// SPDX-License-Identifier: Apache-2.0 OR MIT OR Unlicense

// This is "kernel 4" in a 4-kernel pipeline. It renders the commands
// in the per-tile command list to an image.

// Right now, this kernel stores the image in a buffer, but a better
// plan is to use a texture. This is because of limited support.

#version 450
#extension GL_GOOGLE_include_directive : enable
#extension GL_EXT_nonuniform_qualifier : enable

#include "mem.h"
#include "setup.h"

#define CHUNK 8
#define CHUNK_DY (TILE_HEIGHT_PX / CHUNK)
layout(local_size_x = TILE_WIDTH_PX, local_size_y = CHUNK_DY) in;

layout(set = 0, binding = 1) readonly buffer ConfigBuf {
    Config conf;
};

layout(rgba8, set = 0, binding = 2) uniform writeonly image2D image;

#if GL_EXT_nonuniform_qualifier
layout(rgba8, set = 0, binding = 3) uniform readonly image2D images[];
#else
layout(rgba8, set = 0, binding = 3) uniform readonly image2D images[1];
#endif

#include "ptcl.h"
#include "tile.h"

#define BLEND_STACK_SIZE 4

// Layout of a clip scratch frame:
// Each frame is WIDTH * HEIGHT 32-bit words, then a link reference.

// Link offset and frame size in 32-bit words.
#define CLIP_LINK_OFFSET (TILE_WIDTH_PX * TILE_HEIGHT_PX)
#define CLIP_BUF_SIZE (CLIP_LINK_OFFSET + 1)

shared MallocResult sh_clip_alloc;

// Allocate a scratch buffer for clipping.
MallocResult alloc_clip_buf(uint link) {
    if (gl_LocalInvocationID.x == 0 && gl_LocalInvocationID.y == 0) {
        MallocResult m = malloc(CLIP_BUF_SIZE * 4);
        if (!m.failed) {
            write_mem(m.alloc, (m.alloc.offset >> 2) + CLIP_LINK_OFFSET, link);
        }
        sh_clip_alloc = m;
    }
    barrier();
    return sh_clip_alloc;
}

vec3 tosRGB(vec3 rgb) {
    bvec3 cutoff = greaterThanEqual(rgb, vec3(0.0031308));
    vec3 below = vec3(12.92)*rgb;
    vec3 above = vec3(1.055)*pow(rgb, vec3(0.41666)) - vec3(0.055);
    return mix(below, above, cutoff);
}

vec3 fromsRGB(vec3 srgb) {
    // Formula from EXT_sRGB.
    bvec3 cutoff = greaterThanEqual(srgb, vec3(0.04045));
    vec3 below = srgb/vec3(12.92);
    vec3 above = pow((srgb + vec3(0.055))/vec3(1.055), vec3(2.4));
    return mix(below, above, cutoff);
}

// unpacksRGB unpacks a color in the sRGB color space to a vec4 in the linear color
// space.
vec4 unpacksRGB(uint srgba) {
    vec4 color = unpackUnorm4x8(srgba).wzyx;
    return vec4(fromsRGB(color.rgb), color.a);
}

// packsRGB packs a color in the linear color space into its 8-bit sRGB equivalent.
uint packsRGB(vec4 rgba) {
    rgba = vec4(tosRGB(rgba.rgb), rgba.a);
    return packUnorm4x8(rgba.wzyx);
}

vec4[CHUNK] fillImage(uvec2 xy, CmdImage cmd_img) {
    vec4 rgba[CHUNK];
    for (uint i = 0; i < CHUNK; i++) {
        ivec2 uv = ivec2(xy.x, xy.y + i * CHUNK_DY) + cmd_img.offset;
#ifdef ENABLE_IMAGE_INDICES
        vec4 fg_rgba = imageLoad(images[cmd_img.index], uv);
#else
        vec4 fg_rgba = imageLoad(images[0], uv);
#endif
        fg_rgba.rgb = fromsRGB(fg_rgba.rgb);
        rgba[i] = fg_rgba;
    }
    return rgba;
}

