Use a lower precision buffer for bloom calculation.

Slightly (~100 µs) improves performance of bloom, but might open up more detailed (→less jitter) bloom implementations.
This commit is contained in:
IntegratedQuantum 2024-12-02 20:43:58 +01:00
parent 5f722274b1
commit 1455605387

View File

@ -366,9 +366,9 @@ const Bloom = struct { // MARK: Bloom
if(width != currentWidth or height != currentHeight) {
width = currentWidth;
height = currentHeight;
buffer1.updateSize(width/4, height/4, c.GL_RGB16F);
buffer1.updateSize(width/4, height/4, c.GL_R11F_G11F_B10F);
std.debug.assert(buffer1.validate());
buffer2.updateSize(width/4, height/4, c.GL_RGB16F);
buffer2.updateSize(width/4, height/4, c.GL_R11F_G11F_B10F);
std.debug.assert(buffer2.validate());
}
gpu_performance_measuring.startQuery(.bloom_extract_downsample);