Start with stm32f4 floating point operation (FPU) function + use DSP Library

The parameters involved ： Here we use stm32F4  MDK5.2 For example

（1） Configuration on FPU Functional logo

#define   __FPU_PRESENT       1U   // stay stm32f4xx.h The system has been configured

#define   __FPU_USED          1U  // This is in core_cm4.h in , Need configuration

As you can see from the picture , want First Judge __CC_ARM Whether to define , Again Judge __TARGET_FPU_VFP Whether to define , Last Judge __FPU_PRESENT Has been configured to 1, front 3 When both conditions are true , only To configure __FPU_USED=1; It can be simplified , Configure... Directly in the user program __FPU_USED=1 that will do ;

__CC_ARM：  #define   __CC_ARM     //1 Some systems will configure themselves

__TARGET_FPU_VFP： The defined equipment has FPU function , So add the next sentence ：

  #define   __TARGET_FPU_VFP    //1

Of course __CC_ARM、__TARGET_FPU_VFP It can also be in MDK->Option->C/C++->… In the definition of , Here's the picture

(2) Floating point configuration function at void SystemInit(void) That is to say system_stm32f4xx.c in

void SystemInit(void)

{

  /* FPU settings ------------------------------------------------------------*/

  #if (__FPU_PRESENT == 1) && (__FPU_USED == 1)

    SCB->CPACR |= ((3UL << 10*2)|(3UL << 11*2));  /* set CP10 and CP11 Full Access */

  #endif

That's it    In fact, it can not be configured __FPU_PRESENT and  __FPU_PRESENT, Directly compile the condition to delete , This procedure works ; If some libraries do not have these configurations , Then the user can add

…………………..

}

(3) Add header file

#include “arm_math.h”  Use STM32 Its own mathematical processing function , Simultaneous addition arm_math.h Header file path , At this time, the compiler will report an error , Tips ：#error "Define according the used Cortex core ARM_MATH_CM7, ARM_MATH_CM4, ARM_MATH_CM3, ARM_MATH_CM0PLUS or ARM_MATH_CM0", It means that there is no definition of kernel mathematical type , solve ,2 A way ：

Method 1： stay MDK Configure... In the software , I'm using M4, choice ARM_MATH_CM4

Method 2： The user can define ARM_MATH_CM4 that will do

2 There are two ways to see which users like

（4）MDK The configuration is as follows , Select... From the floating point options ：Use Single Precision

If you choose Not Used Error will be reported at compile time

The above configuration is already available STM32 Floating point operations for FPU The function of , test result , use 2 Multiply decimals , Online simulation , Look at what appears in disassembly xxx.F32 sentence , It means that FPU function

Others will be in MDK To configure C++ define Medium plus ARM_MATH_MATRIX_CHECK,ARM_MATH_ROUNDING These two configurations

ARM_MATH_ROUNDING It's the rounding function , Can not add ;

ARM_MATH_MATRIX_CHECK Is the matrix parameter check function , Can not add

（5） If you use STM32 Self contained dsp The operation function also needs to add the corresponding library function

The above is just open STM32 Internal hardware FPU Floating point operation function , There is no mention of DSP

Want to use DSP There's still some work to be done

1. Join in dsp Math library （ You can add it to any folder ）

 

 

 

arm_cortexM4lf_math.lib What is the function ？

This is the following specific dsp An assembly of functions or a library compiled from them , The following specific functions are c Compiling , If you make your employees do this directly , It will lead to low efficiency , So use the corresponding assembly function to realize , Don't pay attention to this lib file , If you want to know the specific function , You can refer to the following instructions , Look at specific .C File to find the corresponding function

arm_cortexM4lf_math.lib Already included all dsp Specific function , In fact, just add this , This .lib The file cannot be opened ;  It is not necessary to add the following corresponding .C file , Provided here .C Files just let designers know how to call functions , And internal program implementation steps

But not vice versa , If you add the corresponding c file , Instead of joining lib The file will compile and report an error

2. The specific functions of the official tape dsp function library （ This does not have to be added , It's just that the designer knows , Know the function name ）

3. Usage method

For example, use sine sin To join arm_sin_f32.c（ No need ）

cosine cos To join  arm_cos_f32.c call （ No need ）

There is no corresponding header file , To add other header files , You can see the specific c file

Open the corresponding library .c file , There is a function name , When actually using it, just press this , You can't do as before math.h The corresponding function name in

CosVal=arm_cos_f32(radians);// Calculate the cosine function

You can only join arm_cortexM4lf_math.lib file , Then, the function of the corresponding function c File removal project , Continue to call the corresponding dsp function , Same operation

exception handling ：

Sometimes configuration FPU Incorrect , There will be HardFault_Handler, as a result of : The compiler generates floating-point instructions , But the hardware does not turn on the floating-point operation function , Such an operation is wrong , Check whether the configuration is correct