Merge pull request 'perf(ui): subgroup prefix-sum stream compaction in the UI compute shaders (#124)' (#145) from claude/issue-124 into master

This commit is contained in:
catbot 2026-06-18 15:27:30 +02:00
commit 9046562a90
5 changed files with 167 additions and 43 deletions

View file

@ -1,5 +1,7 @@
#version 460
#extension GL_GOOGLE_include_directive : enable
#extension GL_KHR_shader_subgroup_basic : enable
#extension GL_KHR_shader_subgroup_arithmetic : enable
#include "ui-shared.glsl"
// One workgroup per 8×8 screen tile. The workgroup cooperatively streams the
@ -18,6 +20,34 @@ shared vec4 s_outline[UI_CHUNK];
shared uint s_keep[UI_CHUNK];
shared uint s_order[UI_CHUNK];
shared uint s_count;
shared uint s_subTotals[UI_CHUNK]; // per-subgroup survivor totals (carry step)
// Stable in-order compaction of one chunk's survivors via a per-subgroup
// exclusive prefix-sum plus a carry across subgroups, so a survivor's slot in
// s_order[] equals the number of survivors with a smaller local index — buffer
// (draw) order preserved. Replaces the old lane-0 serial scan. See
// ui-fused.comp.glsl for the full rationale; the body is identical.
void uiCompactChunk() {
uint lid = gl_LocalInvocationIndex;
uint keep = s_keep[lid];
uint subPrefix = subgroupExclusiveAdd(keep);
uint subTotal = subgroupAdd(keep);
if (subgroupElect())
s_subTotals[gl_SubgroupID] = subTotal;
barrier();
uint base = 0u;
uint total = 0u;
for (uint s = 0u; s < gl_NumSubgroups; ++s) {
uint t = s_subTotals[s];
if (s < gl_SubgroupID) base += t;
total += t;
}
if (lid == 0u) s_count = total;
if (keep != 0u)
s_order[base + subPrefix] = lid;
}
void main() {
ivec2 screenPx;
@ -56,13 +86,7 @@ void main() {
s_keep[lid] = keep ? 1u : 0u;
barrier();
if (lid == 0u) {
uint n = 0u;
uint lim = min(UI_CHUNK, pc.hdr.itemCount - base);
for (uint k = 0u; k < lim; ++k)
if (s_keep[k] != 0u) s_order[n++] = k;
s_count = n;
}
uiCompactChunk();
barrier();
if (valid) {