govaerog

Impulse users include 8 of the top 10 government contractors, most Agencies, NASA, JPL and most of the National Labs. Working typically under NDA, Impulse provides high-rel solutions using COTS tools and FPGA boards. Impulse solutions are built on industry standards like C, OpenCL, VHDL and Verilog. They are finely documented and easy to maintain over many years. To get started use the Contact tab or the e-mail icon above to arrange a no-pressure technical call.

Image & Video Processing. 
Implement efficient application specific algorithms. Processing, reconstructing or converting high throughput semi-custom data can be extensively pipelined to run 10- to 100-times faster in FPGA. Our tool flow enables streams, signals or memories to be used to synch data. We also let you easily experiment with frame rates and resolution; as well as partitioning logic between FPGA and CPU. 
Complex Analytics. We excel at straight-ahead acceleration of huge data set analytics like Monte Carlo, Smith Waterman or math driven like BLAS. We have achieved 4x acceleration over CPU on applications that used to run in 20+ hours. Teams turn to Impulse to help them shave large chunks of processing time off of analytics as they grow. Impulse development tools help rapidly understand the flow.Encryption, Conversion and Transforms. These are often deployed as a heterogeneous design with an on board processor used for control logic, FPGA fabric programmed from an HLL as the infrastructure and the actual encryptor or transform running from VHDL or Verilog code. This is one of the quickest to iterate and most powerful prototyping methodologies you can work with.Wire-speed Ethernet Filtering. Streams processing is an effective way to monitor Ethernet feeds in real time. Using the ability of an FPGA to expand into flexible multi-core processors, each executing an independent streaming process, latencies are held to under a microsecond. Your custom trigger, filter or other logic is dropped into the middle of the stream, with minor effort, in a way that shaves weeks off of iteration times.