#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <strings.h>
#include <math.h>
#include <sys/time.h>

#include "yolov8_seg.h"
#if defined(YOLO_RUNTIME_SHARED_PREPROCESS)
#include "yolo_rknn_preprocess.h"
#endif
#include "common.h"
#include "file_utils.h"
#include "image_utils.h"

static int64_t runtime_time_us()
{
    struct timeval value;
    gettimeofday(&value, nullptr);
    return static_cast<int64_t>(value.tv_sec) * 1000000 + value.tv_usec;
}

static thread_local float runtime_confidence_threshold = -1.0f;
static thread_local float runtime_nms_threshold = -1.0f;
static thread_local int runtime_class_agnostic_nms = -1;
static thread_local double runtime_preprocess_ms = 0.0;
static thread_local double runtime_inference_ms = 0.0;
static thread_local double runtime_postprocess_ms = 0.0;

extern "C" void yolo_adapter_set_thresholds(float confidence, float nms)
{
    runtime_confidence_threshold = confidence;
    runtime_nms_threshold = nms;
}

extern "C" void yolo_adapter_set_class_agnostic_nms(int enabled)
{
    runtime_class_agnostic_nms = enabled != 0 ? 1 : 0;
}

extern "C" int yolo_adapter_get_class_agnostic_nms(void)
{
    if (runtime_class_agnostic_nms >= 0) return runtime_class_agnostic_nms;
    const char *value = getenv("YOLO_CLASS_AGNOSTIC_NMS");
    if (!value) return 0;
    return strcmp(value, "1") == 0 || strcasecmp(value, "true") == 0 ||
           strcasecmp(value, "yes") == 0 || strcasecmp(value, "on") == 0;
}

extern "C" void yolo_adapter_get_performance(double *preprocess_ms, double *inference_ms, double *postprocess_ms)
{
    if (preprocess_ms) *preprocess_ms = runtime_preprocess_ms;
    if (inference_ms) *inference_ms = runtime_inference_ms;
    if (postprocess_ms) *postprocess_ms = runtime_postprocess_ms;
}

static float runtime_threshold(const char *name, float configured, float fallback)
{
    if (configured >= 0.0f) return configured;
    const char *value = getenv(name);
    return value ? strtof(value, nullptr) : fallback;
}

static void dump_tensor_attr(rknn_tensor_attr *attr)
{
    printf("  index=%d, name=%s, n_dims=%d, dims=[%d, %d, %d, %d], n_elems=%d, size=%d, fmt=%s, type=%s, qnt_type=%s, "
           "zp=%d, scale=%f\n",
           attr->index, attr->name, attr->n_dims, attr->dims[0], attr->dims[1], attr->dims[2], attr->dims[3],
           attr->n_elems, attr->size, get_format_string(attr->fmt), get_type_string(attr->type),
           get_qnt_type_string(attr->qnt_type), attr->zp, attr->scale);
}

int init_yolov8_seg_model(const char *model_path, rknn_app_context_t *app_ctx)
{

    int ret;
    int model_len = 0;
    char *model;
    rknn_context ctx = 0;

    // Load RKNN Model
    model_len = read_data_from_file(model_path, &model);
    if (model == NULL)
    {
        printf("load_model fail!\n");
        return -1;
    }

    ret = rknn_init(&ctx, model, model_len, 0, NULL);
    free(model);
    if (ret < 0)
    {
        printf("rknn_init fail! ret=%d\n", ret);
        return -1;
    }

    // Get Model Input Output Number
    rknn_input_output_num io_num;
    ret = rknn_query(ctx, RKNN_QUERY_IN_OUT_NUM, &io_num, sizeof(io_num));
    if (ret != RKNN_SUCC)
    {
        printf("rknn_query fail! ret=%d\n", ret);
        return -1;
    }
    printf("model input num: %d, output num: %d\n", io_num.n_input, io_num.n_output);

