On Tue, 28 Jul 2020, Brad Chamberlain wrote:
I'm surprised your serial version would still be 60% behind Fortran; it
seems like we ought to be able to reach parity without too much trouble.
Silly me. I wrote code as
y += a * x
where y and x are vectors. I redirect it with iterators and the elapsed
time went down 20%. That is a serious drop.
Also, Fortran does things as
do i=1,n
a = ddot(x, y) // i.e. a = x.y
axpy(x, y, a) // y += a * x
end do
do i=1,n
t(i) = ddot(x, y)
end do
do i=1,n
axpy(x, y, a) // y += a * x
end do
Regards - Damian
Pacific Engineering Systems International, 277-279 Broadway, Glebe NSW 2037
Ph:+61-2-8571-0847 .. Fx:+61-2-9692-9623 | unsolicited email not wanted here
Views & opinions here are mine and not those of any past or present employer
_______________________________________________
Chapel-developers mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/chapel-developers
