Files
dorf/full.patch
T
popertots 4c14f79cd2 perf: eliminate HashMap lookups on z-scroll with fast recolor + targeted boundary dirty
Three optimizations for z-scroll performance:

Option 1 — Fast in-place recolor (no HashMap lookups):
- TilemapChunkStates stores solid_bits + underground_bits per chunk
- recolor_chunk_for_depth iterates flat Vec<Option<TileData>> — pure cache-friendly sequential reads
- tile_fade_color computes color from tileset_idx + z_diff without any HashMap access
- vs repopulate_chunk_tiles: 32K Vec reads vs 32K HashMap lookups per frame

Option 2 — Targeted boundary dirty (from 9×chunks to 2×chunks):
- on_camera_z_changed now only marks the 4 boundary z-levels:
  old_entered, old_exited, new_entered, new_exited
- Registry.z_change_keys tracks z-changed keys separately from occlusion dirty_keys
- populate_tilemap_chunk_data uses fast recolor for z_change_keys,
  full repopulate for dirty_keys (occlusion — rare)

Option 3 — Separate budgets:
- z_change_keys processed with MAX_POPULATE_PER_FRAME budget (fast recolor)
- dirty_keys processed with MAX_POPULATE_PER_FRAME budget (full repopulate)
- Both pipelines tracked separately in benchmark

Also:
- populate_chunk_tiles now returns underground_bits for state storage
- compute_solid_bits helper extracts solid tiles from TileData Vec
- spawn_tilemap_chunks computes and stores both bitmasks on spawn
- despawn_tilemap_chunks clears all state for despawned chunks
2026-03-20 15:21:19 +00:00

