chunk gen now tries nodes first, also messed around a lot w rendering

Cargo.lock

@@ -84,9 +84,9 @@ dependencies = [
 
 [[package]]
 name = "arrayref"
-version = "0.3.8"
+version = "0.3.9"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "9d151e35f61089500b617991b791fc8bfd237ae50cd5950803758a179b41e67a"
+checksum = "76a2e8124351fda1ef8aaaa3bbd7ebbcb486bbcd4225aca0aa0d84bb2db8fecb"
 
 [[package]]
 name = "arrayvec"
@@ -405,9 +405,9 @@ dependencies = [
 
 [[package]]
 name = "cc"
-version = "1.1.18"
+version = "1.1.19"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b62ac837cdb5cb22e10a256099b4fc502b1dfe560cb282963a974d7abd80e476"
+checksum = "2d74707dde2ba56f86ae90effb3b43ddd369504387e718014de010cec7959800"
 dependencies = [
  "jobserver",
  "libc",
@@ -1029,9 +1029,9 @@ checksum = "78ca9ab1a0babb1e7d5695e3530886289c18cf2f87ec19a575a0abdce112e3a3"
 
 [[package]]
 name = "memmap2"
-version = "0.9.4"
+version = "0.9.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "fe751422e4a8caa417e13c3ea66452215d7d63e19e604f4980461212f3ae1322"
+checksum = "fd3f7eed9d3848f8b98834af67102b720745c4ec028fcd0aa0239277e7de374f"
 dependencies = [
  "libc",
 ]
@@ -2113,9 +2113,9 @@ checksum = "e91b56cd4cadaeb79bbf1a5645f6b4f8dc5bde8834ad5894a8db35fda9efa1fe"
 
 [[package]]
 name = "unicode-segmentation"
-version = "1.11.0"
+version = "1.12.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d4c87d22b6e3f4a18d4d40ef354e97c90fcb14dd91d7dc0aa9d8a1172ebf7202"
+checksum = "f6ccf251212114b54433ec949fd6a7841275f9ada20dddd2f29e9ceea4501493"
 
 [[package]]
 name = "unicode-width"

src/client/mod.rs

@@ -17,8 +17,8 @@ pub use state::*;
 use system::render::add_grid;
 
 use crate::{
-    server::Server,
     common::{ClientMessage, ServerHandle, ServerMessage},
+    server::Server,
 };
 
 use self::{input::Input, render::Renderer, ClientState};
@@ -46,8 +46,9 @@ pub struct Client<'a> {
     server: ServerHandle,
     server_id_map: HashMap<Entity, Entity>,
     systems: ClientSystems,
-    target: Instant,
+    frame_target: Instant,
     frame_time: Duration,
+    second_target: Instant,
     the_thing: bool,
 }
 
@@ -91,8 +92,9 @@ impl Client<'_> {
             world,
             server,
             server_id_map: HashMap::new(),
-            target: Instant::now(),
+            frame_target: Instant::now(),
             frame_time: FRAME_TIME,
+            second_target: Instant::now(),
             the_thing: false,
         }
     }
@@ -124,8 +126,8 @@ impl Client<'_> {
             }
         }
 
-        if now >= self.target {
+        if now >= self.frame_target {
-            self.target += self.frame_time;
+            self.frame_target += self.frame_time;
             let mut commands = std::mem::take(&mut self.render_commands);
             let world_cmds = std::mem::take(&mut self.world.resource_mut::<RenderCommands>().0);
             commands.extend(world_cmds);
@@ -133,6 +135,17 @@ impl Client<'_> {
             self.renderer.draw();
         }
 
+        if now >= self.second_target {
+            self.second_target += Duration::from_secs(1);
+            // let timer = self.renderer.timer();
+            // println!(
+            //     "avg: {:4?}; max: {:4?}; fps: {:4?}",
+            //     timer.avg(),
+            //     timer.max(),
+            //     timer.per_sec(),
+            // );
+        }
+
         if self.exit {
             event_loop.exit();
         }

src/client/render/mod.rs

@@ -4,7 +4,7 @@ pub mod voxel;
 pub use command::*;
 
 use super::camera::Camera;
-use crate::client::rsc::CLEAR_COLOR;
+use crate::{client::rsc::CLEAR_COLOR, util::timer::Timer};
 use nalgebra::Vector2;
 use util::DepthTexture;
 use voxel::VoxelPipeline;
@@ -114,6 +114,7 @@ impl<'a> Renderer<'a> {
         let view = output
             .texture
             .create_view(&wgpu::TextureViewDescriptor::default());
+
         let mut compute_pass = encoder.begin_compute_pass(&wgpu::ComputePassDescriptor {
             label: None,
             timestamp_writes: None,

src/client/render/voxel/ray_oct/mod.rs

@@ -34,7 +34,7 @@ pub struct VoxelPipeline {
 }
 
 const RENDER_SHADER: wgpu::ShaderModuleDescriptor<'_> = include_wgsl!("shader/render.wgsl");
-const COMPUTE_SHADER: wgpu::ShaderModuleDescriptor<'_> = include_wgsl!("shader/compute.wgsl");
+const COMPUTE_SHADER: wgpu::ShaderModuleDescriptor<'_> = include_wgsl!("shader/compute_working.wgsl");
 
 impl VoxelPipeline {
     pub fn new(device: &wgpu::Device, config: &wgpu::SurfaceConfiguration) -> Self {
@@ -82,7 +82,7 @@ impl VoxelPipeline {
 
     pub fn update_shader(&mut self, device: &wgpu::Device) {
         let Ok(shader) = std::fs::read_to_string(
-            env!("CARGO_MANIFEST_DIR").to_owned() + "/src/client/render/voxel/ray_oct/shader/compute.wgsl",
+            env!("CARGO_MANIFEST_DIR").to_owned() + "/src/client/render/voxel/ray_oct/shader/compute_working.wgsl",
         ) else {
             println!("Failed to reload shader!");
             return;
@@ -198,7 +198,7 @@ impl VoxelPipeline {
             zoom: camera.scale,
             transform,
         };
-        self.layout.view.update(device, encoder, belt, data)
+        self.layout.view.update(device, encoder, belt, data);
     }
 
     pub fn draw<'a>(&'a self, render_pass: &mut wgpu::RenderPass<'a>) {

src/client/render/voxel/ray_oct/shader/compute.wgsl

@@ -42,34 +42,7 @@ fn main(@builtin(global_invocation_id) cell: vec3<u32>) {
     let pos = view.transform * vec4<f32>(pixel_pos, 1.0);
     let dir = view.transform * vec4<f32>(normalize(pixel_pos), 0.0);
 
