#include "Rendering/Renderer.h"
#include "Algorithm/PathTracing.h"

#define FLIP_Y

static inline void ensure_camera_aspect_ratio(camera_t* camera, const rendering_config_t* config)
{
    float aspect_ratio = (float)config->width / config->height;
    if (fabsf((float)config->width / config->height - camera->size_x / camera->size_y) > 0.001f)
    {
        camera->size_y = camera->size_x / aspect_ratio;
        camera->fov_y = 2.0f * (float)atan(camera->size_x / (2.0f * camera->focal_length * aspect_ratio));
    }
}

static inline vec2s compute_ndc(float x, float y, uint32_t width, uint32_t height)
{
    return (vec2s){
        .x = x / (float)width,
#ifdef FLIP_Y
        .y = 1.0f - y / (float)height
#else
        .y = y / (float)height
#endif
    };
}

static inline uint16_t get_sample_count(uint16_t sample_count, int flag)
{
    switch (flag)
    {
        case DEBUG_BVH:
        case DEBUG_SOBOL:
        case DEBUG_UV:
            return 1;
        default:
            return sample_count;
    }

    return sample_count;
}

static void render_pixel(const rendering_config_t* config, scene_t* scene, vec3s coord,  uint32_t x,  uint32_t y, debug_flag_t flag, vec4s* pixel_color)
{
    vec4s accumulated_color = glms_vec4_zero();

    uint32_t pixel_id = y * config->width + x;
    uint16_t sample_count = get_sample_count(config->sample_count, flag);
    float inv_sample = 1.0f / (float)sample_count;

    vec3s camera_right = quat_get_right(scene->camera.rotation);
    vec3s camera_up = quat_get_up(scene->camera.rotation);

    for (uint16_t k = 0; k < sample_count; k++)
    {
        // TODO: Hash it
        uint32_t sobol_idx = pixel_id * config->sample_count + (k + 1);

        // Apply AA
        float du = sobol_sample(sobol_idx, PRNG_FILTER_U);
        float dv = sobol_sample(sobol_idx, PRNG_FILTER_V);
        vec2s position_ndc = compute_ndc((float)x + du, (float)y + dv, config->width, config->height);

        float screen_x = position_ndc.x * 2.0f - 1.0f;
        float screen_y = position_ndc.y * 2.0f - 1.0f;
        float sensor_offset_x = screen_x * scene->camera.size_x * 0.5f;
        float sensor_offset_y = screen_y * scene->camera.size_y * 0.5f;

        vec3s image_plane_point = coord;
        image_plane_point = glms_vec3_add(image_plane_point, glms_vec3_scale(camera_right, sensor_offset_x));
        image_plane_point = glms_vec3_add(image_plane_point, glms_vec3_scale(camera_up, sensor_offset_y));

        ray_t ray = ray_create(scene->camera.position, glms_vec3_normalize(glms_vec3_sub(image_plane_point, scene->camera.position)));

        vec4s out_color = flag == 0 ? path_trace(scene, ray, sobol_idx, config->max_depth) : render_debug(scene, ray, sobol_idx, flag);
        accumulated_color = glms_vec4_add(accumulated_color, out_color);
    }

    *pixel_color = glms_vec4_scale(accumulated_color, inv_sample);
}

// TODO: Progressive rendering
void renderer_start(render_job_t* job)
{
    ensure_camera_aspect_ratio(&job->scene->camera, job->config);

    uint32_t tile_count_x = (job->config->width + job->config->bucket_size - 1) / job->config->bucket_size;
    uint32_t tile_count_y = (job->config->height + job->config->bucket_size - 1) / job->config->bucket_size;
    uint32_t tile_count = tile_count_x * tile_count_y;

    float inv_sample = 1.0f / job->config->sample_count;
    vec3s coord = glms_vec3_add(job->scene->camera.position, glms_vec3_scale(quat_get_forward(job->scene->camera.rotation), job->scene->camera.focal_length));

    int64_t x, y, tile_index; // OpenMP requires these to be declared outside the parallel region. Also, they need to be signed integers. To avoid overflow, we need to use int64_t

#pragma omp parallel for schedule(dynamic, 1) default(none) \
    shared(tile_count_x, tile_count_y, tile_count, coord, inv_sample, job) \
    private(x, y, tile_index)
    for (tile_index = 0; tile_index < tile_count; tile_index++)
    {
        uint32_t tile_x_0 = (uint32_t)tile_index % tile_count_x * job->config->bucket_size;
        uint32_t tile_y_0 = (uint32_t)tile_index / tile_count_x * job->config->bucket_size;
        uint32_t tile_x_1 = (uint32_t)fmin(tile_x_0 + job->config->bucket_size, job->config->width);
        uint32_t tile_y_1 = (uint32_t)fmin(tile_y_0 + job->config->bucket_size, job->config->height);

        for (y = tile_y_0; y < tile_y_1; y++)
        {
            for (x = tile_x_0; x < tile_x_1; x++)
            {
                if (job->is_done)
                {
                    goto tile_done;
                }

                vec4s pixel_color;
                render_pixel(job->config, job->scene, coord, (uint32_t)x, (uint32_t)y, job->debug_flag, &pixel_color);
                render_target_set_pixel(job->render_target, (uint32_t)x, (uint32_t)y, pixel_color);
            }
        }
    tile_done:;
    }

    job->is_done = true;
}