2. OmpSs User Guide

The information included in this document is provided “as is”, with no warranties whatsoever, including any warranty of merchantability, fitness for any particular purpose, or any warranty otherwise arising out of any proposal, specification, or sample. The document is not guaranteed to be complete and/or error-free at this stage and it is subject to changes without furthernotice. Barcelona Supercomputing Center will not assume any responsibility for errors or omissions in this document. Please send comments, corrections and/or suggestions to pm-tools at bsc.es. This document is provided for informational purposes only.

Note

There is a PDF version of this document at http://pm.bsc.es/ompss-docs/user-guide/OmpSsUserGuide.pdf

• 2.1. Installation of OmpSs
  • 2.1.1. Preparation
  • 2.1.2. Installation of Extrae (optional)
  • 2.1.3. Installation of Nanos++
    • 2.1.3.1. Nanos++ build requirements
    • 2.1.3.2. Nanos++ configure flags
  • 2.1.4. Installation of Mercurium C/C++/Fortran source-to-source compiler
• 2.2. Compile OmpSs programs
  • 2.2.1. Drivers
    • 2.2.1.1. Common flags accepted by Mercurium drivers
    • 2.2.1.2. Help
    • 2.2.1.3. Passing vendor-specific flags
  • 2.2.2. Compiling an OmpSs program for shared-memory
    • 2.2.2.1. Your first OmpSs program
  • 2.2.3. Compiling an OmpSs program with CUDA tasks
  • 2.2.4. Compiling an OmpSs program with OpenCL tasks
  • 2.2.5. Problems during compilation
    • 2.2.5.1. How can you help us to solve the problem quicker?
• 2.3. Running OmpSs Programs
  • 2.3.1. Runtime Options
    • 2.3.1.1. General runtime options
    • 2.3.1.2. CUDA specific options
    • 2.3.1.3. Cluster specific options
  • 2.3.2. Running on Specific Architectures
    • 2.3.2.1. CUDA Architecture
    • 2.3.2.2. Offload Architecture
  • 2.3.3. Runtime Modules (plug-ins)
    • 2.3.3.1. Scheduling policies
    • 2.3.3.2. Throttling policies
    • 2.3.3.3. Instrumentation modules
    • 2.3.3.4. Barrier algorithms
    • 2.3.3.5. Dependence managers
    • 2.3.3.6. Thread Manager
  • 2.3.4. Extra Modules (plug-ins)
• 2.4. Installation of OmpSs from git
  • 2.4.1. Additional requirements when building from git
  • 2.4.2. Nanos++ from git
  • 2.4.3. Mercurium from git
• 2.5. FAQ: Frequently Asked Questions
  • 2.5.1. What is the difference between OpenMP and OmpSs?
    • 2.5.1.1. Initial team and creation
    • 2.5.1.2. Worksharings
  • 2.5.2. How to create burst events in OmpSs programs
  • 2.5.3. How to execute hybrid (MPI+OmpSs) programs
    • 2.5.3.1. Compilation
    • 2.5.3.2. Execution
    • 2.5.3.3. Instrumentation
  • 2.5.4. How to exploit NUMA (socket) aware scheduling policy using Nanos++
    • 2.5.4.1. Automatic NUMA node discovery
    • 2.5.4.2. Using programmer hints
    • 2.5.4.3. Nesting
    • 2.5.4.4. Other
  • 2.5.5. My application crashes. What can I do?
    • 2.5.5.1. Does it crash at the beginning or at the end?
    • 2.5.5.2. Get a backtrace
    • 2.5.5.3. I cannot get a backtrace
  • 2.5.6. I am trying to use regions, but tasks are serialised and I think they should not
  • 2.5.7. My application does not run as fast as I think it could
  • 2.5.8. How to run OmpSs on Blue Gene/Q?
    • 2.5.8.1. Installation
    • 2.5.8.2. Compiling your application
    • 2.5.8.3. Instrumenting
  • 2.5.9. Why macros do not work in a #pragma?
    • 2.5.9.1. Alternatives
    • 2.5.9.2. Why rely on the native preprocessor, then?
  • 2.5.10. How to track dependences for a given task using paraver
    • 2.5.10.1. Functionality

• Index