I was hoping that Google Glass would catch on a long time ago. Alas, that didn’t happen. Though I feel that battery performance plays a big role in that, it isn’t just processor performance. I think that there are some killer apps for a head-mounted device with a camera that is “always-on” indexing things. It doesn’t have to record if you don’t want it to, just index. So that you can always ask when was the last time and place you saw something. Perhaps the next-gen lithium-ion batteries with solid electrolytes can make that feasible.

I played with Aparapi and TornadoVM just a bit to run JVM code on GPUs. I’m more interested in doing it with Scala than Java, but they have some interesting potential. The challenge seems to be that you still need to know a fair bit about the memory layout of capabilities of the target platform to really make things work well. On the plus side, I believe that these are areas where there is a fair bit of active research, but I haven’t really stayed on top of what is happening.

I believe that there is reasonable support for LLVM to output code that runs on GPUs. Focusing work on that level could potentially benefit a lot of languages. Rust is obvious there, but things like Scala Native mean that it would boost other languages as well. Indeed, most languages could probably have an LLVM back-side for the compiler written fairly quickly if there were the right motivation.