Google buys startup doing multicore programming

Google on Tuesday announced it has bought PeakStream, a startup that specializes in software programming tools for high-performance, multicore and parallel processors.

In its statement, Google said, "We believe the PeakStream team's broad technical expertise can help build products and features that will benefit our users. We look forward to providing them with additional resources as they continue developing high-performance applications for modern multicore systems."

A Google spokesperson declined to detail the price.

PeakStream's Website was down following the announcement. A company spokesperson wasn't immediately available to comment on the purchase.

PeakStream advertises that it "makes it easy to program new high-performance, multicore and parallel processors and convert them into radically powerful computing engines for computationally intense applications."

The fledgling company launched last September with $17 million in Series B funding, helmed by executives from graphics chipmaker Nvidia, Sun Microsystems and VMware.

Veiled plans
It's not known what immediate plans Google has for PeakStream, but programming software for multiple cores or multiple threads is widely viewed as more difficult than single-core or single-threaded programming.

In a press release last month, Faisal Saied, senior research scientist for information technology at Purdue University, warned that multicore chips weren't being fully utilized because of lack of programming expertise.

"High-performance computing experts have learned to deal with this, but they are a fraction of the programmers," said Saied. "In the future, you won't be able to get a computer that's not multicore, and as multicore chips become ubiquitous, all programmers will have to learn new tricks."

In an interview just prior to Google Developer Day last week, Chris DiBona, open-source program manager for Google, acknowledged that multicore programming isn't for everyone.

"A lot of developers are in fact pretty well served by concentrating on what a single thread of execution looks like, because when you start introducing multiple threads, it can add a level of complexity to your code that's pretty daunting," DiBona said. "And some developers aren't going to get a lot of benefit from taking advantage of an extra core or an extra CPU."

Tools of the trade
DiBona nonetheless said both Google's engineers and open-source programmers have some good tools to deal with the challenge, including some that Google contributed: Perf Tools, a collection of a high-performance multithreaded malloc() implementation and other performance analysis tools, and Coredumper, a tool for multithreaded core dumps. And he pointed to the Apache Server's facility with multiple threads as evident that the problem is far from insurmountable.

"Cores should be taken with a grain of salt like anything else," said DiBona.

- Thomas Claburn
EE Times