Commit Graph

35 Commits

Author SHA1 Message Date
Guillaume Abadie
0734868f53 Implements DynamicRenderScaling API with GPU timing measurement integrated within RDG
This allows to define a new dynamic scaling in the renderer with low amount of boiler plate:

DynamicRenderScaling::FHeuristicSettings GetDynamicTranslucencyResolutionSettings()
{
	RenderingDynamicScaling::FHeuristicSettings BucketSetting;
	BucketSetting.Model = RenderingDynamicScaling::EHeuristicModel::Quadratic;
	BucketSetting.bModelScalesWithPrimaryScreenPercentage = true;
	BucketSetting.MinResolutionFraction = ...
	...
	return BucketSetting;
}

DynamicRenderScaling::FBudget GDynamicTranslucencyResolution(TEXT("DynamicTranslucencyResolution"), &GetDynamicTranslucencyResolutionSettings);


And then simply define a scope to measure the GPU timing as such:

{
	DynamicRenderScaling::FRDGScope DynamicTranslucencyResolutionScope(GraphBuilder, GDynamicTranslucencyResolution);

	// add passes to GraphBuilder
}

#rb zach.bethel
#jira UE-152561
#preflight 628f1219bb14235aa38c904c

[CL 20376428 by Guillaume Abadie in ue5-main branch]
2022-05-26 01:58:36 -04:00
zach bethel
e37a08177c Submit async compute work prior to a fence from graphics -> async compute so that prior work isn't grouped with the submission to work around an issue in D3D12, which currently brute forces a fence from graphics to async compute any time graphics work is submitted.
#preflight 627c359a9f7ad2a14b8388d0
#fyi sebastien.hillaire

[CL 20152556 by zach bethel in ue5-main branch]
2022-05-11 18:40:35 -04:00
john huelin
95c1680b6e Fix DrawCallCount per pass not working when multithreaded RHI command recording is enabled.
[REVIEW]
#rnx

#ROBOMERGE-OWNER: john.huelin
#ROBOMERGE-AUTHOR: john.huelin
#ROBOMERGE-SOURCE: CL 20105654 via CL 20105671 via CL 20107113 via CL 20107146
#ROBOMERGE-BOT: UE5 (Release-Engine-Staging -> Main) (v943-19904690)

[CL 20110543 by john huelin in ue5-main branch]
2022-05-09 16:35:01 -04:00
zach bethel
4eedc02f37 Fixes to external access RDG feature to handle async compute without validation failures.
#preflight 6270551191629533ec2b4bc9

[CL 20017395 by zach bethel in ue5-main branch]
2022-05-02 18:31:37 -04:00
zach bethel
f457a69101 Added RHI tracked access API to remove Unknown transitions.
- New RHI command list SetTrackedAccess method for the user to supply a current whole-resource state.
 - New RHI command context GetTrackedAccess method for querying the tracked access in RHIBeginTransitions / RHIEndTransitions on the RHI thread.
 - Hooked RHICmdList.Transition and FRHICommandListExecutor::Transition to assign tracked state automatically.
 - Refactored RDG and resource pools to use new RHI tracking.
      - FRDGPooledBuffer / FRDGPooledTexture no longer contain tracked state. RDG temp-allocates state through the graph allocator instead.
      - All prologue transitions are 'Unknown', and all epilogue transitions coalesce into a whole resource state.
 - Implemented platform support for patching the 'before' state with the tracked state.
 - Implemented various RHI validation checks:
      - Asserts that the user assigned tracked state matches RHI validation tracked state, for all subresources.
      - Asserts that tracked state is not assigned or queried from a parallel translation context.
 - Added FRHIViewableResource and FRHIView base classes to RHI. FRHIView contains a pointer to an FRHIViewableResource. This is currently a raw pointer, but should be extended to a full reference in a later CL.

