Forward alpha / particles / captures: use the froxel light grid #817

New issue

Open

opened 2026-10-17 18:10:01 +00:00 by sky_line · 0 comments

sky_line commented

2026-10-17 18:10:01 +00:00

Owner

Context: measured on a GTX 1080 (Pascal) at 4K with the test-control channel and the shader clock (RenderVKShaderClockLighting) / pipeline stats (RenderVKPipelineStats). The frame is GPU-bound but not bandwidth-bound (memory controller ~36% busy on the 1080, ~14% on a 5090): it is latency / occupancy bound, so trading a little memory traffic for fewer registers and fewer dependent loads should help every GPU generation.

Once the froxel grid lands: alpha, particles and probe / mirror captures switch from the coarse 16x9x24 clusters (cap 64, lossy on overflow) to the exact froxel masks. Must use the light-only froxels, not the RenderVKLightGrid tile masks (those drop lights by opaque-depth bins, wrong for alpha in front of the opaque surface). Context: measured on a GTX 1080 (Pascal) at 4K with the test-control channel and the shader clock (RenderVKShaderClockLighting) / pipeline stats (RenderVKPipelineStats). The frame is GPU-bound but not bandwidth-bound (memory controller ~36% busy on the 1080, ~14% on a 5090): it is latency / occupancy bound, so trading a little memory traffic for fewer registers and fewer dependent loads should help every GPU generation.

sky_line added the

labels

2026-10-17 18:10:01 +00:00

sky_line added this to the Slipstream project

2026-10-17 18:10:01 +00:00

sky_line referenced this issue

2026-10-17 21:14:16 +00:00

Probe captures: capture-only forward shader (no RT/SSR/mirror/debug) + light grid #829