void main() {
    if (mem_error != NO_ERROR) {
        return;
    }

    uint tile_ix = gl_WorkGroupID.y * conf.width_in_tiles + gl_WorkGroupID.x;
    Alloc cmd_alloc = slice_mem(conf.ptcl_alloc, tile_ix * PTCL_INITIAL_ALLOC, PTCL_INITIAL_ALLOC);
    CmdRef cmd_ref = CmdRef(cmd_alloc.offset);

    uvec2 xy_uint = uvec2(gl_GlobalInvocationID.x, gl_LocalInvocationID.y + TILE_HEIGHT_PX * gl_WorkGroupID.y);
    vec2 xy = vec2(xy_uint);
    vec3 rgb[CHUNK];
    float mask[CHUNK];
    uint blend_stack[BLEND_STACK_SIZE][CHUNK];
    uint blend_spill = 0;
    uint blend_sp = 0;
    Alloc clip_tos = new_alloc(0, 0);
    for (uint i = 0; i < CHUNK; i++) {
        rgb[i] = vec3(0.5);
#ifdef ENABLE_IMAGE_INDICES
        if (xy_uint.x < 1024 && xy_uint.y < 1024) {
            rgb[i] = imageLoad(images[gl_WorkGroupID.x / 64], ivec2(xy_uint.x, xy_uint.y + CHUNK_DY * i)/4).rgb;
        }
#endif
        mask[i] = 1.0;
    }

