Stone Ridge Technology and Impulse Accelerated Technologies teamed together to integrate Impulse C to FPGA toolset with Stone Ridge’s RDX-11 FPGA board and development kit. The integration enables software developers to write HLL (high level language) algorithms that rapidly compile to optimized RTL (run time language). Developers can purchase Impulse C or Stone Ridge boards from the respective manufacturers.
The integration provides C based links to hardware features and busses on the FPGA and RDX-11 board such that software developers can accelerate on the board without writing hardware controllers. Processes run as custom streaming cores in dedicated FPGA hardware without an operating system, to maximize throughput and reduce power consumption. Packet data integrity is increased when compared to microprocessor based single- or dual-stream solutions.
The Stone Ridge/Impulse collaboration is an example of hardware/software co-design where the computational resources are optimized for the software problem being addressed. For designs with significant non-sequential logic the speed improvements can be 10 – 100x. Compared to hand coded RTL methodologies, the design entry can take two thirds the time and iterations one eighth the time.
Applications currently in use at Stone Ridge using ImpulseC center around high-speed feeds of data, images or signals. The integration to the new RDX development kit brings to reconfigurable computing C programmable cutting-edge FPGA technology with high bandwidth and ample memory to achieve maximum processing power. Specifically regarding accelerated network processing, Stone Ridge boards work with Impulse C to enable software developers to move critical path network interface processes to hardware, where they run faster in multiple streams without the overhead of an operating system.
Impulse Accelerated Technologies and Stone Ridge Technology have preconfigured systems available that include all the hardware and software a development team will need to create a first application and connect it up to a datastream. Systems come with 8 hours of remote training and the option of having algorithms professionally refactored for multi-streaming parallelism. On-site training and installation are available.