NVENC perf bug

[Sulik]

cuCtxCreate() should be called with CU_CTX_SCHED_BLOCKING_SYNC, otherwise CUDA will use CPU spins for CPU<->GPU synchronization (bad for perf & concurrency)

 

[dodgepong]

Feel free to submit a pull request with the changes you believe need to be made.

 

[Lain]

Do indeed make a pull request if you happen to have time.