Snapdragon Math Libraries
If you are an app developer or an algorithm developer dealing with low level math routines, you know that squeezing the peak performance out of any processor is difficult and time consuming. And it’s even worse, when you have to start from scratch to support the newer generation processor. To address this portability across generations and to give you superior performance while maintaining the numerical correctness for your mathematical routines, Qualcomm is providing developers with Snapdragon Math Libraries.
Snapdragon Math Libraries are high performance numerical libraries that support all generations of Qualcomm® Snapdragon™ processors. By using Snapdragon Math Libraries, not only can you take advantage of the performance offered by Snapdragon processors, but also depend on the numerical correctness and portability across Snapdragon processors. In addition, Snapdragon Math Libraries help your application run well across different generations of Snapdragon processors.
Snapdragon Math Libraries supports almost 2000 primitives (full BLAS levels 1, 2 and 3 support, and full LAPACK 3.6.0 support), including all precisions (single, double, single complex, and double complex). This means it can be used for anything from machine learning and deep neural networks to simulating electromagnetic wave propagation.
Snapdragon Math Libraries are designed to allow developers to meet the ever increasing demand for performance, decrease development time, while abstracting away the chipset differences between each generation of Qualcomm Technologies processors. This allows you to focus on higher level algorithms and user experience without worrying if your application will run well on an individual processor.
Snapdragon Math Libraries support Qualcomm Snapdragon mobile processors, allowing developers to use a single library and scale their algorithms from mobile phones.
With support for standard BLAS and LAPACK interfaces, the Qualcomm Snapdragon Math Libraries make it easy to port frameworks and applications to Qualcomm silicon.