Page 118 - Computer_Science_F5

Exercise 2.8 2. How could register renaming be
used to eliminate the dependency
1. Imagine you are a computer you identified? Show how the
detective investigating a case of instructions might be rewritten with
slow program execution. The culprit temporary registers assigned by the
FOR ONLINE READING ONLY
is inefficient use of registers in the renaming process. Chapter Two: Performance and optimisation of computer processor
processor! You have identified a 3. Explain how register renaming, by
segment of instructions that seem eliminating dependencies, helps
to be creating a problem: improve overall program execution
mul r1, r2, r3 speed. Hint: Think about how it
sub r1, r1, 4 allows for more efficient instruction
add r4, r1, r5 scheduling.
(a) Identify the data dependency 4. Modern processors employ
between these instructions. various techniques to optimise
What type of dependency is performance. Research another
it (RAW, WAR, or WAW)? technique (e.g., pipelining or out-
(b) Explain why this dependency of-order execution) and explain
might create a bottleneck in how it works alongside register
program execution. renaming for a performance boost.

Thread and data Level parallelism

Scenario 2.3: Kitchen as a simple processor

Imagine planning a pizza party with friends involves ordering pizzas, preparing
drinks, decorating, and setting up games. It can be overwhelming for one
person to manage everything simultaneously. A program acts like a recipe
for this party—an instruction set understood by the computer, detailing each
step of the planning process. When executed, this becomes a process—a live
instance of the program, similar to following a recipe while cooking. Multiple
processes can run concurrently, like baking cookies while making pizza dough.
Threads within a process represent individual tasks, such as preparing dough
and toppings concurrently, as they are lightweight and efficient. Multithreading
enables simultaneous handling of multiple tasks in party planning, like
decorating while a friend prepares drinks, significantly enhancing efficiency. In
contrast, parallelism involves multiple chefs collaborating on different dishes
simultaneously, reflecting concurrent execution of various party planning tasks
using multithreading.

109
Student’s Book Form Five

Computer Science Form 5.indd 109 23/07/2024 12:33