Skip to content
CampusPrep
AL-802 (B) · High Performance computing/Important Questions

High Performance computing (AL-802 (B)) - Important Questions

  1. Unit 47 Marks High Priority

    Explain false sharing in shared-memory parallel programs. Describe how it impacts performance and list at least four techniques to mitigate false sharing.

    High-frequency core topic on cache contention in shared-memory parallelism; practical mitigation techniques are routinely examined.

  2. Unit 410 Marks Low Priority

    Apply Amdahl's law to compute theoretical speedup for a parallel program and discuss limits to scaling. Given the parallel fraction $p$ and $N$ processors compute the speedup $$S = \frac{1}{\left(1-p\right) + \frac{p}{N}}$$ and explain the implications as $N$ grows.

    Core performance-modeling derivation frequently used to evaluate scalability limits of parallel programs.

  3. Unit 47 Marks Low Priority

    Explain the memory hierarchy and compute the Average Memory Access Time (AMAT) given hit rates and latencies for L1, L2 and main memory. Include a numerical example and show the AMAT calculation.

    Standard question on memory hierarchy performance; tests AMAT computation and understanding of cache hit/miss effects.

  4. Unit 410 Marks Low Priority

    Describe OpenMP parallel for, data scoping clauses (shared, private, reduction) and common pitfalls (race conditions, false sharing). Provide a short code example and explain how to fix a race condition.

    Practical programming question assessing knowledge of OpenMP directives, data scoping and common concurrency pitfalls.

  5. Unit 47 Marks Low Priority

    Compare MPI point-to-point and collective communications. When should one use non-blocking (e.g., MPI_Isend/MPI_Irecv) versus blocking communication? Discuss examples and benefits of overlapping computation with communication.

    Typical distributed-memory communication question testing understanding of MPI communication patterns and overlap strategies.

  6. Unit 410 Marks Low Priority

    Perform a roofline / balance model calculation: given FLOP count and memory traffic, compute the arithmetic intensity $AI = \frac{\text{FLOPs}}{\text{Bytes}}$, and estimate whether the kernel is compute-bound or memory-bound. Show calculations and conclude attainable performance.

    Performance-analysis problem requiring application of the roofline / balance model to determine bottleneck (compute vs memory).

  7. Unit 47 Marks Medium Priority Asked: 2024

    What are the design considerations for implementing a permissioned blockchain?

    Exact transcription of a past-paper question from May 2024.

Go to where you left off?

Quick Add to Notes

Save questions, your own notes and screenshots into notes filed by unit. It takes a free account.

Create free account

Have an account? Log in