FYI, I managed to rebuild the Bufius cpuminer with optimizations set to -O1 and the resulting binary has zero rejects. Looks like there is a clang optimization problem with cpuminer after all.
Ah, I get it. Even though my Mac Pro (5,1) has the same roughly 40 KH/s for the GPU and CPU, the CPU actually has 24 1.7 KH/s workers. And it's because these workers are slow that they can't keep up.
I went ahead and tried the sgminer that you provided. I'm getting similar KH/s out of the GPU as with cpuminer, but with no stales pointed at the same pool.
So it's looking more and more likely that the optimization level expla…
No problem. I'll review the code (which is I think what you're asking). I probably can't spare the time till this weekend though. I'll let you know if I spot anything.
Thanks for the autoreconf suggestion. It got me further, but…
I took a look at the github site, and from there I found the corresponding bitcointalk forum thread. According to this post, there's a problem when building with optimization levels of -O2 or higher under the Lion version of clang that could result …
One more quick question. When I run vertminer from the command line it says it's running in "scrypt" mode. Am I supposed to somehow specify that the algorithm is scrypt 2048? Or is scrypt N automatically baked into this version?
I'm an experienced scrypt miner (mostly with ASICs) and new to scrypt N, so I'm afraid that I don't know of a better backend miner. Just getting my feet wet here...