|
|
|
|
Publication Year: |
|
2017
|
|
|
Authors |
|
Wen-mei Hwu, Izzat El Hajj, Simon Garcia de Gonzalo, Carl Pearson, Nam Sung Kim, Deming Chen, Jinjun Xiong, Zehra Sura
|
|
|
Published: |
|
In IEEE International Conference on Rebooting Computing (ICRC), 2017.
|
|
|
Abstract: |
|
We have been experiencing two very importantmovements in computing. On the one hand, a tremendousamount of resource has been invested into innovativeapplications such as first-principle-based methods, deeplearning and cognitive computing. On the other hand, theindustry has been taking a technological path whereapplication performance and energy efficiency vary by morethan two orders of magnitude depending on their parallelism,heterogeneity, and locality. We envision that a perfect stormis coming because of the interaction between these twomovements. Many of these new and high-valued applicationsneed to touch a very large amount of data with little data reuseand data movement has become the dominating factor for bothpower and performance of these applications. It will be criticalto match the compute throughput to the data access bandwidthand to locate the compute near data. Much has been andcontinuously needs to be learned about algorithms, languages,compilers and hardware architecture in this movement. Whatare the killer applications that may become the new driver forfuture technology development? How hard is it to programexisting systems to address the data movement issues today?How will we program these systems in the future? How willinnovations in memory devices present further opportunitiesand challenges in designing new systems? What is the impacton long-term software engineering cost of applications? In thispaper, we present some lessons learned as we design the IBMIllinoisC3SR (Center for Cognitive Computing SystemsResearch) Erudite system inside this perfect storm.
|
|
|
|
|
|