Parallella Community

To calculate the speedup of an algorithm (for example the Matrix Multiplication) for different matrix sizes (128 x 128, 256 x 256, 512 x 512, ...) I have to run the code on 16 core and the on only 1 core of the Epyphany, right? Is this possible? As I have seen on the examples on github, at none of the algorithms we count the speedup but the execution time. Why?
Thank you!!!

Scalability and performance/throughput are independent measures. If you just want to show how fast your architecture is, the scalability number is not relevant.