News

Open|SpeedShop new feature work

The O|SS team is working on an overview experiment which gathers performance information consisting of several types of metrics in one experiment run and creates csv (comma separated data) files for each thread of execution. The summary/overview experiment functionality is accessed via the cbtfsummary command and is available in the development tree of O|SS. cbtfsummary “how you run your experiment normally” is all that is needed to gather the metrics and create the csv files. For more details see the O|SS Reference guide available on https://openspeedshop.org/documentation.

Another feature that is in development is AVX512 and vector instruction detection in general for Intel platforms. This feature will allow vector instruction detection data gathered by O|SS in a CLI optional view. The view shows the time spent, percentage of time attributed to the instruction, name of the library or executable, vector instruction opcode and operands, and the maximum operand size for this instruction. The maximum operand size is the physical machine size, not the actual vector size at run time. That is an extension that is not available at this time. This feature in a development branch and will be available soon.

Open|SpeedShop at SC17

SC17.4CBlackRedTextOutline
Members of the Open|SpeedShop team will be at Super Computing in 2017. This year we will, once again, have booth (833) and will be giving demonstrations on-demand throughout the show at the Open|SpeedShop booth: 833.

To schedule a meeting with Jim or Don please send email to jeg AT krellinst.org or stop by the booth.

Members of our team are presenting the “How to Analyze the Performance of Parallel Codes 101” tutorial on Monday, 11/13/17 (8:30am-5:00pm). The slides are not available at this time, but here are the slides from last years tutorial: Here is the URL.

Open|SpeedShop at SC16

SC16.4CBlackRedTextOutline
Members of the Open|SpeedShop team will be at Super Computing in 2016. This year we will, once again, have booth and will be giving demonstrations on-demand throughout the show at the Open|SpeedShop booth: .

To schedule a meeting with Jim or Don please send email to jeg AT krellinst.org.

Members of our team are presenting the “How to Analyze the Performance of Parallel Codes 101” tutorial on Monday, 11/14/16 (8:30am-5:00pm). The slides are not available at this time, but here are the slides from last years tutorial: Here is the URL.

Open|SpeedShop Now on github

The Open|SpeedShop and Component Based Tool Framework (CBTF) sources are now available on github.
The repositories may be found at these locations:

The Open|SpeedShop release tarballs will be moving to github in the future. The release tarballs are currently still being hosted on sourceforge.

Open|SpeedShop GPU Support

We continue to work on performance analysis of NVIDIA GPU based applications. We are in the midst of a NASA Phase II SBIR project that focuses on providing both GPU and Intel MIC performance analysis capabilities. A good portion of this functionality is included in the 2.2.3 update release. More will be available in the coming months.

Open|SpeedShop at SC15

Members of the Open|SpeedShop team will be at Super Computing in 2015. This year we will, once again, have booth and will be giving demonstrations on-demand throughout the show at the Open|SpeedShop booth: 3302.

To schedule a meeting with Jim or Don please send email to jeg AT krellinst.org.

Members of our team are presenting the “How to Analyze the Performance of Parallel Codes 101” tutorial on Monday, 11/16/14 (8:30am-5:00pm). URL for slides will be posted prior to SC15.

Open|SpeedShop at SC14

We will be at Super Computing in 2014. This year we will not have a dedicated booth, but will be giving demonstrations and will be available for discussion at the DOE booth (1939):

  • TUE 11/18 2:00-3:00pm
  • WED 11/19 1:00-2:00pm
  • THU 11/20 10:00-11:00am

and at the Emerging Technologies booth (Show Floor: 233):

  • TUE 11/18 11:00am-12:00pm
  • WED 11/19 12:00-1:00pm and 3:00pm-4:00pm
  • THU 11/20 12:00-1:00pm

We will also be available for informal demonstrations any of the days starting Monday morning. Please send email to jeg at krellinst.org to schedule.

Members of our team are presenting the “How to Analyze the Performance of Parallel Codes 101” tutorial on Sunday, 11/16/14 Half-Day (8:30am-12pm). See this page for the slides: http://openspeedshop.org/?p=942

GPGPU performance analysis

We continue to work toward releasing a version of Open|SpeedShop that supports performance analysis of applications built with CUDA or OpenCL using NVIDIA GPUs. We had anticipated this being available much earlier but funding was not continued on the original project. We have been funded again and are working toward our goals of a full CUDA/GPU experiment. Availability will be shared here as we get closer to releasing this functionality.

