Impulse Accelerated Technologies will extend their C-to-FPGA compiler to support Pico Computing’s M-Series modules. The Pico modules feature the Xilinx Virtex-6 and Spartan-6 FPGA devices in a configuration that enables engineers to scale up from one to multiple modules. As a result, algorithms can be optimized for lower power, higher parallel-streaming operation on an FPGA, and then scaled to increase performance.
The Pico Computing’s M-Series modules are ideal for security and bioinformatics applications. Algorithms in both cases involve large amounts of non-sequential logic that accelerates well by massive parallelization. The Impulse tools enhance these solutions by enabling designers to mix VHDL and Verilog with ANSI C. In this methodology the system logic that will need to incur many iterations can be developed in C. The portions, such as actual encryption engines, that use the most processing logic can be isolated as machine generated VHDL and further refined by hand if needed. Impulse C can easily call external functions in HDL or via optional math.h, image processing and other libraries.
FPGAs are becoming more popular as hardware accelerators in high performance computing applications. For certain logic types, offloading computationally intensive algorithms to an FPGA enables them to be accelerated, outperforming CPU/DSP/GPUs at lower power though parallel streaming operation.
Using algorithmic derived hardware acceleration, developers partition logic among processing elements with a single line of C-code. This method also compiles to hardware logic, embedded cores and even FPGA-enabled boards without requiring software developers to learn how to program in HDL. Impulse C tools support automatic pipeline generation, instruction scheduling and other optimizations for increased algorithm throughput.