New GPU Optimization & Parallel Programming series on Microway Blog

We're launching a brand new blog post series on parallel programming for accelerators (GPUs and beyond). The series begins with posts on high-performance CUDA GPU programming. Start with the post below:

CUDA Parallel Thread Management

Abstract:  Regardless of the environment or architecture you are using, one thing is certain: you must properly manage the threads running in your application to optimize performance. This post will discuss how to get the most out of your threads in a CUDA application.

Read the full blog post

Accelerate Your Geospatial Intelligence Applications

Microway is working with NVIDIA to speed up defense and geo intelligence applications with GeoInt Accelerator products. This community heavily relies on accurate and timely information in its strategic and day-to-day operations. Intelligence gathering and assessment are essential to these applications that analyze data coming from a number of diverse sources including satellites, UAVs, surveillance cameras, drones, sensors and radar.

Converting the raw data from these sources into actionable information requires a significant infrastructure. The NVIDIA GeoInt Accelerator is the first accelerated platform that enables defense and homeland security analysts to accurately generate actionable intelligence from petabytes of data, images and video. Powered by GPUs, Image and Video Processing can run up to 10x-75x faster compared to CPU only systems.

Now you can experience these speed ups first hand, FREE!

NVIDIA Releases CUDA 5.5 Toolkit

The production release of CUDA 5.5 is now available from NVIDIA. Features of note include:

Optimized For MPI Applications

• Enhanced Hyper-Q support for multiple MPI processes via the new Multi-Process Service (MPS) on Linux systems
• MPI Workload Prioritization enabled by CUDA stream prioritization
• Multi-process MPI debugging and profiling

Guided Performance Analysis

• Step-by-step guidance helps you identify performance bottlenecks and apply optimizations in the NVIDIA Visual Profile and Nsight, Eclipse Edition

Support For ARM Platforms

• Native compilation, for easy application porting
• Fast cross-compile on x86 for large applications

NVIDIA CUDA 5.5 Toolkit

"Big Data" Hadoop meets HPC?

As Apache Hadoop moves from version 1 to version 2, the structure is changing: no longer will MapReduce be the sole data processing framework. Other frameworks, such as MPI, will plug-in to Hadoop, allowing them to take advantage of the powerful capabilities of Hadoop, such as the HDFS distributed filesystem. The module which makes this possible is named YARN (Yet Another Resource Negotiator).

The introduction of YARN does not make traditional HPC obsolete; nor is MapReduce going away. Instead, this development gives more flexibility to groups with "Big Data" requirements. Data can be accessed and analyzed using their existing MapReduce workflows. Further, applications which had no access to the data managed by Hadoop can run as well.

Read the full article by Douglas Eadline in Admin Magazine