Microsoft  and Intel are giving a $20 million boost to parallel computing researchers who are developing better ways of writing applications that can take advantage of multicore processors.

The $20 million will be distributed over five years to the University of California, Berkeley  and the University of Illinois at Urbana-Champaign (UIUC). Another $15 million is expected to come from a California state-grant program and UIUC.

Dual- and quad-core processors are becoming increasingly common, and the number of cores in each processor will only grow in coming years. Intel already is developing an 80-core processor.

Software designed for single-core processors, however, can't take full advantage of multicore computing, which can execute multiple threads and processes simultaneously; many programmers find it difficult to write applications suited for the multicore world. The goal of this initiative, announced this week, is to eventually make "parallel programming . . . synonymous with programming."

"There have been no major efforts on how to make parallel programming easier for the average programmer," says Marc Snir, a professor of computer science and electrical and computer engineering at UIUC. Laptop computers and PDAs, in particular, have been unable to leverage all the benefits of parallel computing, he says.

Snir and his colleagues aim to develop new programming languages and other technologies to help programmers take advantage of multicore processing. They also are developing new ideas for how vendors like Intel and Advanced Micro Devices can design multicore processing to make programming easier, he says.

One problem with parallel computing today occurs when simultaneous threads run at different speeds, Snir says. In a simple example, say two people are trying to book a plane reservation at the same time, and there is only one seat left. If the requests are being handled by two threads, the person whose request is being processed by the faster thread will get the seat.

In a complex research environment, the outcome of a computation using multiple threads may depend on calculations being solved in an exact order. If one thread unexpectedly goes faster than another, running the same program twice could yield two different answers. "How you avoid this problem is for us a major research issue," Snir says. The goal is to make sure that "parallel codes are deterministic, repeatable and always give you the same answer for the same input."

Whitepapers

• The Trend from UNIX to Linux in SAP(r) Data Centers
• Secure Wireless Printing Options
• Improve Virtual Storage TCO by Comparing Vendors
• Making IT Invisible
• Best Practices for VMware ESX Server
• Virtualizing Storage for VMware Success

View more DATA CENTER whitepapers

Webcasts

• Branch Office Trends and Best Practices
• Data Protection and Application Availability for Your VMware Environment with Veritas NetBackup and Cluster Server for VMware ESX
• Making NetBackup Highly Available with Storage Foundation HA
• Best Practices for Deploying WAN Optimization for Disaster Recovery
• Tips and Techniques for Remote Data Backup and Delivery
• WAN Optimization Landscape

View more DATA CENTER webcasts

Special Reports

• Executive Guide: Storage Heats Up
• Symantec State of the Data Center Report
• Data Centers: At the center of the action
• Executive Guide: Virtualization Reality Check
• Executive Guide: Data Center Decisions
• Get More From Your WAN

View more DATA CENTER special reports