    // Get Model Input Info
    printf("input tensors:\n");
    rknn_tensor_attr input_attrs[io_num.n_input];
    memset(input_attrs, 0, sizeof(input_attrs));
    for (int i = 0; i < io_num.n_input; i++)
    {
        input_attrs[i].index = i;
        ret = rknn_query(ctx, RKNN_QUERY_INPUT_ATTR, &(input_attrs[i]), sizeof(rknn_tensor_attr));
        if (ret != RKNN_SUCC)
        {
            printf("rknn_query fail! ret=%d\n", ret);
            return -1;
        }
        dump_tensor_attr(&(input_attrs[i]));
    }

    // Get Model Output Info
    printf("output tensors:\n");
    rknn_tensor_attr output_attrs[io_num.n_output];
    memset(output_attrs, 0, sizeof(output_attrs));
    for (int i = 0; i < io_num.n_output; i++)
    {
        output_attrs[i].index = i;
        ret = rknn_query(ctx, RKNN_QUERY_OUTPUT_ATTR, &(output_attrs[i]), sizeof(rknn_tensor_attr));
        if (ret != RKNN_SUCC)
        {
            printf("rknn_query fail! ret=%d\n", ret);
            return -1;
        }
        dump_tensor_attr(&(output_attrs[i]));
    }

    // Set to context
    app_ctx->rknn_ctx = ctx;

    // Select quantized decoding from the output tensor attributes.
    if (output_attrs[0].qnt_type == RKNN_TENSOR_QNT_AFFINE_ASYMMETRIC && output_attrs[0].type != RKNN_TENSOR_FLOAT16)
    {
        app_ctx->is_quant = true;
    }
    else
    {
        app_ctx->is_quant = false;
    }


    app_ctx->io_num = io_num;
    app_ctx->input_attrs = (rknn_tensor_attr *)malloc(io_num.n_input * sizeof(rknn_tensor_attr));
    memcpy(app_ctx->input_attrs, input_attrs, io_num.n_input * sizeof(rknn_tensor_attr));
    app_ctx->output_attrs = (rknn_tensor_attr *)malloc(io_num.n_output * sizeof(rknn_tensor_attr));
    memcpy(app_ctx->output_attrs, output_attrs, io_num.n_output * sizeof(rknn_tensor_attr));

    if (input_attrs[0].fmt == RKNN_TENSOR_NCHW)
    {
        printf("model is NCHW input fmt\n");
        app_ctx->model_channel = input_attrs[0].dims[1];
        app_ctx->model_height = input_attrs[0].dims[2];
        app_ctx->model_width = input_attrs[0].dims[3];
    }
    else
    {
        printf("model is NHWC input fmt\n");
        app_ctx->model_height = input_attrs[0].dims[1];
        app_ctx->model_width = input_attrs[0].dims[2];
        app_ctx->model_channel = input_attrs[0].dims[3];
    }
    printf("model input height=%d, width=%d, channel=%d\n",
           app_ctx->model_height, app_ctx->model_width, app_ctx->model_channel);

    return 0;
}

int release_yolov8_seg_model(rknn_app_context_t *app_ctx)
{
    if (app_ctx->input_attrs != NULL)
    {
        free(app_ctx->input_attrs);
        app_ctx->input_attrs = NULL;
    }
    if (app_ctx->output_attrs != NULL)
    {
        free(app_ctx->output_attrs);
        app_ctx->output_attrs = NULL;
    }
    if (app_ctx->rknn_ctx != 0)
    {
        rknn_destroy(app_ctx->rknn_ctx);
        app_ctx->rknn_ctx = 0;
    }
    return 0;
}

int inference_yolov8_seg_model(rknn_app_context_t *app_ctx, image_buffer_t *img, object_detect_result_list *od_results)
{
    int ret;
    image_buffer_t dst_img;
    letterbox_t letter_box;
    rknn_input inputs[app_ctx->io_num.n_input];
    rknn_output outputs[app_ctx->io_num.n_output];
    const float nms_threshold = runtime_threshold("YOLO_NMS_THRESHOLD", runtime_nms_threshold, NMS_THRESH);
    const float box_conf_threshold = runtime_threshold("YOLO_CONF_THRESHOLD", runtime_confidence_threshold, BOX_THRESH);
    int bg_color = 114; // pad color for letterbox
    int64_t started_us = 0;
    int64_t elapsed_us = 0;