-    let start = start_ray(pos, dir);
+    var color = trace_full(pos, dir);
-    var color = vec4<f32>(0.0);
-    if start.hit {
-        var res = ray_next(start.ray, LEAF_BIT);
-        var safe = 0;
-        var normals = start.normals;
-        while res.data != 0 {
-            safe += 1;
-            if safe > 100 {break;}
-            let data = res.data & LEAF_MASK;
-            let vcolor = get_color(data);
-            let diffuse = max(dot(global_lights[0].dir, normals[res.ray.axis]) + 0.1, 0.0);
-            let ambient = 0.2;
-            let lighting = max(diffuse, ambient);
-            let new_color = min(vcolor.xyz * lighting, vec3<f32>(1.0));
-            color += vec4<f32>(new_color.xyz * vcolor.a, vcolor.a) * (1.0 - color.a);
-            if color.a > FULL_ALPHA { break; }
-            let old_t = res.ray.t;
-            res = ray_next(res.ray, res.data);
-            let dist = res.ray.t - old_t;
-            if data == 3 {
-                let a = min(dist / 128.0, 1.0);
-                color += vec4<f32>(vec3<f32>(0.0) * a, a) * (1.0 - color.a);
-            }
-        }
-        // color = vec4<f32>(dir.xyz * res.ray.t / 2048.0, 1.0);
-    }
-    // var color = trace_full(pos, dir);
     let light_mult = clamp((-dot(dir.xyz, global_lights[0].dir) - 0.99) * 200.0, 0.0, 1.0);
     let sun_color = light_mult * vec3<f32>(1.0, 1.0, 1.0);
     let sky_bg = vec3<f32>(0.3, 0.6, 1.0);
@@ -89,210 +62,7 @@ const EPSILON = 0.00000000001;
 const MAX_ITERS = 2000;
 // NOTE: CANNOT GO HIGHER THAN 23 due to how floating point
 // numbers are stored and the bit manipulation used
-const MAX_SCALE: u32 = 12;
+const MAX_SCALE: u32 = 10;
-
-struct Ray {
-    t: f32,
-    vox_pos: vec3<f32>,
-    t_inc: vec3<f32>,
-    scale: u32,
-    min_adj: vec3<f32>,
-    child: u32,
-    axis: u32,
-    node_start: u32,
-    group_offset: u32,
-    inv_dir_bits: u32,
-    parents: array<u32, MAX_SCALE>,
-};
-
-struct RayResult {
-    ray: Ray,
-    data: u32,
-}
-
-struct RayStart {
-    hit: bool,
-    ray: Ray,
-    normals: mat3x3<f32>,
-}
-
-fn start_ray(pos_view: vec4<f32>, dir_view: vec4<f32>) -> RayStart {
-    let gi = 0;
-    let group = voxel_groups[gi];
-    if group.scale == 0 {
-        return RayStart(false, Ray(), mat3x3<f32>());
-    }
-    let dimensions = vec3<u32>(1u << group.scale);
-    let dim_f = vec3<f32>(dimensions);
-    let dim_i = vec3<i32>(dimensions);
-
-    // transform so that group is at 0,0
-    let pos = (group.transform_inv * pos_view).xyz;
-    var dir = (group.transform_inv * dir_view).xyz;
-    if dir.x == 0 {dir.x = EPSILON;}
-    if dir.y == 0 {dir.y = EPSILON;}
-    if dir.z == 0 {dir.z = EPSILON;}
-
-    let dir_if = sign(dir);
-    let dir_uf = max(dir_if, vec3<f32>(0.0));
-
-
-
-    // calculate normals
-    var normals = mat3x3<f32>(
-        (group.transform * vec4<f32>(dir_if.x, 0.0, 0.0, 0.0)).xyz,
-        (group.transform * vec4<f32>(0.0, dir_if.y, 0.0, 0.0)).xyz,
-        (group.transform * vec4<f32>(0.0, 0.0, dir_if.z, 0.0)).xyz,
-    );
-    var axis = 0u;
-
-    // find where ray intersects with group
-    let pos_min = (vec3<f32>(1.0) - dir_uf) * dim_f;
-    // time of intersection; x = td + p, solve for t
-    var t_min = (pos_min - pos) / dir;
-    if outside3f(pos, ZERO3F, dim_f) {
-        // points of intersection
-        let px = pos + t_min.x * dir;
-        let py = pos + t_min.y * dir;
-        let pz = pos + t_min.z * dir;
-
-        // check if point is in bounds
-        let hit = vec3<bool>(
-            inside2f(px.yz, ZERO2F, dim_f.yz),
-            inside2f(py.xz, ZERO2F, dim_f.xz),
-            inside2f(pz.xy, ZERO2F, dim_f.xy),
-        ) && (t_min > ZERO3F);
-        if !any(hit) {
-            return RayStart(false, Ray(), mat3x3<f32>());
-        }
-        axis = select(select(2u, 1u, hit.y), 0u, hit.x);
-    }
-    t_min *= f32(1u << (MAX_SCALE - group.scale));
-    // time to move 1 unit in each direction
-    let full = f32(1u << MAX_SCALE);
-    let t_inc = abs(1.0 / dir) * full;
-    let t_offset = max(max(t_min.x, t_min.y), t_min.z);
-    let t = max(0.0, t_offset);
-
-    let dir_i = vec3<i32>(dir_if);
-    let dir_u = vec3<u32>((dir_i + vec3<i32>(1)) / 2);
-    let dir_bits = vec_to_dir(dir_u);
-    let inv_dir_bits = 7 - dir_bits;
-
-    let node_start = 1u;
-    let scale = MAX_SCALE - 1;
-    let scale_exp2 = 0.5;
-    let parents = array<u32, MAX_SCALE>();
-
-    var child = 0u;
-    var vox_pos = vec3<f32>(1.0);
-    let t_center = t_min + scale_exp2 * t_inc;
-    if t > t_center.x { vox_pos.x = 1.5; child |= 4u; }
-    if t > t_center.y { vox_pos.y = 1.5; child |= 2u; }
-    if t > t_center.z { vox_pos.z = 1.5; child |= 1u; }
-    let min_adj = t_min - t_inc;
-
-    return RayStart(
-        true,
-        Ray(
-            t,
-            vox_pos,
-            t_inc,
-            scale,
-            min_adj,
-            child,
-            axis,
-            node_start,
-            group.offset,
-            inv_dir_bits,
-            parents,
-        ),
-        normals
-    );
-}
-
-fn ray_next(ray: Ray, skip: u32) -> RayResult {
-    let group_offset = ray.group_offset;
-    let t_inc = ray.t_inc;
-    let min_adj = ray.min_adj;
-    let inv_dir_bits = ray.inv_dir_bits;
-    var scale = ray.scale;
-    var scale_exp2 = bitcast<f32>((scale + 127 - MAX_SCALE) << 23);
-    var vox_pos = ray.vox_pos;
-    var t = ray.t;
-    var node_start = ray.node_start;
-    var child = ray.child;
-    var parents = ray.parents;
-    var axis: u32;
-    var data = 0u;
-    loop {
-        let t_corner = vox_pos * t_inc + min_adj;
-        let node = voxels[group_offset + node_start + (child ^ inv_dir_bits)];
-        if node >= LEAF_BIT {
-            if node != skip {
-                data = node;
-                break;
-            }
-
-            // move to next time point and determine which axis to move along
-            let t_next = t_corner + scale_exp2 * t_inc;
-            t = min(min(t_next.x, t_next.y), t_next.z);
-            axis = select(select(0u, 1u, t == t_next.y), 2u, t == t_next.z);
-            let move_dir = 4u >> axis;
-
-            // check if need to pop stack
-            if (child & move_dir) > 0 {
-                // calculate new scale; first differing bit after adding
-                let axis_pos = vox_pos[axis];
-                // AWARE
-                let differing = bitcast<u32>(axis_pos) ^ bitcast<u32>(axis_pos + scale_exp2);
-                scale = (bitcast<u32>(f32(differing)) >> 23) - 127 - (23 - MAX_SCALE);
-                scale_exp2 = bitcast<f32>((scale + 127 - MAX_SCALE) << 23);
-                if scale >= MAX_SCALE { break; }
-
-                // restore & recalculate parent
-                let parent_info = parents[scale];
-                node_start = parent_info >> 3;
-                child = parent_info & 7;
-                let scale_vec = vec3<u32>(scale + 23 - MAX_SCALE);
-                // remove bits lower than current scale
-                vox_pos = bitcast<vec3<f32>>((bitcast<vec3<u32>>(vox_pos) >> scale_vec) << scale_vec);
-            }
-            // move to next child and voxel position
-            child += move_dir;
-            vox_pos[axis] += scale_exp2;
-        } else {
-            // push current node to stack
-            parents[scale] = (node_start << 3) + child;
-            scale -= 1u;
-
-            // calculate child node vars
-            scale_exp2 *= 0.5;
-            child = 0u;
-            let t_center = t_corner + scale_exp2 * t_inc;
-            if t > t_center.x { vox_pos.x += scale_exp2; child |= 4u; }
-            if t > t_center.y { vox_pos.y += scale_exp2; child |= 2u; }
-            if t > t_center.z { vox_pos.z += scale_exp2; child |= 1u; }
-            node_start += 8 + node;
-        }
-    }
-    return RayResult(
-        Ray(
-            t,
-            vox_pos,
-            t_inc,
-            scale,
-            min_adj,
-            child,
-            axis,
-            node_start,
-            group_offset,
-            inv_dir_bits,
-            parents,
-        ),
-        data
-    );
-}
 
 fn trace_full(pos_view: vec4<f32>, dir_view: vec4<f32>) -> vec4<f32> {
     let gi = 0;
@@ -326,14 +96,13 @@ fn trace_full(pos_view: vec4<f32>, dir_view: vec4<f32>) -> vec4<f32> {
 
     // find where ray intersects with group
     let pos_min = (vec3<f32>(1.0) - dir_uf) * dim_f;
-    var pos = pos_start;
     // time of intersection; x = td + p, solve for t
-    var t_min = (pos_min - pos) / dir;
+    var t_min = (pos_min - pos_start) / dir;
-    if outside3f(pos, ZERO3F, dim_f) {
+    if outside3f(pos_start, ZERO3F, dim_f) {
         // points of intersection
-        let px = pos + t_min.x * dir;
+        let px = pos_start + t_min.x * dir;
-        let py = pos + t_min.y * dir;
+        let py = pos_start + t_min.y * dir;
-        let pz = pos + t_min.z * dir;
+        let pz = pos_start + t_min.z * dir;
 
         // check if point is in bounds
         let hit = vec3<bool>(
@@ -344,10 +113,10 @@ fn trace_full(pos_view: vec4<f32>, dir_view: vec4<f32>) -> vec4<f32> {
         if !any(hit) {
             return vec4<f32>(0.0);
         }
-        pos = select(select(pz, py, hit.y), px, hit.x);
         axis = select(select(2u, 1u, hit.y), 0u, hit.x);
     }
-    t_min *= f32(1u << (MAX_SCALE - group.scale));
+    let t_mult = f32(1u << (MAX_SCALE - group.scale));
+    t_min *= t_mult;
     // time to move 1 unit in each direction
     let full = f32(1u << MAX_SCALE);
     let inc_t = abs(1.0 / dir) * full;
@@ -362,6 +131,7 @@ fn trace_full(pos_view: vec4<f32>, dir_view: vec4<f32>) -> vec4<f32> {
     var node_start = 1u;
     var scale = MAX_SCALE - 1;
     var scale_exp2 = 0.5;
+    var skip = LEAF_BIT;
     var color = vec4<f32>(0.0);
     var parents = array<u32, MAX_SCALE>();
 