NOTE on RHI thread constraint:

Transition evaluation is now restricted to the RHI thread (i.e. no parallel translation contexts). Transitions aren't performed in parallel translate contexts anyway, so this is not a problem. If, however, we decide to refactor parallel translation to be more general, this implementation could be extended to track the state per context and update from the 'dispatch' thread.

#preflight 6233b4396666d7e753a16aaf
#rb kenzo.terelst

[CL 19513316 by zach bethel in ue5-main branch]
2022-03-25 11:19:10 -04:00
aleksander netzel
9a8aa76713 Add ERDGPassFlags::NeverParallel to allow RDG passes to stay on render thread.
Both Immediate/AsyncCompute passes are always executed in order on the render thread but other RDG passes can run in parallel.
With my previous change, RayTracingScene pass was changed to Compute/AsyncCompute to have the correct order on the GPU.
But when it was set to Compute (when AsyncCompute was disabled) it was running in parallel with passes it depends on which led to random crashes.
Because the dependency between passes was implicit, we needed a new flag to express the intent and get the desired order between execution lambdas.

On behalf of Zach Bethel.

#rb zach.bethel
#jira UE-141003
#preflight 61fb0d9e923ac18db7015160
#lockdown michal.valient

#ROBOMERGE-AUTHOR: aleksander.netzel
#ROBOMERGE-SOURCE: CL 18841970 in //UE5/Release-5.0/... via CL 18841984 via CL 18842341
#ROBOMERGE-BOT: UE5 (Release-Engine-Test -> Main) (v910-18824042)

[CL 18842374 by aleksander netzel in ue5-main branch]
2022-02-03 09:15:49 -05:00
zach bethel
2f790c0cc8 Removed r.RenderTargetPool.Events visualizer. It has been replaced by RDG insights.
#rb none
#preflight 61e6f8bb837b79f7cced82ee

#ROBOMERGE-AUTHOR: zach.bethel
#ROBOMERGE-SOURCE: CL 18642485 in //UE5/Release-5.0/... via CL 18644124 via CL 18644421
#ROBOMERGE-BOT: UE5 (Release-Engine-Test -> Main) (v900-18638592)

[CL 18644664 by zach bethel in ue5-main branch]
2022-01-18 14:18:07 -05:00
jason hoerner
144df66f84 UE5_RELEASE: MGPU, include GPU mask per pass in DumpGPU. The GPU mask is added at the front of the pass name, making it easy to visually see in the pass browser, and also to text search for globally. The mask is only added when running on a system with multiple GPUs present (very few users), and is enabled by default for discoverability.
#rb Guillaume.Abadie
#rnx
#jira none
#preflight 61d83b345d522c9b40a54579

#ROBOMERGE-AUTHOR: jason.hoerner
#ROBOMERGE-SOURCE: CL 18541852 in //UE5/Release-5.0/... via CL 18541856
#ROBOMERGE-BOT: STARSHIP (Release-Engine-Staging -> Release-Engine-Test) (v899-18417669)

[CL 18541859 by jason hoerner in ue5-release-engine-test branch]
2022-01-07 08:18:56 -05:00
guillaume abadie
dd1c4774d5 Improves DumpGPU command
Dumps improvements:
1) Bring up for consoles
2) Better out of memory resiliency during the dumping process
3) Dumps console variables in CSV
4) Dumps process' log after completion
5) Dumps mip chains through FDumpTextureCS compute shader
6) Dumps depth & stencil texture formats through the FDumpTextureCS compute shader
7) Dumps at draw granularity with FRDGBuilder::DumpDraw(); (experimental)
8) Dumps final png screenshot to the dump directory
9) Adds & Dumps the FRDGBufferDesc::Metadata for viewer to decode buffer binary automatically
10) Dumps the PassParameters with structure metadata to decode shader parameters automatically
11) Adds CTRL+SHIFT+/ shortcut

