Python, scipy.weave and openMP - overclocking the code
Hello% username%, this article is devoted to the problem of increasing the speed of mathematical calculations based on the python language using scipy.weave and openMP .
Many may ask: “Why use python for mathematical calculations at all?”, But we will not answer “eternal” questions, nor will we consider many other solutions to this problem, such as, for example, psyco .
As described above, our tool is the scipy.weave library, as well as the openMP library.
scipy - a set of libraries for computing in applied mathematics and science. openMP is an open standard for parallelizing programs in C, C ++ and Fortran.
On Debian-like Linux systems, you need to run:
To increase the computation speed, you need to implement the “narrow” part of the python code (usually a cycle in which some actions with the matrix occur) in C and add openMP directives for parallelization.
')
I think that there is nothing better than to be convinced of this method by the example of solving the following problem:
In python c, the use of numpy this task, not taking into account various preparatory operations, like matrix generation and other things, solves in a couple of lines of code:
scipy.weave is part of the scipy library, which allows you to use C / C ++ code inside python code.
It happens as follows:
those. The C code itself is stored as a multiline string , and the python code variables are passed to C list, where the elements are text-like constants of the same name. Also, numpy arrays are transmitted to C not in the form of a matrix, but in the form of a vector, that is why there is one cycle in the code, not two.
By the way, the resulting C code can be searched in / tmp /% user% / python2x_intermediate / compiler_x
Now it is necessary to add openMP directives to the added version and in the inline call add the missing parameters, namely:
The above source code can be run and make sure that scipy.weave really gives an increase in speed:
The following resources were used in the code writing:
Many may ask: “Why use python for mathematical calculations at all?”, But we will not answer “eternal” questions, nor will we consider many other solutions to this problem, such as, for example, psyco .
Instruments
As described above, our tool is the scipy.weave library, as well as the openMP library.
scipy - a set of libraries for computing in applied mathematics and science. openMP is an open standard for parallelizing programs in C, C ++ and Fortran.
Package installation
On Debian-like Linux systems, you need to run:
apt-get install python-scipy apt-get install libgomp1
Method
To increase the computation speed, you need to implement the “narrow” part of the python code (usually a cycle in which some actions with the matrix occur) in C and add openMP directives for parallelization.
')
Example
I think that there is nothing better than to be convinced of this method by the example of solving the following problem:
- there is a matrix of size n by n, a vector of size n and an integer;
- it is necessary to subtract from each row of the matrix the vector multiplied by an integer (from the simplex method).
Python implementation
In python c, the use of numpy this task, not taking into account various preparatory operations, like matrix generation and other things, solves in a couple of lines of code:
Generation of a random matrix x on y, in our case x = y:
- # cycle through the rows of the matrix, where i is the row number
- # s - integer, randRow - random vector
- for i in xrange ( N ) :
- matrix [ i,: ] - = c * randRow
- # generate random x by y matrix
- # Matrix elements - random numbers from 0 to 99 inclusive
- def randMat ( x, y ) :
- randRaw = lambda a : [ randint ( 0 , 100 ) for i in xrange ( 0 , a ) ]
- randConst = lambda x, y : [ randRaw ( x ) for e in xrange ( 0 , y ) ]
- return array ( randConst ( x, y ) )
Implement scipy.weave without openMP
scipy.weave is part of the scipy library, which allows you to use C / C ++ code inside python code.
It happens as follows:
- #C code
- codeC =
- "" "
- int i = 0;
- for (i = 0; i <N * M; i ++) {
- matrix [0, i] = matrix [0, i] - (c * randRow [i% M]);
- }
- " " "
- weave. inline ( codeC, [ 'matrix' , 'c' , 'randRow' , 'N' , 'M' ] , compiler = 'gcc' )
those. The C code itself is stored as a multiline string , and the python code variables are passed to C list, where the elements are text-like constants of the same name. Also, numpy arrays are transmitted to C not in the form of a matrix, but in the form of a vector, that is why there is one cycle in the code, not two.
By the way, the resulting C code can be searched in / tmp /% user% / python2x_intermediate / compiler_x
Implementation of scipy.weave with openMP
Now it is necessary to add openMP directives to the added version and in the inline call add the missing parameters, namely:
Full source code with all implementations can be downloaded here.
- # C and openMP code
- codeOpenMP =
- "" "
- int i = 0;
- omp_set_num_threads (2);
- #pragma omp parallel shared (matrix, randRow, c) private (i)
- {
- #pragma omp for
- for (i = 0; i <N * M; i ++) {
- matrix [0, i] = matrix [0, i] - (c * randRow [i% M]);
- }
- }
- " " "
- ...
- weave . inline ( codeOpenMP, [ 'matrix' , 'c' , 'randRow' , 'N' , 'M' ] ,
- extra_compile_args = [ '-O3 -fopenmp' ] ,
- compiler = 'gcc' ,
- libraries = [ 'gomp' ] ,
- headers = [ '<omp.h>' ] )
Comparison of results
The above source code can be run and make sure that scipy.weave really gives an increase in speed:
Test on size: 100x100 Pure python: 0.0725984573364 Pure C: 0.303888320923 C plus OpenMP: 0.109100341797 Test - ok Test on size: 1000x1000 Pure python: 1.00839138031 Pure C: 0.506997108459 C plus OpenMP: 0.333213806152 Test - ok Test on size: 2000x2000 Pure python: 3.24151515961 Pure C: 2.10800170898 C plus OpenMP: 1.17690563202 Test - ok Test on size: 3000x3000 Pure python: 5.54490089417 Pure C: 4.61800098419 C plus OpenMP: 2.56960391998 Test - ok
Literature
The following resources were used in the code writing:
Source: https://habr.com/ru/post/135857/
All Articles