AMD CodeAnalyst Performance Analyzer is a free download that works great with Microsoft® Visual Studio®. Frank Swehosky, 10 year CodeAnalyst team veteran, gives a personal introduction to how CodeAnalyst integrates with Microsoft Visual Studio 2010 and gives some insight into the type of performance analysis you can do with native C/C++ code and managed code.

AMD recently announced the creation of a new program designed to accelerate the ecosystem and software applications that take advantage of AMD Fusion APUs – Accelerated Processing Units. How would you architect your software differently, knowing that the GPU could handle some of the general purpose computing tasks and free up the CPU to do more? Perhaps you're on the bleeding edge and already parallelize your code to do exactly that. Either way, you'll want to check out the AMD Fusion Fund.

This month AMD adds to its support of key Open Source Software communities by becoming a Gold Level Sponsor of the Apache Software Foundation (ASF). AMD has been actively supporting the Linux®, GCC, Xen, and KVM communities for the last several years, contributing code and optimizations to help take advantage of the innovative hardware features in our CPU and GPU technologies. We're expanding that effort to encompass technologies further up the software stack, namely the open source middleware and applications that help form the foundation of cloud computing and "big data" problem solving.

This article outlines a new method for round-to-even rounding with large-scale data processing and illustrates its application with a typical computational problem in which two integer numbers are added and scaled down (division by another integer), rendering a floating-pointing number that needs to be rounded. The new technique can be implemented through streaming single-instruction/multiple-data (SIMD) extensions (SSE) instructions, and can be expanded for the general division (no scaling) case. We compare results between traditional implementation and our optimized version running on x86 and x86-64 architectures using C++ on a Microsoft® Windows® platform.

This latest release includes the following:

• Updated default P-State values for many Family10h BSDs to correlate closely with real parts
• USB Mouse now supports absolute mouse movements for better mouse tracking with the host
• Significantly improved AVX performance
• and more

GPUPerfAPI is AMD's library for accessing GPU performance counters on ATI Radeon graphics cards. It is used by GPU PerfStudio 2 and the ATI Stream Profiler and is now available to third party developers who wish to incorporate it within their own applications. GPUPerfAPI supports DirectX10, DirectX11, OpenGL, and OpenCL applications.
OpenCL™ and the OpenCL™ logo are trademarks of Apple Inc. used by permission by Khronos.

Allen Ly, AMD Senior Application Engineer, conducted a live course on OpenCL™ programming in May 2010.  The presentation, training guide, and programming exercise material from his course are now available for your self-paced learning.  The course provides an overview of GPU compute, introduces the OpenCL programming model, and discusses optimization tips.
OpenCL™ and the OpenCL™ logo are trademarks of Apple Inc. used by permission by Khronos.

• Introductory Tutorial to OpenCL™ for HPC at SAAHPC'10
  July 12: Knoxville, Tennessee
• VMWorld
  August 30 – September 2: San Francisco, California
• Oracle OpenWorld
  September 19-23: San Francisco, California