10 Chapter 1: Introduction
Processing capabilities
Matrox RadientPro CL features an on-board real-time processing FPGA device,
which can be configured to offload and even accelerate the most
compute-intensive part of typical image processing applications, without
generating additional data traffic within the host computer (Host).
Processing FPGA
The Processing FPGA on Matrox RadientPro CL is a highly customizable Altera
Stratix V FPGA. The operations performed on-board are controlled using the
Matrox Imaging Library (MIL) application development software. Using MIL,
the processing units (PUs) of an FPGA configuration can be rearranged to perform
the operations in the required sequence, without having to necessarily generate a
new FPGA configuration. When the need arises, Matrox’s FPGA design services
can be employed to generate an application-specific FPGA configuration.
Before the Processing FPGA can process grabbed images, they must be stored in
main on-board memory. If images stored in Host memory are required, they can
be streamed directly to the Processing FPGA for processing. Images resulting from
Processing FPGA processing can be stored in main on-board memory or streamed
to the Host.
On-board acquisition and processing memory
Matrox RadientPro CL has two types of on-board memory: main on-board
memory and dedicated Processing FPGA memory (optional).
Main on-board memory is used to store acquired images, and images for or
resulting from processing. As main on-board memory, Matrox RadientPro CL is
equipped with 2 Gbytes of DDR3 SDRAM. Main on-board memory is accessed
through the memory controller of the Processing FPGA. The memory controller
has multiple input ports and a data transfer rate of up to 12.8 Gbytes/sec.
Matrox RadientPro CL can optionally feature dedicated Processing FPGA
memory to store intermediate data while the Processing FPGA performs
operation. As Processing FPGA memory, Matrox RadientPro CL uses 4 banks of
QDR-II SRAM that total 16 or 32 Mbytes of memory and have a total transfer
rate of up to 4.0 Gbytes/sec in each direction.
