serpent: add parallel processing for CFB decryption
* cipher/cipher.c (gcry_cipher_open): Add bulf CFB decryption function for Serpent. * cipher/serpent-sse2-amd64.S (_gcry_serpent_sse2_cfb_dec): New function. * cipher/serpent.c (_gcry_serpent_sse2_cfb_dec): New prototype. (_gcry_serpent_cfb_dec) New function. (selftest_cfb_128) New function. (selftest) Call selftest_cfb_128. * src/cipher.h (_gcry_serpent_cfb_dec): New prototype.
Patch makes Serpent-CFB decryption 4.0 times faster on Intel Sandy-Bridge and
2.7 times faster on AMD K10.
- Signed-off-by: Jussi Kivilinna <jussi.kivilinna@iki.fi>