Tools for HPC development

Parallel applications development

• Data set preparation - dealing with data sets that tend to be much larger than in conventional computing.
• Code debugging - need special tools to debug code that runs on many processors.
• Code analysis and profiling - helps in load balancing.
• Program simulation - cheaper way to applications development.
• Interpretation and presentation of results - again, HPC applications tend to output vast amounts of data that needs to be displayed effectively.

• Data set preparation
  
  • Often dealing with data sets that tend to be much larger than in conventional computing.
  • Use of Graphical User Interfaces (GUI's) to set input parameters.
• Code debugging
  
  • Need special tools to debug code that runs on several processors.
  • Allows examination of the program source whilst it is executing. Alternatively, perform a post-mortem.
  • Facilities such as 'breakpoints' permit inspection and modification of variables in each of the processes.
  • Ideally, there should be a central facility to enable control of all processes.
• Code performance analysis and profiling
  A typical profiler will show:
  • Colour-coded graphics displays indicating when processors are computing, communicating, or idle.
  • Time axis to follow each processor's activities during execution.
  • Graphics to indicate communication patterns as a function of time.
  • Subroutine usage to guide where parallelism should be concentrated.
• Program simulation
  • A simulator allows development of parallel code on a serial machine.
  • Assures program will execute but will not predict runtime on parallel target.
  • For example, Cray T3D simulator on Cray YMP.
• Visualisation
  HPC applications tend to output vast quantities of data.
  • Visualization provides an effective means for presentation of results.
  • Creative uses of colour, texture, translucency, and others allow complicated models to be displayed in a form that can be more easily understood e.g. AVS, PVwave.

• Packages in areas such as advanced visualisation, CFD, Molecular Modelling, Geographical Information Systems (GIS).
• Grand Challenges as suggested by the HPC Initiative of the U.S. and U.K. governments.
• Database and transaction processing required by banks, airline reservation systems, and others.
• Virtual Reality for both work and leisure activities.