Simple Parallel Computer Architectures
 
 
 
 
 


 


 
 
 
 
 
 
 
 
 
 
 
 
 

- each instruction broadcast to all processing elements

- all processors must operate in lock-step (such as add two numbers)

- good for vector and array problems

- comes in both shared and distributed memory configurations
 
 
 


 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 


 

- Shared Memory Multiprocessor

- each processor can run at its own speed

- parallel execution of threads, one per processor

- shared memory space is an advantage for ease of programming

- contention is possible on the shared interconnect and on the shared memory

- maintaining cache coherency
 


 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 


 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 


 

Shared Bus Interconnection

 


 

 
 
 
 
 
 
 
 
 


 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Ring Interconnection Network
 


 

 
 
 
 
 
 
 
 
 
 
 


 


 
 






















































Mesh Interconnection Network

 


 

 
 
 
 
 
 
 
 
 


 
 


























































3D Mesh Interconnection

 


 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 


 


 
 














































A Torus Interconnection System

 


 

 
 
 
 
 
 
 
 
 


 
 


















































3D Cube Interconnection

 


 

 
 
 
 
 
 
 
 
 
 
 
 
 


 
 






















































4D Hypercube Interconnection