* PipelineLibrary (PSO Caching) support for Metal
- API changes to handle Metal drivers implictly doing save/load of PipelineLibrary data
- Fixed up code related to Metal device selection
- PipelineLibrary support for Mac and ios
Signed-off-by: moudgils <moudgils@amazon.com>
* Fix compile errors for Dx12, Vulkan backend + Unit tests
Signed-off-by: moudgils <47460854+moudgils@users.noreply.github.com>
* Fixed errors related to M1 GPU
Signed-off-by: moudgils <moudgils@amazon.com>
* Fix a minor 'tab' validation issue
Signed-off-by: moudgils <moudgils@amazon.com>
* Addressed feedback
Signed-off-by: moudgils <moudgils@amazon.com>
* Minor feedback
Signed-off-by: moudgils <moudgils@amazon.com>
* Added a few asserts
Signed-off-by: moudgils <47460854+moudgils@users.noreply.github.com>
* Fix a typo
Signed-off-by: moudgils <moudgils@amazon.com>
* Updated all array_view uses with the C++20 span.
The updates were done in the following order
1. `AZStd::array_view<([^>].+)\* ?>` -> `AZStd::span<\1 const>`
2. `AZStd::array_view<(?:const )(.+)>` -> `AZStd::span<const \1>`
3. `AZStd::array_view` -> `AZStd::span`
Removed the implementation of array_view.
Signed-off-by: lumberyard-employee-dm <56135373+lumberyard-employee-dm@users.noreply.github.com>
* Added missing whitespace between `const` and the typename for spans.
Updated the ShaderTest comparison of the ShaderResourceGroupLayout span
to compare the sizes as well
Updated comments on some of the methods that stated that they return "an
array" to mention they return "a span".
Signed-off-by: lumberyard-employee-dm <56135373+lumberyard-employee-dm@users.noreply.github.com>
- Each resource type is tracked and updated separately
- Added caching ability for Raytracing srg to save ~2ms for a scene containing 100 x 50 vegetation patch
Signed-off-by: moudgils <moudgils@amazon.com>
This change is a preparation for moving the CPU profiler/visualization system from Atom into its own Gem by removing the dependency on local time tracking object AZ::RHI::CpuTimingStatistics
Full changes include:
- Removed all usage of AZ::RHI::CpuTimingStatistics
-- Replaced with pushing to AZ::Statistics::StatisticalProfilerProxy global instance
- Promoted VariableTimer from AZ::RHI to AZ::Debug
- Removed now unused CpuTimingStatistics.h
Signed-off-by: AMZN-ScottR 24445312+AMZN-ScottR@users.noreply.github.com
* Intial attempt to convert the Atom/RHI/FrameScheduler to use the new TaskGraph api
Signed-off-by: rgba16f <82187279+rgba16f@users.noreply.github.com>
* Avoid enqueuing work on the active task thread if the submitted task
graph is waitable
When submitting a task graph, supplying a wait event implies that
dependent jobs must occur on threads that do not wait on the event (in
the absence of work stealing). This change prevents this by adding a
notion of a task thread enable/disable state, and prohibiting dependent
jobs from being enqueued on waiting threads.
Signed-off-by: Jeremy Ong <jcong@amazon.com>
* Convert RPI/Scene to use TaskGraph pass 1, Culling jobs remain on the old system
Signed-off-by: rgba16f <82187279+rgba16f@users.noreply.github.com>
* RemoveTask Graph changes from the FrameScheduler::ExecuteGroups, use old job system instead
Signed-off-by: rgba16f <82187279+rgba16f@users.noreply.github.com>
* Per review, removing commented out code
Signed-off-by: rgba16f <82187279+rgba16f@users.noreply.github.com>
* Cleanup debug code, & build fix
Signed-off-by: rgba16f <82187279+rgba16f@users.noreply.github.com>
* Add a cvar & interface to query whether to use jobs or task graph
Signed-off-by: rgba16f <82187279+rgba16f@users.noreply.github.com>
* Make TaskGraph assert if you try to wait inside a job
Signed-off-by: rgba16f <82187279+rgba16f@users.noreply.github.com>
* Fix TaskTest SpawnSubgraph to account for the new TaskGraphEvent assert on wait in a running task
Signed-off-by: rgba16f <82187279+rgba16f@users.noreply.github.com>
* 3 minor cleanups. 1) Events always store a ptr to their executor 2) Fix clang compile error 3) remove an early out.
Signed-off-by: rgba16f <82187279+rgba16f@users.noreply.github.com>
* Fix double group end that was causing assert/crash plus misc minor diff's with development
Signed-off-by: rgba16f <82187279+rgba16f@users.noreply.github.com>
* Fix deallocation failure on deactivation of the TaskGraphSystemComponent. Also make the system component account for multiple creation in Unit Tests.