202 lines
7.3 KiB
Diff

diff --git a/config.toml b/config.toml
index 9e6f89b..fbb1c42 100644
--- a/config.toml
+++ b/config.toml
@@ -1,4 +1,4 @@
-initial_chunk_radius = 5
+initial_chunk_radius = 15
[spawn_counts]
dorfs = 5
diff --git a/src/world/mod.rs b/src/world/mod.rs
index 9698168..f597c34 100644
--- a/src/world/mod.rs
+++ b/src/world/mod.rs
@@ -27,6 +27,7 @@ impl Plugin for WorldPlugin {
.init_resource::<tiles::TilemapBenchmark>()
.init_resource::<tiles::TilemapChunkRegistry>()
.init_resource::<tiles::TilemapChunkSpawner>()
+ .init_resource::<tiles::PreviousZIndex>()
.add_message::<GenerateChunkEvent>()
.add_message::<ChunkTerrainEvent>()
.add_message::<ChunkWeatheringAndPrecipitationEvent>()
diff --git a/src/world/tiles/benchmark.rs b/src/world/tiles/benchmark.rs
index bac3570..8541a71 100644
--- a/src/world/tiles/benchmark.rs
+++ b/src/world/tiles/benchmark.rs
@@ -37,6 +37,11 @@ pub struct TilemapBenchmark {
pub dirty_keys_total: usize,
/// CPU memory estimate for tile data in MB.
pub tile_data_mb: f64,
+
+ pub last_z_change_dirty_ms: f64,
+ pub last_z_change_dirty_count: usize,
+ pub last_z_change_populate_ms: f64,
+ pub z_change_history_ms: Vec<f64>,
}
impl Default for TilemapBenchmark {
@@ -53,6 +58,10 @@ impl Default for TilemapBenchmark {
dirty_keys_last: 0,
dirty_keys_total: 0,
tile_data_mb: 0.0,
+ last_z_change_dirty_ms: 0.0,
+ last_z_change_dirty_count: 0,
+ last_z_change_populate_ms: 0.0,
+ z_change_history_ms: Vec::with_capacity(100),
}
}
}
@@ -69,7 +78,7 @@ impl TilemapBenchmark {
let p50_idx = (self.frame_times.len() as f32 * 0.50) as usize;
let p99_idx = (self.frame_times.len() as f32 * 0.99) as usize;
let mut sorted = self.frame_times.clone();
- sorted.sort();
+ sorted.sort_by(|a, b| a.partial_cmp(b).unwrap());
let p50 = sorted
.get(p50_idx.min(sorted.len().saturating_sub(1)))
.copied()
@@ -107,6 +116,35 @@ impl TilemapBenchmark {
);
println!("║ Tile data mem: {:.2} MB", self.tile_data_mb);
println!("║ Cumulative: {:.2}s", self.populate_ms / 1000.0);
+ println!("╠══════════════════════════════════════════════════════════════╣");
+ let z_p50 = if !self.z_change_history_ms.is_empty() {
+ let mut sorted = self.z_change_history_ms.clone();
+ sorted.sort_by(|a, b| a.partial_cmp(b).unwrap());
+ sorted[sorted.len() / 2]
+ } else {
+ 0.0
+ };
+ let z_p95 = if !self.z_change_history_ms.is_empty() {
+ let mut sorted = self.z_change_history_ms.clone();
+ sorted.sort_by(|a, b| a.partial_cmp(b).unwrap());
+ let idx = ((sorted.len() as f32) * 0.95) as usize;
+ sorted[idx.min(sorted.len().saturating_sub(1))]
+ } else {
+ 0.0
+ };
+ println!("║ Z-scroll spike:");
+ println!(
+ "║ Last dirty marking: {:.2}ms ({} keys)",
+ self.last_z_change_dirty_ms, self.last_z_change_dirty_count
+ );
+ println!(
+ "║ Last z-change populate: {:.2}ms",
+ self.last_z_change_populate_ms
+ );
+ println!(
+ "║ Z-change populate p50: {:.2}ms p95: {:.2}ms",
+ z_p50, z_p95
+ );
println!("╚══════════════════════════════════════════════════════════════╝");
println!();
}
diff --git a/src/world/tiles/tilemap_chunk.rs b/src/world/tiles/tilemap_chunk.rs
index 6074dc5..bf8f459 100644
--- a/src/world/tiles/tilemap_chunk.rs
+++ b/src/world/tiles/tilemap_chunk.rs
@@ -50,6 +50,17 @@ impl Default for TilemapChunkSpawner {
}
}
+pub const MAX_POPULATE_PER_FRAME: usize = 512;
+
+#[derive(Resource)]
+pub struct PreviousZIndex(pub i32);
+
+impl Default for PreviousZIndex {
+ fn default() -> Self {
+ Self(0)
+ }
+}
+
impl TilemapChunkSpawner {
pub fn queue_chunk(&mut self, chunk_pos: IVec2) {
for z in 0..=(Z_TOTAL as usize) {
@@ -284,18 +295,27 @@ pub fn populate_tilemap_chunk_data(
if registry.dirty_keys.is_empty() {
return;
}
+
+ let was_z_change = bench.last_z_change_dirty_count > 0;
let now = Instant::now();
let camera_z = z_index.0 as i32;
- let current_z = (camera_z as f32 + Z_BELOW) as usize;
+ let current_z_index = (camera_z as f32 + Z_BELOW) as usize;
+
+ let mut to_process: Vec<ChunkLayerKey> = registry.dirty_keys.iter().cloned().collect();
+ to_process.sort_by_key(|k| (k.z_index as i32 - current_z_index as i32).unsigned_abs());
+ to_process.truncate(MAX_POPULATE_PER_FRAME);
+
+ for key in &to_process {
+ registry.dirty_keys.remove(key);
+ }
- let dirty: Vec<ChunkLayerKey> = registry.dirty_keys.drain().collect();
let mut processed = 0usize;
- for key in dirty {
+ for key in to_process {
let Some(&entity) = registry.entities.get(&key) else {
continue;
};
- let Ok((mut tile_data, chunk_key)) = chunk_data_query.get_mut(entity) else {
+ let Ok((mut tile_data, _)) = chunk_data_query.get_mut(entity) else {
continue;
};
@@ -320,15 +340,48 @@ pub fn populate_tilemap_chunk_data(
bench.dirty_keys_total += processed;
bench.tile_data_mb = (registry.entities.len() as f64 * (CHUNK_SIZE * CHUNK_SIZE) as f64 * 4.0)
/ (1024.0 * 1024.0);
+
+ if was_z_change {
+ bench.last_z_change_populate_ms = elapsed;
+ bench.z_change_history_ms.push(elapsed);
+ if bench.z_change_history_ms.len() > 100 {
+ bench.z_change_history_ms.remove(0);
+ }
+ }
}
-pub fn on_camera_z_changed(z_index: Res<ZIndex>, mut registry: ResMut<TilemapChunkRegistry>) {
- if z_index.is_changed() {
- let keys: Vec<ChunkLayerKey> = registry.entities.keys().cloned().collect();
- for key in keys {
+pub fn on_camera_z_changed(
+ z_index: Res<ZIndex>,
+ mut registry: ResMut<TilemapChunkRegistry>,
+ mut prev_z: ResMut<PreviousZIndex>,
+ mut bench: ResMut<super::TilemapBenchmark>,
+) {
+ if !z_index.is_changed() {
+ return;
+ }
+
+ let old_camera_z = prev_z.0;
+ let new_camera_z = z_index.0 as i32;
+ prev_z.0 = new_camera_z;
+
+ let now = Instant::now();
+ let mut count = 0;
+
+ let keys: Vec<ChunkLayerKey> = registry.entities.keys().cloned().collect();
+ for key in keys {
+ let tile_z = camera_z_for_z_index(key.z_index);
+ let old_diff = old_camera_z - tile_z;
+ let new_diff = new_camera_z - tile_z;
+ let was_in_range = old_diff >= 0 && old_diff <= 8;
+ let is_in_range = new_diff >= 0 && new_diff <= 8;
+ if was_in_range || is_in_range {
registry.dirty_keys.insert(key);
+ count += 1;
}
}
+
+ bench.last_z_change_dirty_ms = now.elapsed().as_secs_f64() * 1000.0;
+ bench.last_z_change_dirty_count = count;
}
pub fn despawn_tilemap_chunks(