STM32F4有DSP和FPU,FPU很容易就能开启了,在system_stm32f4xx.c就有设置:
/* FPU settings ------------------------------------------------------------*/
#if (__FPU_PRESENT == 1) && (__FPU_USED == 1)
SCB->CPACR |= ((3UL << 10*2)|(3UL << 11*2)); /* set CP10 and CP11 Full Access */
#endif
__FPU_PRESENT就是说芯片有没有FPU功能,明显F4是有的,然后__FPU_USED就是说编译器有没有用FPU,明显我们一直都有开启.编译器都很聪明的,知道哪些要FPU哪些不用.所以只要这么设置了,就可以了,万事大吉.
但是DSP库就不一样了,在以前用TI的片子,叫IQMath库,感觉就是浮点转定点做,牺牲精度换取时间.这个浮点转定点,在于其他芯片,也是可以用的,只要是ARM核,在M4可以用DSP.[怀疑态度]
stsw-stm32068STM32F4-Discovery_FW_V1.1.0LibrariesCMSISDSP_Lib
数学库是给我们预先编译好的,当然我们也可以自己编译.但是好像并没有必要.对于我们工程,做以下添加,就可以开始测试Math的东西了.
然后,添加库:
添加声明:
另外还可以做Q标定,详细,看IQMath吧,这不是讨论的范围.在需要Math的地方添加arm_math.h就可以了.示例程序,以及打包:
#include "stm32f4xx.h"
#include "arm_math.h"
#define USE_STATIC_INIT
/* ----------------------------------------------------------------------
** Global defines
** ------------------------------------------------------------------- */
#define TEST_LENGTH_SAMPLES (20*4)
/* ----------------------------------------------------------------------
** List of Marks scored by 20 students for 4 subjects
** ------------------------------------------------------------------- */
const float32_t testMarks_f32[TEST_LENGTH_SAMPLES] =
{
42.000000, 37.000000, 81.000000, 28.000000,
83.000000, 72.000000, 36.000000, 38.000000,
32.000000, 51.000000, 63.000000, 64.000000,
97.000000, 82.000000, 95.000000, 90.000000,
66.000000, 51.000000, 54.000000, 42.000000,
67.000000, 56.000000, 45.000000, 57.000000,
67.000000, 69.000000, 35.000000, 52.000000,
29.000000, 81.000000, 58.000000, 47.000000,
38.000000, 76.000000, 100.000000, 29.000000,
33.000000, 47.000000, 29.000000, 50.000000,
34.000000, 41.000000, 61.000000, 46.000000,
52.000000, 50.000000, 48.000000, 36.000000,
47.000000, 55.000000, 44.000000, 40.000000,
100.000000, 94.000000, 84.000000, 37.000000,
32.000000, 71.000000, 47.000000, 77.000000,
31.000000, 50.000000, 49.000000, 35.000000,
63.000000, 67.000000, 40.000000, 31.000000,
29.000000, 68.000000, 61.000000, 38.000000,
31.000000, 28.000000, 28.000000, 76.000000,
55.000000, 33.000000, 29.000000, 39.000000
};
/* ----------------------------------------------------------------------
* Number of subjects X 1
* ------------------------------------------------------------------- */
const float32_t testUnity_f32[4] =
{
1.000, 1.000, 1.000, 1.000
};
/* ----------------------------------------------------------------------
** f32 Output buffer
** ------------------------------------------------------------------- */
static float32_t testOutput[TEST_LENGTH_SAMPLES];
/* ------------------------------------------------------------------
* Global defines
*------------------------------------------------------------------- */
#define NUMSTUDENTS 20
#define NUMSUBJECTS 4
/* ------------------------------------------------------------------
* Global variables
*------------------------------------------------------------------- */
uint32_t numStudents = 20;
uint32_t numSubjects = 4;
float32_t max_marks, min_marks, mean, std, var;
uint32_t student_num;
/* ----------------------------------------------------------------------------------
* Main f32 test function. It returns maximum marks secured and student number
* ------------------------------------------------------------------------------- */
int32_t main()
{
#ifndef USE_STATIC_INIT
arm_matrix_instance_f32 srcA;
arm_matrix_instance_f32 srcB;
arm_matrix_instance_f32 dstC;
/* Input and output matrices initializations */
arm_mat_init_f32(&srcA, numStudents, numSubjects, (float32_t *)testMarks_f32);
arm_mat_init_f32(&srcB, numSubjects, 1, (float32_t *)testUnity_f32);
arm_mat_init_f32(&dstC, numStudents, 1, testOutput);
#else
/* Static Initializations of Input and output matrix sizes and array */
arm_matrix_instance_f32 srcA = {NUMSTUDENTS, NUMSUBJECTS, (float32_t *)testMarks_f32};
arm_matrix_instance_f32 srcB = {NUMSUBJECTS, 1, (float32_t *)testUnity_f32};
arm_matrix_instance_f32 dstC = {NUMSTUDENTS, 1, testOutput};
#endif
/* ----------------------------------------------------------------------
*Call the Matrix multiplication process function
* ------------------------------------------------------------------- */
arm_mat_mult_f32(&srcA, &srcB, &dstC);
/* ----------------------------------------------------------------------
** Call the Max function to calculate max marks among numStudents
** ------------------------------------------------------------------- */
arm_max_f32(testOutput, numStudents, &max_marks, &student_num);
/* ----------------------------------------------------------------------
** Call the Min function to calculate min marks among numStudents
** ------------------------------------------------------------------- */
arm_min_f32(testOutput, numStudents, &min_marks, &student_num);
/* ----------------------------------------------------------------------
** Call the Mean function to calculate mean
** ------------------------------------------------------------------- */
arm_mean_f32(testOutput, numStudents, &mean);
/* ----------------------------------------------------------------------
** Call the std function to calculate standard deviation
** ------------------------------------------------------------------- */
arm_std_f32(testOutput, numStudents, &std);
/* ----------------------------------------------------------------------
** Call the var function to calculate variance
** ------------------------------------------------------------------- */
arm_var_f32(testOutput, numStudents, &var);
while(1); /* main function does not return */
}
程序下载:DSP_FPU测试
想知道速度提升,可以通过翻转IO,下断点观察,就不详细说了.