@@ -382,13 +152,25 @@ fn trace_full(pos_view: vec4<f32>, dir_view: vec4<f32>) -> vec4<f32> {
         let t_corner = vox_pos * inc_t + min_adj;
         let node = voxels[group.offset + node_start + (child ^ inv_dir_bits)];
         if node >= LEAF_BIT {
-            if node != LEAF_BIT {
+            if node != skip {
-                let vcolor = get_color(node & LEAF_MASK);
+                skip = node;
-                let diffuse = max(dot(global_lights[0].dir, normals[axis]) + 0.1, 0.0);
+                let normal = normals[axis];
+                let sun_dir = global_lights[0].dir;
+                let new_pos = pos_view + dir_view * t / t_mult - vec4<f32>(normals[axis] * 0.001, 0.0);
+
+                let light = trace_light(new_pos, vec4<f32>(-sun_dir, 0.0));
+                let diffuse = max(dot(sun_dir, normal) + 0.1, 0.0);
                 let ambient = 0.2;
-                let lighting = max(diffuse, ambient);
+                let specular = (exp(max(
-                let new_color = min(vcolor.xyz * lighting, vec3<f32>(1.0));
+                    -(dot(reflect(dir_view.xyz, normal), sun_dir) + 0.90) * 4.0, 0.0
-                color += vec4<f32>(new_color.xyz * vcolor.a, vcolor.a) * (1.0 - color.a);
+                )) - 1.0) * light;
+                let lighting = max(diffuse * light.a, ambient);
+
+                let vcolor = get_color(node & LEAF_MASK);
+                let new_rgb = min(vcolor.xyz * lighting + specular.xyz + light.xyz * vcolor.xyz, vec3<f32>(1.0));
+                let new_a = min(vcolor.a + specular.a, 1.0);
+                let new_color = vec4<f32>(new_rgb, new_a);
+                color += vec4<f32>(new_color.xyz * new_color.a, new_color.a) * (1.0 - color.a);
                 if color.a > FULL_ALPHA { break; }
             }
 
@@ -438,6 +220,164 @@ fn trace_full(pos_view: vec4<f32>, dir_view: vec4<f32>) -> vec4<f32> {
     return color;
 }
 
+fn trace_light(pos_view: vec4<f32>, dir_view: vec4<f32>) -> vec4<f32> {
+    let gi = 0;
+    let group = voxel_groups[gi];
+    if group.scale == 0 {
+        return vec4<f32>(0.0);
+    }
+    let dimensions = vec3<u32>(1u << group.scale);
+    let dim_f = vec3<f32>(dimensions);
+    let dim_i = vec3<i32>(dimensions);
+
+    // transform so that group is at 0,0
+    let pos_start = (group.transform_inv * pos_view).xyz;
+    var dir = (group.transform_inv * dir_view).xyz;
+    if dir.x == 0 {dir.x = EPSILON;}
+    if dir.y == 0 {dir.y = EPSILON;}
+    if dir.z == 0 {dir.z = EPSILON;}
+
+    let dir_if = sign(dir);
+    let dir_uf = max(dir_if, vec3<f32>(0.0));
+
+
+
+    // calculate normals
+    var normals = mat3x3<f32>(
+        (group.transform * vec4<f32>(dir_if.x, 0.0, 0.0, 0.0)).xyz,
+        (group.transform * vec4<f32>(0.0, dir_if.y, 0.0, 0.0)).xyz,
+        (group.transform * vec4<f32>(0.0, 0.0, dir_if.z, 0.0)).xyz,
+    );
+    var axis = 0u;
+
+    // find where ray intersects with group
+    let pos_min = (vec3<f32>(1.0) - dir_uf) * dim_f;
+    // time of intersection; x = td + p, solve for t
+    var t_min = (pos_min - pos_start) / dir;
+    if outside3f(pos_start, ZERO3F, dim_f) {
+        // points of intersection
+        let px = pos_start + t_min.x * dir;
+        let py = pos_start + t_min.y * dir;
+        let pz = pos_start + t_min.z * dir;
+
+        // check if point is in bounds
+        let hit = vec3<bool>(
+            inside2f(px.yz, ZERO2F, dim_f.yz),
+            inside2f(py.xz, ZERO2F, dim_f.xz),
+            inside2f(pz.xy, ZERO2F, dim_f.xy),
+        ) && (t_min > ZERO3F);
+        if !any(hit) {
+            return vec4<f32>(0.0);
+        }
+        axis = select(select(2u, 1u, hit.y), 0u, hit.x);
+    }
+    let t_mult = f32(1u << (MAX_SCALE - group.scale));
+    t_min *= t_mult;
+    // time to move 1 unit in each direction
+    let full = f32(1u << MAX_SCALE);
+    let inc_t = abs(1.0 / dir) * full;
+    let t_offset = max(max(t_min.x, t_min.y), t_min.z);
+    var t = max(0.0, t_offset);
+    var old_t = t;
+
+    let dir_i = vec3<i32>(dir_if);
+    let dir_u = vec3<u32>((dir_i + vec3<i32>(1)) / 2);
+    let dir_bits = vec_to_dir(dir_u);
+    let inv_dir_bits = 7 - dir_bits;
+
+    var node_start = 1u;
+    var scale = MAX_SCALE - 1;
+    var scale_exp2 = 0.5;
+    var mask = vec4<f32>(0.0);
+    var skip = LEAF_BIT;
+    var parents = array<u32, MAX_SCALE>();
+
+    var child = 0u;
+    var vox_pos = vec3<f32>(1.0);
+    let t_center = t_min + scale_exp2 * inc_t;
+    if t > t_center.x { vox_pos.x = 1.5; child |= 4u; }
+    if t > t_center.y { vox_pos.y = 1.5; child |= 2u; }
+    if t > t_center.z { vox_pos.z = 1.5; child |= 1u; }
+    let min_adj = t_min - inc_t;
+    var data = 0u;
+
+    var iters = 0;
+    loop {
+        if iters == MAX_ITERS {
+            return vec4<f32>(1.0, 0.0, 1.0, 1.0);
+        }
+        iters += 1;
+        let t_corner = vox_pos * inc_t + min_adj;
+        let node = voxels[group.offset + node_start + (child ^ inv_dir_bits)];
+        if node >= LEAF_BIT {
+            if node != skip {
+                skip = node;
+                if data == 3 {
+                    let dist = (t - old_t) / t_mult;
+                    let vcolor = vec4<f32>(vec3<f32>(0.0), min(dist / 12.0, 1.0));
+                    mask += vec4<f32>(vcolor.xyz * vcolor.a, vcolor.a) * (1.0 - mask.a);
+                }
+                data = node & LEAF_MASK;
+                if data != 3 && data != 0 {
+                    let vcolor = get_color(data);
+                    mask += vec4<f32>(vcolor.xyz * vcolor.a, vcolor.a) * (1.0 - mask.a);
+                }
+                old_t = t;
+                if mask.a > FULL_ALPHA { break; }
+            }
+
+            // move to next time point and determine which axis to move along
+            let t_next = t_corner + scale_exp2 * inc_t;
+            t = min(min(t_next.x, t_next.y), t_next.z);
+            axis = select(select(0u, 1u, t == t_next.y), 2u, t == t_next.z);
+            let move_dir = 4u >> axis;
+
+            // check if need to pop stack
+            if (child & move_dir) > 0 {
+                // calculate new scale; first differing bit after adding
+                let axis_pos = vox_pos[axis];
+                // AWARE
+                let differing = bitcast<u32>(axis_pos) ^ bitcast<u32>(axis_pos + scale_exp2);
+                scale = (bitcast<u32>(f32(differing)) >> 23) - 127 - (23 - MAX_SCALE);
+                scale_exp2 = bitcast<f32>((scale + 127 - MAX_SCALE) << 23);
+                if scale >= MAX_SCALE { break; }
+
+                // restore & recalculate parent
+                let parent_info = parents[scale];
+                node_start = parent_info >> 3;
+                child = parent_info & 7;
+                let scale_vec = vec3<u32>(scale + 23 - MAX_SCALE);
+                // remove bits lower than current scale
+                vox_pos = bitcast<vec3<f32>>((bitcast<vec3<u32>>(vox_pos) >> scale_vec) << scale_vec);
+            }
+            // move to next child and voxel position
+            child += move_dir;
+            vox_pos[axis] += scale_exp2;
+        } else {
+            // push current node to stack
+            parents[scale] = (node_start << 3) + child;
+            scale -= 1u;
+
+            // calculate child node vars
+            scale_exp2 *= 0.5;
+            child = 0u;
+            let t_center = t_corner + scale_exp2 * inc_t;
+            if t > t_center.x { vox_pos.x += scale_exp2; child |= 4u; }
+            if t > t_center.y { vox_pos.y += scale_exp2; child |= 2u; }
+            if t > t_center.z { vox_pos.z += scale_exp2; child |= 1u; }
+            node_start += 8 + node;
+        }
+    }
+    if data == 3 {
+        let dist = (t - old_t) / t_mult;
+        let vcolor = vec4<f32>(vec3<f32>(0.0), min(dist / 12.0, 1.0));
+        mask += vec4<f32>(vcolor.xyz * vcolor.a, vcolor.a) * (1.0 - mask.a);
+    }
+    mask.a = 1.0 - mask.a;
+    mask = vec4<f32>(mask.a * mask.xyz, mask.a);
+    return mask;
+}
+
 fn dir_to_vec(bits: u32) -> vec3<u32> {
     return vec3<u32>(bits >> 2, (bits & 2) >> 1, bits & 1);
 }

src/client/render/voxel/ray_oct/shader/compute_struct.wgsl

@@ -42,7 +42,56 @@ fn main(@builtin(global_invocation_id) cell: vec3<u32>) {
     let pos = view.transform * vec4<f32>(pixel_pos, 1.0);
     let dir = view.transform * vec4<f32>(normalize(pixel_pos), 0.0);
 
