Instruction Level Parallelism

Digital Principles and Computer Organization:

Chapter 10: Parallel Processing

Instruction Level Parallelism

• Instruction–Level Parallelism (ILP) is a technique used in computer architecture to improve processor performance by executing multiple instructions from a program in parallel. It focuses on finding and exploiting parallelism within a single instruction stream, rather than using multiple processors.

Concept :

• Traditional programs are written sequentially, meaning instructions execute one after another.

• ILP aims to identify independent instructions within a program that can be executed simultaneously, improving speed.

• This is achieved through a combination of compiler techniques and processor design.

Benefits :

• Increased instruction throughput : By executing multiple instructions concurrently, ILP can significantly improve the number of instructions completed per unit time.

• Reduced execution time : Programs can potentially run faster as instructions don't have to wait in line for execution.

Challenges :

• Dependencies between instructions : Not all instructions are independent. Some instructions might rely on the results of previous ones, preventing parallel execution.

• Resource limitations : Processors have a limited number of functional units (e.g., ALU, multiplier) that can execute instructions simultaneously.

• Control complexity : Identifying parallelism and managing instruction execution order requires sophisticated compiler and processor design.

Techniques for ILP :

• Pipelining : Divides the instruction cycle into stages, allowing for multiple instructions to be processed concurrently in different stages.

• Out–of–order execution : Analyzes instructions for dependencies and reorders them for parallel execution when possible, potentially exceeding the program's original order.

• Data forwarding : Bypasses the write–back stage of an instruction if its result is needed by a subsequent instruction, allowing for faster execution.

• Superscalar execution : Processors with multiple execution units can handle multiple instructions per cycle, as long as they don't rely on the same resources.

Review Questions

1. What is ILP ?

2. Write a note on instruction–level parallelism.

3. State benefits and challenges of instruction–level parallelism.

4. State the techniques for ILP.