GPU is all about performance. Nearly all the time, very high level languages have nothing to do there. The CPU part of high level user applications will probab

GPU is all about performance. Nearly all the time, very high level languages have nothing to do there. The CPU part of high level user applications will probably be written in very high level languages/runtimes with, sometimes, some other parts being bare metal accelerated (GPU or CPU). Devs wanting hardcore performance should write their stuff directly in GPU assembly (I think you can do that only with AMD) or at best with a SPIR-V assembler. Not to mention doing complex stuff around the linux closed source nvidia driver is just asking for trouble. Namely, either you deploy hardware/software nvidia did validade, or just prepare to suffer... it means 'middle-men' deploying nvidia validaded solutions have near 0 added value.