Open|SpeedShop

1
OpenSpeedShopA Dyninst Based Performance Tool
Overview Status

• Jim Galarowicz
• Manager - SGI Tools group
• Bill Hachfeld
• Steve Collins

2
Trademark Acknowledgements

• Intel, Intel Inside (logos) and Itanium are
  trademarks of Intel Corporation in the United
  States, other countries, or both.
• Linux is a trademark of Linus Torvalds in the
  United States, other countries or both.
• Qt and the Qt logo are trademarks of Trolltech in
  Norway, the United States and other countries.
• OpenSpeedShop, SGI SpeedShop, IRIX, SGI and SGI
  Altix are trademarks of Silicon Graphics Inc.
• AIX and PowerPC are trademarks or registered
  trademarks of International Business Machines
  Corporation in the United States, other
  countries, or both.
• Tru64 is a trademark of  Hewlett-Packard
  Corporation.
• AMD, the AMD Arrow logo,AMDOpteron, and
  combinations thereof, are trademaks of Advanced
  Micro Devices, Inc.
• All other trademarks mentioned herein are the
  property of their respective owners

3
Presentation Agenda

• OpenSpeedShop Presentation Agenda
• Project Overview
• Design Goals
• Design Overview

4
Project Overview

• What is OpenSpeedShop?
• Open Source Performance Tool Comprised of
• Based on SGI Speedshop software concepts for
  the IRIX operating system
• Developed for Multi-Platform Support
• Implemented Using Open Source Components
• DOE/NNSA co-funded via ASC (Advanced Simulation
  and Computing Program) PathForward
• Partnering with Universities of Wisconsin
  Maryland
• Schedule
• Project Began July, 2004
• Quarterly Milestone Releases
• First Development Phase Ends mid-2006

5
Design Goals

• Easy to Use
• Dont have to be a Computer Scientist or Computer
  Engineer
• What on earth is call stack sampling? Why would
  I use it?
• GUI, Command Line, Batch Scheduled -- Your
  Choice!
• Portable
• Platform Independent
• Use components that already support the platforms
  of interest
• Some recoding may be necessary if using unique
  features of a particular platform
• Scalable
• Thousands of Processors
• ASC laboratory systems already use 8192
  processors. Future systems even larger.
• SGI Altix system installed at a customer site
  has 10240 processors.
• Hundreds of Shared Libraries
• Some SGI customers load 300 shared libraries.

6
Design Goals (Continued)

• Extensible
• Gathering New Types of Performance Data
• New Views of Performance Data
• Supportable
• Tested Stable
• Testing done from the beginning. Not an
  afterthought!
• Test suites delivered with tool
• Simple Installation
• Troubleshooting tools assist in installation and
  configuration.
• Reuse
• Use Open Source Tools When Possible
• Dont reinvent the wheel!
• Keeping in Mind Licensing Requirements

7
Design Overview

• Foundation of Open Source Components
• Core Components (shown here)
• Build System (Autoconf, Automake, Libtool, Make
  GCC)
• Source Control (CVS)
• Testing (Python)
• Layered Approach
• Open Source OS-Dependent Layer
• Non-Monolithic to Encourage Reuse
• Higher Levels of Abstraction Going Up
• Extensible via Plugin Modules
• Support New Data Views
• Vendor Differentiation

8
Design Overview Dyninst

• Universities of Wisconsin Maryland
• http//www.dyninst.org/
• Provides
• Process Control
• Instrumentation of Unmodified Binaries
• Symbol Table Information
• Status
• Supports Most Platforms of Interest
• AIX /PowerPC,
• IRIX /MIPS,
• Linux /Intel IA-32,
• Linux /Intel IA64,
• Tru64/Alpha
• Additional Platform Support Planned
• Additional Compiler Support Planned

9
Design Overview DPCL

• Open Source Project
• http//sourceforge.net/projects/dpcl
• Provides
• Dyninsts Capabilities Across a Cluster
• Transport of Performance Data
• Status
• Supports Some Platforms of Interest
• AIX/Power, Linux/IA32, Linux/IA64
• Needs Better Per-Line Symbol Information
• Needs Better Shared Library Support
• Needs Scalability Enhancements
• Needs Integration of Hybrid changes
• Needs Integration of SGI changes/fixes

