The CPU we usually use is a fixed-point CPU, meaning the CPU of the entire point numerical operation. When you encounter a floating point of 1.1 + 1.1, the fixed-point CPU encounters a big problem. For 32-bit single-chip microcomputer, use q-based processing to play his own performance, but accuracy and speed still will not increase much.
A new CPU is now designed, called FPU, this chip dedicates the operation of floating point numbers, so that the processor is separated from the number of points and floating point numbers, and the total number of dotted CPUs is handed over, the floating point number is handed over. We have seen Ti's DSP, as well as a Microcontroller with a DSP function. The former is not used, does not comment, and the latter must make some necessary settings if you need to use the FPU floating point calculation function.
First, since the floating point operation is carried out in the FPU, the FPU should be enabled first. In system_init (), define __fpu_present and __fpu_used
/ * FPU SETTINGS ---------------------------------------------------------------------------------------------------------------------------------------- -------------- * /
#If (__fpu_present == 1) && (__fpu_used == 1)
SCB- "CPACR | = ((3 ul" "10 * 2) | (3 ul" "11 * 2)); / * SET CP10 and CP11 Full Access * /
#ENDIF
This will enable the FPU.
For the above changes, this simple addition and subtraction and division operation FPU appears in the program works. But for complex, such as triangular operations, the opening operation, etc., we need to join the Math.h header file. But if you add him simple, Keil will automatically call the internal Math.h, which is designed for the ARM processor, specifically for the fixed-point CPU and standard algorithm (IEEE-754). There is no effect on STM32F4 using the FPU. Therefore, it is necessary to change Math.h to the ST library, namely arm_math.h. In this header file, it involves another file Core_cmx.h (x = 0, 3, 4), of course, as in the STM32F1 series, add CORE_CM4.H in the project.
When it is here, it is all set, the difference is the last step, call! But don't underestimate this step, because if the call is incorrect, the settings in front are white. Do not write as above, because the triangle function is used as Sin (), because the name of their function comes from Math.h, so you call the standard Math.h in the Keil library. To use the ARM_SIN_F32 () function in ARM_Math.h (see line.5780, see DSP_LIB \ Source \ FastMathFunctions), you can see that he uses three spline interpolation methods to quickly evaluate (see line.263 / * cubic Interpolation process * /).
Note the exception function, SQRT (), in ARM_Math.H is ARM_SQRT_F32 (). When you use him, you need to open #if (__fpu_used == 1) && defined (__cc_arm), remember! It can also be found that the top function is also divided into Q15 and Q31. I think their difference is the accuracy problem, but they don't have the FPU to calculate. It is to use the 0x5f3759df to quickly and open. If you are very Unfamiliar, view http://en.wikipedia.org/wiki/fast_inverse_square_root. However, his handling may be different.
There are also many DSP functions in DSP_LIB \ SOURCE, interested in research. , Reading the full text, the technology area
MYBATIS Dynamic SQL Detailed
Design the HDB3 encoder in the data transmission system with VHDL language
How does the bare metal program drive hardware? What is the seniors say?
Typical examples of object-oriented programming SOC principles
Introduction to Java Common Development Tools for Embedded Development
Our other product:
