Pre-skin: stage the skin palette in shared memory #822

New issue

Closed

opened 2026-10-17 18:10:02 +00:00 by sky_line · 1 comment

sky_line commented

2026-10-17 18:10:02 +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.

c_preskin (~1.1 ms) makes four palette loads per vertex that depend on its weight load. Copying an item's palette (110 x 48 B) into shared memory removes that dependent chain; only pays for items with many vertices. Needs measurement: preskin time split by item size. 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 18:10:02 +00:00

sky_line added this to the Slipstream project

2026-10-17 18:10:02 +00:00

sky_line referenced this issue from a commit

2026-10-22 19:32:55 +00:00

#822: pre-skin stages the skin palette in shared memory (RenderVKPreskinSharedPalette)