Open|SpeedShop at SC13

While in Denver for SC13, stop by the Open|SpeedShop booth (#3722) for a demonstration of the latest features available with the release of Open|SpeedShop version 2.1, developers Jim Galarowicz and Don Maghrak will be on hand throughout the show to present the tool and to answer questions.

When: Demos also available whenever is convenient for you to visit Jim and Don at Booth #3722
Tuesday, November 19 @ 10:30A & 2:30P
Wednesday, November 20 @ 10:30A & 2:30P
Thursday, November 21 @ 10:30A & 1:30P

Jim and Don will also be available to discuss two other projects:
Open|speedShop’s new tool creation infrastructure, the Component Based Tool Framework (CBTF). This framework allows for the rapid creation of tools connecting existing CBTF tool infrastructure components into highly scalable HPC performance or debugger tools. Open|SpeedShop’s existing runtime support libraries and collectors are being replaced by the more modular and scalable CBTF components and libraries.
A SBIR research project for adding GPU support to CBTF and Open|SpeedShop named: Open|SpeedShop Ease of Use Performance Analysis for Heterogeneous Processor Systems. The project prototypes new Open|SpeedShop experiments to gather GPU related statistics for GPU based applications.

Dane Gardner, of Argo Navis Technologies LLC, will be demonstrating various tools utilizing the Parallel Tools GUI Framework (PTGF), an easy-to-use toolkit designed especially for new and existing parallel tool projects requiring a graphical user interface.

Where:
Supercomputing Conference 2013
Colorado Convention Center, Denver, Colorado
Open|SpeedShop – Booth #3722

Tutorial:
Monday, November 18, 8:30 am – 12:00 pm
How to Analyze the Performance of Parallel Codes 101
Room: 301, Colorado Convention Center

New User Guide

We have begun the process of rewriting our Open|SpeedShop Users Guide. See the Documentation section of this web site and choose the Open|SpeedShop Version 2.0.2 Users Guide URL. The previous Users Guide is still available, re-labled as Previous/Old Users Guide.

GPGPU performance analysis

We continue to work toward releasing a version of Open|SpeedShop that supports performance analysis of applications built with CUDA or OpenCL using NVIDIA GPUs. Availability will be shared here as we get closer to releasing this functionality.

SuperComputing 2012

Thanks for visiting our booth at SC2012 in Salt Lake City. It was good to meet with you. Please let us know your successes or if you are having any difficulties building or using Open|SpeedShop.

SuperComputing 2012

While in Salt Lake City for SC2012, please stop by the Open|SpeedShop booth (#1337) for a demonstration of the latest features available with the release of Open|SpeedShop version 2.0.2. Open|SpeedShop developers Jim Galarowicz and Don Maghrak will be on hand throughout the show to present the tool and to answer questions.

When: Demos also available whenever is convenient for you to visit Jim and Don at Booth #1337
Tuesday, November 13 @ 10:30A & 2:30P
Wednesday, November 14 @ 10:30A & 2:30P
Thursday, November 15 @ 10:30A & 1:30P

Jim and Don will also be available to discuss two other projects:
• The Krell Institute’s new tool creation infrastructure, the Component Based Tool Framework (CBTF). This framework allows for the rapid creation of tools connecting existing CBTF tool infrastructure components into highly scalable HPC performance or debugger tools. OpenSpeedShop’s existing runtime support libraries and collectors are being replaced by the more modular and scalable CBTF components and libraries. For more information see: http://ft.ornl.gov/doku/cbtfw/start.
• A SBIR research project for adding GPU support to CBTF and Open|SpeedShop named: Open|SpeedShop Ease of Use Performance Analysis for Heterogeneous Processor Systems. The project prototypes new Open|SpeedShop experiments to gather GPU related statistics for GPU based applications.

Dane Gardner of Argo Navis Technologies LLC, will be available to demonstrate a prototype version of a commercial, powerful, lightweight, debugging tool, called SWAT. The University of Wisconsin team, partners with Argo Navis, will also be available to discuss the SWAT debugger project at the University of Wisconsin’s Paradyn/Dyninst booth #126 .

Where:
Supercomputing Conference 2012
Salt Palace Convention Center, Salt Lake City, Utah
Open|SpeedShop – Booth 1337

Tutorial:
Sunday, November 11, 8:30 pm – 12:00 pm
How to Analyze the Performance of Parallel Codes 101
Room: not determined, will add when it becomes available

Please check SC12 website for information on how to register for the above tutorial.

