Introduction to Parallel Computing: Lecture Notes – Module 2

Summary

Parallel computing speeds up tasks by using multiple processors. Students learn why it is important, how it works and what it is used for.

Motivating Parallelism

Parallelism speeds up computing and allows more data to be processed at once. It is used in many applications. It is important because single-processor speeds cannot keep increasing.

Superscalar Execution

Superscalar processors schedule and execute multiple instructions at once. They do this by looking at many instructions and choosing which ones can run together. This is limited by the complexity of the hardware.

Memory System Performance

Memory performance is important for parallel computing. It can be improved by using data in a way that reduces delays. Memory layout and how computations are organised also affect performance.

Shared-Address-Space Platforms

Shared-address-space platforms allow multiple processors to access the same memory. They can be non-uniform memory access or uniform memory access. Programming these platforms requires coordinating access to shared data.

Key terms

Moore's law

The idea that the number of transistors on a chip doubles roughly every 18 months.

Superscalar execution

A way of executing multiple instructions at once in a processor.

Memory latency

The delay before a memory operation can be completed.

NUMA

A type of shared-address-space platform where memory access time depends on the location of the memory.

Shared address space

A way of programming where multiple processors can access the same memory.

Message-passing platforms

A type of parallel computing platform where each processor has its own memory.

Questions this document helps with

• What is the role of parallelism in computing?
• How do superscalar processors work?
• What are the differences between NUMA and UMA platforms?

Summary written with AI from the document's text.