Reference
- UIUC ECE 408 Lecture 27
- UC Berkeley CS267 Project 1
time ( sequential ) / time ( parallel )
在measure sequential的时候，一般measure最好的，找到的baseline，而不是自己写的baseline
speedup on P processor / P
常用于衡量多cpu core的程序
希望是接近1，大多数情况都是小于1的
superlinear speedup：由于使用了extra resoruce (例如cache等等) 所以efficency大于1
对于GPU来说，efficency衡量的就不是很有效了。
对于GPU，一般衡量efficency是 compare resource used with GPU's peak value
对于多少个processor，efficency是接近1。对于多少个processor efficency是减小的
speed up <-> num of processor的图
good scalability : speed up curve not fall of for max measureable value of P
下图的这个例子里，就不是scalable code
In strong scaling we keep the problem size constant but increase the number of processors
In weak scaling we increase the problem size proportionally to the number of processors so the work/processor stays the same