Featured

Is Next-Gen Xeon Phi a Supercomputer-GPU Killer?

Nvidia GPUs held their own against Xeon Phi in 2013; today the Xeon E5/Phi combo smokes everything in its path and may end the need for supercomputing GPUs altogether. Benchmarks praising the power of the latest version of Intel’s Xeon E5 chips did more than just prove new silicon is …

Read Full Post Posted in Design

Mixing C++11 and TBB Threads

Threading Building Blocks (TBB) includes a set of threading class that live in the standard namespace. How well do these fit together with the C++11 classes? Some cases work, and some don’t. Jeff Cogswell clears it up. When you’re using Threading Building Blocks (TBB) with a C++11 compiler, you can …

Read Full Post Posted in Build

Paired With Xeon Phi, ‘Haswell’ Shatters Speed Records

A benchmark using intense risk calculator simulating the toughest challenge of finance-market analysis shows the new Xeon E5 leaping across long list of compute-intensive challenges. The recently released third edition of Intel’s “Haswell EP” Xeon E5 chalked up marks 30 percent faster than an identical system running the earlier “Ivy …

Read Full Post Posted in Design

Faster Parallel Code with ‘Lockless’ Programming

When you read about parallel programming, you’re likely to hear about the importance of lockless programming. What exactly is this? Why is it important? Jeff Cogswell walks you through the answers to these questions and more. As you learn more about parallel programming, you’re going to occasionally see people talking …

Read Full Post Posted in Build

Most Recent Posts Blog Archive

Testing Out OpenMP 4.0 Features in LLVM/Clang

The Clang compiler, which is part of the LLVM tools, includes a special branch that adds OpenMP support. In this blog, Jeff Cogswell tries out the OpenMP 4.0 features of Clang, and gets some interesting results. Let’s test out the OpenMP 4.0 features in clang. But before getting started, I …

Read Full Post Posted in Build

DoD Enlists Cray to Build New ‘Haswell’-Powered Military Supercomputer

The Department of Defense’s High Performance Computing Modernization Program (HPCMP) is purchasing a new $26 million supercomputer and storage system for the DoD Supercomputing Resource Center at the U.S. Army Research Laboratory. The next-generation Cray supercomputer will feature the new Intel Xeon processor E5-2600 v3 processor, formerly code-named “Haswell,” according …

Read Full Post Posted in Design

How to Use Debugging, Exception Features in TBB

Threading Building Blocks (TBB) includes a set of exception handlers that are carefully crafted for multithreaded support, as well as code for debugging. In this blog, Jeff Cogswell gives a brief introduction to both features. Writing parallel code can sometimes be difficult and can result in bugs and strange errors. …

Read Full Post Posted in Build

Code Teams Selected to Prep for Next-Gen DoE Supercomputer

The Department of Energy’s National Energy Research Scientific Computing Center (NERSC) is collaborating with an initial group of research teams to prepare them for its forthcoming $70 million, next-generation supercomputer. NERSC, located at Lawrence Berkeley National Laboratory, has chosen 20 research projects for a code optimization effort, where research teams …

Read Full Post Posted in Design

‘High Performance Parallelism Pearls’: Lessons from the Field

The latest book from James Reinders and Jim Jeffers packs the real-world parallel programming experiences of 69 experts into 28 chapters. “High Performance Parallelism Pearls: Multicore and Many-core Programming Approaches” aims to show readers how to “leverage parallelism on processors and coprocessors with the same programming – illustrating the most …

Read Full Post Posted in Design

Is Next-Gen Xeon Phi a Supercomputer-GPU Killer?

Nvidia GPUs held their own against Xeon Phi in 2013; today the Xeon E5/Phi combo smokes everything in its path and may end the need for supercomputing GPUs altogether. Benchmarks praising the power of the latest version of Intel’s Xeon E5 chips did more than just prove new silicon is …

Read Full Post Posted in Design

Mixing C++11 and TBB Threads

Threading Building Blocks (TBB) includes a set of threading class that live in the standard namespace. How well do these fit together with the C++11 classes? Some cases work, and some don’t. Jeff Cogswell clears it up. When you’re using Threading Building Blocks (TBB) with a C++11 compiler, you can …

Read Full Post Posted in Build

Paired With Xeon Phi, ‘Haswell’ Shatters Speed Records

A benchmark using intense risk calculator simulating the toughest challenge of finance-market analysis shows the new Xeon E5 leaping across long list of compute-intensive challenges. The recently released third edition of Intel’s “Haswell EP” Xeon E5 chalked up marks 30 percent faster than an identical system running the earlier “Ivy …

Read Full Post Posted in Design