Load Balancing

Perfectly Load-Balanced Program

In a perfectly balanced parallel program no set of processors is idle while other set of processors is doing some computation

Amdahl’s Law

• Assume that code has serial fraction
• Let T(1) be the execution time in one processor
• T(P) is execution time on P processors
  
• Then speed up on P processors is given by

About Data

• In a shared memory parallel program variables are either “shared” or “private”

“private” Variables

• Visible to one thread only
• Changes made to these variable are not visible to other threads
• Example : Local variables in a function that is executed in parallel

“shared” Variables

• Visible to all threads
• Changes made to these variable by one thread are visible to other threads
• Example : Global data