AXISLib is part of the copmpany's software development environment for HPEC (high performance embedded computing) applications.
Release of AXISLib-AVX 2.6 delivers faster FFTs, new matrix math functions, and new 'half float' conversion functions.
The latest release delivers double the performance of the previous version of AXISLib for 32 point FFTs, and now outperforms the Intel Math Kernel Library (MKL) – the gold standard – by 20 percent. This can be vital for radar applications such as pulse compression and other compression and pattern matching algorithms.
Also targeted at advanced radar applications, the new matrix math functions can lead to a 300 µs (microsecond) saving in compute time when working with 128 x 128 matrices.
The new 'half float' conversion functions make data translation from sensor to CPU more efficient, and can double the data throughput from sensor to CPU, reducing latency and increasing data processing capability.
AXISLib 2.6 enables users of Abaco's SBC347D – based on the Intel Xeon-D processor – as well as Abaco's entire family of platforms based on latest generation Intel Core™ i7 processors – to take maximum advantage of the performance these processors offer while reducing development time, cost and risk.
"In today's demanding defense applications, saving even a few microseconds in processing time can make the difference between mission success and mission failure," said Mrinal Iyengar, Vice President, Product Management at Abaco Systems. "These enhancements to AXISLib provide those savings in several ways that, cumulatively, add up to a very worthwhile increase in system performance."