Why is python slow?

Python is a dynamically typed, interpreted language. For many developers, it is well known that Python runs slowly. Its characteristic that everything is an object is one of the reasons for its slow running. The following article explains Let me introduce to you some reasons why python is slow. I hope it will be helpful to you.

Python is a dynamic language, not a static language

This means that when the python program is executed, it is compiled The compiler does not know the type of the variable. In C, the compiler knows the type of a variable when it is defined, but in python it only knows that it is an object when executed.

So if you write the following in C:

/ * C代码* /
int  a  =  1 ;
int  b  =  2 ;
int  c  =  a  +  b ;

The C compiler knows from the start that a and b are integers: they simply cannot be anything else! With this knowledge, it can call a routine that adds two integers, returning another integer that is just a simple value in memory.

The process executed in C is roughly as follows:

1. Assign 1 to a;

2. Assign 2 to b;

3. Call binary addition binary_add(a, b)(a, b);

4. Assign the structure to a c variable

python medium The effective code is as follows:

# python code
a = 1
b = 2
c = a + b

Here the interpreter only knows that 1 and 2 are objects, but does not know what type of objects they are. So the interpreter must check each variable's PyObject_HEAD to find the type information, and then call the appropriate summation routine for both types. Finally, it must create and initialize a new Python object to hold the return value.

The execution process is roughly as follows:

1. Assign 1 to a

(1) Set a->PyObject_HEAD->typecode to an integer

(2) Set Seta->val = 1

2, assign 2 to b

(1) Set b->PyObject_HEAD->typecode to an integer

(2) Set b->val = 2

3. Call binary addition binary_add(a, b)

(1) Find the type code a->PyObject_HEAD

(2) a is an integer, the value is a->val

(3) Find the type code b->PyObject_HEAD

(4) b is an integer, the value is b ->val

(5) Call binary addition binary_add(a->val, b->val)

(6) The result is result, which is an integer.

4. Create a new object c

(1) Set c->PyObject_HEAD->typecode to an integer

(2) Assign c->val Giving the result

dynamic typing means any operation requires more steps. This is the main reason why Python is slower than C when it comes to numerical data operations.

Python is an interpreted language rather than a compiled language

The differences between interpreted languages ​​and compiled languages ​​will also cause differences in the speed of program execution. . An intelligent compiler can predict and optimize for repetitive and unnecessary operations. This will also increase the speed of program execution.

Python’s object model will bring inefficient memory access

In the above example, compared to the C language, operating on integers in Python will cause An additional layer of type information. When there are a lot of integers and you want to perform some kind of batch operation, a list is often used in python, and a buffer-based array is used in C. In its simplest form, a Numpy array is a Python object built around an array in C. That is to say, Numpy has a pointer pointing to the value of the continuous cache area data, while in python, the python list has a pointer that only wants to cache the area. Each pointer points to a python cache object, and each object is bound to a data (an integer in this case).

Schematic diagrams of these two situations:

It can be clearly seen from the above figure that when operating on data (such as sorting, calculation , search, etc.), Numpy is more efficient than python in terms of survival cost and access cost.

The above is the detailed content of Why is python slow?. For more information, please follow other related articles on the PHP Chinese website!