Precompute point multiplication results in pedersen

[omerfirmak]

Provides 5x performance improvement

[omerfirmak]

@yelhousni, any chance we get an eye on this?

[yelhousni]

@yelhousni, any chance we get an eye on this?

So the point is instead of computing highPart*p2 + lowPart * p1 you precompute a table of multiples of p1 and p2 that you later lookup?

[omerfirmak]

@yelhousni, any chance we get an eye on this?

So the point is instead of computing highPart*p2 + lowPart * p1 you precompute a table of multiples of p1 and p2 that you later lookup?

Yeah, pretty much.

16 values are precomputed for each nibble in highPart and lowPart.

Then according to values of each nibble one of those values is selected per nibble and they are all summed up for the end result.

So we are replacing 2 scalar multiplications with multiple point additions.

[yelhousni]

Right! I remember proposing something in these lines: #327 (comment)

[yelhousni]

LGTM

[omerfirmak]

Thanks! Anything I can help with to get this merged?