    float area[CHUNK];
    while (true) {
        uint tag = Cmd_tag(cmd_alloc, cmd_ref).tag;
        if (tag == Cmd_End) {
            break;
        }
        switch (tag) {
        case Cmd_Stroke:
            // Calculate distance field from all the line segments in this tile.
            CmdStroke stroke = Cmd_Stroke_read(cmd_alloc, cmd_ref);
            float df[CHUNK];
            for (uint k = 0; k < CHUNK; k++) df[k] = 1e9;
            TileSegRef tile_seg_ref = TileSegRef(stroke.tile_ref);
            do {
                TileSeg seg = TileSeg_read(new_alloc(tile_seg_ref.offset, TileSeg_size), tile_seg_ref);
                vec2 line_vec = seg.vector;
                for (uint k = 0; k < CHUNK; k++) {
                    vec2 dpos = xy + vec2(0.5, 0.5) - seg.origin;
                    dpos.y += float(k * CHUNK_DY);
                    float t = clamp(dot(line_vec, dpos) / dot(line_vec, line_vec), 0.0, 1.0);
                    df[k] = min(df[k], length(line_vec * t - dpos));
                }
                tile_seg_ref = seg.next;
            } while (tile_seg_ref.offset != 0);
            for (uint k = 0; k < CHUNK; k++) {
                area[k] = clamp(stroke.half_width + 0.5 - df[k], 0.0, 1.0);
            }
            break;
        case Cmd_Fill:
            CmdFill fill = Cmd_Fill_read(cmd_alloc, cmd_ref);
            for (uint k = 0; k < CHUNK; k++) area[k] = float(fill.backdrop);
            tile_seg_ref = TileSegRef(fill.tile_ref);
            // Calculate coverage based on backdrop + coverage of each line segment
            do {
                TileSeg seg = TileSeg_read(new_alloc(tile_seg_ref.offset, TileSeg_size), tile_seg_ref);
                for (uint k = 0; k < CHUNK; k++) {
                    vec2 my_xy = vec2(xy.x, xy.y + float(k * CHUNK_DY));
                    vec2 start = seg.origin - my_xy;
                    vec2 end = start + seg.vector;
                    vec2 window = clamp(vec2(start.y, end.y), 0.0, 1.0);
                    if (window.x != window.y) {
                        vec2 t = (window - start.y) / seg.vector.y;
                        vec2 xs = vec2(mix(start.x, end.x, t.x), mix(start.x, end.x, t.y));
                        float xmin = min(min(xs.x, xs.y), 1.0) - 1e-6;
                        float xmax = max(xs.x, xs.y);
                        float b = min(xmax, 1.0);
                        float c = max(b, 0.0);
                        float d = max(xmin, 0.0);
                        float a = (b + 0.5 * (d * d - c * c) - xmin) / (xmax - xmin);
                        area[k] += a * (window.x - window.y);
                    }
                    area[k] += sign(seg.vector.x) * clamp(my_xy.y - seg.y_edge + 1.0, 0.0, 1.0);
                }
                tile_seg_ref = seg.next;
            } while (tile_seg_ref.offset != 0);
            for (uint k = 0; k < CHUNK; k++) {
                area[k] = min(abs(area[k]), 1.0);
            }
            break;
        case Cmd_Solid:
            for (uint k = 0; k < CHUNK; k++) {
                area[k] = 1.0;
            }
            break;
        case Cmd_Alpha:
            CmdAlpha alpha = Cmd_Alpha_read(cmd_alloc, cmd_ref);
            for (uint k = 0; k < CHUNK; k++) {
                area[k] = alpha.alpha;
            }
            break;
        case Cmd_Color:
            CmdColor color = Cmd_Color_read(cmd_alloc, cmd_ref);
            vec4 fg_rgba = unpacksRGB(color.rgba_color);
            for (uint k = 0; k < CHUNK; k++) {
                rgb[k] = mix(rgb[k], fg_rgba.rgb, mask[k] * area[k] * fg_rgba.a);
            }
            break;
        case Cmd_Image:
            CmdImage fill_img = Cmd_Image_read(cmd_alloc, cmd_ref);
            vec4 rgba[CHUNK] = fillImage(xy_uint, fill_img);
            for (uint k = 0; k < CHUNK; k++) {
                rgb[k] = mix(rgb[k], rgba[k].rgb, mask[k] * area[k] * rgba[k].a);
            }
            break;
        case Cmd_BeginClip:
            uint blend_slot = blend_sp % BLEND_STACK_SIZE;
            if (blend_sp == blend_spill + BLEND_STACK_SIZE) {
                // spill to scratch buffer
                MallocResult m = alloc_clip_buf(clip_tos.offset);
                if (m.failed) {
                    return;
                }
                clip_tos = m.alloc;
                uint base_ix = (clip_tos.offset >> 2) + gl_LocalInvocationID.x + TILE_WIDTH_PX * gl_LocalInvocationID.y;
                for (uint k = 0; k < CHUNK; k++) {
                    write_mem(clip_tos, base_ix + k * TILE_WIDTH_PX * CHUNK_DY, blend_stack[blend_slot][k]);
                }
                blend_spill++;
            }
            for (uint k = 0; k < CHUNK; k++) {
                blend_stack[blend_slot][k] = packsRGB(vec4(rgb[k], clamp(abs(area[k]), 0.0, 1.0)));
            }
            blend_sp++;
            break;
        case Cmd_EndClip:
            blend_slot = (blend_sp - 1) % BLEND_STACK_SIZE;
            if (blend_sp == blend_spill) {
                uint base_ix = (clip_tos.offset >> 2) + gl_LocalInvocationID.x + TILE_WIDTH_PX * gl_LocalInvocationID.y;
                for (uint k = 0; k < CHUNK; k++) {
                    blend_stack[blend_slot][k] = read_mem(clip_tos, base_ix + k * TILE_WIDTH_PX * CHUNK_DY);
                }
                clip_tos.offset = read_mem(clip_tos, (clip_tos.offset >> 2) + CLIP_LINK_OFFSET);
                blend_spill--;
            }
            blend_sp--;
            for (uint k = 0; k < CHUNK; k++) {
                vec4 rgba = unpacksRGB(blend_stack[blend_slot][k]);
                rgb[k] = mix(rgba.rgb, rgb[k], area[k] * rgba.a);
            }
            break;
        case Cmd_Jump:
            cmd_ref = CmdRef(Cmd_Jump_read(cmd_alloc, cmd_ref).new_ref);
            cmd_alloc.offset = cmd_ref.offset;
            continue;
        }
        cmd_ref.offset += Cmd_size;
    }

    for (uint i = 0; i < CHUNK; i++) {
        imageStore(image, ivec2(xy_uint.x, xy_uint.y + CHUNK_DY * i), vec4(tosRGB(rgb[i]), 1.0));
    }
}