    if ((!app_ctx) || !(img) || (!od_results))
    {
        return -1;
    }

    memset(od_results, 0x00, sizeof(*od_results));
    memset(&letter_box, 0, sizeof(letterbox_t));
    memset(&dst_img, 0, sizeof(image_buffer_t));
    memset(inputs, 0, sizeof(inputs));
    memset(outputs, 0, sizeof(outputs));
    runtime_preprocess_ms = 0.0;
    runtime_inference_ms = 0.0;
    runtime_postprocess_ms = 0.0;

    // Pre Process
    started_us = runtime_time_us();
    app_ctx->input_image_width = img->width;
    app_ctx->input_image_height = img->height;
    dst_img.width = app_ctx->model_width;
    dst_img.height = app_ctx->model_height;
    dst_img.format = IMAGE_FORMAT_RGB888;
    dst_img.size = get_image_size(&dst_img);
    dst_img.virt_addr = (unsigned char *)malloc(dst_img.size);
    if (dst_img.virt_addr == NULL)
    {
        printf("malloc buffer size:%d fail!\n", dst_img.size);
        return -1;
    }

    // letterbox
#if defined(YOLO_RUNTIME_SHARED_PREPROCESS)
    ret = yolo_convert_image_with_letterbox(img, &dst_img, &letter_box, bg_color);
#else
    ret = convert_image_with_letterbox(img, &dst_img, &letter_box, bg_color);
#endif
    if (ret < 0)
    {
        printf("convert_image_with_letterbox fail! ret=%d\n", ret);
        goto out;
    }
    // Set Input Data
    inputs[0].index = 0;
    inputs[0].type = RKNN_TENSOR_UINT8;
    inputs[0].fmt = RKNN_TENSOR_NHWC;
    inputs[0].size = app_ctx->model_width * app_ctx->model_height * app_ctx->model_channel;
    inputs[0].buf = dst_img.virt_addr;

    ret = rknn_inputs_set(app_ctx->rknn_ctx, app_ctx->io_num.n_input, inputs);
    if (ret < 0)
    {
        printf("rknn_input_set fail! ret=%d\n", ret);
        goto out;
    }
    runtime_preprocess_ms = (runtime_time_us() - started_us) / 1000.0;

    // Run
    printf("rknn_run\n");
    started_us = runtime_time_us();
    ret = rknn_run(app_ctx->rknn_ctx, nullptr);
    elapsed_us = runtime_time_us() - started_us;
    runtime_inference_ms = elapsed_us / 1000.0;
    printf("rknn_run time=%.2fms, FPS = %.2f\n", elapsed_us / 1000.f,
           elapsed_us > 0 ? 1000.f * 1000.f / elapsed_us : 0.f);
    if (ret < 0)
    {
        printf("rknn_run fail! ret=%d\n", ret);
        goto out;
    }

    // Get Output
    memset(outputs, 0, sizeof(outputs));
    for (int i = 0; i < app_ctx->io_num.n_output; i++)
    {
        outputs[i].index = i;
        outputs[i].want_float = (!app_ctx->is_quant);
    }
    ret = rknn_outputs_get(app_ctx->rknn_ctx, app_ctx->io_num.n_output, outputs, NULL);
    if (ret < 0)
    {
        printf("rknn_outputs_get fail! ret=%d\n", ret);
        goto out;
    }

    // Post Process
    started_us = runtime_time_us();
    post_process(app_ctx, outputs, &letter_box, box_conf_threshold, nms_threshold, od_results);
    elapsed_us = runtime_time_us() - started_us;
    runtime_postprocess_ms = elapsed_us / 1000.0;
    printf("post_process time=%.2fms\n", elapsed_us / 1000.f);

    // Remeber to release rknn output
    rknn_outputs_release(app_ctx->rknn_ctx, app_ctx->io_num.n_output, outputs);

out:
    if (dst_img.virt_addr != NULL)
    {
        free(dst_img.virt_addr);
    }

    return ret;
}