The win executables took some time to compile, they are linkes in the announcement thread, and here:
Win64: http://www.share-online.biz/download.php?id=E7OXBNRMUU
Win32: http://www.share-online.biz/download.php?id=L78K8NRMGUG

Since the quark-algo is pure plain c, i compiled it for different processor/instructionsets, just pick one that works best for you. AVX better than SSE and so on. Win64 will definatly gain heavy speed improvements: On my old core2 laptop, I had a win32_sse3 run with around 9-12 khash/s per core. With w64_sse4 i got around 22-26 khash/s per core.

Yes definatly, one core on wallet: ~11kHash/s, w64_sse4 one core: ~23kHash/s
So at least on my testing machine and win64: around 100% gain

Github will come hopefully after the weekend, and hopefully coreectly setup as fork.


Yeah nice, taking it out of the loop, need to try out whats the speedgain in the cpuminer.
I wonder if we get even more gain (since memory is not an issue) if we create for multiple times used hash functions in one hash run, distinct contexts, and then combine them in one big struct, so thats one big continous block of memory which can be copied with one single memcpy call, instead now 9 calls for smaller blocks.