By Amdahl's Law in previous slide, we can say that it is hard or almost impossible to achieve 5x speedup with given resources. However, the inherently serial operations could run faster due to lot's of cache locality, ILP and etc and we might be able to achieve the desired speed-up.
This comment was marked helpful 0 times.
aew
I agree, by the previous slide, because $s = 0.25$ and $p = 6$, the maximum speedup by Amdahl's law is $\leq \frac{1}{.25 + \frac{.75}{6}} \approx 2.667$. However, as @taegyunk mentioned, it may be possible to achieve the 5x speedup through other methods.
By Amdahl's Law in previous slide, we can say that it is hard or almost impossible to achieve 5x speedup with given resources. However, the inherently serial operations could run faster due to lot's of cache locality, ILP and etc and we might be able to achieve the desired speed-up.
This comment was marked helpful 0 times.
I agree, by the previous slide, because $s = 0.25$ and $p = 6$, the maximum speedup by Amdahl's law is $\leq \frac{1}{.25 + \frac{.75}{6}} \approx 2.667$. However, as @taegyunk mentioned, it may be possible to achieve the 5x speedup through other methods.
This comment was marked helpful 0 times.