Viewer improvements:
1) Tips display onload to spread some knowledge to the user
2) Supports for opening any pass/resources in new web browser tab
3) Emulates 16 and 32 bits UINT texture visualization with multiple webgl 8bit UINT textures
4) Fixes the webpage's tab going out of memory after visualizing many large resources.
5) Fixes the webpage's tab going out of memory after loading large buffer.
6) Adds support for more texture format with RGB channel reswizzling
7) Implements UI color-sheme based on UE5's editor theme
8) Implements texel color picker capabable of decoding every pixel format.
9) Implements texture viewer zooming with the mouse wheel
10) Implements a r.DumpGPU.Viewer.Visualize to open a specific RDG output resource when opening the viewer

#rb juan.canada
#preflight 619bb638fa0b360c406c42c5
[FYI] juan.canada, zach.bethel

#ROBOMERGE-AUTHOR: guillaume.abadie
#ROBOMERGE-SOURCE: CL 18260079 via CL 18372399 via CL 18372914
#ROBOMERGE-BOT: STARSHIP (Release-Engine-Staging -> Release-Engine-Test) (v895-18170469)

[CL 18373039 by guillaume abadie in ue5-release-engine-test branch]
2021-12-03 16:04:00 -05:00
christopher waters
5953546bfb Adding RHIUniformBufferLayoutInitializer and RHIUniformBufferLayout is a proper FRHIResource. This brings proper RHI lifetime support to RHIUniformBufferLayout so we don't delete the layouts while RHI commands that use them are still in flight.
This should fully address UniformBufferLayout/UniformBuffer crashes that randomly happen during shader compiling in the Editor.

#jira none
#rb arciel.rekman, ben.ingram, mihnea.balta, stu.mckenna, will.damon
#preflight 611eb6c6008be90001f8b031

#ROBOMERGE-SOURCE: CL 17243608 in //UE5/Main/...
#ROBOMERGE-BOT: STARSHIP (Main -> Release-Engine-Test) (v855-17104924)

[CL 17243630 by christopher waters in ue5-release-engine-test branch]
2021-08-19 23:56:04 -04:00
zach bethel
89fc4cf833 Parallel RDG execution improvements.
- Added ERDGBuilderFlags::AllowParallelExecute to tag specific builders to attempt parallel execution. This avoids cases where small graphs fork tasks and end up causing contention. Only the main scene render graphs are tagged.
 - Moved RHI transition creation to an async task.
 - Moved parallel execute setup and dispatch to an async task.
 - Fixed RDG draining asserts using a short-term workaround by tagging relevant scene textures as non-transient.
 - Deprecated RDG AddPass utilities without names and fixed up last remnants.
 - Enabled parallel RDG execution by default.

[FYI] christopher.waters

#ROBOMERGE-SOURCE: CL 16925941 in //UE5/Main/...
#ROBOMERGE-BOT: STARSHIP (Main -> Release-Engine-Test) (v836-16769935)

[CL 16925957 by zach bethel in ue5-release-engine-test branch]
2021-07-22 12:43:00 -04:00
zach bethel
5ec769f91e RDG Parallel Execution (disabled by default)
- Refactored RDG to support free-threaded execution of passes.
 - Refactored renderer to use specific RHI command list variants in pass lambda. Immediate command list passes are forced to stay on the render thread, while other variants can be parallelized.

#rb christopher.waters

#ROBOMERGE-SOURCE: CL 16838717 in //UE5/Main/...
#ROBOMERGE-BOT: STARSHIP (Main -> Release-Engine-Test) (v836-16769935)

[CL 16838724 by zach bethel in ue5-release-engine-test branch]
2021-07-13 12:38:37 -04:00
zach bethel
db4d19f641 Fix for culled RDG uniform buffer being created.
#rb andrew.lauritzen

[CL 16523561 by zach bethel in ue5-main branch]
2021-06-01 17:25:18 -04:00
zach bethel
b843dabf06 Minor RDG fixes and support for read-only system textures.
[CL 16456650 by zach bethel in ue5-main branch]
2021-05-25 17:11:34 -04:00
zach bethel
90bc2efada RDG refactor to enable draining of work after issuing occlusion queries.
- New Drain() method on FRDGBuilder; will flush all pending work.
 - Drained passes are not culled; resource lifetimes are extended; async compute fences are optimized as best as possible but fence joining may occur after the drain.
 - Batch up and pre-build all resource transitions. This is a prerequisite for parallel command lists.
 - Removed ServiceLocalQueue passes with built-in RDG AddDispatchHint().