-    var color = trace_full(pos, dir);
+    let start = start_ray(pos, dir);
+    var color = vec4<f32>(0.0);
+    let ambient = 0.2;
+
+    if start.hit {
+        var res = ray_next(start.ray, LEAF_BIT);
+        var normals = start.normals;
+        let specular = (exp(max(
+            -(dot(reflect(dir.xyz, normals[res.ray.axis]), global_lights[0].dir) + 0.90) * 4.0, 0.0
+        )) - 1.0);
+        while res.data != 0 {
+            let data = res.data & LEAF_MASK;
+            let vcolor = get_color(data);
+            let diffuse = max(dot(global_lights[0].dir, normals[res.ray.axis]) + 0.1, 0.0);
+            let light = max(diffuse, ambient);
+            let new_color = min(vcolor.xyz * light, vec3<f32>(1.0));
+            color += vec4<f32>(new_color.xyz * vcolor.a, vcolor.a) * (1.0 - color.a);
+            if color.a > FULL_ALPHA { break; }
+            let old_t = res.ray.t;
+            res = ray_next(res.ray, res.data);
+            if data == 3 {
+                let dist = (res.ray.t - old_t) / start.t_mult;
+                let a = min(dist / 12.0, 1.0);
+                color += vec4<f32>(vec3<f32>(0.0) * a, a) * (1.0 - color.a);
+            }
+        }
+        if color.a != 0 {
+            let pos = pos + dir * res.ray.t / start.t_mult - vec4<f32>(normals[res.ray.axis] * 0.001, 0.0);
+            let dir = vec4<f32>(-global_lights[0].dir, 0.0);
+            let start = start_ray(pos, dir);
+            res = ray_next(start.ray, LEAF_BIT);
+            var light = 1.0;
+            while res.data != 0 {
+                let data = res.data & LEAF_MASK;
+                let vcolor = get_color(data);
+                if data != 3 { light -= vcolor.a * light; }
+                if light <= 0 { break; }
+                let old_t = res.ray.t;
+                res = ray_next(res.ray, res.data);
+                if data == 3 {
+                    let dist = (res.ray.t - old_t) / start.t_mult;
+                    let a = min(dist / 12.0, 1.0);
+                    light -= a;
+                }
+            }
+            color = vec4<f32>(color.xyz * max(light, ambient), color.a) + vec4<f32>(vec3<f32>(specular * light), 0.0);
+        }
+        // color = vec4<f32>(pos.xyz / 128.0, 1.0);
+    }
+    // var color = trace_full(pos, dir);
     let light_mult = clamp((-dot(dir.xyz, global_lights[0].dir) - 0.99) * 200.0, 0.0, 1.0);
     let sun_color = light_mult * vec3<f32>(1.0, 1.0, 1.0);
     let sky_bg = vec3<f32>(0.3, 0.6, 1.0);
@@ -52,25 +101,56 @@ fn main(@builtin(global_invocation_id) cell: vec3<u32>) {
     textureStore(output, cell.xy, color);
 }
 
+const LEAF_BIT = 1u << 31u;
+const LEAF_MASK = ~LEAF_BIT;
+
 const ZERO3F = vec3<f32>(0.0);
 const ZERO2F = vec2<f32>(0.0);
 const FULL_ALPHA = 0.999;
 const EPSILON = 0.00000000001;
-const MAX_ITERS = 1000;
+const MAX_ITERS = 2000;
+// NOTE: CANNOT GO HIGHER THAN 23 due to how floating point
+// numbers are stored and the bit manipulation used
 const MAX_SCALE: u32 = 10;
 
-fn trace_full(pos_view: vec4<f32>, dir_view: vec4<f32>) -> vec4<f32> {
+struct Ray {
+    t: f32,
+    vox_pos: vec3<f32>,
+    t_inc: vec3<f32>,
+    scale: u32,
+    min_adj: vec3<f32>,
+    child: u32,
+    axis: u32,
+    node_start: u32,
+    group_offset: u32,
+    inv_dir_bits: u32,
+    parents: array<u32, MAX_SCALE>,
+};
+
+struct RayResult {
+    ray: Ray,
+    data: u32,
+}
+
+struct RayStart {
+    hit: bool,
+    ray: Ray,
+    normals: mat3x3<f32>,
+    t_mult: f32,
+}
+
+fn start_ray(pos_view: vec4<f32>, dir_view: vec4<f32>) -> RayStart {
     let gi = 0;
     let group = voxel_groups[gi];
     if group.scale == 0 {
-        return vec4<f32>(0.0);
+        return RayStart();
     }
     let dimensions = vec3<u32>(1u << group.scale);
     let dim_f = vec3<f32>(dimensions);
     let dim_i = vec3<i32>(dimensions);
 
     // transform so that group is at 0,0
-    let pos_start = (group.transform_inv * pos_view).xyz;
+    let pos = (group.transform_inv * pos_view).xyz;
     var dir = (group.transform_inv * dir_view).xyz;
     if dir.x == 0 {dir.x = EPSILON;}
     if dir.y == 0 {dir.y = EPSILON;}
@@ -91,7 +171,6 @@ fn trace_full(pos_view: vec4<f32>, dir_view: vec4<f32>) -> vec4<f32> {
 
     // find where ray intersects with group
     let pos_min = (vec3<f32>(1.0) - dir_uf) * dim_f;
-    var pos = pos_start;
     // time of intersection; x = td + p, solve for t
     var t_min = (pos_min - pos) / dir;
     if outside3f(pos, ZERO3F, dim_f) {
@@ -107,57 +186,81 @@ fn trace_full(pos_view: vec4<f32>, dir_view: vec4<f32>) -> vec4<f32> {
             inside2f(pz.xy, ZERO2F, dim_f.xy),
         ) && (t_min > ZERO3F);
         if !any(hit) {
-            return vec4<f32>(0.0);
+            return RayStart();
         }
-        pos = select(select(pz, py, hit.y), px, hit.x);
         axis = select(select(2u, 1u, hit.y), 0u, hit.x);
     }
-    t_min *= f32(1u << (MAX_SCALE - group.scale));
+    let t_mult = f32(1u << (MAX_SCALE - group.scale));
+    t_min *= t_mult;
     // time to move 1 unit in each direction
     let full = f32(1u << MAX_SCALE);
-    let inc_t = abs(1.0 / dir) * full;
+    let t_inc = abs(1.0 / dir) * full;
     let t_offset = max(max(t_min.x, t_min.y), t_min.z);
-    var t = max(0.0, t_offset);
+    let t = max(0.0, t_offset);
 
     let dir_i = vec3<i32>(dir_if);
     let dir_u = vec3<u32>((dir_i + vec3<i32>(1)) / 2);
     let dir_bits = vec_to_dir(dir_u);
     let inv_dir_bits = 7 - dir_bits;
 
-    var node_start = 1u;
+    let node_start = 1u;
-    var scale = MAX_SCALE - 1;
+    let scale = MAX_SCALE - 1;
-    var scale_exp2 = 0.5;
+    let scale_exp2 = 0.5;
-    var color = vec4<f32>(0.0);
+    let parents = array<u32, MAX_SCALE>();
-    var parents = array<u32, MAX_SCALE>();
 
     var child = 0u;
-    var vox_pos = vec3<f32>(0.0);
+    var vox_pos = vec3<f32>(1.0);
-    let t_center = t_min + scale_exp2 * inc_t;
+    let t_center = t_min + scale_exp2 * t_inc;
-    if t > t_center.x { vox_pos.x = 0.5; child |= 4u; }
+    if t > t_center.x { vox_pos.x = 1.5; child |= 4u; }
-    if t > t_center.y { vox_pos.y = 0.5; child |= 2u; }
+    if t > t_center.y { vox_pos.y = 1.5; child |= 2u; }
-    if t > t_center.z { vox_pos.z = 0.5; child |= 1u; }
+    if t > t_center.z { vox_pos.z = 1.5; child |= 1u; }
+    let min_adj = t_min - t_inc;
 
