On Sat, Feb 19, 2011 at 00:24, Barry Smith <[email protected]> wrote:

Even adjusting the array passed in will be worth it.

Yes, the work to futz around with the small array is tiny compared to the cost of streaming the vector through extra times.

Should MAXPY have an optional constant scale factor, does it come up enough?

It looks like about half of the uses in PETSc are preceded by a loop to apply some constant factor, so I think such a parameter would be worthwhile.