#jira UE-114622

[CL 16393495 by zach bethel in ue5-main branch]
2021-05-19 17:54:58 -04:00
zach bethel
804d775535 Minor RDG optimizations.
- Simplified texture subresource tracking.
 - Removed map lookup for each resource in SetupPass.
 - Improved Compile / CollectPassResources to reduce cache misses.
 - Added some container reservations to reduce reallocation costs.
 - Added snapping of buffers to page boundaries to improve re-use.

#rb none

[CL 16208311 by zach bethel in ue5-main branch]
2021-05-05 11:58:15 -04:00
zach bethel
8c1793ad5a Integration of RHI transient allocator into RDG.
- Replaced legacy transient support from RDG and replaced with new API.
 - Reworked acquire / discard operations a bit and added RHI validation to track correctness.
 - Reworked RDG barrier batching to include acquire / discard operations.
 - Hardened render pass merging logic and expanded to support lifetime extension of transient resources.
 - Added transient tag to RDG insights to track which resources are transient.

#rb luke.thatcher, kenzo.terelst

[CL 15726534 by zach bethel in ue5-main branch]
2021-03-17 12:44:59 -04:00
zach bethel
26dddad59e Fixed clang build break.
#rb trivial

[CL 15466212 by zach bethel in ue5-main branch]
2021-02-18 20:28:08 -04:00
zach bethel
3054b6c388 Added tag tracking whether a pass uses the immediate command list to RDG insights.
#rnx
#rb none

[CL 15465640 by zach bethel in ue5-main branch]
2021-02-18 19:44:41 -04:00
zach bethel
50d368cbf4 Improved sorting and coloration of resources in RDG insights.
#rb none
#rnx

[CL 15338594 by zach bethel in ue5-main branch]
2021-02-05 12:11:16 -04:00
zach bethel
274c3b2ba7 RDG Insights Plugin
#jira none
#rb catalin.dragoiupruna

[CL 15296346 by zach bethel in ue5-main branch]
2021-02-03 13:17:04 -04:00
zach bethel
0a145b23fe Improved RDG state merging to better handle subresources. Subresource tracking is used for the rest of the graph if subresource states are encountered. Improved error reporting for edge case where transitions are inserted out of order. Added explicit merge check for depth read / write to make sure that it doesn't merge with anything else.
#rb none

[CL 14874576 by zach bethel in ue5-main branch]
2020-12-07 18:45:31 -04:00
zach bethel
7ae3cdb31f Implemented support for multi-pipe UAV access. Implemented command list stats on RDG builder to avoid helper passes. Added {Begin, End}UAVOverlap to RDG UAVs.
#rb none

[CL 14821597 by zach bethel in ue5-main branch]
2020-11-30 13:27:08 -04:00
zach bethel
4e959189a6 Reworking RDG allocator to include container memory. This removes the dependency on the thread local memstack, and MemMark can no longer release RDG memory. All RDG memory is now released after execution.
#rb christopher.waters

[CL 14783232 by zach bethel in ue5-main branch]
2020-11-18 18:25:03 -04:00
zach bethel
5630bcfd89 Added Multi-pipeline Transition Support to platforms and RDG. Fixed uniform buffer hack in composition lighting.
#rb luke.thatcher, kenzo.terelst
#jira none

[CL 14727873 by zach bethel in ue5-main branch]
2020-11-11 19:22:36 -04:00