Logo
Pattern

Discover published sets by community

Explore tens of thousands of sets crafted by our community.

Message Passing vs Shared Memory

10

Flashcards

0/10

Still learning
StarStarStarStar

Data Locality

StarStarStarStar

Advantages: Reduces memory access times, Improves cache performance, Minimizes communication overhead; Disadvantages: Programming complexity, May require explicit data distribution control, Affected by system's memory hierarchy

StarStarStarStar

Distributed Memory

StarStarStarStar

Advantages: Scales well to large systems, Natural for geographically distributed computation, Facilitates modular programming; Disadvantages: High communication cost for non-local data, Requires explicit data decomposition, Complex error handling due to system size

StarStarStarStar

Shared Memory

StarStarStarStar

Advantages: Faster data access, Simpler communication model, Efficient for fine-grained communication; Disadvantages: Limited scalability, Potential for data races, Complex data synchronization

StarStarStarStar

NUMA (Non-Uniform Memory Access)

StarStarStarStar

Advantages: Can scale to a large number of processors, Locality of reference can lead to performance gains, Flexible; Disadvantages: Complex memory hierarchy, Performance can vary significantly, Non-trivial programming model

StarStarStarStar

Message Passing

StarStarStarStar

Advantages: Scalability, No data races, Explicit communication; Disadvantages: Communication overhead, Complex programming model, Less efficient for fine-grained communication

StarStarStarStar

Cache Coherence

StarStarStarStar

Advantages: Simplifies programming by providing a consistent view of memory; Disadvantages: Protocol overhead, Can limit scalability, Complexity in multiprocessor systems

StarStarStarStar

False Sharing

StarStarStarStar

Advantages: --; Disadvantages: Reduces performance due to unnecessary invalidation of cache lines, Hard to detect and debug, Can deteriorate scalability

StarStarStarStar

Synchronization Overhead

StarStarStarStar

Advantages: Ensures data consistency and order; Disadvantages: Can lead to performance bottlenecks, Deadlocks, Complicates programming

StarStarStarStar

Load Balancing

StarStarStarStar

Advantages: Improves resource utilization, Maximizes throughput, Minimizes response time; Disadvantages: Overhead of balancing decisions, May not be optimal due to dynamic workloads, Complexity in heterogeneous environments

StarStarStarStar

Amdahl's Law

StarStarStarStar

Advantages: Provides a theoretical speedup limit, Easy to apply and understand, Helps in analyzing the benefits of parallelization; Disadvantages: Assumes constant workload, Doesn't account for overhead, Can be overly pessimistic

Know
0
Still learning
Click to flip
Know
0
Logo

© Hypatia.Tech. 2024 All rights reserved.