Pregunta de entrevista de NVIDIA
How do you handle multi-threading issues on a machine with multiple CPUs and GPUs? Followed by many detailed, pointed follow up questions.
Respuesta de la entrevista
Anónimo
21 ene 2023
Had to describe OS-provided things like mutexes, condition variables, etc; NUMA and thread affinity; data dependencies; data fences/barriers; shared memory (CPU+GPU); Load/Store and memory models; etc.