Accelerator Architecture
Abstract
The efficient solving of scientific problems on supercomputers requires a sufficient amount of computational resources, and more often than not this involves performing floating-point calculations. Curiously, the fraction of silicon die area dedicated to supporting these computations is frequently quite modest compared to a modern processor's chip size. Providing specialized devices with an increased count of floating-point units may offer substantial speed advantage over conventional CPUs. Such accelerator devices need not necessarily be limited only to floating-point arithmetic, but may serve other purposes such as input/output management, cryptography, multimedia encoding and decoding, and signal processing. ...
Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.
Read now
Unlock full access