GMP 4.3 multiplication performance

[Niels Möller]

Torbjorn Granlund <tg at gmplib.org> writes:

> If using submul_1 removes the last scratch space needs, then perhaps use
> it unconditionally, instead of lshift.

This is the case for toom_interpolate_7pts. But on the other hand,
when called by toom44_mul, there's free scratch space available (the
same space that was used for scratch for the last recursive
multiplication), so even though it is esthetically pleasing to remove
the scratch parameter from the interpolation function, it does not
result in any overall reduction of memory consumption.

Regards,
/Niels