Fixes GPU crashes when releasing Indirect draw and query related dx12 objects
Fixes imgui profiler related to viewing buffer allocations
Signed-off-by: moudgils <moudgils@amazon.com>
* Fixed log2 shaper equations. Added bspline sampling for lut. Added options for custom log2 or linear lut with custom exposure ranges.
Signed-off-by: Ken Pruiksma <pruiksma@amazon.com>
* Added support for PQ shaper. Added shader option & cvar for lut sampling quality. Fixed issues in the blend lut shader that were causing considerable quality loss. No longer always changing to the log2 1000 nit shaper when blending luts - if the source luts all use the same shaper, keep using that shaper in the blended lut.
Signed-off-by: Ken Pruiksma <pruiksma@amazon.com>
* Fixed an integer -> float
Signed-off-by: Ken Pruiksma <pruiksma@amazon.com>
* Minor PR reveiw updates
Signed-off-by: Ken Pruiksma <pruiksma@amazon.com>
* Profiler: implement loading from saved capture
Adds functionality for finding a saved capture on disk and then
deserializing it using rapidjson's built-in buffered stream reader. This
does require use of raw file pointers since saved captures can be
hundreds of megabytes.
Actually showing the data in the visualizer is TODO.
* Profiler: use heap buffer over stack buffer
* Profiler: move deserialization logic to ImGuiCpuProfiler
Signed-off-by: Jacob Hilliard <jhlliar@amazon.com>
* Profiler: Implement continuous capture internals
Signed-off-by: Jacob Hilliard <jhlliar@amazon.com>
* Profiler: extend ProfileCaptureSystemComponent
Implements dumping of saved CPU profiling data to a local file, blocking
call.
Signed-off-by: Jacob Hilliard <jhlliar@amazon.com>
* Profiler: Working IO thread for serialization
Signed-off-by: Jacob Hilliard <jhlliar@amazon.com>
* Profiler: switch to AZ::JobFunction for IO
Signed-off-by: Jacob Hilliard <jhlliar@amazon.com>
* Profiler: move to a ring buffer for storage
Signed-off-by: Jacob Hilliard <jhlliar@amazon.com>
* Profiler: switch back to IO thread
Signed-off-by: Jacob Hilliard <jhlliar@amazon.com>
* Profiler: add TODO
Signed-off-by: Jacob Hilliard <jhlliar@amazon.com>
* Profiler: correct thread safety issues
Signed-off-by: Jacob Hilliard <jhlliar@amazon.com>
* Add CaptureCpuFrameTime method to ProfilingCaptureSystemComponent for monitoring CPU performance.
Signed-off-by: Cynthia Lin <cyntlin@amazon.com>
* ly_test_tools: Refactor benchmark data aggregator in preparation for CPU frame times.
Signed-off-by: Cynthia Lin <cyntlin@amazon.com>
* performance benchmarks: Aggregate and report CPU frame times based on JSON data.
Signed-off-by: Cynthia Lin <cyntlin@amazon.com>
* AutomatedTesting: Capture CPU frame time in AtomFeatureIntegrationBenchmark.
Signed-off-by: Cynthia Lin <cyntlin@amazon.com>
Repro steps:
- Create two entities.
- Entity 1
- Add a mesh component and assign a model with multiple sub-meshes
- Add a material component. The material component looks correct.
- Entity 2
- Add a mesh component and assign the same model as the other entity
- Add a material component. The material component shows "<unknown>" for all material slot names
The problem was that ReflectedPropertyEditor creates a new Asset<> reference with the correct ID but does not load it. This asset is passed to EditorMaterialComponent, MaterialComponentController, and MeshFeatureProcessor and none of these tell the Asset to load. The MeshFeatureProcessor was not loading the Asset or connecting to the AssetBus because the instance already existed in the InstanceDatabse so from the FP's perspecive there was no need. But for the FP's GetModelAsset() API to function correctly it needs to have the asset initialized to the available AssetData pointer. So we updated the MeshFeatureProcessor to always connect to the AssetBus so it will find the available AssetData via the OnAssetReady callback.
Signed-off-by: santorac <55155825+santorac@users.noreply.github.com>
I had to update the ModelLod class to take in both the ModelLodAsset and ModelAsset for initialization so it can fetch the slots for each mesh.
Signed-off-by: santorac <55155825+santorac@users.noreply.github.com>
Here we inttroduce a new struct ModelMaterialSlot which formalizes the concept of material slot, with an ID, display name, and default material assignment. The ID still comes from the MaterialUid like before. The display name is built-in, rather than being parsed out from the asset file name. And the default material assignment can be any material asset, it doesn't have to come from the FBX (or other scene file).
This commit is just the preliminary set of changes. Cursory testing shows that it works pretty well but more testing is needed (and likely some fixes) before merging.
Here is what's left to do...
Add serialization version converters to preserve prior prefab data.
See if we can get rid of GetLabelByAssetId function only rely on the display name inside ModelMaterialSlot.
I'm not sure if the condition for enabling the "Edit Material Instance..." context menu item is correct.
Test actors
Lots more testing in general
Signed-off-by: santorac <55155825+santorac@users.noreply.github.com>
* Expose shadow bias to component & feature processors. Shadow bias now works more consistently with various near / far shadow planes and caster positions. Bias now also affects esm shadows which helps eliminate acne in certain situations.
Signed-off-by: Ken Pruiksma <pruiksma@amazon.com>
* Adding jira comment to light configuration serialization version. Improved comment on final adjustment to bias before its sent to the shader.
Signed-off-by: Ken Pruiksma <pruiksma@amazon.com>
* Hooking up bias to behavior context.
Signed-off-by: Ken Pruiksma <pruiksma@amazon.com>
* fix: Correct typo in profiling capture system.
Signed-off-by: Cynthia Lin <cyntlin@amazon.com>
* Move benchmark data aggregation from ASV into ly_test_tools module.
Signed-off-by: Cynthia Lin <cyntlin@amazon.com>
Added an initialization delay before starting the DiffuseProbeGrid texture readbacks, to allow the textures to settle.
Signed-off-by: dmcdiar <dmcdiar@amazon.com>