-    var iters = 0;
+    return RayStart(
+        true,
+        Ray(
+            t,
+            vox_pos,
+            t_inc,
+            scale,
+            min_adj,
+            child,
+            axis,
+            node_start,
+            group.offset,
+            inv_dir_bits,
+            parents,
+        ),
+        normals,
+        t_mult,
+    );
+}
+
+fn ray_next(ray: Ray, skip: u32) -> RayResult {
+    let group_offset = ray.group_offset;
+    let t_inc = ray.t_inc;
+    let min_adj = ray.min_adj;
+    let inv_dir_bits = ray.inv_dir_bits;
+    var scale = ray.scale;
+    var scale_exp2 = bitcast<f32>((scale + 127 - MAX_SCALE) << 23);
+    var vox_pos = ray.vox_pos;
+    var t = ray.t;
+    var node_start = ray.node_start;
+    var child = ray.child;
+    var parents = ray.parents;
+    var axis: u32;
+    var data = 0u;
     loop {
-        if iters == MAX_ITERS {
+        let t_corner = vox_pos * t_inc + min_adj;
-            return vec4<f32>(1.0, 0.0, 1.0, 1.0);
+        let node = voxels[group_offset + node_start + (child ^ inv_dir_bits)];
-        }
-        iters += 1;
-        let t_corner = vox_pos * inc_t + t_min;
-        let node = voxels[group.offset + node_start + (child ^ inv_dir_bits)];
         if node >= LEAF_BIT {
-            if node != LEAF_BIT {
+            if node != skip {
-                let vcolor = get_color(node & LEAF_MASK);
+                data = node;
-                let diffuse = max(dot(global_lights[0].dir, normals[axis]) + 0.1, 0.0);
+                break;
-                let ambient = 0.2;
-                let lighting = max(diffuse, ambient);
-                let new_color = min(vcolor.xyz * lighting, vec3<f32>(1.0));
-                color += vec4<f32>(new_color.xyz * vcolor.a, vcolor.a) * (1.0 - color.a);
-                if color.a > FULL_ALPHA { break; }
             }
 
             // move to next time point and determine which axis to move along
-            let t_next = t_corner + scale_exp2 * inc_t;
+            let t_next = t_corner + scale_exp2 * t_inc;
             t = min(min(t_next.x, t_next.y), t_next.z);
             axis = select(select(0u, 1u, t == t_next.y), 2u, t == t_next.z);
             let move_dir = 4u >> axis;
@@ -165,22 +268,23 @@ fn trace_full(pos_view: vec4<f32>, dir_view: vec4<f32>) -> vec4<f32> {
             // check if need to pop stack
             if (child & move_dir) > 0 {
                 // calculate new scale; first differing bit after adding
-                let axis_pos = u32(vox_pos[axis] * full);
+                let axis_pos = vox_pos[axis];
-                let differing = axis_pos ^ (axis_pos + u32(scale_exp2 * full));
+                // AWARE
-                scale = u32(firstLeadingBit(differing));
+                let differing = bitcast<u32>(axis_pos) ^ bitcast<u32>(axis_pos + scale_exp2);
-                if scale == MAX_SCALE { break; }
+                scale = (bitcast<u32>(f32(differing)) >> 23) - 127 - (23 - MAX_SCALE);
+                scale_exp2 = bitcast<f32>((scale + 127 - MAX_SCALE) << 23);
+                if scale >= MAX_SCALE { break; }
 
                 // restore & recalculate parent
                 let parent_info = parents[scale];
                 node_start = parent_info >> 3;
                 child = parent_info & 7;
-                let scale_vec = vec3<u32>(scale);
+                let scale_vec = vec3<u32>(scale + 23 - MAX_SCALE);
-                // remove lower scale bits
+                // remove bits lower than current scale
-                vox_pos = vec3<f32>((vec3<i32>(vox_pos * full) >> scale_vec) << scale_vec) / full;
+                vox_pos = bitcast<vec3<f32>>((bitcast<vec3<u32>>(vox_pos) >> scale_vec) << scale_vec);
-                scale_exp2 = 1.0 / f32(1u << (MAX_SCALE - scale));
             }
             // move to next child and voxel position
-            child ^= move_dir;
+            child += move_dir;
             vox_pos[axis] += scale_exp2;
         } else {
             // push current node to stack
@@ -190,20 +294,31 @@ fn trace_full(pos_view: vec4<f32>, dir_view: vec4<f32>) -> vec4<f32> {
             // calculate child node vars
             scale_exp2 *= 0.5;
             child = 0u;
-            let t_center = t_corner + scale_exp2 * inc_t;
+            let t_center = t_corner + scale_exp2 * t_inc;
             if t > t_center.x { vox_pos.x += scale_exp2; child |= 4u; }
             if t > t_center.y { vox_pos.y += scale_exp2; child |= 2u; }
             if t > t_center.z { vox_pos.z += scale_exp2; child |= 1u; }
             node_start += 8 + node;
         }
     }
-    // return vec4<f32>(f32(iters) / f32(MAX_ITERS), 0.0, 0.0, 1.0);
+    return RayResult(
-    return color;
+        Ray(
+            t,
+            vox_pos,
+            t_inc,
+            scale,
+            min_adj,
+            child,
+            axis,
+            node_start,
+            group_offset,
+            inv_dir_bits,
+            parents,
+        ),
+        data
+    );
 }
 
-const LEAF_BIT = 1u << 31u;
-const LEAF_MASK = ~LEAF_BIT;
-
 fn dir_to_vec(bits: u32) -> vec3<u32> {
     return vec3<u32>(bits >> 2, (bits & 2) >> 1, bits & 1);
 }

src/client/render/voxel/ray_oct/shader/compute_working.wgsl

@@ -52,11 +52,16 @@ fn main(@builtin(global_invocation_id) cell: vec3<u32>) {
     textureStore(output, cell.xy, color);
 }
 
+const LEAF_BIT = 1u << 31u;
+const LEAF_MASK = ~LEAF_BIT;
+
 const ZERO3F = vec3<f32>(0.0);
 const ZERO2F = vec2<f32>(0.0);
 const FULL_ALPHA = 0.999;
 const EPSILON = 0.00000000001;
-const MAX_ITERS = 1000;
+const MAX_ITERS = 2000;
+// NOTE: CANNOT GO HIGHER THAN 23 due to how floating point
+// numbers are stored and the bit manipulation used
 const MAX_SCALE: u32 = 10;
 
 fn trace_full(pos_view: vec4<f32>, dir_view: vec4<f32>) -> vec4<f32> {
@@ -91,16 +96,13 @@ fn trace_full(pos_view: vec4<f32>, dir_view: vec4<f32>) -> vec4<f32> {
 
     // find where ray intersects with group
     let pos_min = (vec3<f32>(1.0) - dir_uf) * dim_f;
-    let pos_max = dir_uf * dim_f;
-    var pos = pos_start;
     // time of intersection; x = td + p, solve for t
-    let t_min = (pos_min - pos) / dir;
+    var t_min = (pos_min - pos_start) / dir;
-    let t_max = (pos_max - pos) / dir;
+    if outside3f(pos_start, ZERO3F, dim_f) {
-    if outside3f(pos, ZERO3F, dim_f) {
         // points of intersection
-        let px = pos + t_min.x * dir;
+        let px = pos_start + t_min.x * dir;
-        let py = pos + t_min.y * dir;
+        let py = pos_start + t_min.y * dir;
-        let pz = pos + t_min.z * dir;
+        let pz = pos_start + t_min.z * dir;
 
         // check if point is in bounds
         let hit = vec3<bool>(
@@ -111,29 +113,36 @@ fn trace_full(pos_view: vec4<f32>, dir_view: vec4<f32>) -> vec4<f32> {
         if !any(hit) {
             return vec4<f32>(0.0);
         }
-        pos = select(select(pz, py, hit.y), px, hit.x);
         axis = select(select(2u, 1u, hit.y), 0u, hit.x);
     }
+    let t_mult =f32(1u << (MAX_SCALE - group.scale));
+    t_min *= t_mult;
     // time to move 1 unit in each direction
-    let inc_t = abs(1.0 / dir);
+    let full = f32(1u << MAX_SCALE);
+    let inc_t = abs(1.0 / dir) * full;
     let t_offset = max(max(t_min.x, t_min.y), t_min.z);
     var t = max(0.0, t_offset);
 
     let dir_i = vec3<i32>(dir_if);
-    let dir_u = vec3<u32>((dir_i + vec3<i32>(1)) / 2);
+    let dir_u = vec3<u32>(dir_uf);
     let dir_bits = vec_to_dir(dir_u);
     let inv_dir_bits = 7 - dir_bits;
 
     var node_start = 1u;
-    var scale = group.scale - 1;
+    var scale = MAX_SCALE - 1;
-    var half_t_span = f32(1u << scale) * inc_t;
+    var scale_exp2 = 0.5;
-    var t_center = t_min + half_t_span;
     var color = vec4<f32>(0.0);
     var parents = array<u32, MAX_SCALE>();
+    var prev = LEAF_BIT;
+    var old_t = t;
 