About Open|SpeedShop

Open|SpeedShop is the result of a multiple year partnership between the Krell Institute, the Department or Energy (DOE) National Nuclear Security Administration, the DOE’s Office of Science; Lawrence Livermore National Laboratory, Los Alamos National Laboratory and Sandia National Laboratories. Open|SpeedShop builds on top of a broad list of community infrastructures, most notably Dyninst and MRNet from the Universities of Wisconsin and Maryland, libmonitor from Rice University, and PAPI from University of Tennessee, Knoxville.
Open|SpeedShop is an open source multi platform Linux performance tool which is targeted to support performance analysis of applications running on both single node and large scale IA64, IA32, EM64T, AMD64, PPC, Blue Gene (L, P, and Q) and Cray (XT, XE, and XK) platforms.

Open|SpeedShop is explicitly designed with usability in mind and is for application developers and computer scientists. The base functionality includes:

* Sampling Experiments
* Support for Callstack Analysis
* Hardware Performance Counter Access and Sampling
* MPI Profiling and Tracing
* I/O Profiling and Tracing
* Floating Point Exception Analysis

In addition, Open|SpeedShop is designed to be modular and extensible. It supports several levels of plug-ins which allow users to add their own performance experiments.

Open|SpeedShop development is hosted by the Krell Institute. The infrastructure and base components of Open|SpeedShop are released as open source code primarily under LGPL. For more information, visit: osstransfer.wpengine.com
About Component Based Tool Framework (CBTF):
This project creates a community tool infrastructure (CBTF) that provides individual tool building components to allow others to add or replace modules, to integrate parts into their own tools, or to customize existing solutions. CBTF builds on the existing Open|SpeedShop infrastructure and decomposes it into individual modules that match the necessary tool components. At the same time the project addresses the challenges found in performance tools for high processor count systems in each module. This community tool infrastructure is currently being used to create an enhanced version of Open|SpeedShop, which, while completely different in its architecture, will provide scalable performance analysis for high processor count applications through a familiar interface. This project also builds upon and enhances capabilities and reusability of project partner components. For more information see: http://ft.ornl.gov/doku/cbtfw/start.

About: Scalable Targeted Debugger for Scientific and Commercial Computing (SWAT):
SWAT is a commercial, powerful, lightweight, debugging tool, which will be of use to both supercomputer application programmers and to programmers of cluster- and cloud-based parallel e-commerce and engineering systems and middleware. The tool is based on STAT, a stack-trace debugging tool produced jointly by the University of Wisconsin and Lawrence Livermore National Laboratory. The feasibility of STAT has been proven in demonstrations on the largest supercomputers, up to 200,000 processes on both the IBM Blue Gene and Cray XT systems. Argo Navis Technologies LLC and the University of Wisconsin are partnering to bring SWAT to the HPC and commercial market place through a DOE funded Small Business Technology Transfer (STTR) program.

About: Open|SpeedShop Ease of Use Performance Analysis for Heterogeneous Processor Systems

This project is focused on the research and integration of heterogeneous processor support into Open|SpeedShop, concentrating specifically on GPUs. Areas of research include:
• Finding the sections of code that are candidates to become GPU kernels.
• Reporting the time spent in the GPU device.
• Reporting the cost and size of data transferred to and from the GPU.
• Reporting information to understand the balance of CPU versus GPU utilization.
• Reporting information to understand the balance transfer of data between the host and device memory with the execution of computational kernels.
• Reporting information to understand the performance of the internal computational kernel code running on the GPU device.

This project is a DOE funded Small Business Innovation Research (SBIR) program.

Open|Speedshop Forums moved to Google Groups
Please see the forum page for more information.

Beta Version of Open|SpeedShop 2.0.2 is available on sourceforge

A new beta version of Open|SpeedShop (2.0.2) that includes better support for the Blue Gene and Cray XT/XE platforms. New features were added to the osscompare convenience script. There are also updates to the Command Line Interface views which allow users to do arithmetic on columns to create new derived metric values/columns.

How-To-Use Open|SpeedShop HPC-Admin Magazine Article

In this article, we will describe how to use Open|SpeedShop through step-by-step examples illustrating how to find a number of different performance bottlenecks. Additionally, we will describe the tool’s most common usage model (workflow) and provide several performance data viewing options.

