Codelets for ToomN1 (for N=2, 3, 4, 6, 8) should be added and here's why. (Also: a significant non-triviality on where cut-off points should be).
tg at gmplib.org
Fri Apr 6 11:24:02 UTC 2018
> By the way, in the last message I forgot to point to some graphs we
> elaborated to show where each multiplication variant is better. They can
> be found at https://gmplib.org/devel/, under the title "New Toom
> multiplication code for unbalanced operands".
> Those graphs are old, they did not yet take into account either Toom6h or
I think these graphs are reasonably valid even if they date back to the
big Toom push (which is soon 10 years ago).
indeed, those graphs clearly show that a simple threshold mechanism is not
optimal. I'd like to see a similar graph for 0 < n < 256.
Look at the log/log graphs; they start at n = 12 but otherwise provide
excellent detail for small n.
> Maybe a more fine-grained table for small sizes... then another table for
> larger values.. We shall thinkabout that.
I believe log/log provide just that (at least if the measured sizes are
mae in a geometric series).
I could perhaps re-measure these things, just need to locate the program
which did it. :-(
Please encrypt, key id 0xC8601622
More information about the gmp-discuss