-    var child = (u32(t > t_center.x) << 2) + (u32(t > t_center.y) << 1) + u32(t > t_center.z);
+    var child = 0u;
-    var child_pos = dir_to_vec(child);
+    var vox_pos = vec3<f32>(1.0);
-    var vox_pos = child_pos * (1u << scale);
+    let t_center = t_min + scale_exp2 * inc_t;
+    if t > t_center.x { vox_pos.x = 1.5; child |= 4u; }
+    if t > t_center.y { vox_pos.y = 1.5; child |= 2u; }
+    if t > t_center.z { vox_pos.z = 1.5; child |= 1u; }
+    let min_adj = t_min - inc_t;
 
     var iters = 0;
     loop {
@@ -141,20 +150,34 @@ fn trace_full(pos_view: vec4<f32>, dir_view: vec4<f32>) -> vec4<f32> {
             return vec4<f32>(1.0, 0.0, 1.0, 1.0);
         }
         iters += 1;
+        let t_corner = vox_pos * inc_t + min_adj;
         let node = voxels[group.offset + node_start + (child ^ inv_dir_bits)];
         if node >= LEAF_BIT {
-            if node != LEAF_BIT {
+            if node != prev {
-                let vcolor = get_color(node & LEAF_MASK);
+                if node != LEAF_BIT {
-                let diffuse = max(dot(global_lights[0].dir, normals[axis]) + 0.1, 0.0);
+                    let dist = (t - old_t) / t_mult;
-                let ambient = 0.2;
+                    old_t = t;
-                let lighting = max(diffuse, ambient);
+                    let filt = min(dist / 64.0, 1.0);
-                let new_color = min(vcolor.xyz * lighting, vec3<f32>(1.0));
+                    if prev == LEAF_BIT + 3 {
-                color += vec4<f32>(new_color.xyz * vcolor.a, vcolor.a) * (1.0 - color.a);
+                        color.a += filt * (1.0 - color.a);
-                if color.a > FULL_ALPHA { break; }
+                        if color.a > FULL_ALPHA { break; }
+                    }
+                    let pos = (vox_pos - 1.5) * (dir_if) + 0.5 - scale_exp2 * (1.0 - dir_uf);
+                    // let pos = t / t_mult;
+                    // if true {return vec4<f32>(pos, 1.0);}
+                    let vcolor = get_color(node & LEAF_MASK, pos);
+                    let diffuse = max(dot(global_lights[0].dir, normals[axis]) + 0.1, 0.0);
+                    let ambient = 0.2;
+                    let lighting = max(diffuse, ambient);
+                    let new_color = min(vcolor.xyz * lighting, vec3<f32>(1.0));
+                    color += vec4<f32>(new_color.xyz * vcolor.a, vcolor.a) * (1.0 - color.a);
+                    if color.a > FULL_ALPHA { break; }
+                }
+                prev = node;
             }
 
             // move to next time point and determine which axis to move along
-            let t_next = t_center + half_t_span * vec3<f32>(child_pos);
+            let t_next = t_corner + scale_exp2 * inc_t;
             t = min(min(t_next.x, t_next.y), t_next.z);
             axis = select(select(0u, 1u, t == t_next.y), 2u, t == t_next.z);
             let move_dir = 4u >> axis;
@@ -163,45 +186,44 @@ fn trace_full(pos_view: vec4<f32>, dir_view: vec4<f32>) -> vec4<f32> {
             if (child & move_dir) > 0 {
                 // calculate new scale; first differing bit after adding
                 let axis_pos = vox_pos[axis];
-                let differing = axis_pos ^ (axis_pos + (1u << scale));
+                // AWARE
-                scale = firstLeadingBit(differing);
+                let differing = bitcast<u32>(axis_pos) ^ bitcast<u32>(axis_pos + scale_exp2);
-                if scale == group.scale { break; }
+                scale = (bitcast<u32>(f32(differing)) >> 23) - 127 - (23 - MAX_SCALE);
+                scale_exp2 = bitcast<f32>((scale + 127 - MAX_SCALE) << 23);
+                if scale >= MAX_SCALE { break; }
 
                 // restore & recalculate parent
                 let parent_info = parents[scale];
                 node_start = parent_info >> 3;
                 child = parent_info & 7;
-                let scale_vec = vec3<u32>(scale + 1);
+                let scale_vec = vec3<u32>(scale + 23 - MAX_SCALE);
-                vox_pos = (vox_pos >> scale_vec) << scale_vec; // remove lower scale bits
+                // remove bits lower than current scale
-                half_t_span = f32(1u << scale) * inc_t;
+                vox_pos = bitcast<vec3<f32>>((bitcast<vec3<u32>>(vox_pos) >> scale_vec) << scale_vec);
-                t_center = vec3<f32>(vox_pos) * inc_t + t_min + half_t_span;
             }
             // move to next child and voxel position
-            child ^= move_dir;
+            child += move_dir;
-            child_pos = dir_to_vec(child);
+            vox_pos[axis] += scale_exp2;
-            vox_pos |= child_pos << vec3<u32>(scale);
-            // vox_pos[axis] += (1u << scale);
         } else {
             // push current node to stack
             parents[scale] = (node_start << 3) + child;
             scale -= 1u;
 
             // calculate child node vars
-            half_t_span /= 2.0;
+            scale_exp2 *= 0.5;
-            t_center += half_t_span * (vec3<f32>(child_pos * 2) - 1.0);
+            child = 0u;
-            child_pos = vec3<u32>(vec3<f32>(t) > t_center);
+            let t_center = t_corner + scale_exp2 * inc_t;
-            child = (child_pos.x << 2) + (child_pos.y << 1) + child_pos.z;
+            if t > t_center.x { vox_pos.x += scale_exp2; child |= 4u; }
-            vox_pos += child_pos * (1u << scale);
+            if t > t_center.y { vox_pos.y += scale_exp2; child |= 2u; }
+            if t > t_center.z { vox_pos.z += scale_exp2; child |= 1u; }
             node_start += 8 + node;
         }
     }
+    // let fog = min(t / t_mult / 1000.0, 1.0);
+    // return vec4<f32>(color.xyz * (1.0 - fog) + vec3<f32>(fog), color.a * (1.0 - fog) + fog);
     // return vec4<f32>(f32(iters) / f32(MAX_ITERS), 0.0, 0.0, 1.0);
     return color;
 }
 
-const LEAF_BIT = 1u << 31u;
-const LEAF_MASK = ~LEAF_BIT;
-
 fn dir_to_vec(bits: u32) -> vec3<u32> {
     return vec3<u32>(bits >> 2, (bits & 2) >> 1, bits & 1);
 }
@@ -210,19 +232,24 @@ fn vec_to_dir(vec: vec3<u32>) -> u32 {
     return vec.x * 4 + vec.y * 2 + vec.z * 1;
 }
 
-fn get_color(id: u32) -> vec4<f32> {
+fn get_color(id: u32, pos: vec3<f32>) -> vec4<f32> {
+    let random = random(pos);
+    let random2 = random(pos + vec3<f32>(0.0001));
     switch id {
         case 0u: {
             return vec4<f32>(0.0);
         }
         case 1u: {
-            return vec4<f32>(0.5, 0.5, 0.5, 1.0);
+            let color = vec3<f32>(0.5, 0.5, 0.5 + random * 0.2) * (random2 * 0.4 + 0.8);
+            return vec4<f32>(color, 1.0);
         }
         case 2u: {
-            return vec4<f32>(0.5, 1.0, 0.5, 1.0);
+            let color = vec3<f32>(0.4 + random * 0.2, 0.9, 0.4 + random * 0.2) * (random2 * 0.2 + 0.9);
+            return vec4<f32>(color, 1.0);
         }
         case 3u: {
-            return vec4<f32>(0.5, 0.5, 1.0, 0.5);
+            let color = vec3<f32>(0.5, 0.5, 1.0) * (random2 * 0.2 + 0.8);
+            return vec4<f32>(color, 0.5);
         }
         default: {
             return vec4<f32>(1.0, 0.0, 0.0, 1.0);
@@ -230,6 +257,10 @@ fn get_color(id: u32) -> vec4<f32> {
     }
 }
 
+fn random(pos: vec3<f32>) -> f32 {
+    return fract(sin(dot(pos,vec3<f32>(12.9898,78.233,25.1279)))*43758.5453123);
+}
+
 fn outside3f(v: vec3<f32>, low: vec3<f32>, high: vec3<f32>) -> bool {
     return any(v < low) || any(v > high);
 }

src/common/component/chunk/mod.rs

@@ -8,7 +8,7 @@ use bevy_derive::{Deref, DerefMut};
 use bevy_ecs::{bundle::Bundle, component::Component, entity::Entity, system::Resource};
 use nalgebra::Vector3;
 
-pub const SCALE: u32 = 8;
+pub const SCALE: u32 = 10;
 pub const SIDE_LENGTH: usize = 2usize.pow(SCALE);
 pub const SHAPE: (usize, usize, usize) = (SIDE_LENGTH, SIDE_LENGTH, SIDE_LENGTH);
 pub const DIMENSIONS: Vector3<usize> = Vector3::new(SIDE_LENGTH, SIDE_LENGTH, SIDE_LENGTH);

src/server/chunk/load.rs

@@ -1,17 +1,10 @@
 use std::collections::{HashMap, HashSet};
 
 use bevy_ecs::{entity::Entity, system::Commands};
-use nalgebra::Vector3;
-use ndarray::{s, Array3, Axis};
-use simdnoise::NoiseBuilder;
 
 use crate::{
-    client::render::voxel::VoxelColor,
+    common::component::{ChunkBundle, ChunkData, ChunkMesh, ChunkPos}, server::generation::generate_tree, util::
-    common::component::{chunk, ChunkBundle, ChunkData, ChunkMesh, ChunkPos},
+        thread::{ExitType, ThreadChannel, ThreadHandle}
-    util::{
-        oct_tree::OctTree,
-        thread::{ExitType, ThreadChannel, ThreadHandle},
-    },
 };
 
 pub struct ChunkManager {
@@ -114,32 +107,7 @@ fn chunk_loader_main(channel: ThreadChannel<ServerChunkMsg, ChunkLoaderMsg>) {
                 let tree = ChunkData::from_tree(generate_tree(pos));
                 let tree_time = std::time::Instant::now() - start;
 
