ECE 412 Computer Architecture Lectures 22 and 23: Multiprocessor Architecture
Lecture 22: Multiprocessor Architecture Basics
Parallel Programming Models
Multiprocessor Architecture Models
Shared Memory Architectural Model
Message Passing Architectural Model a.k.a. Distributed Memory
Dataflow Architectural Model
Systolic Array Architectural Model
Shared Memory Multiprocessor Software: Loop-level parallelism
FORTRAN DOALL Parallel Loop Construct
FORTRAN DOACROSS Parallel Loop Construct
DOACROSS Example
Parallel Loop Execution
Parallel Loop Execution (continued)
Parallel Loop Execution (continued): Iteration Scheduling
Parallel Loop Execution (continued): Barrier Synchronization
Limits of Parallel Processing
Mutual Exclusion
Hardware Synchronization
Implementing lock() and unlock()
Implementing lock() and unlock() (cont.)
lock() and unlock() Example
Need for atomic fetch_and_add()
Lecture 23: System Synchronization Implementation Issues
I. Where are synchronization variables?
I. Where are synchronization variables? (continued)
I. Where are synchronization variables? Relation to other issues
II. How do caches react to synchronization?
II. How do caches react to synchronization? (continued)
III. How do buffers react to synchronization?
III. How do buffers react to synchronization? (continued)
IV. How does the bus support synchronization?
IV. How does the bus support synchronization? (continued)
V. What is the scope of synchronization?
V. What is the scope of synchronization? (continued)
Email: sias@crhc.uiuc.edu