If I have a multi-core CPU and a normal, serial C program, will the program compile to only be utilized on one of the processors? Or will the system try to distribute the work load across the other processors?
This might be the wrong place for your question. CUDA is related to parallel processing on GPUs, not CPUs.
I’m asking this question in regards to comparing execution times between programs on the CPU and GPU. If I’m converting an algorithm to utilize the GPU, I want to know how much of an acceleration it would achieve over a serial algorithm. Therefore I would like to know if any serial code I write is purely serialized, i.e. it utilizes only one core on a multi-core processor.
Yes, a serial code will run on a single CPU core.