-                // let start = std::time::Instant::now();
+                println!("gen time: {:<5?}; size: {}", tree_time, tree.raw().len());
-                // let mut data = generate(pos);
-                // let data_time = std::time::Instant::now() - start;
-                //
-                // let start = std::time::Instant::now();
-                // let shape = s![
-                //     1..data.len_of(Axis(0)) - 1,
-                //     1..data.len_of(Axis(1)) - 1,
-                //     1..data.len_of(Axis(2)) - 1
-                // ];
-                // let mut slice = data.slice_mut(shape);
-                // let mut iter = tree.into_iter();
-                // slice.assign(&Array3::from_shape_fn(chunk::SHAPE, |_| {
-                //     iter.next().unwrap()
-                // }));
-                // let convert_time = std::time::Instant::now() - start;
-                //
-                // let start = std::time::Instant::now();
-                // let mesh = ChunkMesh::from_data(data.map(|i| COLOR_MAP[*i as usize]).view());
-                // let mesh_time = std::time::Instant::now() - start;
-                //
-                // println!(
-                //     "data: {:<5?} mesh: {:<5?} convert: {:<5?} tree: {:<5?}",
-                //     data_time, mesh_time, convert_time, tree_time
-                // );
-                println!("gen time: {:<5?}", tree_time);
 
                 channel.send(ServerChunkMsg::ChunkGenerated(GeneratedChunk {
                     pos,
@@ -154,83 +122,3 @@ fn chunk_loader_main(channel: ThreadChannel<ServerChunkMsg, ChunkLoaderMsg>) {
     }
 }
 
-fn generate(pos: ChunkPos) -> Array3<u32> {
-    let shape = [chunk::SIDE_LENGTH + 2; 3];
-    if pos.y > 0 || pos.y < -1 {
-        return Array3::from_elem(shape, 0);
-    }
-    let posf: Vector3<f32> = (pos.cast() * chunk::SIDE_LENGTH as f32) - Vector3::from_element(1.0);
-    let (noise, min, max) = NoiseBuilder::gradient_2d_offset(
-        posf.x,
-        chunk::SIDE_LENGTH + 2,
-        posf.z,
-        chunk::SIDE_LENGTH + 2,
-    )
-    .with_seed(0)
-    .with_freq(0.005)
-    .generate();
-    Array3::from_shape_fn(shape, |(x, y, z)| {
-        generate_at(Vector3::new(x, y, z), posf, &noise, min, max)
-    })
-}
-
-fn generate_tree(pos: ChunkPos) -> OctTree {
-    if pos.y > 0 || pos.y < -1 {
-        return OctTree::from_leaf(0, 8);
-    }
-    let posf: Vector3<f32> = pos.cast() * chunk::SIDE_LENGTH as f32;
-    let (noise, min, max) =
-        NoiseBuilder::gradient_2d_offset(posf.x, chunk::SIDE_LENGTH, posf.z, chunk::SIDE_LENGTH)
-            .with_seed(0)
-            .with_freq(1.0 / (chunk::SIDE_LENGTH as f32))
-            .generate();
-    OctTree::from_fn_rec(&mut |p| generate_at(p, posf, &noise, min, max), chunk::SCALE)
-}
-
-fn generate_at(p: Vector3<usize>, posf: Vector3<f32>, noise: &[f32], min: f32, max: f32) -> u32 {
-    // 0 air 1 stone 2 "sand" 3 water
-    let y = p.y as f32 + posf.y;
-    // highest heights, 0.0 .. 1.0 relative to chunk size
-    let [water, grass, top] = [0.18, 0.35, 0.5].map(|f| chunk::SIDE_LENGTH as f32 * f);
-    let n = ((noise[p.x + p.z * chunk::SIDE_LENGTH] - min) / (max - min) * 2.0).exp2() * top * 0.25;
-    if y < n {
-        if y < water {
-            1
-        } else if y < grass {
-            2
-        } else {
-            1
-        }
-    } else if y <= water {
-        3
-    } else {
-        0
-    }
-}
-
-const COLOR_MAP: [VoxelColor; 4] = [
-    VoxelColor {
-        r: 0,
-        g: 0,
-        b: 0,
-        a: 0,
-    },
-    VoxelColor {
-        r: 150,
-        g: 150,
-        b: 150,
-        a: 255,
-    },
-    VoxelColor {
-        r: 100,
-        g: 255,
-        b: 100,
-        a: 255,
-    },
-    VoxelColor {
-        r: 100,
-        g: 100,
-        b: 255,
-        a: 200,
-    },
-];

src/server/generation/mod.rs

@@ -0,0 +1,145 @@
+use nalgebra::Vector3;
+use simdnoise::NoiseBuilder;
+
+use crate::{
+    common::component::{chunk, ChunkPos},
+    util::oct_tree::OctTree,
+};
+
+pub fn generate_tree(pos: ChunkPos) -> OctTree {
+    if pos.y > 0 || pos.y < -1 {
+        return OctTree::from_leaf(0, 8);
+    }
+    let posf: Vector3<f32> = pos.cast() * chunk::SIDE_LENGTH as f32;
+    let noise1 = generate_noise_map(0, 1.0, posf, chunk::SCALE, &mut |v: f32| {
+        (v * 2.0).exp2() * TOP * 0.25
+    });
+    let noise2 = generate_noise_map(1, 50.0, posf, chunk::SCALE, &mut |v: f32| v * 20.0 + GRASS);
+    OctTree::from_fn_rec(
+        &mut |p| generate_leaf(p, posf, (&noise1.base, &noise2.base)),
+        &mut |p, lvl| generate_node(p, lvl, posf, (&noise1, &noise2)),
+        chunk::SCALE,
+    )
+}
+
+const WATER: f32 = 0.18 * chunk::SIDE_LENGTH as f32;
+const GRASS: f32 = 0.35 * chunk::SIDE_LENGTH as f32;
+const TOP: f32 = 0.5 * chunk::SIDE_LENGTH as f32;
+
+// 0 air 1 stone 2 grass 3 water
+fn generate_leaf(p: Vector3<usize>, posf: Vector3<f32>, noise: (&[f32], &[f32])) -> u32 {
+    let y = p.y as f32 + posf.y;
+    let n = noise.0[p.x + p.z * chunk::SIDE_LENGTH];
+    let n2 = noise.1[p.x + p.z * chunk::SIDE_LENGTH];
+    if y < n {
+        if y < WATER {
+            1
+        } else if y < n2 {
+            2
+        } else {
+            1
+        }
+    } else if y <= WATER {
+        3
+    } else {
+        0
+    }
+}
+
+// 0 air 1 stone 2 grass 3 water
+fn generate_node(
+    p: Vector3<usize>,
+    scale: u32,
+    posf: Vector3<f32>,
+    noise: (&NoiseMap, &NoiseMap),
+) -> Option<u32> {
+    let side_len = 2usize.pow(scale);
+    let y = NumRange {
+        min: p.y as f32 + posf.y,
+        max: (p.y + side_len - 1) as f32 + posf.y,
+    };
+    let l = scale as usize - 1;
+    let i = (p.x >> scale) + (p.z >> scale) * (chunk::SIDE_LENGTH / side_len);
+    let n = &noise.0.levels[l][i];
+    let n2 = &noise.1.levels[l][i];
+    Some(if y.max < n.min {
+        if y.max < WATER {
+            1
+        } else if y.max < n2.min && y.min >= WATER {
+            2
+        } else if y.min > n2.max {
+            1
+        } else {
+            return None;
+        }
+    } else if y.max <= WATER && y.min > n.max {
+        3
+    } else if y.min > WATER && y.min > n.max {
+        0
+    } else {
+        return None;
+    })
+}
+
+fn generate_noise_map(
+    seed: i32,
+    freq: f32,
+    posf: Vector3<f32>,
+    levels: u32,
+    adjust: &mut impl FnMut(f32) -> f32,
+) -> NoiseMap {
+    let mut size = 2usize.pow(levels);
+    let (mut base, min, max) = NoiseBuilder::gradient_2d_offset(posf.x, size, posf.z, size)
+        .with_seed(seed)
+        .with_freq(freq / (size as f32))
+        .generate();
+    for v in &mut base {
+        *v = adjust((*v - min) / (max - min));
+    }
+    let first_len = base.len() / 4;
+    let mut first = Vec::with_capacity(first_len);
+    for y in (0..size).step_by(2) {
+        for x in (0..size).step_by(2) {
+            let a = base[x + y * size];
+            let b = base[x + 1 + y * size];
+            let c = base[x + (y + 1) * size];
+            let d = base[x + 1 + (y + 1) * size];
+            first.push(NumRange {
+                min: a.min(b).min(c).min(d),
+                max: a.max(b).max(c).max(d),
+            })
+        }
+    }
+    let mut arr = vec![first];
+    for l in 1..levels as usize {
+        size /= 2;
+        let prev = &arr[l - 1];
+        let mut new = Vec::with_capacity(prev.len() / 4);
+        for y in (0..size).step_by(2) {
+            for x in (0..size).step_by(2) {
+                let a = &prev[x + y * size];
+                let b = &prev[x + 1 + y * size];
+                let c = &prev[x + (y + 1) * size];
+                let d = &prev[x + 1 + (y + 1) * size];
+                new.push(NumRange {
+                    min: a.min.min(b.min).min(c.min).min(d.min),
+                    max: a.max.max(b.max).max(c.max).max(d.max),
+                })
+            }
+        }
+        arr.push(new);
+    }
+    NoiseMap { base, levels: arr }
+}
+
+#[derive(Debug)]
+pub struct NoiseMap {
+    levels: Vec<Vec<NumRange>>,
+    base: Vec<f32>,
+}
+
+#[derive(Debug)]
+pub struct NumRange {
+    min: f32,
+    max: f32,
+}

src/server/mod.rs

@@ -3,6 +3,7 @@ mod client;
 mod rsc;
 mod system;
 mod test;
+mod generation;
 
 pub use client::*;
 

src/util/mod.rs

@@ -1,2 +1,3 @@
 pub mod thread;
 pub mod oct_tree;
+pub mod timer;

src/util/oct_tree.rs

@@ -1,6 +1,5 @@
 use std::fmt::Debug;
 
