ref: 76c290f92d5e011dce678bf281e9767d1a9b2bc4
dir: /sys/src/ape/lib/openssl/crypto/cast/asm/readme/
There is a ppro flag in cast-586 which turns on/off generation of pentium pro/II friendly code This flag makes the inner loop one cycle longer, but generates code that runs %30 faster on the pentium pro/II, while only %7 slower on the pentium. By default, this flag is on.