Open-Source Scientific Algorithm Library (SAL) for High-Performance Embedded Computing
For many years, optimized versions of Mercury’s Scientific Algorithm Library (SAL) and API have been the de facto standard library implementation for PowerPC®-enabled radar applications. Using SAL has often resulted in orders-of-magnitude increases in vector math execution performance over other implementations. As technology has advanced to higher capabilities on single and many-core x86 and GPU platforms, Mercury has decided to contribute this rich API and the C version reference design of its math functions to the user community for evaluation, use, and augmentation.
Ease of Application Integration while Achieving More Computations Faster
Libraries are often tightly coupled with target hardware compute platforms, making optimization very difficult. The OpenSAL API, on the other hand, provides a level of abstraction from the hardware, making the application programmer’s job easier.
Scientific algorithm libraries are used as accelerators to improve size, weight, and power (SWaP) metrics for HPEC applications. Functions like fast Fourier transforms (FFTs) and fast convolutions are often optimized to execute on target platforms, enabling the target compute engine to execute real- time-constrained computations faster.
By offering OpenSAL to the open-source community, Mercury hopes that its capabilities will expand, so that it may become an open-architecture de facto standard for scientific algorithm libraries on many more compute platforms. By leveraging the wider community’s contributions, expertise, and support, we can all get there faster and better.
OpenSAL V1.0 is available now for download under open-source license GNU GPLv3 as a Sourceforge.net project.