Game Development Reference
In-Depth Information
Today, the use of GPGPU is prevalent on most modern architectures. With the
parallel nature of the GPU, any algorithm that can process data in parallel can
generally run orders of magnitudes faster than their CPU counterparts. Therefore
it makes sense to take advantage of this power, considering that suitable hard-
ware can now be found on most common hardware configurations. This section
will covers articles that present techniques that go beyond the normal pixel and
triangle scope of GPUs and take advantage of the parallelism of modern graphic
processors to accomplish such tasks.
The first article, “Bit-Trail Traversal for Stackless LBVH on DirectCompute”
by Sergio Murgu ıa, Francisco Avila, Leo Reyes, and Arturo Garc ıa, describes an
improvement to existing stackless bounding volume hierarchy (BVH) methods
that enables the fast construction of the structure while maintaining traversal
performance sucient for real-time ray tracing. The algorithm can be used for
basic ray tracing but can also be extended for use in global illumination and
other algorithms that can depend on rapid BVH traversal on the GPU. The
article covers an implementation of the stackless linear BVH running solely on
the GPU as well as the various traversal algorithms that may be needed.
And finally, in “Real-Time JPEG Compression Using DirectCompute,” Ste-
fan Petersson describes a complete JPEG encoder implementation on the GPU
using DirectCompute. This allows for the real-time storage and encoding of not
only image data but also video sequences. With the close integration of the en-
coder with the application, the latency and compression cost is reduced, making
real-time video streaming from 3D applications more feasible. In the article, he
covers all aspects from color quantization to color-space conversions and entropy
coding—and how each of the different elements can be implemented on the GPU.
—Sebastien St-Laurent
Search Nedrilad ::

Custom Search