10
Design Overview SQLite

• Open Source Project
• http//www.sqlite.org
• Provides
• In-Process Relational Database Engine
• Single-File Databases
• Up to 2Tb
• Queries Specified Using SQL
• No Complex Configuration
• Status
• Supports Major Platforms of Interest
• AIX/Power, IRIX/MIPS, Linux/IA32, Linux/IA64,
  Linux/AMD Opteron, Tru64/Alpha
• Needs Little to No Enhancement

11
Design Overview PAPI

• University of Tennessee
• http//icl.cs.utk.edu/papi
• Provides
• Access to CPU Performance Counters
• Cross-Platform API
• Status
• Supports Major Platforms of Interest
• AIX/Power, IRIX/MIPS, Linux/IA32, Linux/IA64,
  Linux/AMD Opteron, Tru64/Alpha
• May Need Better Per-Thread Support

12
Design Overview Python Boost.Python

• Open Source Projects
• http//www.python.org
• http//www.boost.org/libs/python
• Provides
• Object-Oriented Scripting Language
• Extensibility via C/C
• Simple Generation of Wrappers (Boost.Python)
• Script-Driven Testing Environment
• Status
• Supports Major Platforms of Interest
• AIX/Power, IRIX/MIPS, Linux/IA32, Linux/IA64,
  Linux/AMD Opteron, Tru64/Alpha
• Needs Little to No Enhancement

13
Design Overview Qt

• Trolltech AS
• http//www.trolltech.com/products/qt
• Provides
• Object-Oriented GUI Toolkit
• GUI Builder
• HTML Help System
• Internationalization Support
• Cross-Platform API
• Status
• Supports Major Platforms of Interest
• AIX/Power, IRIX/MIPS, Linux/IA32, Linux/IA64,
  Linux/AMD Opteron, Tru64/Alpha
• Needs Little to No Enhancement

Qt is trademark of Trolltech in Norway, the
United States and other countries.
14
Design Overview Framework

• Provides
• Process Control
• Performance Data Collection Abstractions
• Collectors, Parameters, Metrics
• Symbol Table Caching
• Data Management Caching
• Extensibility via Collector Plugins
• Simplified Instrumentation
• Asynchronous Operation
• C API
• Status
• Design Completed
• Mostly Implemented

15
Design Overview UI Semantic Routines

• Provides
• Common Performance Tool Operations
• Create Process, Attach to Thread
• Start/Stop Thread
• Experiment Management/Selection
• Get/Set Parameters
• Start/Stop Data Collection
• Save/Restore Performance Data
• Framework State Management
• Status
• Most of Command Set Designed
• Partially Implemented

16
Design Overview CLI

• Provides
• Parsing of Complex Commands
• Logging of Commands
• Replay of Logged Commands
• Extensibility via Command Plugins
• Status
• Command Set Partially Defined
• Needs Usability Study
• Partially Implemented

17
Design Overview GUI

• Provides
• Display Panel Management
• Process Selection Control
• Experiment Selection Control
• Annotated Source Views
• Sorted Lists (e.g. Top 5 Functions)
• Data Comparisons
• Thread vs. Thread
• Pre-Optimization vs. Post-Optimization
• Wizards to Guide Novice User
• Extensibility via Panel Plugins
• Status
• Initial Usability Study completed
• Wizards Workflow Prototyped
• Needs Additional Usability Study
• Partially Implemented

18
Milestones

• Open/SpeedShop Milestones
• Infrastructure deliveries in September and
  December, 2004
• PC sampling experiment delivered to DOE
  partners on 3/14/05
• Additional Experiments to delivered one per
  quarter proposed schedule
• User Time 2nd Quarter 2005
• Hardware Counter 3rd Quarter 2005
• Floating Point Exception 4th Quarter 2005
• Input/Output tracing 1st Quarter 2006
• MPI call tracing 2nd Quarter 2006
• Support for additional platforms

19
Questions

• Questions
• Jim Galarowicz
• jeg_at_sgi.com
• Steve Collins
• slc_at_sgi.com