From: Josh Simmons <josh@nega.tv>
Date: Sat, 1 Jun 2024 14:36:49 +0000 (+0200)
Subject: shark: Improve performance of binning shader
X-Git-Url: https://git.nega.tv//gitweb.cgi?a=commitdiff_plain;h=483d031bca775327fd9db2b9c8bba2c1ff0011ba;p=josh%2Fnarcissus

shark: Improve performance of binning shader
---

diff --git a/title/shark-shaders/shaders/primitive_2d_bin.comp.glsl b/title/shark-shaders/shaders/primitive_2d_bin.comp.glsl
index 894ebb5..277d272 100644
--- a/title/shark-shaders/shaders/primitive_2d_bin.comp.glsl
+++ b/title/shark-shaders/shaders/primitive_2d_bin.comp.glsl
@@ -10,21 +10,32 @@
 
 #include "primitive_2d.h"
 
+layout(std430, set = 0, binding = 3) readonly coherent buffer glyphInstanceBufferCoherent {
+    GlyphInstance glyph_instances_coherent[];
+};
+
+#define SUBGROUP_SIZE 64
+#define NUM_PRIMS_WG (SUBGROUP_SIZE * 32)
+
 // TODO: Spec constant support for different subgroup sizes.
-layout (local_size_x = 64, local_size_y = 1, local_size_z = 1) in;
+layout (local_size_x = SUBGROUP_SIZE, local_size_y = 1, local_size_z = 1) in;
 
-shared uint bitmap_0[64];
+shared uint bitmap_0[SUBGROUP_SIZE];
 
 void main() {
     const uvec2 bin_coord = gl_GlobalInvocationID.yz;
     const uvec2 bin_min = bin_coord * TILE_SIZE * 8;
     const uvec2 bin_max = min(bin_min + TILE_SIZE * 8, primitive_uniforms.screen_resolution);
 
-    for (uint i = 0; i < 2048; i += 64) {
-        const uint prim_index = gl_WorkGroupID.x * 2048 + i + gl_SubgroupInvocationID;
+    for (uint i = 0; i < NUM_PRIMS_WG; i += gl_SubgroupSize.x) {
+        const uint prim_index = gl_WorkGroupID.x * NUM_PRIMS_WG + i + gl_SubgroupInvocationID;
         bool intersects = false;
         if (prim_index < primitive_uniforms.num_primitives) {
-            intersects = test_glyph(prim_index, bin_min, bin_max);
+            const GlyphInstance gi = glyph_instances_coherent[prim_index];
+            const Glyph gl = glyphs[gi.index];
+            const vec2 glyph_min = gi.position + gl.offset_min;
+            const vec2 glyph_max = gi.position + gl.offset_max;
+            intersects = !(any(lessThan(bin_max, glyph_min)) || any(greaterThan(bin_min, glyph_max)));
         }
         const uvec4 ballot = subgroupBallot(intersects);
         bitmap_0[i / 32 + 0] = ballot.x;
@@ -33,54 +44,39 @@ void main() {
 
     memoryBarrierShared();
 
-    uint bitmap_1[2];
-    {
-        const uvec4 ballot = subgroupBallot(bitmap_0[gl_SubgroupInvocationID] != 0);
-        bitmap_1[0] = ballot.x;
-        bitmap_1[1] = ballot.y;
-    }
-
-    for (uint y = 0; y < 8; y++) {
-        for (uint x = 0; x < 8; x++) {
-            const uvec2 tile_coord = gl_GlobalInvocationID.yz * 8 + uvec2(x, y);
-            const uvec2 tile_min = tile_coord * TILE_SIZE;
-            const uvec2 tile_max = min(tile_min + TILE_SIZE, primitive_uniforms.screen_resolution);
-            [[branch]]
-            if (any(greaterThanEqual(tile_min, tile_max))) {
-                continue;
-            }
-
-            const uint tile_index = tile_coord.y * primitive_uniforms.tile_stride + tile_coord.x;
-
-            for (uint i = 0; i < 2; i++) {
-                uint out_1 = 0;
-
-                uint word_1 = bitmap_1[i];
-                while (word_1 != 0) {
-                    const uint bit_1 = findLSB(word_1);
-                    word_1 ^= word_1 & -word_1;
-
-                    uint out_0 = 0;
-                    uint index_0 = i * 32 + bit_1;
-                    uint word_0 = bitmap_0[index_0];
-                    while (word_0 != 0) {
-                        const uint bit_0 = findLSB(word_0);
-                        word_0 ^= word_0 & -word_0;
-
-                        const uint prim_index = gl_WorkGroupID.x * 2048 + index_0 * 32 + bit_0;
-                        if (test_glyph(prim_index, tile_min, tile_max)) {
-                            out_0 |= 1 << bit_0;
-                        }
+    const uint x = gl_SubgroupInvocationID.x & 7;
+    const uint y = gl_SubgroupInvocationID.x >> 3;
+    const uvec2 tile_coord = gl_GlobalInvocationID.yz * 8 + uvec2(x, y);
+    const uvec2 tile_min = tile_coord * TILE_SIZE;
+    const uvec2 tile_max = min(tile_min + TILE_SIZE, primitive_uniforms.screen_resolution);
+
+    if (all(lessThan(tile_min, tile_max))) {
+        const uint tile_index = tile_coord.y * primitive_uniforms.tile_stride + tile_coord.x;
+
+        for (uint i = 0; i < 2; i++) {
+            uint out_1 = 0;
+
+            for (uint j = 0; j < 32; j++) {
+                uint out_0 = 0;
+                uint index_0 = i * 32 + j;
+                uint word_0 = bitmap_0[index_0];
+                while (word_0 != 0) {
+                    const uint bit_0 = findLSB(word_0);
+                    word_0 ^= word_0 & -word_0;
+
+                    const uint prim_index = gl_WorkGroupID.x * NUM_PRIMS_WG + index_0 * 32 + bit_0;
+                    if (test_glyph(prim_index, tile_min, tile_max)) {
+                        out_0 |= 1 << bit_0;
                     }
+                }
 
-                    if (out_0 != 0) {
-                        out_1 |= 1 << bit_1;
-                    }
+                if (out_0 != 0) {
+                    out_1 |= 1 << j;
                     tile_bitmap_wo[tile_index * TILE_STRIDE + TILE_BITMAP_L0_OFFSET + gl_WorkGroupID.x * 64 + index_0] = out_0;
                 }
-
-                tile_bitmap_wo[tile_index * TILE_STRIDE + TILE_BITMAP_L1_OFFSET + gl_WorkGroupID.x * 2 + i] = out_1;
             }
+
+            tile_bitmap_wo[tile_index * TILE_STRIDE + TILE_BITMAP_L1_OFFSET + gl_WorkGroupID.x * 2 + i] = out_1;
         }
     }
 }