For multi-core parallelism, multiple threads of instructions are spawn and can be run simultaneously using the various cores. And spawning and joining threads has a cost. Is there a similar sort of cost for SIMD or is it safe to only consider the clock cycles for their execution?
For multi-core parallelism, multiple threads of instructions are spawn and can be run simultaneously using the various cores. And spawning and joining threads has a cost. Is there a similar sort of cost for SIMD or is it safe to only consider the clock cycles for their execution?