Common Parallelization Strategies

Parallelization Strategies: challenge

 Selecting the best Parallelization Strategy: using proper parallelization strategy is crucial in
order to achieve the optimal performance, such that the calculation takes the shortest
possible time.
1. Crucial: Very Important
2. Optimal Performance: Best performance using minimum resources in minimum time.
 It is very difficult to give a universal method for finding the optimal strategy, since it
depends not only on the platform, but also on the application itself.

Parallelization Strategies:

 Parallelization Strategies defines how to implement parallelization

opportunities.
 How many levels?
 How many granular? (Depth of details)
1. Granularity:
Parallelization Granularity refers to the size of parallel entities at which we
can divide an application.
Granularity is concerned with depth or levels of details.
CGPA Full Degree GPA Per Semester  GPA Per Subject

1 Year 12 Months
1 Month 30 Days
1 Day 24 Hour
1 Hour 60 minutes
1 minute 60 Seconds
Granularity => Depth of details