Use a giant UBO to optimize performance in 2D

This removes the countless small UBO writes we had before
and replaces them with a single large write per render pass.

This results in much faster rendering on low-end devices
but improves speed on all devices.
This commit is contained in:
clayjohn
2022-06-20 21:56:26 -07:00
parent 0c23a2cfe3
commit 154b9c1c91
12 changed files with 744 additions and 507 deletions
+3
View File
@@ -1925,6 +1925,9 @@
<member name="rendering/gl_compatibility/driver.windows" type="String" setter="" getter="" default="&quot;opengl3&quot;">
Windows override for [member rendering/gl_compatibility/driver].
</member>
<member name="rendering/gl_compatibility/item_buffer_size" type="int" setter="" getter="" default="16384">
Maximum number of canvas items commands that can be drawn in a single viewport update. If more render commands are issued they will be ignored. Decreasing this limit may improve performance on bandwidth limited devices. Increase this limit if you find that not all objects are being drawn in a frame.
</member>
<member name="rendering/global_illumination/gi/use_half_resolution" type="bool" setter="" getter="" default="false">
If [code]true[/code], renders [VoxelGI] and SDFGI ([member Environment.sdfgi_enabled]) buffers at halved resolution (e.g. 960×540 when the viewport size is 1920×1080). This improves performance significantly when VoxelGI or SDFGI is enabled, at the cost of artifacts that may be visible on polygon edges. The loss in quality becomes less noticeable as the viewport resolution increases. [LightmapGI] rendering is not affected by this setting.
[b]Note:[/b] This property is only read when the project starts. To set half-resolution GI at run-time, call [method RenderingServer.gi_set_use_half_resolution] instead.