Brochure

29
multiprocessing
What started out as a technology to enable graphics capa-
bilities for video games has turned out to be a revolutionary
way to effectively use hundreds of cores on a single device for
certain military and aerospace applications.
General Purpose computation on Graphics Processing
Units (GPGPUs) were developed as graphics chips for pixel
processing, but subsequent development environments such
as Compute Unified Device Architecture (CUDA) and OpenCL
have leveraged the cores’ Single Precision (SP) floating-point
data processing capabilities for radar processing and other
military and aerospace applications.
GPGPU devices do not have the large need for transistors and
die space of their GPP counterparts. Thus, GPGPUs have the
ability to provide a large number of processing cores on a
single device at a minimum of size, weight and power (SWaP).
GE Intelligent Platforms provides GPGPU devices designed to
withstand the harsh environments encountered in military and
aerospace situations while delivering greatly improved systems
performance. For instance, GPGPU ruggedized devices in radar
systems can deliver greater range, number of targets, and
speed while standing up to extreme temperatures, dust, shock
and vibration.
Developing new military and aerospace GPGPU products with
minimal SWaP is made easier as a result of a unique partnership
GPGPU
many cores on a single device
between GE and NVIDIA, whose graphics processing units are
based on the CUDA architecture.
In addition to mainstream libraries, GPGPU application
development is enhanced by GE’s AXIS advanced integrated
multiprocessor software. The AXISLib-GPU software provides
the industry-standard Vector, Signal and Image Processing
Library (VSIPL) for NVIDIA-based GPU platforms. DSP application
developers are able to quickly develop applications through the
VSIPL API model without the need to learn GPU programming.
AXISLib-GPU supports the development of high-performance
DSP and multiprocessing applications on GE’s platforms
such as the IPN250, NPN250 and GRA111 graphics boards.
These platforms have been deployed for radar, sonar, image
processing, signals intelligence and intelligence, surveillance
and reconnaissance (ISR) applications.
GE Intelligent Platforms’ AXIS multiprocessing software suite
also facilitates the development of complex applications
over multiple clusters of GPU platforms. AXISFlow provides a
communications API for multi-threaded, multi-core and multi-
processor communications. Our AXISView provides a set of
GUI tools for system visualization, application instrumentation,
debugging and monitoring, allowing the developer to speed the
process of timing the applications for the best real-time perfor-
mance on the chosen system.
Form Factor Chip Set I/O Cooling Conformal Coating
Extended
Temperature Range
IPN250 VPX 6U NVIDIA GT 240 CUDA Capable Graphics Processor
DVI, HDMI, VGA, GbE, PCIe, USB, SATA,
GPIO, Audio, TV Input , Serial
Air, Spray & Conduction Option
■
NPN240 VPX 6U NVIDIA GT 240 CUDA Capable Graphics Processor DVI, HDMI or DV1, VGA Air, Spray & Conduction
■
MAGIC1
3U Boxed,
VPX 3U
NVIDIA G73 with an Intel Core 2 Duo @ 2.16 GHz or 1.5 GHz,
NVIDIA GT 240 CUDA Capable Graphics Processor
DVI, Gigabit Ethernet, USB, VGA Base-plate Yes
■
GRA111 VPX 3U NVIDIA GT 240 CUDA Capable Graphics Processor DVI, NTSC, PAL, RS-170, VGA Air & Conduction Option
■
photo courtesy of u.s. navy