May be replaced by

INTRODUCTION TO COMPILERS FOR
HIGH PERFORMANCE COMPUTERS

Supercomputing Applications

• Used in scientific computing requiring very large compute time
• Weather prediction: For a days prediction at 109 fpo/sec requires 3 hours
• Used in biology, genetic engineering, astrophysics, aerospace, nuclear and particle physics, medicine, tomography ...

Power of supercomputers comes from

• Hardware technology: Faster machines
• Multilevel architectural parallelism
  Vector handles arrays with a single instruction
  Parallel lot of processors each capable of executing an independent instruction stream
  VLIW handles many instructions in a single cycle

Programming Paradigms

Vector Machines very close to sequential machines. Different in use of arrays
Parallel Machines deal with a system of processes. Has individual threads execution. Synchronization is a very important issue.

• Important things to remeber are
  • Aavoid deadlocks
  • Prevent race conditions
  • Avoid too many parallel threads
• Performance evaluation criterion
  • Speed up vs number of processors
  • Synchronization over heads
  • Number of processors which can be kept busy