to apply to the heap before submission.
- We have to roll descriptors back to their values before the incoming view updates were applied so that we can leverage the CPU heap copy at all times. This isn't deferring the updates; it's storing the values before the updates and making sure the heap is used with those values.
- When a context encounters a draw/dispatch and there were any descriptor updates, the previously used heap is updated with the correct set of descriptors before a new heap is created for the subsequent draws/dispatches.
Additional changes
- Allowing RHIs to override FRHITextureReference for custom bindless implementations.
- Adding FRHIDescriptorAllocator::GetAllocatedRange to allow managers to find the smallest range of descriptors that need to be copied to new heaps.
- DescriptorCache now holds reference counted pointers to the bindless heaps to avoid potential use after free scenarios.
- Adding ED3D12DescriptorHeapFlags to mirror D3D12_DESCRIPTOR_HEAP_FLAGS while adding new flags.
- Adding ability to pool descriptor heaps to avoid high OS overhead when constantly allocating new ones.
- Pooling descriptor heaps required more descriptor heap managers to implement CleanupResources.
#jira UE-162014
#rb Luke.Thatcher
[CL 30183702 by christopher waters in ue5-main branch]
- Adding EShaderCodeFeatures::Barycentrics
- Adding GRHIGlobals.SupportsBarycentricsSemantic
- Adding FDataDrivenShaderPlatformInfo::GetSupportsBarycentricsIntrinsics to control COMPILER_SUPPORTS_BARYCENTRIC_INTRINSICS
- Adding FDataDrivenShaderPlatformInfo::GetSupportsBarycentricsSemantic to control PLATFORM_SUPPORTS_BARYCENTRICS_SEMANTIC
#jira UE-193429
#rb graham.wihlidal, mihnea.balta
[CL 29771745 by christopher waters in ue5-main branch]
OpenGL ES and Metal use framebuffer fetch.
Vulkan uses dual source blending.
For Vulkan and OpenGL ES there is a fallback shader permutation for drivers that don't support this. The fallback is the same as the existing solution that uses regular blending (i.e. looks different).
Others uses dual source blending and we force use of DXC for those shaders.
#rb Dmitriy.Dyomin, Florin.Pascu
[CL 29245271 by florian penzkofer in ue5-main branch]
This functionality is now well covered by Lumen reflections which are the supported path going forward.
#jira UE-198247
#rb Aleksander.Netzel,Yuriy.ODonnell
[CL 28866153 by chris kulla in ue5-main branch]
- Added support for up to 32 samplers to the D3D12 RHI.
- Added 'MaxSamplers' to ShaderPlatform in DataDrivenPlatformInfo, this is defaulted to 16 for all shader platforms and can be modified the the DDPI ini files. This value will set the shader compiler define 'PLATFORM_MAX_SAMPLERS'.
- Added a 'RequiredSamplersSwitch' material editor node to branch based on a given shader platform's maximum sampler count.
- To support more than 16 samplers on feature level SM5 platforms Dxc and sm6.0 are used to compile the shader. On platforms that don't support Dxc and sm6.0 the preivew menu item will be disabled.
#jira UE-191404
#rb florin.pascu
[CL 27850133 by michael wanderson in ue5-main branch]
Allocating and freeing slots can be contended, getting and setting a slot value is fast but has an additional indirection compared with OS TLS. Used only for windows editor for now, because it's a bit slower than OS TLS implementation.
#rb dmytro.vovk, francis.hurteau
[CL 27827939 by andriy tylychko in ue5-main branch]
Platforms that do support uniform buffer objects can now provide batched primitive data through UBO. There is a limit UBO range that can be accessed in shaders, so we group instances into batches that fit into this limit. Switch uses 64KB views, other platfroms16KB views. For each primitive we allocate 512Bytes and 256Bytes for instance. Mobile platforms that do not support UBO will use a desktop version of GPUScene.
There are a few things that still missing: Dynamic mesh passes, static lighting
#rb ola.olsson, benjamin.rouveyrol
[CL 26354848 by marc audy in ue5-main branch]
* This should be used to guard any features that require explicitly accessing buffer or textures SRVs in vertex shaders
#rb Dmitriy.Dyomin
#rb Kenzo.Terelst
[CL 26207274 by yuriy odonnell in ue5-main branch]
