Optimization issue

[OhGodAPet]

This isn't the place for this, but I could find no other way to contact you, since you're never on IRC. I've been trying to optimize the aesb_pseudo_round_mut() code, but it seems obfuscated, and I'm not entirely clear on what it does. I attempted to replace it with this linux only code:

http://pastebin.com/qczpTnjY

but it failed - hashing, yet not submitting valid shares.

[lucasjones]

I made some changes your code, let me know how this works for you.

http://pastebin.com/haB7FsWb

[OhGodAPet]

I think this commit causes a bad pointer to be passed to aesb_pseudo_round_mut(). Suddenly, the optimizations that worked before segfault when moving the data pointed at by rdi into xmm1.

927ecbe

Neeeeever mind, I'm an idiot. I have another optimization idea, but first, I have to figure out why it's now crashing in fast_aesb_single_round(). Segfaults... yet gdb can read the address just fine...

[OhGodAPet]

Got it. You fuck the alignment in that commit. Changing it to 16 fixes the issue!