Fall 2009 IDF session on Intel(R) Advanced Vector Extensions (Intel(R) AVX)

Are you looking to write the most cutting edge SIMD code? Do you want to learn about Intel® AVX, Intel’s newest instruction set? Would you like to know what tools are available to develop and analyze AVX code for the newest architecture? If so, come to the ARCS002 and ACQ001 sessions at Fall 2009 Intel Developer Forum.

Last year Intel announced Intel® Advanced Vector Extensions (Intel® AVX) at the Spring IDF forum which is slated to be a part of platforms ranging from notebooks to servers.  Intel® AVX is a new 256-bit SIMD FP vector extension of Intel Architecture which is targeted for the Sandy Bridge processor family in the 2010 timeframe. Intel® AVX accelerates the trend towards FP intensive computation in general purpose applications like image, video, and audio processing and engineering applications such as 3D modeling and analysis, scientific simulation, and financial analytics. The enhancements in Intel® AVX allows for improved performance due to larger vectors, new extensible syntax, and rich functionality including the ability to better manage, rearrange and sort data.

At this year’s fall IDF in San Francisco Intel will be conducting two sessions dedicated to Intel® AVX. The first presentation will include an overview of the AVX instructions, an in-depth discussion of the tools that are available today to develop AVX code and will explore performance of some AVX based kernels. The second presentation will be a chalk-talk/Q&A session on the same topic where Intel’s Senior Microprocessor Architects will be at hand to answer your questions on the Sandy Bridge architecture and the new Intel® AVX instruction set.

The Intel® AVX session will first delve into a brief overview of the benefits and key features of Intel® AVX such as wider register and vector widths, newer instructions like broadcast, maskmov and permutes and introduction of three/four operand instructions that allow for non-destructive destinations.

Are you ready to develop an AVX application today? If so, the next part of the presentation will give you the information that you need today to get a head start on Intel AVX software development. Intel engineers have been busy developing tools that can be used today to develop and analyze Intel® AVX applications.First and foremost, there is the Intel Profession Compiler Suite, Version 11.1 which contains an Intel compiler, Intel Performance Primitives and Math Kernel Libraries. The tools in the compiler suite have been extended to support AVX. Intel also provides a software emulation tool called SDE that allows users to run AVX code on any Intel platform. SDE allows developers to do functional validation of AVX code. SDE also packages other tools like xed, a disassembler utility for AVX code.  If the developer is looking extract more performance from their AVX code, Intel provide a tool called Intel Architecture Code Analyzer that can be used to analyze basic blocks for latency, throughput, execution port usage and critical path information. Developers can use that information to modify, fine-tune, and optimize their application for best performance.

The talk will then explore several AVX code kernels to demonstrate conversion of legacy SSE code to optimized AVX code, showcasing new AVX instructions and programming paradigms. The session will also discuss some AVX tunings tips and best known methods that will aid developers when developing with AVX instructions for Sandy Bridge architecture. We want you to begin your AVX experience armed with the right tools and methods that you need to succeed.

We hope to see you at the Fall IDF 2009 and our sessions, ARCS002 – Intel® Advanced Vector Extensions (Intel® AVX) – Intel’s Next Major Instruction Set Architecture Extensions and ARCQ001 - Q&A: Intel® Advanced Vector Extensions (Intel® AVX) on September 22, at 4:00pm and 6:00pm respectively.


Pallavi Mehrotra, Senior Software Engineer, Intel Corp.
Richard Hubbard, Senior Software Engineer, Intel Corp.