Page 131 - Computer_Science_F5
P. 131
Computer Science 1. For the following code snippet, a code segment for a modern
Revision Exercise 2
processor. The code involves
multiple loops and calculations.
How would you approach this task,
trace the execution steps (such as
FOR ONLINE READING ONLY
fetch, decode, execute, memory
unrolling, instruction scheduling,
access, write back) in both a non- considering the concepts of loop
and data dependence analysis?
pipelined and a pipelined processor Briefly outline your strategy.
(assuming no hazards). Compare
the number of clock cycles it takes 6. Investigate the impact of cache size
to complete the instructions in both and memory access patterns on the
scenarios. Explain the speedup effectiveness of loop unrolling.
achieved through pipelining. 7. Research real-world compiler
Code snippet optimisation techniques related to
data dependence analysis (beyond
lw $t0, 4($s3) # Load value from the basic example provided).
memory to $t0 Explain, in your own words, how
sub $t1, $t0, $s4 # Subtract $s4 data dependence analysis helps
from loaded value to $t1 compilers exploit ILP effectively.
2. Research the concept of forwarding 8. Discuss two potential benefits and
in pipelined processors. Briefly two trade-offs associated with using
explain how forwarding can be superscalar processors compared to
used to address data hazards and scalar processors.
potentially improve performance 9. Imagine you are a computer
compared to stalling. architect tasked with designing the
3. Imagine you are a computer next generation of processors. How
architect. How might future could you leverage the concepts
processor designs enhance ILP of superscalar execution and ILP
techniques? beyond simply increasing the
4. Research and compare two different number of execution units? Briefly
architectures (Such as, VLIW and describe your innovative approach.
Superscalar) that exploit ILP. 10. Assume the initial BHT prediction is
Summarise their key differences and “Strongly taken” (11). Analyse the
advantages in a short presentation. differences in prediction behavior
5. Imagine you are a compiler compared to an initial prediction
developer working on optimising of “Strongly not taken” scenario.
122
for Advanced Secondary Schools
Computer Science Form 5.indd 122 23/07/2024 12:33
