Power: The proliferation of mobile devices as well as the power wall necessitate a need for low-power solutions.
The need for saving power will not go away, so there should be ample interest and opportunities in this area.
Aside from energy efficiency, predictability, fault tolerance, accuracy, and security are often at least equally important aspects when designing hardware and software.
Thus, novel concepts as well as long reaching research in the areas of computer architecture design, computation models, parallelization methods, software stacks, and programming and debugging tools are required.
Hardware Security: Security is tremendously important in all computer science disciplines.
However, exploits are far easier to implement at the software level as opposed to the hardware level, and having hardware solutions for security end up being rather impractical while achieving minimal benefit, in my opinion.General Scope “Heterogeneity in Architectures and Systems – From Embedded to HPC” was the specific focus of the International Conference on Architecture of Computing Systems (ARCS 2016).This leitmotif reflected the ongoing progress in semiconductor technology that allows for building fascinating, complex computing systems, including multiple (heterogeneous) microprocessors, large on-chip memory hierarchies, advanced interconnection networks, and peripherals.One of my professors was doing research in , although I haven’t been following the state-of-the-art regarding this topic.Datacenters: While maybe not as “hot” as Deep Learning, this is a hugely important and impactful field that will interest any large tech company that hosts their own data centers (eg. Without industry connections, research in this area is difficult, but any solutions in this area will literally save companies hundreds of millions of dollars.It is a direct evolution of the Monash multiprocessing kernel, developed for the Monash multiprocessor system in 1987, and a major research accomplishment by a long running research group founded and led by the late Professor Chris Wallace.Previous Wal NUT related projects included the addition of a stdio libraries, a compiler port, a network protocol stack design, a shell design and other miscellaneous topics.Still, it won’t stop researches from trying, and I’m pretty sure people are still publishing papers in this area.Unfortunately, a lot of topics that are fundamental to understanding computer architecture are basically dead at this point: Branch Prediction: This field is dead. Single-core optimizations/Instruction-level parallelism: While it would be great to have some innovation into this field again, we’ve unfortunately hit the power wall, and thus we now obtain most of our performance benefits from multi-core.I have had and still maintain a long standing interest in computer system design, encompassing machine architecture and organisation, operating systems with a focus on real time performance, IPC, File Systems and I/O Subsystems.This is a continuing thread of activity which started during the late 1980s, when I maintained VAX architecture machines and peripherals.