GDC09: programming efficiently and simultaneously with Larrabee

By Michael J Huelskoetter (68 posts) on August 18, 2009 at 8:00 am

Today Steve McCalla's second Larrabee session took place as he gave the first lecture yesterday evening about the upcoming Intel GPGPU architecture.

So Steve mentioned again things like "Gather", "Swizzle", "Predication" and other complicated stuff.

The idea behind all these terms is always the same: Due to very wide vector registers Larrabee can execute loops, adds, multiply, maskings and other operations very parallel and thus very fast on hardware level.

This will happen on a greater number of processor cores which are ideal for general purpose tasks as well as graphics oriented calculation.

Summarized this means: The instruction set of Larrabee (LRBni) will compute highly complicated mathematical tasks very quick and very parallel. Moreover, Steve made a really interesting statement:

If Intel was a software company, we would be under the top three of this industry!

This means that Intel engages several of thousands software engineers who develop year after year new software tools and optimize current ones constantly. So that other software developers can make the most out of Intel microprocessors. That's what Steve was asking for!

After his tech session Mister McCalla took the time to answer our questions. We still wanted to know why game developers should looking forward to Larrabee.


Categories: Events, Gaming, Parallel Programming, Visual Computing

Comments (0)

Trackbacks (5)


Leave a comment  

To obtain technical support, please go to Software Support.
Name (required)*

Email (required; will not be displayed on this page)*

Your URL (optional)


Comment*