You said it!

Alright people, new version out! This should bring a little more speedup to almost all CPU's... including AMD ones!

I finally managed to get some decent speed increase (+38% over the previous version) on my Phenom II, and this seems to apply to all processors of the K10 family, including newer Semprons.
Unfortunately you won't notice any additional speedup on older K8's, like Athlon 64's.
I haven't been able to test this version on newer AMD microarchitectures such as Bulldozer or Bobcat, so feedback from FX/Fusion owners will be greatly appreciated.

Diki is now back from his vacation, and was kind enough to provide new binaries for Windows. Here they are:
https://github.com/downloads/pooler/cpuminer/pooler-cpuminer-win32-20120103.zip (32-bit)
https://github.com/downloads/pooler/cpuminer/pooler-cpuminer-win64-20120103.zip  (64-bit)
Source code: https://github.com/pooler/cpuminer