Designed for floating-point intensive applications, the Intel Advanced Vector Extensions (AVX) are a new 256-bit instruction set extension to Intel Streaming SIMD Extensions (Intel SSE). Released to support the 2nd generation Intel Core processor family, AVX improves performance due to wider 256-bit vectors, a new extensible instruction format (Vector Extension or VEX), and by its rich functionality.
“This paper discusses options that developers can choose from to integrate Intel® AVX into their application without explicitly coding in low level assembly language. The most direct way that a C/C++ developer can access the features of Intel® AVX is to use the C-compatible instrinsic instructions. The intrinsic functions provide access to the Intel® AVX instruction set and to higher-level math functions in the Intel® Short Vector Math Library (SVML). These functions are declared in the immintrin.h and ia32intrin.h header files respectively. There are several other ways that an application programmer can utilize Intel® AVX without explicitly adding Intel® AVX instructions to their source code. This document presents a survey of these methods using the Intel® C++ Composer XE 2011 targeting execution on a Sandy Bridge system.
Download the Full Article (PDF).