TotalView Offers Debugging Support for Intel Xeon Phi Coprocessor

TotalView 8.12 Features Support for Cray XC30™ Supercomputer Series and Apple Lion and Mountain Lion Platforms

Rogue Wave Software, the largest independent provider of cross-platform software development tools and embedded components for the next generation of High Performance Computing (HPC) applications, announced the release of its leading parallel debugger, TotalView 8.12. This version of TotalView offers official support for the Intel^® Xeon Phi™ coprocessor, Cray^® XC30™ series supercomputers, and Apple^® OS X Lion and Mountain Lion platforms. The release of TotalView 8.12 is part of Rogue Wave's commitment to improve developer productivity on advanced architectures.

“Intel and Rogue Wave have a long history of working together to ensure customer benefit from the combination of Intel technologies with Rogue Wave software. This continues with our work together on tools specifically optimized for the Intel Xeon Phi,” stated James Reinders, Director of Parallel Programming at Intel. “Being known as an intuitive and easy-to-use debugger, TotalView works to help developers to quickly and confidently take advantage of the massive parallelism available on the Intel Xeon Phi coprocessor.”

TotalView 8.12 gives developers the ability to view, control, and debug codes running on both the host processor and the Intel Xeon Phi coprocessor. This version of TotalView supports host-side applications using the Intel offload directives (LEO) and applications running natively on the Intel Xeon Phi coprocessor. Users can debug scalable MPI applications that are launched from the host environment, but run as native on one or more Intel Xeon Phi coprocessors on a server or across the nodes of an enabled cluster.

The early-access version of TotalView 8.12 has already been deployed and successfully used at a number of strategic customer sites over the past six months, including the National Institute for Computational Sciences (NICS) and Sandia National Laboratories. The experience of these sites was that the porting process was fairly simple due to the similar architecture between the Intel Xeon processor and the Intel Xeon Phi coprocessor; however, the interesting challenge was optimizing the code to fully leverage the many cores of the advanced coprocessor. Within the framework of the Beacon Project, teams from the NICS at the University of Tennessee used the early-access version to assist in the optimization process for a number of different codes, including the kinetic model of computational fluid dynamics and the Gyro tokamak plasma simulation. By using TotalView, the teams succeeded in improving their applications' speed and performance on the Intel Xeon Phi coprocessor. At Sandia National Laboratories, teams have been testing the early-access version of TotalView 8.12 and found the asynchronous thread control to be a crucial functionality in optimizing code for the Intel Xeon Phi coprocessor.

“TotalView is a useful tool for optimizing code in our movement towards increasingly advanced and complex architectures. For codes developed on the Intel Xeon Phi coprocessor and other future architectures, thread control is going to be a particularly critical feature, and it is one that Rogue Wave's TotalView offers,” stated Richard Barrett, Principal Member of the Technical Staff and leader of the Application Performance Modeling and Analysis Team within the Extreme-scale computing group at Sandia National Laboratories. “The consistency of TotalView 8.12 with previous versions meant that there was a quick learning curve for our teams.”

“The initial success stories of TotalView debugging applications on this critical architecture predict even greater achievements in the future,” stated Chris Gottbrath, Principal Product Manager at Rogue Wave Software. “Rogue Wave continues to push TotalView to new bounds with advances in scalability and support for heterogeneous architectures. This investment shows our commitment to increasing developer productivity as technology progresses.”

As part of the full debugging support for the Cray XC30 series supercomputers, TotalView 8.12 provides functionality for Cray Abnormal Termination Processing (ATP), which allows users to troubleshoot problems that occur while using Cray systems in production. Also released in TotalView 8.12 is a new sophisticated debugging session management feature, which helps improve developer productivity by allowing users to save debugging session information and settings. Functionality introduced in TotalView 8.12 includes:

• Improved performance with startup, stepping, variable viewing, and breakpoint setting on C++ programs that make extensive use of templates.
• Support for Intel Xeon Phi coprocessors, Cray XC30 supercomputers, and Apple OS X Lion and Mountain Lion platforms.
• New Sessions Manager framework that saves time from one debugging session to the next.
• Support for Cray Abnormal Termination Processing (ATP).
• Wide compiler, OS, and platform updates.
• Extended STLView support for the STL set, multi-set, and multi-map templates.
• Improved performance at scale.

TotalView is a scalable and intuitive debugger for parallel applications written in C, C++, and Fortran. Designed to improve developer productivity, TotalView simplifies and shortens the process of developing, debugging, and optimizing complex applications. TotalView provides a powerful combination of capabilities for pinpointing and fixing hard-to-find bugs, such as race conditions, memory leaks, and memory overruns. Providing developers the ability to step freely, both forwards and backwards, through program execution, TotalView's unique reverse debugging capabilities drastically reduce the amount of time invested in troubleshooting code. To help developers maximize hardware capabilities, TotalView also provides debugging support for NVIDIA CUDA, OpenACC, and the Intel Xeon Phi coprocessor.