Titlebook: Chip Multiprocessor Architecture; Techniques to Improv Kunle Olukotun,Lance Hammond,James Laudon Book 2007 Springer Nature Switzerland AG 2

TOXIC

https://doi.org/10.1007/978-1-4020-8198-9 measured in terms of the execution latency of individual tasks. Most desktop processor applications still fall in this category, as users are generally more concerned with their computers responding to their commands as quickly as possible than they are with its ability to handle many commands simu

ineptitude

https://doi.org/10.1007/978-1-4020-8198-9 realistically human programmers will always be able to do a better job at dividing up applications into separate tasks that can work efficiently on each of the cores within a CMP. However, historically speaking parallel programming has been so much more difficult than conventional uniprocessor prog

gregarious

kindred

978-3-031-00592-3Springer Nature Switzerland AG 2007

滔滔不絕的人

Lipoprotein(A)

Improving Throughput,over the network. Since individual network requests are typically completely independent tasks, whether those requests are for web pages, database access, or file service, they are typically spread across many separate computers built using high-performance conventional microprocessors, a technique

Encapsulate

Improving Latency Automatically, measured in terms of the execution latency of individual tasks. Most desktop processor applications still fall in this category, as users are generally more concerned with their computers responding to their commands as quickly as possible than they are with its ability to handle many commands simu

全能

extinct

AMulticore World: The Future of CMPs, designers. In addition to the power issues that are making them the only viable design option possible in today’s processor marketplace, CMPs require only a fairly modest engineering effort for each generation of processors, since each member of a family of processors just requires stamping down a

GIST

Improving Throughput, switch hub, or network to shared memory and I/O devices. The overall multiprocessor system can usually be physically smaller and use less power than an equivalent set of uniprocessor systems because physically large components such as memory, hard drives, and power supplies can be shared by some or