Textures, FrameBuffers and VertexBuffers allocated by the Sdl2 Renderer were only being released via finalizers. This could lead to OpenGL out of memory errors since resources may not be cleaned up in a timely manner. To avoid this, IDisposable has been implemented and transitively applied to classes that use these resources.
As a side-effect some static state is no longer static, particularly in Renderer, in order to facilitate this change and just for nicer design in general.
Also dispose some bitmaps.
This enables the map importer to map the .ini video definitions to ours.
The mapping generally is as follows:
Intro => BackgroundInfo
Brief => Briefing
Action => GameStart
Win => GameWon
Lose => GameLost
An issue in some Red Alert maps means that this mapping is not always
quite correct. In those maps that do not have a 'Brief' video defined
(scg03a is an example), Westwood has assigned the video that should
probaby have been the 'Action' video to the 'Intro' slot instead. I can
only assume that that was done due to some limitation in the original
game code. Mappers will have to correct that assignment manually in
those cases.
Reinitializing the initial cell info layer for the path search took a fair bit of time. We cache this initial setup so it only has to be done each time the map size changes. A CopyValuesFrom method in CellLayer is provided which copies values between layers by just copying the internal arrays for super speed.
This speeds up InitCellInfo 10x.
The buffers in SequenceProvider can be freed if Preload is called, since we know everything is loaded. A SequenceProvider is created for each TileSet is use so this saves memory for however many tilesets had been used in the game. This will be at least one for the shellmap, and often more.
The MapCache loading thread is kept alive for 5 seconds after it last generated a map (in anticipation of more requests). Once this time expires the thread is allowed to die, as it is unlikely there will be more requests in the short term. At this time it is ideal to force the changes to be committed to the texture so we can release the buffer. As well as marking the buffer for release, we must access the texture to force the changes stored in the buffer to be written to the texture, after which the buffer can be reclaimed.
Additionally, when starting the MapCache loading thread we must ensure the buffer is created from the main thread since it may query the texture object which has thread affinity. After that the buffer may be modified freely on the loading thread until released.
The index value needs only be big enough to handle all defined terrain types. This is a low number so we can save memory by defining it as a byte. This particularly saves memory for the CustomTerrain field in the Map class, which defines a cell layer for the map using tile indexes, so we can reduce the size of that layer 4x as a result.
This may fix issue #5916.
In any case, it's wanted because this kind of sort is "unstable".
According to the docs:
"This implementation performs an unstable sort; that is, if two
elements are equal, their order might not be preserved."
- Only update shroud within the visible screen area, rather than the whole map. This improves performance on larger maps significantly when scrolling around since large portions of the shroud do not need to be updated.
- Provide methods in Shroud to return delegates to check for explored/visibility for tiles within a certain region. This allows it to return more efficient delegates whenever the region is within the map bounds, or shroud/fog is disabled. In the typical case where the region is in bounds and shroud/fog is enabled, the fast check is almost twice as fast as the slow check.
- Use the Shroud delegate functions in shroud rendering, frozen actors, minimap rendering and resource layer areas to provide a speedup since these areas of code can often take advantage of the fact they perform checks within the map boundary.
- Cache current element in CellRegionEnumerator to prevent repeated work if the element is accessed more than once.
- Decrease the size of elements in some arrays in hopes of reducing memory needs and improving cache hits.
- Extract an enum for edges rather than using magic numbers for everything.
- Remove duplicated code between FoggedEdges and ShroudedEdges by hosting the visibility function into a delegate.
- Make minimap methods more readable.
- Tidy formatting.
- Make some fields readonly.
- Remove unused usings.
The lookup accounts for ~50-60% of the time spent in GetTerrainIndex and GetTerrainInfo, and these methods themselves can account for up to 1.3% of total CPU used so this is a small but measurable win.
