MIMD computers

The proposed standard may be obtained using anonymous file transfer from titan.cs.rice.edu (in public/HPFF/draft). An effort commenced January 1994 to extend the HPF proposal to accommodate irregular data distributions and to support parallel 1/O and more complex data structures, as well as task-driven MIMD computation. Additional information is available by sending electronic mail to hpff-info cs.rice.edu. 

A Transputer-based systolic loop implementation for the simulation of rigid polyatomic molecules described by Craven and Pawley revealed that in contrast to many problems on MIMD computers, the main loss of efficiency arose not from the time spent in communication between processors but, rather, from certain extra calculation and addressing work that was deemed inevitable in the parallel decomposition of the problem, and from a slight load imbalance. This study provided an excellent practical example of the ability of the Transputer to communicate and calculate in parallel, with minimal degradation of the computation rate, and provided very encouraging timing comparisons with the Cray X-MP. 

Each of the five steps demonstrated different parallel efficiencies. Steps two and four were the most intensive in terms of computation. The implementation by Kuppermann and co-workers obtained better than 80% efficiency for up to 64 nodes on the Caltech/JPL Mark Illfp MIMD computer for step two.275 For large global matrix dimensions, step four was 40% efficient on 64 nodes and —80% efficient on 8 nodes. 

Several types of parallel computing architectures have been used over the years, including pipelined processors, specialized SIMD machines and general MIMD computers. Most parallel computing is done on hybrid supercomputers that combine features from several basic architectures. The best way to define parallel processing in detail is to explain how a program executes on a typical hybrid parallel computer. 

Smith, B. 1978. A pipelined, shared resource MIMD computer. International Conference on Parallel Processing, (Bellaire, MI), pp. 6-8. (Aug). 

One can look at society at many different scales one finds parallel computers from neurons in the brain, to bees in a swarm, to people in society. In every case, the underlying methodology is that of asynchronous computing elements (neurons, bees, people) communicating by messages. This so-called message-passing MIMD computer... 

SIMD computers are essentially synchronous models because all PEs execute the same instruction at the same time. The only control the programmer has over an individual PE is to allow or prevent it from executing the instruction. This makes SIMD computers easier to program than MIMD computers (see below) because the control of different programs... 

MIMD computers are essentially asynchronous because each processor or transputer may proceed at its own rate, independently of the behaviour of the others. They may, of course, perform in synchronous mode if required. 

The SIMD category comprises both vector computers, such as the CRAY-1, and so-called processor arrays, such as the Thinking Machines Connection Machine and the classic ILLIAC-IV computer. - SIMD computers ate well suited for applications requiring identical operations on uniform data structures, but SIMD computers cannot necessarily be used efficiently for general-purpose calculations. Presently, MIMD is the most common parallel architecture, and the majority of parallel computers used for quantum chemistry applications are of the MIMD type. Examples of MIMD computers are the... 

---