THE scalable Matrox Odyssey Xpro, available from the Dindima Group , features the premier Motorola G4 PowerPC embedded microprocessor, running at 1GHz.
The PowerPC, combined with Matrox Imaging's state-of-the-art Oasis processing and router ASIC, delivers more than 130 billion operations per second (BOPS).
The single-slot Odyssey Xpro also offers over 5GB per second of memory bandwidth, up to 1GB of DDR SDRAM memory and up to 2GB per second of external I/O bandwidth.
The high-density Matrox Oasis ASIC is the pivotal component on the Odyssey Xpro and integrates a CPU bridge, main memory controller, Pixel Accelerator and Links Controller.
The Pixel Accelerator (PA) is a parallel processor core that accelerates neighbourhood, point-to-point and LUT mapping operations. It consists of an array of 64 processing elements all working in parallel.
Operating at a core frequency of 167MHz enables the PA to process up to two billion pixels per second.
The Links Controller (LINX) is the router that manages all data movement inside and outside the processing node, which consists of the Pixel Accelerator, CPU and main memory.
It can handle several concurrent video and message streams, thereby offloading data management tasks from the CPU and PA and allowing them to focus on image processing tasks.
With more than 30 million transistors, the Oasis also incorporates the 64-bit, 133MHz CPU bus interface to the G4 PowerPC and a 128-bit wide, 166MHz double data rate DRAM interface capable of transferring image data at rates of up to 5.3GB per second.
The Odyssey Xpro is the first vision processor on the market to take full advantage of the new PCI-X bus technology. PCI-X is a high-performance, backwards-compatible enhancement to the conventional PCI bus specification.
Version 1.0 of PCI-X specifies a 64-bit physical connection running at speeds of up to 133MHz, resulting in a peak bandwidth of up to 1GB per second.
The custom-designed Matrox System Interface Bridge (SIB) ASIC, a four-port PCI-X router, handles the PCI-X connections on the Odyssey Xpro, including those between processing nodes, to optional frame grabber modules and to the host PC, including display.
The Odyssey Xpro offers scalability by way of a pair of link ports dedicated to interconnecting multiple boards. These ports provide point-to-point PCI-X connections, delivering up to 1GB per second of bandwidth.
The Odyssey Xpro can capture from the majority of area or line scan video sources, thanks to a standard PCI mezzanine card (PMC) site located on the board that allows for the addition of Camera Link and analogue frame grabber modules.
The Camera Link module, which acquires up to 680MB per second, is available in two versions. The dual-base version enables simultaneous acquisition from two completely independent Camera Link video sources using the Base configuration, while the single-full configuration acquires from a single Camera Link video source using the Base, Medium or Full configuration.
The analogue module, which acquires up to 800MB per second, provides four 10-bit/100MHz channels that can be used independently or synchronised together.
The Odyssey XCL is an entry-level, non-scalable version of the Odyssey Xpro, geared towards cost-sensitive yet computationally intensive applications.
Also featuring the Motorola G4 PowerPC embedded microprocessor, running at 600MHz, the Odyssey XCL offers 256MB of DDR SDRAM memory, up to 1GB per second of external I/O bandwidth, over 5GB per second of memory bandwidth and delivers up to 120 BOPS.
The Odyssey XCL also takes full advantage of PCI-X bus technology and integrates a Camera Link frame grabber that acquires up to 680MB per second and offers the same configuration options as the Odyssey Xpro Camera Link module.
The Odyssey software development tools for both the Odyssey Xpro and Odyssey XCL offer developers a choice of application programming interfaces (APIs) and programming models.
The board can be programmed using either the award-winning, hardware-independent Matrox Imaging Library (MIL) or the Matrox Odyssey Native Library (ONL), which is compatible with the Matrox Genesis family of vision processors.
Both options come with royalty-free run-time environments.
Developers are provided with high-level algorithms and the ability to distribute workload across multiple boards. Highly optimised for the Odyssey family's onboard processors, these software libraries are carefully tuned to exhibit low function call overhead, thereby maximising application efficiency and performance.
In terms of board control, developers can either run their application remotely from the host PC, run it entirely on the board itself, or write their own custom functions for the Odyssey's onboard processors.