Fine Grain Parallelism

• Given a sequence of instructions the issue logic will look at the dependencies between multiple instructions in a window.
  • Dependencies: Data or Control dependency.
  • Dependencies: False or true.
  • Consider add R1, R2 followed by add R3, R4.
  • Consider div R1, R2 followed by div R3, R4
  • Consider load X followed by load Y
• Independent instructions can be executed in parallel if structural resources are available.

Granularity

• Medium Grain Parallelism: Parallelism of an application can be implemented by multiple threads in a single process.
• Usually programmers have to “use” threads in the design.
• Threads are scheduled by user (e.g. pthreads) or by OS (kernel threading).
• Coarse Grain Parallelism: Parallelism in a system by virtue of several concurrent processes
• Need to synchronize using semaphore or other synchronization objects.
• Example: Server side threads for web servers. FTP servers etc.