c，汇编速度的比较

最新推荐文章于 2022-12-07 23:43:33 发布

歌神flaming

最新推荐文章于 2022-12-07 23:43:33 发布

阅读量1.4k

点赞数

分类专栏：工具文章标签： c .s pld neon

本文链接：https://blog.csdn.net/u010684585/article/details/78483138

版权

本文通过对比C语言和汇编代码的执行效率，探讨了使用ARM NEON指令集进行多媒体计算的性能提升。实验中，分别测试了不同NEON指令优化策略，包括单寄存器、多个寄存器以及预加载数据的方式，并分析了它们对运算速度的影响。

摘要由CSDN通过智能技术生成

#include "opencv2/core/core.hpp"
#include "opencv2/highgui/highgui.hpp"
#include <chrono>
#include <arm_neon.h>
#include "opencv2/imgproc/imgproc.hpp"
//extern "C" void fanicCvtBGR2GrayNEON(void *pDst, void *pSrc, unsigned int size);
extern "C" void sfNEONbad1(unsigned short * pDst, short * pSrc, short coeff, short intercept, unsigned int count);//single register
extern "C" void sfNEONbad2(unsigned short * pDst, short * pSrc, short coeff, short intercept, unsigned int count);//serveral registers
extern "C" void sfNEONbad3(unsigned short * pDst, short * pSrc, short coeff, short intercept, unsigned int count);//preload
class q_timer {
public:
void start()
{
m_start = std::chrono::steady_clock::now();
}
double stop()
{
std::chrono::steady_clock::time_point end = std::chrono::steady_clock::now();
return std::chrono::duration_cast<std::chrono::duration<double> >(end - m_start).count();
}
void time_display(const char *disp = "", int nr_frame = 1)
{
printf("Running time (%s) is: %5.5f Seconds.\n", disp, stop() / nr_frame);
}
void fps_display(const char *disp = "", int nr_frame = 1)
{
printf("Running time (%s) is: %5.5f frame per