qfplib次用汇编实现，是一种用来在ARM 单片机上运行的软件浮点库，适用于ARM Cortex- M0/M0+/M1/M23/M3/M4/M7/M33/M55等ARM MCU，特别是适用于M0/M0+，因为这些和是不具备硬件浮点运算的，本期测评主要是在STM32F769I-DISCO上测试。

1。首先下载qfplib

https://www.quinapalus.com/qfplib.html

我们点击M3

进行下载

复制到工程文件夹

2。打开我们之前建立好的串口工程，串口打印这里不在赘述。

添加这两个，上面是用来测量算法时间，下面是qfplib

同时添加文件

3。main文件中进行修改

添加头文件

include "main.h"

/ Private includes ----------------------------------------------------------/ / USER CODE BEGIN Includes /

include "stdio.h"

include "code_time_measurement.h"

include "qfplib-m3.h"

include "math.h"

4。先进性qfplib浮点库性能测试，修改代码如下

printf("======================qfplib==========================\r\n"); CODE_TIME_START("qfplib浮点运算"); printf("qfp_fsqrt = %f\r\n", qfp_fsqrt(3.4f));

CODE_TIME_EVENT("qfp_fsqrt(3.4f)"); printf("qfp_fexp = %f\r\n", qfp_fexp(4.124f));

CODE_TIME_EVENT("qfp_fexp(4.124f)"); printf("qfp_fln = %f\r\n", qfp_fln(88.124f));

CODE_TIME_EVENT("qfp_fln(88.124f)"); printf("qfp_fsin = %f\r\n", qfp_fsin(3.1415926f/4));

CODE_TIME_EVENT("qfp_fsin(3.1415926f/4)"); printf("qfp_fcos = %f\r\n", qfp_fcos(3.1415926f/4));

CODE_TIME_EVENT("qfp_fcos(3.1415926f/4)"); printf("qfp_ftan = %f\r\n", qfp_ftan(3.1415926f/4)); CODE_TIME_EVENT("qfp_ftan(3.1415926f/4)");

printf("qfp_fatan2 = %f\r\n", qfp_fatan2(1.0f, 2.0f)); CODE_TIME_EVENT("qfp_fatan2(1.0f, 2.0f)");

CODE_TIME_STOP();

一定要关闭硬件浮点运算，操作如上图

串口打印如下图

5。进行硬件浮点运算，修改代码如下

printf("======================hardfpu==========================\r\n"); CODE_TIME_START("硬件浮点运算"); printf("sqrtf = %f\r\n", sqrtf(3.4f)); CODE_TIME_EVENT("sqrtf(3.4f)"); printf("expf = %f\r\n", expf(4.124f)); CODE_TIME_EVENT("expf(4.124f)"); printf("logf = %f\r\n", logf(88.124f)); CODE_TIME_EVENT("logf(88.124f)"); printf("sinf = %f\r\n", sinf(3.1415926f/4)); CODE_TIME_EVENT("sinf(3.1415926f/4)"); printf("cosf = %f\r\n", cosf(3.1415926f/4)); CODE_TIME_EVENT("cosf(3.1415926f/4)"); printf("tanf = %f\r\n", tanf(3.1415926f/4)); CODE_TIME_EVENT("tanf(3.1415926f/4)"); printf("atan2f = %f\r\n", atan2f(1.0f, 2.0f)); CODE_TIME_EVENT("atan2f(1.0f, 2.0f)");

CODE_TIME_STOP();

同时使能硬件浮点运算

串口打印输出

6。两者进行对比

发现qfplib的浮点运行性能和硬件浮点运算性能相差无几，几乎达到了硬件浮点性能，这使得在M0/M0+/M1/M2等不具备硬件浮点运算的ARM内核可以运行浮点运算，非常完美的库！！！！

大家可以下载移植测试！！！！！！！！！！！！！！！！！