119 lines
3.2 KiB
C
119 lines
3.2 KiB
C
|
/*
|
|||
|
|
* ===========================================================================================
|
|||
|
|
*
|
|||
|
|
* Filename: neon.c
|
|||
|
|
*
|
|||
|
|
* Description: neon useage.
|
|||
|
|
*
|
|||
|
|
* Version: Melis3.0
|
|||
|
|
* Create: 2020-03-13 10:49:09
|
|||
|
|
* Revision: none
|
|||
|
|
* Compiler: GCC:version 7.2.1 20170904 (release),ARM/embedded-7-branch revision 255204
|
|||
|
|
*
|
|||
|
|
* Author: caozilong@allwinnertech.com
|
|||
|
|
* Organization: BU1-PSW
|
|||
|
|
* Last Modified: 2020-03-20 09:04:00
|
|||
|
|
*
|
|||
|
|
* ===========================================================================================
|
|||
|
|
*/
|
|||
|
|
#include <rtthread.h>
|
|||
|
|
#include <arm_neon.h>
|
|||
|
|
#include <dfs_posix.h>
|
|||
|
|
#include <finsh_api.h>
|
|||
|
|
#include <finsh.h>
|
|||
|
|
#include <debug.h>
|
|||
|
|
#include <log.h>
|
|||
|
|
|
|||
|
|
uint32_t kernel_neon_begin(void);
|
|||
|
|
void kernel_neon_end(uint32_t saved_fpexc);
|
|||
|
|
|
|||
|
|
static float sum_array(float *arr, int len)
|
|||
|
|
{
|
|||
|
|
uint32_t flags;
|
|||
|
|
if (NULL == arr || len < 1)
|
|||
|
|
{
|
|||
|
|
__err("parameter error.");
|
|||
|
|
return 0;
|
|||
|
|
}
|
|||
|
|
|
|||
|
|
flags = kernel_neon_begin();
|
|||
|
|
int dim4 = len >> 2; // 数组长度除4整数
|
|||
|
|
int left4 = len & 3; // 数组长度除4余数
|
|||
|
|
float32x4_t sum_vec = vdupq_n_f32(0.0);//定义用于暂存累加结果的寄存器且初始化为0
|
|||
|
|
for (; dim4 > 0; dim4--, arr += 4) //每次同时访问4个数组元素
|
|||
|
|
{
|
|||
|
|
float32x4_t data_vec = vld1q_f32(arr); //依次取4个元素存入寄存器vec
|
|||
|
|
sum_vec = vaddq_f32(sum_vec, data_vec);//ri = ai + bi 计算两组寄存器对应元素之和并存放到相应结果
|
|||
|
|
}
|
|||
|
|
float sum = vgetq_lane_f32(sum_vec, 0) + vgetq_lane_f32(sum_vec, 1) + vgetq_lane_f32(sum_vec, 2) + vgetq_lane_f32(sum_vec, 3); //将累加结果寄存器中的所有元素相加得到最终累加值
|
|||
|
|
for (; left4 > 0; left4--, arr++)
|
|||
|
|
{
|
|||
|
|
sum += (*arr) ; //对于剩下的少于4的数字，依次计算累加即可
|
|||
|
|
}
|
|||
|
|
kernel_neon_end(flags);
|
|||
|
|
return sum;
|
|||
|
|
}
|
|||
|
|
|
|||
|
|
static void init_array(float *arr, int len)
|
|||
|
|
{
|
|||
|
|
int i = 0;
|
|||
|
|
|
|||
|
|
for (i = 0; i < len; i ++)
|
|||
|
|
{
|
|||
|
|
arr[i] = (i * 1000) / len * 0.38;
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
|
|||
|
|
static void neon1_entry(void *para)
|
|||
|
|
{
|
|||
|
|
float array[123];
|
|||
|
|
|
|||
|
|
init_array(array, 123);
|
|||
|
|
|
|||
|
|
while (1)
|
|||
|
|
{
|
|||
|
|
float sum = sum_array(array, 123);
|
|||
|
|
printf("[DBG]: %s line %d, float sum %f, size %d.\n", __func__, __LINE__, sum, sizeof(__simd64_float32_t));
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
|
|||
|
|
static void neon2_entry(void *para)
|
|||
|
|
{
|
|||
|
|
float array[255];
|
|||
|
|
|
|||
|
|
init_array(array, 255);
|
|||
|
|
|
|||
|
|
while (1)
|
|||
|
|
{
|
|||
|
|
float sum = sum_array(array, 255);
|
|||
|
|
printf("[DBG]: %s line %d, float sum %f, size %d.\n", __func__, __LINE__, sum, sizeof(__simd64_float32_t));
|
|||
|
|
}
|
|||
|
|
}
|
|||
|
|
|
|||
|
|
static void armneon_entry(void)
|
|||
|
|
{
|
|||
|
|
rt_thread_t neon_thread1, neon_thread2;
|
|||
|
|
|
|||
|
|
rt_enter_critical();
|
|||
|
|
neon_thread1 = rt_thread_create("neon1", neon1_entry, RT_NULL, 0x1000, 1, 10);
|
|||
|
|
RT_ASSERT(neon_thread1 != RT_NULL);
|
|||
|
|
rt_thread_startup(neon_thread1);
|
|||
|
|
|
|||
|
|
neon_thread2 = rt_thread_create("neon2", neon2_entry, RT_NULL, 0x1000, 1, 10);
|
|||
|
|
RT_ASSERT(neon_thread2 != RT_NULL);
|
|||
|
|
rt_thread_startup(neon_thread2);
|
|||
|
|
rt_exit_critical();
|
|||
|
|
|
|||
|
|
return;
|
|||
|
|
}
|
|||
|
|
|
|||
|
|
|
|||
|
|
static int cmd_armneon(int argc, const char **argv)
|
|||
|
|
{
|
|||
|
|
armneon_entry();
|
|||
|
|
return 0;
|
|||
|
|
}
|
|||
|
|
|
|||
|
|
FINSH_FUNCTION_EXPORT_ALIAS(cmd_armneon, __cmd_armneon, armneon test);
|
|||
|
|
|