BroadcastTo

将较小张量按 N-D 规则广播到目标形状并写入输出。

输入：

  • input - 输入数据地址。

  • input_shape - 输入形状数组。

  • input_shape_size - 输入形状长度。

  • output_shape - 目标输出形状数组。

  • output_shape_size - 输出形状长度。

  • data_size - 单个元素字节数（例如 sizeof(float)）。

  • core_mask(int, 可选) - 核掩码（仅适用于共享存储版本）。

输出：

  • output - 输出数据地址。

支持平台：

FT78NE MT7004

备注

  • FT78NE 支持的数据类型： fp32、fp64、cplx64、cplx128、int16、int8、int32。

  • MT7004 支持的数据类型： fp32、fp16、cplx64、int16、int32。

共享存储版本:

void i8_broadcastto_s(int8_t *input, int8_t *output, const int *input_shape, int input_shape_size, const int *output_shape, int output_shape_size, int data_size, int core_mask)
void i16_broadcastto_s(int16_t *input, int16_t *output, const int *input_shape, int input_shape_size, const int *output_shape, int output_shape_size, int data_size, int core_mask)
void i32_broadcastto_s(int32_t *input, int32_t *output, const int *input_shape, int input_shape_size, const int *output_shape, int output_shape_size, int data_size, int core_mask)
void hp_broadcastto_s(half *input, half *output, const int *input_shape, int input_shape_size, const int *output_shape, int output_shape_size, int data_size, int core_mask)
void fp_broadcastto_s(float *input, float *output, const int *input_shape, int input_shape_size, const int *output_shape, int output_shape_size, int data_size, int core_mask)
void dp_broadcastto_s(double *input, double *output, const int *input_shape, int input_shape_size, const int *output_shape, int output_shape_size, int data_size, int core_mask)
void c64_broadcastto_s(float *input, float *output, const int *input_shape, int input_shape_size, const int *output_shape, int output_shape_size, int data_size, int core_mask)
void c128_broadcastto_s(double *input, double *output, const int *input_shape, int input_shape_size, const int *output_shape, int output_shape_size, int data_size, int core_mask)

C 调用示例：

 1// FT78NE 多核示例
 2#include <stdio.h>
 3
 4int main(int argc, char *argv[]) {
 5    float *input = (float *)0xA0000000;    // 输入在 DDR 地址 0xA0000000
 6    float *output = (float *)0xB0000000;   // 输出在 DDR 地址 0xB0000000
 7    int input_shape[4] = {1, 50, 1, 1};
 8    int output_shape[4] = {10, 50, 20, 1};
 9    int core_mask = 0xff;
10    fp_broadcastto_s(input, output, input_shape, 4, output_shape, 4, sizeof(float), core_mask);
11    return 0;
12}

私有存储版本:

void i8_broadcastto_p(int8_t *input, int8_t *output, const int *input_shape, int input_shape_size, const int *output_shape, int output_shape_size, int data_size)
void i16_broadcastto_p(int16_t *input, int16_t *output, const int *input_shape, int input_shape_size, const int *output_shape, int output_shape_size, int data_size)
void i32_broadcastto_p(int32_t *input, int32_t *output, const int *input_shape, int input_shape_size, const int *output_shape, int output_shape_size, int data_size)
void hp_broadcastto_p(half *input, half *output, const int *input_shape, int input_shape_size, const int *output_shape, int output_shape_size, int data_size)
void fp_broadcastto_p(float *input, float *output, const int *input_shape, int input_shape_size, const int *output_shape, int output_shape_size, int data_size)
void dp_broadcastto_p(double *input, double *output, const int *input_shape, int input_shape_size, const int *output_shape, int output_shape_size, int data_size)
void c64_broadcastto_p(float *input, float *output, const int *input_shape, int input_shape_size, const int *output_shape, int output_shape_size, int data_size)
void c128_broadcastto_p(double *input, double *output, const int *input_shape, int input_shape_size, const int *output_shape, int output_shape_size, int data_size)

C 调用示例：

 1// FT78NE 单核示例
 2#include <stdio.h>
 3
 4int main(int argc, char *argv[]) {
 5    float *input = (float *)0x10000000;   // 单核版本：输入放在 L2 地址 0x10000000
 6    float *output = (float *)0x10020000;  // 单核版本：输出放在 L2 地址 0x10020000
 7    int input_shape[4] = {1, 50, 1, 1};
 8    int output_shape[4] = {10, 50, 20, 1};
 9    fp_broadcastto_p(input, output, input_shape, 4, output_shape, 4, sizeof(float));
10    return 0;
11}