OpenMP vs trigonmetric transforms

NIcolas_T_ — Tue, 05 Aug 2014 13:00:56 GMT

Hi,

!$omp parallel   private(iy,iz,bx,by) shared(ll,nn,mm,b,im1)
!$omp do
  do iz = 1,ll
    do iy = 1,nn
  
        bx=.0
        by=.0

        !fetch data 
        bx(1:mm)=real(b(1:mm,iy,iz,ik))       
        by(1:mm)= aimag(b(1:mm,iy,iz,ik))

        !compute forward sine transform
        call dst_fwd(im1,bx(0:im1))          
        call dst_fwd(im1,by(0:im1))

       !set data
        b(1:mm,iy,iz,ik) =cmplx( bx(1:mm),by(1:mm))
    enddo
  enddo
!$omp end do
!$omp end parallel

The code snippet above crashes on runtime with no error message when I put OpenMP commands around the the two loops in order to compute Sine Transforms (subroutine dst_fwd ) in parallel.

Does anyone could give me a hint what may cause the program to crash?

Thanks,

Nicolas

I suspect we would need a

TimP — Wed, 06 Aug 2014 12:14:41 GMT

I suspect we would need a full working example.

topic OpenMP vs trigonmetric transforms in Intel® oneAPI Math Kernel Library

OpenMP vs trigonmetric transforms

I suspect we would need a