Actually this is not true. The SIMD width (warp size) is still 32. Divergent kernels won't suffer more with kepler. Maybe you got the wrong impression because nvidia's diagram with its architecture might be oversimplified.

I doubt this kind of code is being generated in *RELEASE* builds. I often check the code being generated in inner loops and most of the time it's the Right Thing (tm). I'm pretty amused to see that the compiler can aggregate calls to sin/cos with the same argument into a single fsincos call, or vectorize some loops over arrays. That's like having the best of two worlds: human readable code that maps directly to the problem at hand AND very well optimized generated code. And given a new CPU and a compiler that understands its architecture and can take advantage of it, my higher level code will profit from it with minimal change. PS: higher level code -> C++

Open source is going well in Brazil because the government is really involved in substituting proprietary software for open source. It's happening wildly in the public sector. I was astonished when my girlfriend (which is doing civil service exams) told me that in her last exam there were questions regarding OpenOffice, instead of Microsoft Office, which was the norm a few years ago.
Being a country with a past (or present) of government corruption, I really don't understand why Microsoft's bribes don't work here (but work elsewhere).