Typo in the Lab2, Fortran, README.md

[AlexisEspinosaGayosso]

First Typo

In the block showing syntax for copying many arrays at once, you are using "c" syntax:

!$acc parallel loop copy(A[0:N], B[0:N])

while it should be

!$acc parallel loop copy(A(1,N), B(1,N))

Second Typo (together with a bug)

In the update section, you use twice:

!acc data copyin( A(n,m), Anew(n,m))

and once:

!acc update self(A(n,m))

While it should be:

!$acc data copyin( A(n,m), Anew(n,m))
.
.
.
!$acc update self(A(n,m))

Now the bug:

In your real code, you are allocating your arrays forcing indexing to start at (0,0) and end at (n-1,m-1). Therefore using the shape indicated above fails to cover the whole array (check your compiling -Minfo). Therefore, you may consider to use A(:,:), Anew(:,:) in your notes. (You are already using :,: in the real code.) Or, even better, modify your indexing in the creation of your arrays to comply with natural fortran indexing starting from (1,1).

Now a suggestion:

I suggest you to keep using the same size of Arrays as in the original example:

integer, parameter :: n=4096, m=4096, iter_max=1000

so that we can compare timings and performance.

And just print a partial size of the total array:

if(mod(iter,100).eq.0 ) then
   !$acc update self(A(:,:))
   write(*,'(i5,f10.6)'), iter, error
   write(*,*) '============='
   do i=1,10
      do j=1,10
         write(*,'(f10.2)', advance="no"), A(i,j)
      enddo
   enddo
   write(*,*) '============='
end if