C++ is way slower than MATLAB

Question

I am trying to generate 5000 by 5000 random number matrix. Here is what I do with MATLAB:

for i = 1:100
    rand(5000)
end

And here is what I do in C++:

#include <iostream>
#include <stdlib.h>
#include <time.h>
#include <ctime>

using namespace std;


int main(){

    int N = 5000;
    double ** A = new double*[N];
    for (int i=0;i<N;i++)
        A[i] = new double[N];


    srand(time(NULL));

    clock_t start = clock();
    for (int k=0;k<100;k++){
        for (int i=0;i<N;i++){
            for (int j=0;j<N;j++){
                A[i][j] = rand();
            }
        }
    }

    cout << "T="<< (clock()-start)/(double)(CLOCKS_PER_SEC/1000)<< "ms " << endl;

}

MATLAB takes around 38 seconds while C++ takes around 90 seconds. In another question, people executed the same code and got same speeds for both C++ and MATLAB.

I am using visual C++ with the following optimizations

I would like to learn what I am missing here? Thank you for all the help.

EDIT: Here is the key thing though... Why MATLAB is faster than C++ in creating random numbers?

In this question, people gave me answers where their C++ speeds are same as MATLAB. When I use the same code I get way worse speeds and I am trying to understand why.

These aren't equivalent. In the C++ case, you have a non-contiguous array, and each call to `rand` requires a subsequent conversion to `double`. At least try running with a contiguous array (i.e. `double *A = new double[5000*5000]`). — Oliver Charlesworth, Aug 02 '15 at 14:00
Also it should read `CLOCKS_PER_SEC/1000.0`. `CLOCKS_PER_SEC` is defined as `(long)1000` on VS2015 but better stay on the safe side in case any other compiler/environment use different values. — Banex, Aug 02 '15 at 14:07
This comparison is _completely_ meaningless. You know nothing about the differences between the `rand()` implementations, you have a gazillion dynamic allocations, you have next to no memory locality... Stop making such comparisons. If you're writing Matlab write Matlab. If you're writing C++ write C++. Then optimise your final product as/when/if required. Nothing more nothing less. — Lightness Races in Orbit, Aug 02 '15 at 14:11
@LightnessRacesinOrbit: Actually, I'm no longer sure about locality. It appears that the OP's code is iterating in a cache-friendly order, and the dimensions are large enough such that the overhead due to non-contiguousness of successive rows should be negligible. I suspect the bulk of the difference is either due to Matlab secretly using multiple threads, or due to significant differences in how rand operates. — Oliver Charlesworth, Aug 02 '15 at 14:14
@LightnessRacesinOrbit well my question is, why in the following link people are getting same speeds on both C++ and MATLAB using the same code? http://stackoverflow.com/questions/31692856/why-matlab-is-faster-than-c-in-creating-random-numbers — Serdar Oztetik, Aug 02 '15 at 14:17
@radres: Wait, why do you have two essentially identical questions? — Oliver Charlesworth, Aug 02 '15 at 14:18
@OliverCharlesworth In one of them I thought it was a problem about C++ and MATLAB. But when I see the answers I realized that I, personally, cannot reach the speeds the other people reach so I created another one. — Serdar Oztetik, Aug 02 '15 at 14:20
Are you running the program without the debugger attached? Regardless of optimization. — Banex, Aug 02 '15 at 14:20
@radres I'm not talking about compilation. Are you running the program through Debug --> Start Without Debugging? — Banex, Aug 02 '15 at 14:35
@Banex Yes, I was building it in release mode and running it. But now that you said it, I tried your way and I still get the slow timing (1/3 speed of matlab) — Serdar Oztetik, Aug 02 '15 at 14:43

score 3 · Answer 1 · answered Aug 02 '15 at 14:04

3

In your C++ code, you are doing 5000 allocations of double[5000] on the heap. You would probably get much better speed if you did a single allocation of a double[25000000], and then do your own arithmetic to convert your 2 indices to a single one.

answered Aug 02 '15 at 14:04

Logicrat

4,438
16
22

but here is the problem, people are using the same code and getting the same speeds on both c++ and matlab. I dont understand that. – Serdar Oztetik Aug 02 '15 at 14:09
5

@radres Why don't you ask those people? – juanchopanza Aug 02 '15 at 14:18

Matt Phillips · Accepted Answer · 2015-08-02T14:49:59.903

Your test is flawed, as others have noted, and does not even address the statement made by the title. You are comparing an inbuilt Matlab function to C++, not Matlab code itself, which in fact executes 100x more slowly than C++. Matlab is just a wrapper around the BLAS/LAPACK libraries in C/Fortran so one would expect a Matlab script, and a competently written C++ to be approximately equivalent, and indeed they are: This code in Matlab 2007b

tic; A = rand(5000); toc

executes in 810ms on my machine and this

#include <iostream>
#include <stdlib.h>
#include <time.h>
#include <ctime>

#define N 5000
int main()
{
    srand(time(NULL));

    clock_t start = clock();

    int num_rows = N,
        num_cols = N;
    double * A = new double[N*N];
    for (int i=0; i<N*N; ++i)
        A[i] = rand();

    std::cout << "T="<< (clock()-start)/(double)(CLOCKS_PER_SEC/1000)<< "ms " << std::endl;

    return 0;
}

executes in 830ms. A slight advantage for Matlab's in-house RNG over rand() is not too surprising. Note also the single indexing. This is how Matlab does it, internally. It then uses a clever indexing system (developed by others) to give you a matrix-like interface to the data.

same codes executes in 830 ms (c++) and the matlab one executes in 343 ms. I am trying to understand this big difference — Serdar Oztetik, Aug 02 '15 at 14:26
@radres Are you using a more recent version of Matlab? I think more recent RNGs preserve more state between calls and are able to achieve greater efficiency. In this case the right C++ equivalent wouldn't be `rand()` but something from [std::random](http://www.cplusplus.com/reference/random/) — Matt Phillips, Aug 02 '15 at 14:49

score 0 · Answer 3 · answered Aug 02 '15 at 14:05

0

I believe MATLAB utilize multiple cpu cores on your machine. Have you try to write a multi-threaded version and measure the difference?

Also, the quality of (pseudo) random would also make slightly difference (but not that much).

answered Aug 02 '15 at 14:05

Non-maskable Interrupt

3,841
1
19
26

I have tried using parallel_for it is of course way faster yet when I run either matlab or c++ they are both using only 1 core of the CPU. And again, there are other people using the same code and getting faster C++ speeds compared to mine. I do not understand what im doing wrong – Serdar Oztetik Aug 02 '15 at 14:10
Like other suggested, your memory layout is not contiguous, you also waste an indirect de-referencing on each double. That can significantly hurts performance and cause cache pollutions. – Non-maskable Interrupt Aug 02 '15 at 14:15
2

Matlab does not use multiple cores here. There is only one state for the random number stream which can be manipulated by only one thread at a time. – Daniel Aug 02 '15 at 14:19

score 0 · Answer 4 · edited May 23 '17 at 11:51

In my experience,

First check that you execute your C++ code in release mode instead of in Debug mode. (Although I see in the picture you are in release mode)
Consider MPI parallelization.
Bear in mind that MATLAB is highly optimized and compiled with the Intel compiler which produces faster executables. You can also try more advanced compilers if you can afford them.
Last you can make a loop aggregation by using a function to generate combinations of i, j in a single loop. (In python this is a common practice given by the function product from the itertools library, see this)

I hope it helps.

C++ is way slower than MATLAB

4 Answers4