The ChipList, by Adrian Offerman; The Processor Portal

Processor Selector

View: show / edit

bookmark bookmark site
bookmark permalink
Mon 9 Mar 2009, 5:05

Maximizing Performance with Fine-Grained Parallelism

On Mon 9 Mar 2009, at 5:05, Intel Software Network Graphics Feed wrote:

Maximizing Performance with Fine-Grained Parallelism

Implementing functional decomposition in games to get the most performance from the processor

By Andrew Binstock

In my previous article [insert link], I described how to use functional decomposition to break up games and visual programming apps into tasks that could be allocated to independent threads. By assigning tasks to individual threads, you enable those tasks to run in parallel. This gives you better overall performance and, because it relieves the main thread from doing all the work, it permits faster and snappier visual presentation.

To review quickly, functional decomposition divides work into discrete tasks. In other words, it breaks (or decomposes) the work by the function being performed, hence the name functional decomposition. The other principal approach, called data decomposition, decomposes work by giving different chunks of data to individual threads. These threads generally perform the same operation on the data chunks and then have some way of integrating the results or the transformed data. I will discuss data decomposition in an upcoming article.

To understand how functional decomposition is handled, we need to examine how threading models work. Once we see the fundamental dynamics, we can look at how threading has traditionally been done and how new thinking, ...

Filed under: GraphicsVisual Computing
Source: Intel Software Network Graphics Feed
Also published by:

ChipList news channel Last Months News

ChipList developers news channel Last Months Developers News

Page viewed 2299 times since Mon 9 Mar 2009, 5:41.