-use bevy_ecs::system::IntoSystem;
 use nalgebra::Vector3;
 use ndarray::ArrayView3;
 
@@ -20,13 +19,13 @@ impl OctNode {
     pub const fn is_leaf(&self) -> bool {
         self.0 >= LEAF_BIT
     }
-    pub fn is_node(&self) -> bool {
+    pub const fn is_node(&self) -> bool {
         self.0 < LEAF_BIT
     }
-    pub fn node_data(&self) -> u32 {
+    pub const fn node_data(&self) -> u32 {
         self.0
     }
-    pub fn leaf_data(&self) -> u32 {
+    pub const fn leaf_data(&self) -> u32 {
         self.0 & !LEAF_BIT
     }
 }
@@ -57,17 +56,22 @@ impl OctTree {
             levels,
         }
     }
-    pub fn from_fn_rec(f: &mut impl FnMut(Vector3<usize>) -> u32, levels: u32) -> OctTree {
+    pub fn from_fn_rec(
-        Self::from_fn_offset(f, levels, Vector3::from_element(0))
+        f_leaf: &mut impl FnMut(Vector3<usize>) -> u32,
+        f_node: &mut impl FnMut(Vector3<usize>, u32) -> Option<u32>,
+        levels: u32,
+    ) -> OctTree {
+        Self::from_fn_offset(f_leaf, f_node, levels, Vector3::from_element(0))
     }
     pub fn from_fn_offset(
-        f: &mut impl FnMut(Vector3<usize>) -> u32,
+        f_leaf: &mut impl FnMut(Vector3<usize>) -> u32,
+        f_node: &mut impl FnMut(Vector3<usize>, u32) -> Option<u32>,
         levels: u32,
         offset: Vector3<usize>,
     ) -> Self {
         let mut data = Vec::new();
         data.push(OctNode::new_node(0));
-        Self::from_fn_offset_inner(f, &mut data, levels, offset);
+        Self::from_fn_offset_inner(f_leaf, f_node, &mut data, levels, offset);
         if data.len() == 2 {
             data.remove(0);
         }
@@ -78,17 +82,18 @@ impl OctTree {
         }
     }
     fn from_fn_offset_inner(
-        f: &mut impl FnMut(Vector3<usize>) -> u32,
+        f_leaf: &mut impl FnMut(Vector3<usize>) -> u32,
+        f_node: &mut impl FnMut(Vector3<usize>, u32) -> Option<u32>,
         accumulator: &mut Vec<OctNode>,
         level: u32,
         offset: Vector3<usize>,
     ) {
         if level == 0 {
-            accumulator.push(OctNode::new_leaf(f(offset)));
+            accumulator.push(OctNode::new_leaf(f_leaf(offset)));
             return;
         } else if level == 1 {
             let leaves: [OctNode; 8] =
-                core::array::from_fn(|i| OctNode::new_leaf(f(offset + CORNERS[i])));
+                core::array::from_fn(|i| OctNode::new_leaf(f_leaf(offset + CORNERS[i])));
             if leaves[1..].iter().all(|l| *l == leaves[0]) {
                 accumulator.push(leaves[0]);
             } else {
@@ -100,20 +105,21 @@ impl OctTree {
         accumulator.resize(i + 8, OctNode::new_node(0));
         let mut data_start = 0;
         for (j, corner_offset) in CORNERS.iter().enumerate() {
-            let sub_start = accumulator.len();
+            let lvl = level - 1;
-            Self::from_fn_offset_inner(
+            let pos = offset + corner_offset * 2usize.pow(lvl);
-                f,
+            if let Some(node) = f_node(pos, lvl) {
-                accumulator,
+                accumulator[i + j] = OctNode::new_leaf(node);
-                level - 1,
-                offset + corner_offset * 2usize.pow(level - 1),
-            );
-            let len = accumulator.len() - sub_start;
-            if len == 1 {
-                accumulator[i + j] = accumulator[sub_start];
-                accumulator.pop();
             } else {
-                accumulator[i + j] = OctNode::new_node(data_start as u32);
+                let sub_start = accumulator.len();
-                data_start += len;
+                Self::from_fn_offset_inner(f_leaf, f_node, accumulator, lvl, pos);
+                let len = accumulator.len() - sub_start;
+                if len == 1 {
+                    accumulator[i + j] = accumulator[sub_start];
+                    accumulator.pop();
+                } else {
+                    accumulator[i + j] = OctNode::new_node(data_start as u32);
+                    data_start += len;
+                }
             }
         }
         if data_start == 0 {
@@ -125,10 +131,7 @@ impl OctTree {
         }
     }
 
-    pub fn from_fn_iter(
+    pub fn from_fn_iter(f: &mut impl FnMut(Vector3<usize>) -> u32, levels: u32) -> Self {
-        f: &mut impl FnMut(Vector3<usize>) -> u32,
-        levels: u32,
-    ) -> Self {
         let mut data = vec![OctNode::new_node(0)];
         let mut level: usize = 1;
         let mut children = Vec::new();
@@ -173,7 +176,7 @@ impl OctTree {
     }
 
     pub fn from_arr(arr: ArrayView3<u32>, levels: u32) -> Self {
-        Self::from_fn_rec(&mut |p| arr[(p.x, p.y, p.z)], levels)
+        Self::from_fn_rec(&mut |p| arr[(p.x, p.y, p.z)], &mut |_, _| None, levels)
     }
     pub fn get(&self, mut pos: Vector3<usize>) -> u32 {
         let mut data_start = 1;
@@ -199,7 +202,6 @@ impl OctTree {
 pub struct OctTreeIter<'a> {
     queue: Vec<OctNode>,
     levels: Vec<u32>,
-    pos: usize,
     cur: u32,
     run: usize,
     data: &'a [OctNode],
@@ -218,9 +220,8 @@ impl<'a> Iterator for OctTreeIter<'a> {
             self.run = 8usize.pow(level);
             self.cur = node.leaf_data();
         } else {
-            let pos = 0;
+            let add = &self.data[..8];
-            let add = &self.data[pos..pos + 8];
+            self.data = &self.data[DATA_OFFSET..];
-            self.data = &self.data[pos + DATA_OFFSET..];
             self.queue.extend(add.iter().rev());
             self.levels.resize(self.levels.len() + 8, level - 1);
         }
@@ -234,7 +235,6 @@ impl<'a> IntoIterator for &'a OctTree {
     fn into_iter(self) -> Self::IntoIter {
         OctTreeIter {
             data: &self.data[1..],
-            pos: 0,
             cur: 0,
             levels: vec![self.levels],
             run: 0,

src/util/timer.rs

@@ -0,0 +1,59 @@
+use std::time::{Duration, Instant};
+
+pub struct Timer {
+    start: Instant,
+    pos: usize,
+    times: Vec<Option<Instant>>,
+    durs: Vec<Option<Duration>>,
+}
+
+impl Timer {
+    pub fn new(len: usize) -> Self {
+        Self {
+            start: Instant::now(),
+            pos: 0,
+            durs: vec![None; len],
+            times: vec![None; len],
+        }
+    }
+    pub fn start(&mut self) {
+        self.start = Instant::now();
+    }
+    pub fn stop(&mut self) {
+        let duration = Instant::now() - self.start;
+        self.durs[self.pos] = Some(duration);
+        self.times[self.pos] = Some(self.start);
+        self.pos = (self.pos + 1) % self.times.len();
+    }
+    pub fn avg(&self) -> Duration {
+        let filtered: Vec<_> = self.durs.iter().filter_map(|d| *d).collect();
+        let len = filtered.len();
+        if len != 0 {
+            let total: Duration = filtered.into_iter().sum();
+            total / len as u32
+        } else {
+            Duration::ZERO
+        }
+    }
+    pub fn max(&self) -> Duration {
+        self.durs
+            .iter()
+            .filter_map(|d| *d)
+            .max()
+            .unwrap_or(Duration::ZERO)
+    }
+    pub fn per_sec(&self) -> usize {
+        let now = Instant::now();
+        let mut count = 0;
+        while count < self.times.len() {
+            let i = (self.pos + count + 1) % self.times.len();
+            let Some(t) = self.times[i] else { break };
+            if now - t <= Duration::from_secs(1) {
+                count += 1;
+            } else {
+                break;
+            }
+        }
+        count
+    }
+}