AddN
=================



对多个输入张量进行 **逐元素相加**，并将结果输出；
支持标量广播。

数学表达式为：

.. math::

    \text{output} = \sum_{i=0}^{tensor\_num-1}\text{inputs[i]}

输入：
    - **inputs** - 输入数据地址数组。
    - **tensor_num** - 输入张量数量。
    - **element_nums** - 输入张量元素数量数组。
    - **core_mask** - 核掩码（仅适用于共享存储版本）。

输出：
    - **output** - 计算结果地址。

支持平台：
    ``FT78NE``
    ``MT7004``

.. note::
    - FT78NE 支持 ``fp32``、``fp64``、``int8``、``int16``、``int32``、``cplx64``、``cplx128`` 类型
    - MT7004 支持 ``fp16``、``fp32``、``int16``、``int32``、``cplx64`` 类型

**共享存储版本:**

.. c:function:: void hp_addn_s(half **inputs, half *output, int tensor_num, int *element_nums, int core_mask)
.. c:function:: void fp_addn_s(float **inputs, float *output, int tensor_num, int *element_nums, int core_mask)
.. c:function:: void dp_addn_s(double **inputs, double *output, int tensor_num, int *element_nums, int core_mask)
.. c:function:: void i8_addn_s(int8_t **inputs, int8_t *output, int tensor_num, int *element_nums, int core_mask)
.. c:function:: void i16_addn_s(int16_t **inputs, int16_t *output, int tensor_num, int *element_nums, int core_mask)
.. c:function:: void i32_addn_s(int32_t **inputs, int32_t *output, int tensor_num, int *element_nums, int core_mask)
.. c:function:: void c64_addn_s(float **inputs, float *output, int tensor_num, int *element_nums, int core_mask)
.. c:function:: void c128_addn_s(double **inputs, double *output, int tensor_num, int *element_nums, int core_mask)

**C调用示例：**

.. code-block:: c
    :linenos:
    :emphasize-lines: 17

    // FT78NE 示例
    #include <stdio.h>
    #include <addn.h>

    int main(int argc, char* argv[]) {
        float *input0 = (float *)0xA0000000;
        float *input1 = (float *)0xA0010000;
        float **inputs = (float **)0xB0000000;
        inputs[0] = input0;
        inputs[1] = input1;
        float *output = (float *)0xC0000000;
        int tensor_num = 2;
        int *element_nums = (int *)0xB0001000;
        element_nums[0] = 1024;
        element_nums[1] = 1024;
        int core_mask = 0xff;
        fp_addn_s(inputs, output, tensor_num, element_nums, core_mask);
        return 0;
    }


**私有存储版本:**

.. c:function:: void hp_addn_p(half **inputs, half *output, int tensor_num, int *element_nums)
.. c:function:: void fp_addn_p(float **inputs, float *output, int tensor_num, int *element_nums)
.. c:function:: void dp_addn_p(double **inputs, double *output, int tensor_num, int *element_nums)
.. c:function:: void i8_addn_p(int8_t **inputs, int8_t *output, int tensor_num, int *element_nums)
.. c:function:: void i16_addn_p(int16_t **inputs, int16_t *output, int tensor_num, int *element_nums)
.. c:function:: void i32_addn_p(int32_t **inputs, int32_t *output, int tensor_num, int *element_nums)
.. c:function:: void c64_addn_p(float **inputs, float *output, int tensor_num, int *element_nums)
.. c:function:: void c128_addn_p(double **inputs, double *output, int tensor_num, int *element_nums)


**C调用示例：**

.. code-block:: c
    :linenos:
    :emphasize-lines: 16

    // FT78NE 示例
    #include <stdio.h>
    #include <addn.h>

    int main(int argc, char* argv[]) {
        float *input0 = (float *)0x10810000;   // L2 空间
        float *input1 = (float *)0x10814000;
        float **inputs = (float **)0x10818000;
        inputs[0] = input0;
        inputs[1] = input1;
        float *output = (float *)0x10820000;
        int tensor_num = 2;
        int *element_nums = (int *)0x1082C000;
        element_nums[0] = 1024;
        element_nums[1] = 1024;
        fp_addn_p(inputs, output, tensor_num, element_nums);
        return 0;
    }
