Hi,
I just learned that for memory-bound codes one can improve the
memory bandwidth significantly using non-temporal stores that are
part of SSE2. However, I could not find out how to get gcc to
generate corresponding code on a Core2.
For the stream triad kernel,
subroutine stream_kernel_triad (a, b, c, n, s)
integer , intent(in) :: n
double precision :: a(*), b(*), c(*)
double precision, intent(in) :: s
integer :: j
do j = 1,n
a(j) = b(j) + s*c(j)
end do
end subroutine stream_kernel_triad
the Intel compiler shows a performance difference of +25%
between "-opt-streaming-stores never" and "... auto (default)"
or "... always". On my system, gcc-4.8's performance for this
fragment matches exactly that one without NT stores from Intel's.
Is there some trick or magic flag I need to specify?
Harald
