Universal SIMD Mathlibrary Math functions for single precision floating point vector datatypes
Download

Universal SIMD Mathlibrary Ranking & Summary

Universal SIMD Mathlibrary Tags

mathematical functions mathematical functions vector datatypes precision floating point SIMD

Universal SIMD Mathlibrary Description

Math functions for single precision floating point vector datatypes Universal SIMD Mathlibrary (libsimdmath) is a library that provides standard math functions for single precision floating point vector datatypes.It is designed for the SIMD platforms x86 (SSE2), PowerPC, and Cell. It implements sin, cos, tan, asin, acos, atan, atan2, sqrt, exp, log, pow, abs, ceil, floor, ldexp, and frexp.Developer commentsFor general x86-linux I could not find a full implementation of the standard C mathfunctions for floating point vectors, which I needed to port my fast panorama stitcher and blender PTStitcherNG. Originally developed for the Cellprocessor of the Sony Playstation 3, I had added a port to MacOS. Both platforms provide builtin support for SIMD-mathfunctions. To unify the codebase of my original project, I wrote versions for all processors, which resulted in this library with a common C-programming interface.For a general introduction to SIMD see the entry in Wikipedia. This article and software only deal with vectors consisting of four single precision floating point numbers (4 · 32 bits = 128 bits). Execution time and accuracy do not depend on the number of floating point elements in the vector; however, I am not aware of existing platforms with more than 4.

Universal SIMD Mathlibrary Related Software