This trades a scalar negation for a group (= field) negation. This is admittedly more for fun because the actual improvement is hardly noticeable, but since it’s in signature verification, I couldn’t resist.
In fact, we could store the xonly pubkey as odd when we parse it, and save the negation entirely, but this adds complexity and I leave this to the next crazy contributor.
This trades a scalar negation for a group (= field) negation. This is
admittedly more for fun because the actual improvement is hardly
noticeable, but since it's in signature verification, I couldn't resist.
In fact, we could store the xonly pubkey as odd when we parse it, and
save the negation entirely, but this adds complexity and I leave this to
the next crazy contributor.
scalar_neg takes about 2 ns on my laptop. fe_neg is much faster.
But damn, while I’m writing this I noticed that I overlooked the fe_normalize_weak call in gej_neg… Well, we could save that one in many cases, but this will add complexity, too. Nevermind.