The Back-Maxwell rangeproof construction we use in dalek-rangeproofs requires scalar inversion. Right now we just have an easy and naive implementation with no optimization, but which is nearly as expensive as a full point*scalar operation.
It could be made faster by some combination of:
[x] using fixed-window exponentiation with a lookup table (see the constant-time scalar-point multiplication for an example)
[x] adding an implementation of squaring for scalars
[x] optimizing the scalar arithmetic: maybe Karatsuba, Montgomery reduction, or some combination?
[x] maybe 64-bit implementation of scalar arithmetic ?
The Back-Maxwell rangeproof construction we use in
dalek-rangeproofs
requires scalar inversion. Right now we just have an easy and naive implementation with no optimization, but which is nearly as expensive as a full point*scalar operation.It could be made faster by some combination of: