Submitted another fix (merged from the klaus_T fork).
quark and x11,x13,x15,x17 should work now, but the performance is down using one thread.
quark with -g 4 is looking good.
INITIAL RESULTS--
I cloned git commit #725 and ran it on my 2 rigs. THE "...does not validate on CPU" ERRORS are GONE! The hash rate still fluctuates. Hash rate is printed to the screen only with the physical GPU number. Using only "-g 4" and default intensity (no -i flag used), the 8 threads on my 2-card rig are reported at launch, as are the 24 threads on my 6 card rig. Then the hash rate fluctuates to higher than normal values, but without error. My 2-card 970 FTW+ rig is averaging near 35Mh/s, and my 6-card 750ti FTW rig is averaging around 60Mh/s. Both rig's hash rates swing wildly.
My total rate for all three machines mining on yaamp was about 80Mh/s before, mining with single threads as reported on the previous page. This was reflected in the pool rate. The pool rate still remains at the same level with commit #725, but the errors are gone. The local rates total over 100Mh/s, including a steady, single-threaded, rate for my 960 SSC of about 9.4Mh/s. The 960 SSC is somewhat like a control sample, I suppose. --scryptr
6x750ti FTW, with build #725
2x970 FTW+, with build #725
Quark hash rate, single-threaded (before)
Quark hash rate, four-threaded (after)
NOTE: Local time is 1 hour behind times on graph.
Threads work with sgminer, they should work here. Dual instances (yielding two threads) worked with Scrypt and CudaMiner.
--scryptr