nisse@lysator.liu.se (Niels Möller) writes:
Resulting patch below. I think I'll check this in soon. Next steps would be:
Add it to the benchmark (probably easiest to do it the same way as salsa20, with a struct in nettle-internal.c).
Adapt the test program to nettle conventions. Possibly convert the round != 20 test cases to tests of the chacha_core function?
Write some assembly.
Finally, pushed into the repo now. I have done (2); (1) and (3) remains. Testing could also be improved a bit more. I used the testcases in https://raw.github.com/secworks/chacha_testvectors/master/doc/draft-stromber...
Regards, /Niels