* Added Async APIs for the various Process*FromList terrain functions.
Please note that we are currently defaulting the number of worker threads to one, because splitting the work over multiple threads causes contention when locking various mutexes, resulting in slower overall wall time for async requests split over multiple threads vs one where all the work is done on a single thread. The latter is still preferable over a regular synchronous call because it is just as quick and prevents the main thread from blocking. This should be changed once the mutex contention issues have been addressed, so that async calls automatically split the work between available job manager worker threads, unless the ProcessAsyncParams specify a different desired number of jobs.
Signed-off-by: bosnichd <bosnichd@amazon.com>
* Fix Linux builds by adding missing #include
Signed-off-by: bosnichd <bosnichd@amazon.com>
* Added a test for cancellation of terrain async requests, and fix it so that it works.
Note that the benchmarks show this implementation to be slightly slower than the previous one, which I presume is because we're now calling a 'perSurfacePointFunction' in the inner loop; this can probably be addressed, but will result in a lot of code duplication, and I think efforts will be better spent on removing the mutex contention to enable running multiple terrain async jobs at the same time.
Signed-off-by: bosnichd <bosnichd@amazon.com>
* Added Async versions for all Process*Region terrain API functions, along with benchmarks.
Signed-off-by: bosnichd <bosnichd@amazon.com>
* Fix the newly added terrain async request benchmarks to actually use the async APIs.
Signed-off-by: bosnichd <bosnichd@amazon.com>
* Revert to the original version which just calls the synchronous API from the job function, along with some other updates in response to review feedback.
Signed-off-by: bosnichd <bosnichd@amazon.com>
* Change the TerrainWorldDebugger to use the async API, along with the following changes:
- TerrainJobContext no longer uses a JobCancelGroup so we can guarantee the completion callbacks of associated jobs will be invoked even if it is cancelled.
- As a result of the above change, the ProcessAsyncCompleteCallback function signature again accepts the associated TerrainJobContext as a param.
- The TerrainProcessAsyncCancellation test has been resurrected and simplified by using binary semaphores instead of condition variables.
- All the async related TerrainSystemBenchmark functions have been simplified by using binary semaphores instead of condition variables.
- Global cancellation of all terrain jobs on deactivation of the TerrainSystem has been reintroduced, but in a different way than before.
- Other miscellaneous changes/fixes made while testing and based on earlier PR feedback.
Signed-off-by: bosnichd <bosnichd@amazon.com>
* Updates based on review feedback:
- Go back to using a vector instead of an array (fixed the original problem by adding custom copy/assignment constructors/operators to the WireframeSector struct).
- When calling WireframeSector::Reset, block until any associated in flight has completed.
- Added the concept of a minimum number of positions per terrain job.
Signed-off-by: bosnichd <bosnichd@amazon.com>
* Use semaphore instead of binary_semaphore in a bunch of places to account for the race condition where a completion callback fires before we started waiting for it.
Signed-off-by: bosnichd <bosnichd@amazon.com>
* Misc SurfaceData Optimizations.
This includes a few different optimizations found while trying to make the bulk query APIs faster:
* Switches mutexes over to shared_lock to optimize for the multi-reader-single-writer pattern
* Surface provider point creation now uses a pre-created set of masks to initialize with, and uses std::move() to move the created point into the output list instead of copying it.
* Splits CombineSortAndFilterNeightboringPoints so that the FilterPoints() can occur separately and efficiently with erase/remove_if, and avoids making a copy of the output points.
* Optimized SurfaceDataShapeComponent::ModifySurfacePoints
* Fixed potential bug where the sort wasn't stable since it only compared the Z value, and could have produced unexpected results for differing points with the exact same Z value.
* Fixed up a couple small bugs and missing checks in the unit tests
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
* Fixed syntax on unit tests.
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
* Making terrain query resolution a single float instead of a Vector2
Signed-off-by: Ken Pruiksma <pruiksma@amazon.com>
* Keeping the concept of different x/y step sizes in region queries since that may be useful and is separate from query resolution. Also keeping the concept of different x/y step sizes in physics since that's independent of the terrain gem.
Signed-off-by: Ken Pruiksma <pruiksma@amazon.com>
* Formatting cleanups
Signed-off-by: Ken Pruiksma <pruiksma@amazon.com>
* A few more minor cleanups
Signed-off-by: Ken Pruiksma <pruiksma@amazon.com>
* Added support to convert serialized Vector2 query resolution to a single float.
Signed-off-by: Ken Pruiksma <pruiksma@amazon.com>
* Switch ray intersection check back to using separate values for x and y resolution
Signed-off-by: Ken Pruiksma <pruiksma@amazon.com>
* Fixing new unit tests added to use float query resolution.
