Discuss parallel algorithm design and complexity measures.

Parallel algorithms divide work across multiple processors. Key measures: work (total operations), span/depth (longest dependency chain), parallelism (work/span). Amdahl's Law limits speedup based on sequential fraction. PRAM models: EREW, CREW, CRCW for memory access. Examples: parallel prefix sum O(n) work O(log n) span, parallel merge sort, matrix multiplication. Practical considerations: communication cost, load balancing, memory hierarchy. GPUs enable massive parallelism for suitable problems.