Signed-off-by: rgba16f <82187279+rgba16f@users.noreply.github.com>
* Update with PR feedback
1) Rename UseTaskGraph to IsTaskGraphActive & update related code
2) prefer TaskExecutor::SetInstance
3) add comments and remove commented out code
Signed-off-by: rgba16f <82187279+rgba16f@users.noreply.github.com>
* Fix incorrect RTTI name for TaskGraphActiveInterface
Signed-off-by: rgba16f <82187279+rgba16f@users.noreply.github.com>
* Move TaskGraphSystemComponent CRC calculation to a shared variable
Signed-off-by: rgba16f <82187279+rgba16f@users.noreply.github.com>
Co-authored-by: Jeremy Ong <jcong@amazon.com>
- Removed all unnecessary includes to Atom CpuProfiler.h
- Added includes to AzCore Profiler.h where necessary
Signed-off-by: AMZN-ScottR 24445312+AMZN-ScottR@users.noreply.github.com
- Added new interface type AZ::Debug::Profiler to externally register profiler systems
- Modified the Atom CPU profiler to register as an AzCore profiler
-- This allows full engine markers to be visualized in the associated ImGui tool
- Converted all AZ_ATOM_PROFILE_* macros to use AZ_PROFILE_* macros instead
Signed-off-by: AMZN-ScottR 24445312+AMZN-ScottR@users.noreply.github.com
- Fixed error in iOS profile non-unity build error: unused variable 'INVALID_OFFSET'. The global variable is static const only visible to the cpp, removed since it's not used.
- Fixed error in windows release 'timeoutSeconds': unreferenced formal parameter
Signed-off-by: moraaar <moraaar@amazon.com>
* Release build fix for Windows
* Release build fix for Android
* Release build fix for Windows
* Release build fix for Android
* Release build fix for Linux
* Release build fix for Mac
* Release build fix for iOS
Signed-off-by: Steve Pham <spham@amazon.com>
NOTE: The memory driller is still intact for now to avoid needing to
modify allocators, but the frame/cpu portions of driller and the
standalone executable are now gone.
Signed-off-by: Jeremy Ong <jcong@amazon.com>
* Various fixes for AtomStarterGame on ios
- Use low end pipeline on ioos by default for BoootStrapComponent
- Track the need to bind null heap within Argument buffers
- Only bind the null heap if its needed
- Track its usage for Vertex/Fragment stages
- Increase null dummy buffer to 1K to address GPU crash oon thee first frame
Signed-off-by: moudgils <moudgils@amazon.com>
* Moved PlatformLimits to setreg. Removed Device PostInit. Some clean up.
Signed-off-by: jiaweig <jiaweig@amazon.com>
* Move setreg loading the PlatformLimitsDescriptor super class.
Signed-off-by: jiaweig <jiaweig@amazon.com>
* Apply same implementation for fake device used in unit tests
Signed-off-by: jiaweig <jiaweig@amazon.com>
* Add implementation for Null renderer. Swap order for register RHI interface in initialization.
Signed-off-by: jiaweig <jiaweig@amazon.com>
* Move back setreg from PlatfromLimitsDescriptor to Device, due to Linux dependency issue.
Signed-off-by: jiaweig <jiaweig@amazon.com>
* Changed the function to take in RHI backend name
Signed-off-by: jiaweig <jiaweig@amazon.com>
* Metal sync interval support
Add shader name to SRGPool for debugging purposes
Set default value of rpi_vsync_interval to 1
Signed-off-by: moudgils <moudgils@amazon.com>
* Update the Mac Editor plist application bundle identifier to be different than Lumberyard
Fix the editor to have runtime dependencies on the scene modules.
Signed-off-by: rgba16f <82187279+rgba16f@users.noreply.github.com>
* Fix Metal RHI calling unsupported timestamp query api on M1 Macs.
Add Query for Timestamps supported on device initialization.
Signed-off-by: rgba16f <82187279+rgba16f@users.noreply.github.com>
* Remove AZ::Scene* runtime dependencies from EmotionFX gem and add AZ::SceneUI as a runtime dependency of SceneProcessing::Editor module.
Signed-off-by: rgba16f <82187279+rgba16f@users.noreply.github.com>
Trace::Error: Failed to load dynamic library at path
Fixes the following harmless errors that occur when running
AssetProcessorBatch on Linux:
1. Module: Unable to locate required entry point
'InitializeDynamicModule' within module
'/GIT/o3de/build/linux_profile/bin/profile/libAtom_RHI_DX12.Builders.so'.
2. Module: Unable to locate required entry point
'InitializeDynamicModule' within module
'/GIT/o3de/build/linux_profile/bin/profile/libAtom_RHI_Metal.Builders.so'.
Signed-off-by: galibzon <garrieta@amazon.com>
* Final update copyright headers to reference license files at the repo root
Signed-off-by: spham <spham@amazon.com>
* Fix copyright validator unit tests to support the stale O3DE header scenario
Signed-off-by: spham <spham@amazon.com>
* Fix mip writes via compute passes by creates subresource views
- Create metal subresource views so that we can write into correct mips for compute passes
- Remove code to set the correct slice/level for gfx passes as the view will be correct now
- Handle cubemap/cubemaparray/3d texture types as the drivers do not allow subresource views.
Fix writing to the correct mip in Metal
Set default m_outputScale to 1 in order to fix a 1/0 error
Add support for logging/printing errors pertaining to GPU crashes.
Setting Release queue's collection latency to MaxFrames.
Fix managed mem synchronization related offset bug
* Enable 16 byte alignment for root constants
* Switching to R8G8B8A8_UINT for dummy stream buffers
* Clamping min values for scissor/viewport for fullscreen pass
* - Fix the second window related tabbing issue
- Merge calls to UseResoources acrooaa all the SRGs
- Move SamplerCache to the device to ensure only one cache to reduce duplication
- Fixes to compute threading numbers getting reset to 0,0,0
- Cleanup withing BufferPoolResolver
- Argument buffers are now queued to be cleaned up upon shutdown
moudgils
lumberyard-employee-dm
rgba16f
Qing Tao
Esteban Papp
galibzon
Scott Romero
moraaar
Steve Pham
Jeremy Ong
jiaweig
pappeste
dmcdiar
Chris Burel
moudgils