Catcrafts/Crafter.Graphics
2
0
Fork 0
Code Issues 1 Pull requests Projects Releases Packages Wiki Activity Actions

perf(ui): cooperative shared-memory tile culling in UI compute shaders

Browse source 
Each 8×8-tile workgroup previously had all 64 threads walk the entire
item list, re-reading every item from the SSBO 64× and running the
per-pixel reject for items that never touch the tile — O(tiles × N)
item loads dominated by SSBO traffic.

The workgroup now streams the item list in chunks of 64: each thread
loads one item, tests its AABB against the whole tile, and the survivors
are compacted — in original buffer order — into shared memory. Every
thread then runs the unchanged per-pixel accumulate over only those
survivors. This drops SSBO traffic ~64× (each item read once per
workgroup instead of once per pixel) and shrinks the inner loop to the
items that actually overlap the tile.

Draw order is preserved exactly: chunks run in array order and the
in-chunk compaction is a stable in-order scan, so later items still
overdraw earlier ones (no atomic-append nondeterminism). The inner loop
body is byte-for-byte the original per-pixel logic, so output is
pixel-identical — the cull only decides which items reach it. Threads
no longer early-return so every thread reaches the workgroup barriers.

Applied to ui-quads, ui-circles, ui-images and ui-text; shared helpers
(uiTileBounds / uiAabbOverlapsTile) live in ui-shared.glsl.

Resolves #46

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
catbot 2026-06-16 15:45:12 +00:00
commit 6eb11fbf71
5 changed files with 337 additions and 107 deletions
Download patch file Download diff file Expand all files Collapse all files

116
shaders/ui-circles.comp.glsl
View file

@ -2,47 +2,99 @@
#extension GL_GOOGLE_include_directive : enable
#include "ui-shared.glsl"
// One workgroup per 8×8 screen tile. The workgroup cooperatively streams the
// CircleItem list in chunks of 64 (see ui-shared.glsl), culling each chunk
// against the tile and compacting survivors — in buffer order — into shared
// memory; every thread then accumulates over only those survivors.
layout(push_constant) uniform PC {
UIDispatchHeader hdr;
} pc;
layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in;
shared vec4 s_centerRadius[UI_CHUNK];
shared vec4 s_color[UI_CHUNK];
shared vec4 s_outline[UI_CHUNK];
shared uint s_keep[UI_CHUNK];
shared uint s_order[UI_CHUNK];
shared uint s_count;
void main() {
ivec2 screenPx;
if (!uiResolveScreenPixel(pc.hdr, screenPx)) return;
bool valid = uiResolveScreenPixel(pc.hdr, screenPx);
vec4 dst = imageLoad(uiImages[pc.hdr.outImage], screenPx);
vec2 sp = vec2(screenPx) + 0.5;
for (uint i = 0u; i < pc.hdr.itemCount; ++i) {
CircleItem it = LoadCircleItem(pc.hdr.itemBuffer, i);
vec2 center = it.centerRadius.xy;
float radius = it.centerRadius.z;
if (radius <= 0.0) continue;
// Cheap bounding-box reject.
if (abs(sp.x - center.x) > radius + 1.0) continue;
if (abs(sp.y - center.y) > radius + 1.0) continue;
float d = length(sp - center) - radius;
float bodyA = clamp(0.5 - d, 0.0, 1.0);
if (bodyA <= 0.0 && it.outline.x <= 0.0) continue;
vec4 src = vec4(it.color.rgb, it.color.a * bodyA);
if (it.outline.x > 0.0) {
float t = abs(d + it.outline.x * 0.5) - it.outline.x * 0.5;
float outlineA = clamp(0.5 - t, 0.0, 1.0);
src.rgb = mix(src.rgb, it.outline.yzw, outlineA);
src.a = max(src.a, outlineA);
}
if (src.a <= 0.0) continue;
dst = uiBlendOver(dst, src);
vec4 dst = vec4(0.0);
vec2 sp = vec2(0.0);
if (valid) {
dst = imageLoad(uiImages[pc.hdr.outImage], screenPx);
sp = vec2(screenPx) + 0.5;
}
imageStore(uiImages[pc.hdr.outImage], screenPx, dst);
vec2 tileMin, tileMax;
uiTileBounds(tileMin, tileMax);
uint lid = gl_LocalInvocationIndex;
for (uint base = 0u; base < pc.hdr.itemCount; base += UI_CHUNK) {
uint idx = base + lid;
bool keep = false;
if (idx < pc.hdr.itemCount) {
s_centerRadius[lid] = uiCircleHeap[pc.hdr.itemBuffer].items[idx].centerRadius;
s_color[lid] = uiCircleHeap[pc.hdr.itemBuffer].items[idx].color;
s_outline[lid] = uiCircleHeap[pc.hdr.itemBuffer].items[idx].outline;
// Cull box matches the in-loop bounding-box reject (radius + 1px).
float radius = s_centerRadius[lid].z;
if (radius > 0.0) {
vec2 c = s_centerRadius[lid].xy;
vec2 r = vec2(radius + 1.0);
keep = uiAabbOverlapsTile(c - r, c + r, tileMin, tileMax);
}
}
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;
}
barrier();
if (valid) {
for (uint j = 0u; j < s_count; ++j) {
uint c = s_order[j];
vec2 center = s_centerRadius[c].xy;
float radius = s_centerRadius[c].z;
if (radius <= 0.0) continue;
// Cheap bounding-box reject.
if (abs(sp.x - center.x) > radius + 1.0) continue;
if (abs(sp.y - center.y) > radius + 1.0) continue;
float d = length(sp - center) - radius;
vec4 outline = s_outline[c];
float bodyA = clamp(0.5 - d, 0.0, 1.0);
if (bodyA <= 0.0 && outline.x <= 0.0) continue;
vec4 col = s_color[c];
vec4 src = vec4(col.rgb, col.a * bodyA);
if (outline.x > 0.0) {
float t = abs(d + outline.x * 0.5) - outline.x * 0.5;
float outlineA = clamp(0.5 - t, 0.0, 1.0);
src.rgb = mix(src.rgb, outline.yzw, outlineA);
src.a = max(src.a, outlineA);
}
if (src.a <= 0.0) continue;
dst = uiBlendOver(dst, src);
}
}
barrier();
}
if (valid) imageStore(uiImages[pc.hdr.outImage], screenPx, dst);
}