Google acquires programming toolmaker PeakStream

Google has acquired PeakStream, a start-up that sells tools for writing software that can take advantage of multicore processors as well as graphics and gaming chips.

Google confirmed the deal but declined to disclose financial terms. The acquisition was first reported Tuesday by The Register and The Wall Street Journal.

Newer computers contain advanced processors, but it's often a challenge to write software that can unlock that power. PeakStream and competitor RapidMind are working on just that problem.

Software is most often designed to run in a linear fashion on a single processing core, but multicore chips can handle two or more tasks simultaneously. At the same time, graphics chips are increasingly suited not just for drawing elaborate videogame scenes or architectural renderings, but also general-purpose programming as well.

And ordinary computers can be spruced up with gaming chips such as the Cell Broadband Engine chip co-developed by IBM, Sony and Toshiba. Such programming tasks are difficult; IBM released a research project called Octopiler to try to make programming Cell easier, for example.

Google's interest in such technology is logical. The company runs thousands of servers and is concerned about getting the maximum use out of each one. Intel custom-designed servers to meet stringent power-efficiency demands, for example. Google also employs numerous programmers who have an interest in such matters as the compilers that convert the source code written by humans into the binary instructions a computer understands.

PeakStream Chief Executive Neil Knox formerly ran Sun Microsystems' x86 and low-end Sparc server business.

Like Google and Sun, PeakStream also has Stanford University roots. Its software began as the Brook Project at Stanford by professor Pat Hanrahan, the company's co-founder and chief scientist. The other company co-founder is Chief Technology Officer Matt Papakipos, who was lead architect for Nvidia's GeForce 6 series of graphics chips, also known as the NV4X products. Varun Mehta, formerly of Network Appliance, is vice president of engineering.

[servermaker]

another SHR meltdown

Another Kleiner/Sequoia company hits the side of a mountain. With "help" like that, who needs competition?!

[thecatch]

A tool to improve a broken down car...

What is needed is a new foundation (new car) with all new parts <br />(software for this discussion).<br /><br />Peakstream, like Mitosis at Intel is still a guessing game/<br />prefetch predicting, what data may be access from memory <br />using the best available math on board to go retrieve that data.<br /><br />So many other problems are critical to the process like <br />scheduling on the multiple cores, file system architecture, data <br />base management tasks etc. etc.<br /><br />Not the best solution possible, (Peakstream), but better then <br />nothing/ and better then yesterday's old way of doing business.<br /><br />But the new foundation is needed with all parts from the 'soup <br />to nuts' included.<br /><br />That foundation is coming.

[chuck_whealton]

Just another example...

Just another example of how that article where Google claims that "foreign workers are the key to their success" is nothing but a joke.<br /><br />The key to their success is having overvalued stock where they can go off and waste lots of stockholder dollars. Kind of like they did on YouTube. Kind of like they do on their upper managements security detail. Kind of like they did with this purchase.<br /><br />Take their overvalued stock away and they're absolutely NOTHING... With the exception of their search engine, have they actually PRODUCED anything ON THEIR OWN that would really categorize them as successful? Or innovative?<br /><br />Charles R. Whealton<br />Charles Whealton @ pleasedontspam.com