array initialization run time comparison ifort vs gfortran

Question

I would like to compare array initialization run time for ifort vs gfortran using this compilation lines with gfortran 10.1.0 and ifort 19.1.3.304 on CentOS Linux 7:

ifort array-initialize.f90 -O3 -init=arrays,zero,minus_huge,snan -g -o intel-array.out

gfortran array-initialize.f90 -O3 -finit-local-zero -finit-integer=-2147483647 -finit-real=snan -finit-logical=True -finit-derived -g -o gnu-array.out

array-initialize.f90:

program array_initialize
    implicit none
    
    integer :: i, j, limit
    real :: my_max
    real :: start, finish

    my_max = -1.0
    limit = 10000

    call cpu_time(start)
    do j=1, limit
        do i=1, limit
            my_max = max(my_max, initializer(i, j))
        end do
    end do
    call cpu_time(finish)

    print *, my_max
    print '("Time = ", f6.3," seconds.")', finish-start

contains
function initializer(i, j)
    implicit none
    real :: initializer
    real :: arr(2)
    integer :: i, j

    arr(1) = -1.0/(2*i+j+1)
    arr(2) = -1.0/(2*j+i+1)
    initializer = max(arr(1), arr(2))
end function
end program array_initialize

Run times for this code:

gnu - 0.096 sec

intel - 0.392 sec

When I remove the init flags:

gnu - 0.098 sec

intel - 0.057 sec

When I replace the array with two variables:

gnu - 0.099 sec

intel - 0.065 sec

What happens here? Does gnu not initialize its arrays? Does intel initialize arrays very slow?

Please note real( 8 ) is not good practice. It is not guaranteed to be double precision, and is not for some compilers. It is not guaranteed to be supported, so the above may fail to compile with certain compilers. Please see https://stackoverflow.com/questions/838310/fortran-90-kind-parameter . Personally I use the iso_fortran_env method mentioned in the comments nowadays. — Ian Bush, May 19 '21 at 13:25
I edited the question with no use of real(8) and updated run times. — nadavhalahmi, May 19 '21 at 14:04
So do the arrays contain the values you requested using the flags. — Vladimir F Героям слава, May 19 '21 at 14:23
@VladimirF I wasn't sure because when I printed the values, I thought it might cause the array to be initialized, because I access its values. — nadavhalahmi, May 19 '21 at 14:31

score 0 · Answer 1 · answered May 19 '21 at 14:30

0

OOPS.

I disabled vectorization using -no-vec on ifort and -fno-tree-vectorize on gfortran, and now the run times are same and about 0.39 sec (just like the original intel time).

answered May 19 '21 at 14:30

nadavhalahmi

101
6

2

What is the point of this exercise? Using options to initialize variables is not a part of the Fortran standard. A compiler can do anything it wants, so compare behavior between any two compilers is pointless. – steve May 19 '21 at 16:05
@steve I have a legacy code which can be compiled with ifort and gfortran. When I compare both performances, I can tell how to optimize each one of them individually. I can tell, where what could have performed better, and help it get there. – nadavhalahmi May 20 '21 at 08:00
1

Don't know about intel, but all of those -finit-* options for gfortran can effect performance. If you're worried about performance with your legacy code, then use a profiler to find hot spots. Artificial benchmarks, such as the one here, have little value. – steve May 20 '21 at 14:53
@steve I use Intel Vtune to profile my code for both ifort and gfortran. Ifort ran about 20% slower and got a major hotspot for `for_array_initialize` function, which is the reason I compared array initializtion in both compilers. – nadavhalahmi May 21 '21 at 08:20

array initialization run time comparison ifort vs gfortran

1 Answers1