TotalView

TotalView
Name	TotalView
Developer	Rogue Wave Software
Released	1995
Latest release version	2021
Programming language	C, C++
Operating system	Linux, macOS, Windows, AIX, Solaris
Genre	Debugger
Website	Rogue Wave Software

TotalView is a commercial debugger designed for debugging parallel, multithreaded, and distributed applications. It is used by developers, researchers, and engineers working with high-performance computing applications on platforms such as Linux, macOS, Windows, AIX, and Solaris. The tool integrates with development environments and middleware used in scientific computing, including support for languages like C, C++, and Fortran.

Overview

TotalView is positioned in the software tools ecosystem alongside products from IBM, Microsoft, Intel, NVIDIA, and Cray, addressing debugging needs in contexts involving MPI, OpenMP, POSIX Threads, and accelerator technologies such as CUDA and OpenCL. It targets users of HPC centers like Argonne National Laboratory, Lawrence Livermore National Laboratory, Oak Ridge National Laboratory, and organizations participating in projects with NERSC and XSEDE. The product competes with and complements tools such as gdb, LLDB, Valgrind, Intel Debugger (IDB), and DTrace (software).

History and Development

TotalView originated in the 1990s as debugging needs expanded in projects funded by entities such as the U.S. Department of Energy and collaborations among research universities like Massachusetts Institute of Technology, Stanford University, and University of California, Berkeley. Over time the product evolved through acquisitions and partnerships involving companies including Etnus, Compaq, SGI, and later Rogue Wave Software, with corporate transactions involving Perforce Software and interactions with vendors like Hewlett-Packard. Development milestones tracked the rise of standards such as MPI-2, MPI-3, and POSIX.1-2008, and adapted to hardware advances from Intel Xeon, AMD Opteron, ARM processors, and interconnects from InfiniBand and Omni-Path.

Features and Capabilities

TotalView provides features for core debugging tasks and advanced introspection used by teams at institutions such as NASA, CERN, Siemens, and Lockheed Martin. Key capabilities include process control and scaling used in simulations at Los Alamos National Laboratory and Sandia National Laboratories, memory debugging and leak detection comparable to AddressSanitizer (ASan), thread-aware breakpoints like those in Intel Inspector, reverse debugging features akin to rr (debugger), and visualization tools similar to interfaces provided by Eclipse and Visual Studio. It supports debugging of message-passing behaviors for stacks using Open MPI and MPICH, and integrates with job schedulers such as SLURM, PBS Professional, and LSF.

Architecture and Design

The architecture of the software uses client-server models resembling arrangements in X Window System clients and SSH-based remote sessions, enabling distributed debugging across network fabrics found in systems from HPE and Cray Inc. It manages symbol information produced by compilers like GCC, Clang, Intel C++ Compiler and PGI Compilers and leverages debugging formats defined by DWARF and ELF. For parallel control it interfaces with runtime libraries implementing MPI and threading libraries implementing POSIX Threads, and telemeters state using protocols influenced by SOAP-like RPCs and modern JSON-RPC patterns used in tools such as Language Server Protocol clients.

Use Cases and Applications

Typical applications are large-scale simulations in domains including computational fluid dynamics at NASA Glenn Research Center, climate modeling used by NOAA, computational chemistry in groups at Pfizer and Merck, and machine learning workloads run on clusters managed by Google and Facebook. Engineers use it to debug distributed consensus algorithms like those in Raft experiments, high-frequency trading prototypes in financial firms such as Goldman Sachs, and embedded systems development by companies like Siemens AG and Bosch. Research groups at universities including University of Cambridge, Imperial College London, and ETH Zurich use it alongside tools such as Matlab, R, and Python bindings for performance analysis and correctness verification.

Licensing and Distribution

TotalView is commercially licensed with tiers for academic, government, and commercial customers; procurement often involves vendor agreements similar to those negotiated with Oracle Corporation, Microsoft, and IBM. Distribution channels include direct enterprise sales, partnerships with original equipment manufacturers such as Dell Technologies and Hewlett Packard Enterprise, and reseller agreements with firms operating HPC services like Atos. Academic licenses and site-wide agreements are common at institutions funded by agencies such as National Science Foundation and Department of Defense.

Category:Debuggers