In addition to our hardware-accelerated products, we also offer custom-design services for those customers with proprietary algorithms and/or special needs. We work with the customer to fully understand their application in order to efficiently map the underlying computations into the most appropriate hardware device and platform. Applications are implemented using FPGAs, GPUs, and DSPs, among others. A variety of form factors are also available, from desktop supercomputers to rack-mounted clustered systems, as well as standalone embedded systems, to enable accelerated solutions for land, sea, sky, and space applications. More information about the underlying hardware devices and form factors is provided below.

Field-programmable gate arrays (FPGAs) are chips that can be completely programmed at the hardware level. They can be thought of as a “sea” of reprogrammable logic gates, connected via configurable routing resources. That is, the fundamental logic functions (e.g., AND, OR, NOT) can be modified and interconnected to build complex logic designs. FPGAs maintain several advantages over microprocessors and digital signal processors (DSPs). First, an FPGA can be ideally configured to perform a given task by designing custom arithmetic pipelines. Additionally, FPGAs are capable of massive computational parallelism within a single chip. Thus, FPGAs allow algorithms to take full advantage of the underlying hardware for optimized performance. In the coming years, FPGA performance is expected to advance much faster than that of microprocessors, thus further increasing their advantage for computationally intense applications. FPGAs are also available in a range of prices and sizes allowing them to be tailored for a wide variety of applications.

EM Photonics harnesses the immense power of commodity graphics processing hardware to accelerate numerical algorithms. A current GPU costs roughly the same amount as a high-end CPU, but is capable of significantly higher floating-point performance. The reason for this disparity is that the GPU needs only to perform specialized calculations, such as those required to render graphics to the screen, while the CPU must offer a complete set of functions. The benefit for high-performance computing is that GPUs are high-powered computation engines, which provide hardware support for many common linear algebra and trigonometric functions. Using this technology, EM Photonics develops very powerful computational engines that can still be run in a desktop environment.

Many times, accelerated solutions in a desktop form factor are appropriate. These designs are implemented via expansion cards in standard desktop PCs. Desktop supercomputer processing is frequently applied to scientific computing applications, where users need to perform computationally intense simulations. This form factor is the most requested, as it is applicable to a broad range of problems.

Although the desktop supercomputer is powerful, it is not always a practical solution. Frequently, customers are looking for an embedded solution that can be deployed as a standalone unit or integrated into an existing framework. Embedded systems require unique designs, as the deployment locations restricts the underlying hardware that can be used, in terms of I/O interconnections, power requirements, and form factor. EM Photonics has experience in developing applications destined for a variety of land, sea, sky, and space platforms.