| Multicore Faces a Long Road |
|
|
|
|
Not only did the electronics industry gain an understanding of the significance of multicore technology in 2006, but it realized just how much work is yet to be done.
Multicore CPUs and embedded multicore systems arrived in 2006, and the number of cores on a die took the place of megahertz as the metric for microprocessors, all of which led to the acceptance by software developers "that programming as they knew it is not going to be the same any more--they will be doing parallel programming going forward," according to parallel processing pioneer and MIT professor of electrical engineering and computer science Anant Agarwal. Computer scientists must address the lack of "algorithms, languages, compilers, and expertise" that faces parallel computing, says Microsoft Research manager of programming and tools Jim Larus, as well as "a lot of practical issues, like developing better support for multithreading, synchronization, debugging, and error detection." Larus also points out that a better understanding is needed of just how people want to use parallel programming. While some in the industry want to create automated tools to transfer legacy code to multicore, parallel programming languages will be needed, since work must get down to the algorithm level. No fundamental methods for parallel programming have been established, and researchers have found that no single technique for parallel programming is universally applicable, but possibilities are being explored, such as lightweight software transaction. The Multicore Association (MCA) has been formed to address the many problems facing the field, and the Embedded Microprocessor Benchmark Consortium (EMBC) has begun developing a standard suite of tests to evaluate the performance of multicore processors. Click Here to View Full Article |
| < Prev | Next > |
|---|
Upcoming Public Events:
With your technical knowledge you are kind of ambidextrous in your domain