As someone who has dabbled in CUDA with some success, I'm going to be a little contrarian here. To me, the difficulty with GPU programming isn't the fact that C

As someone who has dabbled in CUDA with some success, I'm going to be a little contrarian here. To me, the difficulty with GPU programming isn't the fact that CUDA uses C-syntax versus something more readable like Python. GPU programming is fundamentally difficult, and the minor gains from using a familiar language syntax are dwarfed by the need to understand blocks, memory alignment, thread hierarchy, etc. And I don't just say this. I live it. Even though I primarily program in C#, I don't use Hybridizer when I need GPU acceleration. I go straight to CUDA and marshal everything to/from C#. That's not to say that CUDA Python isn't kinda cool, but it's not a magic bullet to finally understanding GPU programming if you've been struggling.