Learning Objectives List sources of parallelization overhead Distinguish between

Question

Learning Objectives List sources of parallelization overhead Distinguish between task vs. data parallelism Apply Foster's methodology for parallel program design to produce a design that minimizes parallel overhead. ° Apply Amdalh's law to calculate the maximum speedup given a fraction of serial processing. . Calculate parallel performance metrics: speedup and efficiency 1. Two types of parallelism task parallelism - split program into major tasks and solve as many tasks in parallel as possible data parallelism - partition the data across the processors with each processor doing the same type of calculations on their own chuck of data a) Which of the above approaches is more scalable (can utilize more processors) as the problems get large (i.e., lots of data)? b) How might a combination of the two be used? 2. Let's think about parallelism, but with real-world examples. Here think of each person as a processor. For each example, determine the "major" sequence of tasks/steps, but also explain what tasks can be "data parallelized" assuming we had many people? a) Building a house b) Consider putting together a 5,000-piece jiggle-saw puzzle with different number of people . 2 people 5 people 25 people 100 people c) Preparing food for a large (1,000 people) banquet with each meal consisting of salad, entree, side-dish, dessert

Explanation / Answer

(a) Data parallelism utilize more processors as there is only one execution thread operating on all sets of data on different processors as compared to task parallelism which utilize less processor as each processor will execute a different thread or process on the same or different set of data. In data parallelism, the amount of parallelization is proportional to the input data size whereas in task parallelism, the amount of parallelization is proportional to the number of independent tasks to be performed.

(b) Data Parallelism is more scalable when there is a huge amount of data as it can run on multiple processors parallelly

Related Questions

Navigate

• Browse (All)
• Browse L
• Subjects

---

---

Integrity-first tutoring: explanations and feedback only — we do not complete graded work. Learn more.