COMP5426 : Parallel and Distributed Computing

Assignment 1-Parallel Matrix Multiplication Optimisation with Pthreads

Designed and implemented a high-performance parallel matrix multiplication system in C using POSIX threads (pthreads) as part of COMP4426/5426 Parallel and Distributed Computing at the University of Sydney. The project focused on analysing execution time, scalability, RAM usage, and low-level optimisation techniques for large-scale matrix operations. :contentReference[oaicite:0]{index=0}

Key Features

• Parallel matrix multiplication using pthreads
• Dynamic multiplication-order optimisation:
  • (A × B) × C
  • A × (B × C)
• 1D row-wise workload partitioning
• Sequential vs parallel performance benchmarking
• Loop unrolling optimisation (factor 4)
• Float vs double precision analysis
• Execution time, speedup, and RAM profiling

Optimisation Techniques

• Multi-threaded computation
• Cache-friendly memory access patterns
• Instruction-level parallelism (ILP)
• Loop unrolling on nested loops
• Pointer arithmetic optimisation
• Cost-based computation ordering

Tools & Technologies

• C
• POSIX Threads (pthreads)
• GCC / Clang
• macOS ARM64 (Apple M1)
• Shell scripting
• Performance benchmarking & profiling

Performance Results

• Achieved over 14× speedup using loop unrolling and multithreading
• Significant reduction in execution time compared to sequential implementation
• Analysed scalability limits caused by CPU core count, cache efficiency, and memory bandwidth

Skills Demonstrated

Parallel programming, performance optimisation, multithreading, low-level systems programming, cache optimisation, benchmarking, algorithm analysis, and concurrent computing.