Fix DMA linear texture copy fast path

[gdkchan]

FindTexture tries to find an existing texture that matches the copy parameters, if found we try to write the data into the texture directly which is more efficient. However, FindTexture was also matching textures that have a height higher than the copy height. This causes problems because linear strided -> linear conversion expects data to be at least as large as the texture itself, and in this case it won't be causing an exception.

The fix here is making FindTexture only match linear textures with exactly the copy height. In addition to that, I removed the RegionY parameter because it is not used for linear copies, only block linear ones (that's an incorrect assumption that was made in the past and corrected, but I missed that code). Since after this only two fields from DmaTexture were accessed, I decided to stop passing the whole struct and just pass those 2 parameters to the function, which is more efficient.

Fixes crash in SD Gundam Battle Alliance Demo with the following exception:

Unhandled exception. System.ArgumentOutOfRangeException: Specified argument was out of the range of valid values.
   at Ryujinx.Graphics.Texture.LayoutConverter.ConvertLinearStridedToLinear(Int32 width, Int32 height, Int32 blockWidth, Int32 blockHeight, Int32 lineSize, Int32 stride, Int32 bytesPerPixel, ReadOnlySpan`1 data) in F:\Ryujinx\Ryujinx.Graphics.Texture\LayoutConverter.cs:line 272
   at Ryujinx.Graphics.Gpu.Engine.Dma.DmaClass.DmaCopy(Int32 argument) in F:\Ryujinx\Ryujinx.Graphics.Gpu\Engine\Dma\DmaClass.cs:line 331
   at Ryujinx.Graphics.Gpu.Engine.Dma.DmaClass.LaunchDma(Int32 argument) in F:\Ryujinx\Ryujinx.Graphics.Gpu\Engine\Dma\DmaClass.cs:line 450
   at Ryujinx.Graphics.Gpu.Engine.GPFifo.GPFifoProcessor.Send(UInt64 gpuVa, Int32 offset, Int32 argument, Int32 subChannel, Boolean isLastCall) in F:\Ryujinx\Ryujinx.Graphics.Gpu\Engine\GPFifo\GPFifoProcessor.cs:line 213
   at Ryujinx.Graphics.Gpu.Engine.GPFifo.GPFifoProcessor.Process(UInt64 baseGpuVa, ReadOnlySpan`1 commandBuffer) in F:\Ryujinx\Ryujinx.Graphics.Gpu\Engine\GPFifo\GPFifoProcessor.cs:line 95
   at Ryujinx.Graphics.Gpu.Engine.GPFifo.GPFifoDevice.DispatchCalls() in F:\Ryujinx\Ryujinx.Graphics.Gpu\Engine\GPFifo\GPFifoDevice.cs:line 188
   at Ryujinx.Ui.RendererWidgetBase.<Render>b__71_0() in F:\Ryujinx\Ryujinx\Ui\RendererWidgetBase.cs:line 394
   at Ryujinx.Graphics.GAL.Multithreading.ThreadedRenderer.<>c__DisplayClass48_0.<RunLoop>b__0() in F:\Ryujinx\Ryujinx.Graphics.GAL\Multithreading\ThreadedRenderer.cs:line 100
   at System.Threading.Thread.StartCallback()

For the record, at that time it was trying to write 256x25 data into a 64x29 RGBA8Srgb texture. With this fix, it tries to write it into a 64x25 texture.

[github-actions]

Download the artifacts for this pull request:

• ryujinx-Release-1.1.0+4ba39cd-linux_x64
• ryujinx-Release-1.1.0+4ba39cd-osx_x64
• ryujinx-Release-1.1.0+4ba39cd-win_x64

Experimental GUI (Avalonia)

• ava-ryujinx-Release-1.1.0+4ba39cd-linux_x64
• ava-ryujinx-Release-1.1.0+4ba39cd-osx_x64
• ava-ryujinx-Release-1.1.0+4ba39cd-win_x64

GUI-less (SDL2)

• sdl2-ryujinx-headless-Release-1.1.0+4ba39cd-linux_x64
• sdl2-ryujinx-headless-Release-1.1.0+4ba39cd-osx_x64
• sdl2-ryujinx-headless-Release-1.1.0+4ba39cd-win_x64

Only for Developers

• ava-ryujinx-Debug-1.1.0+4ba39cd-linux_x64
• ava-ryujinx-Debug-1.1.0+4ba39cd-osx_x64
• ava-ryujinx-Debug-1.1.0+4ba39cd-win_x64
• ryujinx-Debug-1.1.0+4ba39cd-linux_x64
• ryujinx-Debug-1.1.0+4ba39cd-osx_x64
• ryujinx-Debug-1.1.0+4ba39cd-win_x64
• sdl2-ryujinx-headless-Debug-1.1.0+4ba39cd-linux_x64
• sdl2-ryujinx-headless-Debug-1.1.0+4ba39cd-osx_x64
• sdl2-ryujinx-headless-Debug-1.1.0+4ba39cd-win_x64

[marysaka]

lgtm