curve25519-cortex-m4

This repository currently has a 256-bit multiplier modulo 2^256 - 38 optimized for ARM Cortex-M4-based processors (and above), useful to compute arithmetic over djb's Curve25519 (and birrationally equivalent curves).

To use it, compile as:

gcc _mul_coc_umaal_rdc.S -o _mul_coc_umaal_rdc.o

Include the _fe_mul_coc_umaal_rdc.h header in your code, link the _mul_coc_umaal_rdc.o object and call

fe_t operand_a, operand_b;
fe_t result;

_fe_mul_coc_umaal_rdc(result, operand_a, operand_b);

Performance

Running the multiplication routing on a Teensy 3.2 board equipped with an ARM Cortex-M4-based MK20DX256 controller from Freescale clocked at 48MHz takes about 276 CPU cycles.

Code tested was assembled using arm-none-eabi-gcc version 5.4.1.

hayatofujii/curve25519-cortex-m4

curve25519-cortex-m4

Performance