Dict benchmark

[certik]

See here: https://gist.github.com/czgdp1807/e7f7b6ca52c57b16b27ec8d0259c6d4a

There are multiple C++ versions, on my computer this one seems to be the fastest:

#include <unordered_map>
#include <iostream>

size_t size = 100000000;

struct ModuloHash {
    size_t operator() (int32_t key) const
    {
        return key % size;
    }
};

void test_dict() {
    std::unordered_map<int32_t, double, ModuloHash> rollnumber2cpi;
    int32_t i;
    double total = 0;
    rollnumber2cpi.reserve(size);

    for(i = 1000; i < 1000 + size; i++) {
        rollnumber2cpi[i] = double(i/100.0 + 5.0);
    }

    for(i = 1000; i < 1000 + size; i++) {
        total += rollnumber2cpi[i];
    }

    std::cout<<total<<" "<<rollnumber2cpi.size();
}

int main() {
    test_dict();
}

And here is a straightforward LPython version:

from ltypes import i32, f64

def test_dict():
    rollnumber2cpi: dict[i32, f64] = {0: 1.1}
    i: i32
    size: i32 = 100000000
    total: f64 = 0

    for i in range(1000, 1000 + size):
        rollnumber2cpi[i] = float(i/100.0 + 5.0)

    for i in range(1000, 1000 + size):
        total += rollnumber2cpi[i]

    print(total, len(rollnumber2cpi))

test_dict()

This gives me:

$ clang++ -O3 -funroll-loops -std=c++20 -ffast-math a.cpp 
$ time ./a.out
5.00015e+13 100000000./a.out  3.37s user 0.47s system 98% cpu 3.882 total
$ lpython --fast a.py
$ time ./a.out
50001499500000.00000000000000000 100000001
./a.out  0.41s user 0.25s system 96% cpu 0.684 total

[Thirumalai-Shaktivel]

Intel® Core™ i5-8250U CPU @ 1.60GHz × 8 (Ubuntu 22)

$ clang++ -std=c++20 -O3 -funroll-loops -march=native -ffast-math a.cpp
$ time ./a.out
5.00015e+13 100000000./a.out  5.52s user 2.00s system 99% cpu 7.527 total
$ lpython --fast a.py
$ time ./a.out
50001499500000.00000000000000000 100000001
./a.out  1.97s user 1.30s system 99% cpu 3.270 total

[Smit-create]

Apple Mac M1 (2020):

% clang++ -O3 -funroll-loops -std=c++20 -ffast-math a.cpp -o a.out
% time ./a.out
5.00015e+13 100000000./a.out  5.26s user 1.56s system 90% cpu 7.498 total

% lpython --fast expr.py   
% time ./a.out
50001499500000.00000000000000000 100000001
./a.out  0.97s user 0.81s system 88% cpu 2.016 total