Signed-off-by: Ken Pruiksma <pruiksma@amazon.com>
* Updating automated test
Signed-off-by: Ken Pruiksma <pruiksma@amazon.com>
* Initial cleanup of GetSurfacePointsFromRegion in prep for bulk API support.
* Removed the generated lookup point from the output structure. Nothing was using it, and by keeping it separate, I can pass it in as a list of points that can be passed throughout the terrain, gradient, and surface data APIs.
* Clarified on the SurfaceProvider bus that GetSurfacePoints() only gets valid XY values on the inPosition.
* Simplified the TerrainSurfaceDataSystemComponent implementation a bit. The EnumerateHandlers() and the terrain Aabb checks were overkill. Also, the terrain Aabb check assumed that the Z value on the inPosition was valid, which it isn't always.
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
* Remove CryCommon dependency.
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
These perform bulk queries of the underlying gradients which is much faster than performing a lot of individual queries.
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
* Bumped up terrain world limit to allow up to (and including) 4096 x 4096.
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
* Changed loop calculations to handle floating-point math better.
By looping on floating-point values, query resolutions of unstable values like "0.200000007" would sometimes cause the loop to go one more time than it should.
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
* Change flow so that TerrainSystem stops responding during deactivation.
Some systems might accidentally try to call back to the TerrainSystem inside a DestroyBegin notification, so make sure it stops listening before sending out the notification.
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
* Change gradients to cache and use a GradientTransform instance.
In my local test case, calling EBus on every call took 337 ms, using a lambda to wrap the calls took 197 ms, and using the fully cached version took 170 ms.
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
* Remove the wrappingTransform function and go back to the switch statement.
There was a bit of overhead to each function call due to using AZStd::function that just isn't necessary for this use case.
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
* Add profile markers to the heightfield updates so that they're more visible.
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
* Cleared state while component is deactivated.
The state was getting refreshed even while the component was in a deactivated state, which meant that it wasn't properly notifying of state changes when it became active since it wasn't detecting an actual change. By clearing the state when deactivated, and ensuring the state isn't getting refreshed *while* deactivated, the notifications work properly.
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
* Fixed compile warning on unit test.
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
* Addressed PR feedback - changed comments, reduced mutex scope
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
* Optimized wireframe drawing.
As a part of rearranging the code to make use of the upcoming ProcessHeightsFromRegion, the number of calls to GetHeight could be reduced, dropping the refresh time in my test case from 2550 ms to 1068 ms.
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
* Change refresh to only happen on wireframe draws.
This helps ensure that multiple data changes in a single frame don't cause multiple refreshes, and prevents us from taking a refresh penalty when the wireframe isn't being drawn at all.
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
* Change debugger to update wireframes incrementally.
This works by having a fixed NxN grid of sectors. The camera is always considered as being in the center of the grid, so anytime the camera's grid square changes, only a subset of sectors are updated to have the new wireframe data.
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
* Bugfix - sector vertex count was 4x too high.
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
* Bugfixes & comments.
Fixed initial "mark dirty" refresh - the dirty region Z value needed to be ignored.
Added copious comments for the math & logic, and simplified some of the math.
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
* Small update to comment for better readability.
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
* Addressed PR feedback.
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
* Removed "chatty" profile markers.
These three markers were making terrain refreshes take ~200% longer in profile builds.
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
* Terrain surface data wasn't updating its AABB correctly.
The code was making an invalid assumption that changes to the world bounds would result in an invalid dirtyRegion. It actually gets a valid region encapsulating the old and new bounds. The fix is an additional check to see if the terrain bounds changed.
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
* Addressed previous PR feedback that terrainSystem ought to be a local variable in each test.
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
Use Ground Plane non functional
Added functionality to the "Use Ground Plane" toggle. Now, when it's enabled, the Terrain Layer Spawner will always provide valid terrain data at the minimum height of the spawner box, even if no Terrain Height Gradient List exists. If it's disabled, it won't provide any terrain data unless the Terrain Height Gradient List exists.
The "Use Ground Plane" toggle is now functional. When disabled, the terrain layer spawner will say "terrain exists = false" for any point in its bounds unless there's also a Terrain Height Gradient List component with a valid entry. When enabled, it will always say "terrain exists = true", and it will return the min height of the spawner box as the ground plane if there's no valid Terrain Height Gradient List height provider.
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
If a terrain layer spawner went outside the world bounds, the debug wireframe wouldn't update correctly because the heights were outside the wireframe sector AABBs. Adjusted the logic to account for this.
Signed-off-by: Mike Balfour <82224783+mbalfour-amzn@users.noreply.github.com>
Mike Balfour
bosnichd
Sergey Pereslavtsev
jiaweig
Ken Pruiksma
SergeyAMZN
Chris Galvan
amzn-sj
nemerle
sphrose
Junbo Liang
John Jones-Steele