See Look for Bottlenecks with Open|SpeedShop

SuperComputing 2011

SC2010 logo
Thanks for visiting us at SuperComputing 2011 in Seattle! It was very nice to meet with all that stopped by for a demo!

We gave two Open|SpeedShop related tutorials at SC2011.
Monday, November 14, 1:30 pm – 5:00 pm
M15: How to Analyze the Performance of Parallel Codes 101

Friday, November 18, 8:30 am – 12:00 pm
F01: An Introduction into Performance Analysis for HPC Systems with Open|SpeedShop

Our SC2011 tutorial slides are available under the Tutorials menu on this website.

Available Installations

We’ve added a web page that describes where you can access existing Open|SpeedShop installations. The first update is for ORNL and Jaguar.
See Available Installations(Click Here)

SuperComputing 2011

SC2010 logo
Please stop in and visit us at SuperComputing 2011 in Seattle!

Where:
Supercomputing Conference 2011
Washington State Convention Center
Krell Institute – Booth 107, fourth floor

While in Seattle for SC11, stop by the Krell Institute booth (#107) for a demonstration of the latest features available with the release of Open|SpeedShop version 2.0.1. We will be giving demos at our booth during the course of the conference. It will be great to meet with you and discuss your performance tool needs.

We are also giving two Open|SpeedShop related tutorials at SC2011.
Monday, November 14, 1:30 pm – 5:00 pm
M15: How to Analyze the Performance of Parallel Codes 101

Friday, November 18, 8:30 am – 12:00 pm
F01: An Introduction into Performance Analysis for HPC Systems with Open|SpeedShop

Our SC2011 tutorial slides are available under the Tutorials menu on this website.

Beta Version of Open|SpeedShop 2.0.1 is available

A new beta version of Open|SpeedShop (2.0.1) that includes better support for the Blue Gene and Cray XT/XE platforms. Also included is better support for the online portion of the Open|SpeedShop tool. This will allow attaching to running processes and viewing of intermediate performance data results. These features had limited support in Open|SpeedShop version 1.9.3.4 and 2.0.0. New features were added to the osscompare convenience script and to the GUI compare panels (CC). There are also updates to the new hwcsamp experiment which allows users to sample up to 6 hardware counters during an experiment run.

SuperComputing 2010

SC2010 logo
Thanks for visiting with us at SuperComputing 2010 in New Orleans!

We gave a record number of demos at the Krell Institute Booth over the course of the conference. It was great to have met with all who stopped by!

Our SC2010 tutorial slides are available under the Tutorials menu on this website.

Beta Version of Open|SpeedShop 2.0.0 is available

A new beta version of Open|SpeedShop (2.0.0) that will include support for the Blue Gene and Cray XT/XE platforms.

Also included is better support for the online portion of the Open|SpeedShop tool. This will allow attaching to running processes and viewing of intermediate performance data results. These features have limited support in Open|SpeedShop version 1.9.3.4.

New Open|SpeedShop Release: Version 1.9.3.4

A new, 1.9.3.4, update version of Open|SpeedShop was released July 26, 2010. This version improves the support for the Cray-XT5 platform. This version also includes display speed improvements and bug fixes. It also introduces another convenience routine, osscompare, which allows the ability to compare up to five experiments (database files) to each other. osscompare can be run in batch mode. Another set of fixes was added in order to better support the MPI experiments (mpi and mpit) for tracing MPI functions in Fortran applications. These issues were mainly seen on applications using mpich2, mvapich or mvapich2 MPI implementations.

This update also fixes several other issues and bugs. This release is an update to the 1.9.3.3 that was released in February 27, 2010. For help or questions please use the “Contact Us” link on this page.

New Open|SpeedShop Release: Version 1.9.3.3

The new 1.9.3.3 update version of Open|SpeedShop was released February 27, 2010. This version improves the support for call stack unwinding for Intel compiled applications, cointains incremental fixes toward support for PPC platforms, contains iIncremental fixes toward support for the Cray-XT5 and BG/P platforms, includes fixes for better mpich2 and mvapich2 support for OpenSpeedShop experiments on Fortran applications, contains, configuration changes so that MRNet 2.2 and Dyninst 6.1 are the default versions for the online mode of operation, Upgrades the components used by OpenSpeedShop to these new versions:
libdwarf 20091230
libunwind 20100123
PAPI 3.7.0
sqlite 3.6.22
monitor 20091010
Dyninst 6.1
MRNet 2.2

This update also fixes several other issues and bugs. This release is an update to the 1.9.3.2 that was released in January 26, 2010, but contained some call stack unwinding problems. For help or questions please use the “Contact Us” link on this page.

