fjarri

import numpy

import pycuda.autoinit
import pycuda.driver as drv
from pycuda.compiler import SourceModule
from pycuda.gpuarray import to_gpu

mod = SourceModule("""
typedef struct _pair
{

fjarri

function add(A,B,reps)
    for i in 1:reps
        C = A + B
    end
end

function add_bc(A,B,reps)
    for i in 1:reps
        C = A .+ B
    end

fjarri

function test(f)
    tmin = 1e10
    for i in 1:5
        t = time()
        f()
        t = time() - t
        if t < tmin
            tmin = t
        end
    end