Is optimizing as we know it becoming irrelevant?
Many of us have spent our careers amassing knowledge on how to write optimal serial algorithms, and writing code that can be run efficiently on a single core machine. As multi-cores and distributed computing becomes ubiquitous, this knowledge is slowly slipping into arcana.
I am currently working on a new virtual machine called CVML based on the Cat language. While trying to make my VM implementation efficient, it dawned on me that much of my what I know about writing efficient code is becoming irrelevant.
On an 8 core machine, it no longer matters whether my design is efficient in serial, but rather whether computation tasks can be efficiently shared across the different processes.
This makes me think that we may move into a new era where ...