New Open|SpeedShop Release: Version 1.9.3.2

The new 1.9.3.2 update version of Open|SpeedShop was released January 26, 2010. This version improves the support for individual pthread selection when viewing performance data in the GUI, improves cluster analysis support for pthreaded applications, improves the configuration support for Intel MPI and SGI MPT, and improves the SGI MPT, mpich2, and mvapich application recognition at execution time.. The base open source components that Open|SpeedShop uses have also been updated to new and improved versions, most notably, libunwind. This update also fixes several other issues and bugs. This release is an update to the 1.9.3.1 that was released in November, 2009. For help or questions please use the “Contact Us” link on this page.

SuperComputing 2009

Thanks to everyone that stopped by our booth for demos and discussion at our booth at SC 2009 in Portland, Oregon November 16-19, 2009. It was great to meet you and have the opportunity to discuss your performance tool needs and requirements. For those that couldn’t attend and would like to ask a question, please use the “Contact Us” link on the right side of our website to send your question or comment. Thanks!

New Open|SpeedShop Release: Version 1.9.3.1

The new 1.9.3.1 update version of Open|SpeedShop was released November 10, 2009. This version fixes execution and build issues found when building and using Open|SpeedShop on Fedora 11. This update also fixes an issue with building the Dyninst component on Ubuntu. Otherwise, this release is the same as the 1.9.3 that was released shortly before this update.

New Open|SpeedShop Release: Version 1.9.3

The new 1.9.3 version of Open|SpeedShop was released October 31, 2009. In this version, there are enhancements the build system in order to improve the building static versions of the collectors and static versions of the supporting runtime libraries. This change allows users to include these static libraries into their application in order to allow collection of performance data on machines that are more unique than the typical standard cluster. To support this change, Open|SpeedShop supporting components needed to be also targeted to be built statically for the non-standard platforms. So, a new build script which targets building supporting components for the non-standard platforms was also added. This release also contains bug fixes for the 1.9.2 version of Open|SpeedShop.

Click on the “DOWNLOAD Open|SpeedShop 1.9.3” button on this home page to download the release. Click on the Download link under the “Bookmarks” section to explore all the release files, including release notes and the build and install guide.

Open|SpeedShop Release: Version 1.9.2

The new 1.9.2 version of Open|SpeedShop was released August 16, 2009. This release contains many improvements to the offline version of Open|SpeedShop. There were changes to openss command syntax to increase the ease of use for the offline default mode of operation. Several GUI enhancements have been made in the StatsPanel to allow easier creation of optional views of the performance data and also in the custom comparison view, where short cut icons have been added. This release also includes more build, configuration, and install updates. More updates to the install script to check for more required packages and let the user know which packages are missing before the script invokes the build. This release includes enhanced configuration support for mpich2, mvapich and mvapich2. Several bugs fixes are also included in this version.

Click on the “DOWNLOAD Open|SpeedShop 1.9.2” button on this home page to download the release. Click on the Download link under the “Bookmarks” section to explore all the release files, including release notes and the build and install guide.

Try the Open|SpeedShop 1.9.2 Live CD

Create a live CD from the downloadable OpenSpeedShop-Live.iso image available on sourceforge.net. This allows you to try Open|SpeedShop without having to do a native install on your machine. Here are the instructions on how to try it out:

  • Download the iso image (Click Here)
  • Burn ISO image onto a CD
  • Boot from CD
  • Wait until “Getting Started” web page appears
  • Use the Users Guide or the Quick Start Guide from this website or the LiveCD to run Open|SpeedShop

Latest Tutorial Slides Available:

Open|SpeedShop IEEE Cluster 2009 Tutorial Slides (PDF) August 27, 2009

Previous release information:

A new 1.9.1 version of Open|SpeedShop was released December 23, 2008. This release includes enhanced support for Ubuntu 8.10. This release also includes more build, configuration, and install updates. The new install script will check for more required packages and let the user know which packages are missing before the script invokes the build. Several bugs fixes and performance speedups in Open|SpeedShop itself are also included in this version.