/*
 * Copyright (c) 2026 Mykhailo Mamedov. All rights reserved.
 *
 * RESEARCH PREVIEW / REFERENCE ONLY:
 * This source code is provided solely for the purpose of reviewing
 * the author's research methods and implementation.
 *
 * NO LICENSE GRANTED:
 * This code is NOT for distribution, modification, or use in any
 * project (commercial or otherwise). Unauthorized copying or
 * use of this code is strictly prohibited.
 *
 * For inquiries regarding use or licensing, contact: ua.modin@gmail.com
 */

#version 460
#extension GL_EXT_debug_printf : enable
#extension GL_EXT_shader_atomic_float : enable

layout(local_size_x = 64) in; 

struct CurveHeader {
    uint poolOffset;
    uint pointCount;
    uint maxPoints;
    uint _padding;
    vec4 color;
};

struct PointData {
    vec4 pos; // xyz = worldPos, w = radius
};

layout(std430, binding = 0) readonly buffer HeaderBuffer { CurveHeader headers[]; };
layout(std430, binding = 1) readonly buffer PointPool    { PointData pool[]; };
layout(std430, binding = 2) writeonly buffer RenderBuffer { vec4 smoothPoints[]; };
layout(r32f, binding = 3) uniform image3D shadowVolume;

layout(push_constant) uniform Push {
    mat4 viewProj;
    vec3 gridOrigin;
    float gridCellSize;
    float gridExtent;
    uint frameCount;
    float _pad[2];
} pc;

shared vec4 s_Points[100];
shared CurveHeader s_Header;

void main() {
    uint strandIdx = gl_WorkGroupID.x;
    uint segmentIdx = gl_LocalInvocationID.x; // 0 to 63
    
    if (segmentIdx == 0) s_Header = headers[strandIdx];
    barrier();

    if (s_Header.pointCount == 0) return;

    for (uint i = segmentIdx; i < 100; i += 64) {
        if (i < s_Header.pointCount) {
            s_Points[i] = pool[s_Header.poolOffset + i].pos;
        }
    }
    barrier();

    float t_global = float(segmentIdx) / 63.0;
    float segProgress = t_global * float(s_Header.pointCount - 1);
    int i = int(floor(segProgress));
    float t_local = fract(segProgress);

    // Fetch from shared memory
    vec3 p0 = s_Points[max(i - 1, 0)].xyz;
    vec3 p1 = s_Points[i].xyz;
    vec3 p2 = s_Points[min(i + 1, int(s_Header.pointCount - 1))].xyz;
    vec3 p3 = s_Points[min(i + 2, int(s_Header.pointCount - 1))].xyz;

    vec3 finalPos = catmullRom(p0, p1, p2, p3, t_local);
    float radius = mix(s_Points[i].w, s_Points[min(i + 1, int(s_Header.pointCount - 1))].w, t_local);
    uint outIdx = (strandIdx * 64) + segmentIdx;
    smoothPoints[outIdx] = vec4(finalPos, radius);
	ivec3 voxel = ivec3(((finalPos - pc.gridOrigin) / pc.gridExtent)*pc.gridSize);
	if (all(greaterThanEqual(voxel, ivec3(0))) && all(lessThan(voxel, ivec3(pc.gridSize)))) {
		imageAtomicAdd(shadowVolume, voxel, 0.01); 
	}
}