linux cores vs nvidia cores

Accelerated Computing CUDA CUDA Programming and Performance
#1
  1. have an app written in c/c++ designed to run in parallel
  2. has a master exe and 250 say child exes
  3. launching the master exe sequentially invokes the 250 child exes to process some child specific data,
    write a child specific result file, and signal back to the master it has finished
  4. when the the master determines all child exes have finished it imports all the child response files
    and does some analysis

all the above is processed under the protocol of one core managing one exe - no programming language
other than c\c++ is required under the linux or windows os.

can nvidia cores supplant the linux/windows cores in a straightforward transparent manor?
ie can i avoid any nvidia programming or protocalls.

thx in advance - j

#2

Cores are a feature of processors, not operating systems. It is not possible for “nvidia cores [to] supplant the linux/windows cores in a straightforward transparent [manner]”.

When programming NVIDIA GPUs, you may be able to avoid some “nvidia programming or [protocols]”, e.g. by using OpenACC, but not all of them.

#3

thx for your insight - very helpful