Sign up for our newsletter and get the latest HPC news and analysis.

Convey Unveils Financial Analytics Personality

convey_logo_smConvey Computer has announced the latest in a series of ‘personalities’ to adapt their hybrid-core computing platform to a new set of algorithms.  The financial analytics personality [FAP] for the Convey HC-1 is aimed at, you guessed it, the financial industry.  According to industry sources, 73 percent of US equity trading volume is based on advanced algorithmic strategies.  Many of these algorithms are based upon the same notional equations, but tweaked per the individual organization’s specific needs.  Convey now has a personality toolkit to accelerate these algorithms.

One of the major problems facing developers of financial applications is that the instruction set of commodity processors doesn’t map well to many financial algorithms,” states Tony Brewer, Convey co-founder and chief technology officer. “Even a simple intrinsic function can take hundreds of instructions and, in many of these applications, that intrinsic might be executed millions of times.”

These application specific instructions are hard-coded into the HC-1’s FPGA’s thanks to the Convey compilers.  As such, the respective developers aren’t required to know anything specific about gate arrays to successfully accelerate their applications.

Functions like pseudorandom number generation (Mersenne Twister), quasirandom sequence generation (Sobol sequence), and cumulative normal distribution are all packaged into what we call a personality,” said Brewer. “The personality carries out these functions at the gate level, which opens up opportunities for parallelism and vastly increased performance. But the real key to the Convey system is that you don’t need exotic language variations or non-standard programming extensions to access the hardware.”

If you’re interested in taking a look at exactly how these financial primitives perform on the Convey platform, check out their booth [#2589] on the SC09 show floor.

Resource Links: