编程注意（Programming attention）.doc

编程注意（Programming attention） C++ level code optimization Http:/// When it comes to optimization, many people think of assembly directly. Can optimization be at the level of compilation? Of course not, as in the C++ layer, you can do code optimizations, some of which are often unexpected. Determine floating point variables and expressions that are type float In order for the compiler to produce better code (for example, to produce 3DNow or SSE instructions), floating-point variables and expressions must be determined to be of type float. It is important to note that the floating point constant is float with the suffix F or F (for example, 3.14f), otherwise the default is double. To avoid the automatic conversion of type float parameters to double, use float when function declarations are used. Use 32 bit data type Use 32 bit data type. There are many compilers, 32 types of typical but are included are: int, signed, signed, int, unsigned, unsigned, int, long, signed long, long int, signed long int, unsigned long, unsigned long int. Consider whether an integer variable has symbols In many cases, you need to consider whether an integer variable is a signed or unsigned type. For example, there is no negative number to preserve a persons weight data, so there is no need to use symbolic types. However, if you want to preserve the temperature data, you must use the signed variable. In many places, it is necessary to consider whether or not to use symbolic variables. In some cases, symbolic operations are faster, but in some cases the opposite is true. Integer to floating point transformation, Using symbolic integers greater than 16 bits is faster. Because the x86 architecture provides instructions from Symbolic Integration to floating point, no unsigned integer conversion to floating point instructions is provided. The compiler generates assembly code as follows: A bad example: Compile before compile Double x; mov [temp + 4], 0 Unsigned, int, I; MOV, eax, I X = I; MOV, [temp