SIMD

Types of Computers. Computers can be classified by Flynn s taxonomy (19). The three important classes are SISD single instmction, single data SIMD single instmction, multiple data and MIMD multiple instmctions, multiple data. 

Transputers. At higher levels of coimectedness there is a wide variety of parallel computers. A great many parallel computers have been built using INMOS Transputer chips. Individual Transputer chips mn at 2 MELOPS or greater. Transputer chips have four communication channels, so the chips can readily be intercoimected into a two-dimensional mesh network or into any other interconnection scheme where individual nodes are four-coimected. Most Transputer systems have been built as additions to existing host computers and are SIMD type. Each Transputer has a relatively small local memory as well as access to the host s memory through the interconnection network. Not surprisingly, problems that best utilize local memory tend to achieve better performance than those that make more frequent accesses to host memory. Systems that access fast local memory and slower shared memory are often referred to as NUMA, nonuniform memory access, architecture. 

Jose A. Martinho Simdes Manuel E. Minas da Piedade... 

Dudek P and Hicks PJ 2001 A general-purpose cmos vision chip with a processor-per-pixel simd array Proceedings of the 27th European Solid-State Circuits Conference. IEEE, pp. 213-216. 

For data-parallel environments (SIMD programming model), performance can be represented using standard profiling techniques that associate fraction of overall execution time with particular pieces of code. As noted earlier, this programming model has limited utility for the applications we address, and we do not dwell on performance evaluation. 

Whereas much of the algorithmic work to be described has been implemented on a range of parallel hardware, including both SIMD machines such as... 

Boltjes and de Leeuw presented a new MD algorithm for the SIMD CM-2 and compared it to an algorithm known from the literature. 2 Both algorithms were implemented in CM FORTRAN, and their performance was compared... 

SIMD Single instruction, multiple data a model for concurrent computing in which the same instruction is applied simultaneously to multiple data items. 

Hardware can be classified various ways. We choose two classifications single instruction, multiple data (SIMD) and multiple instruction multiple data (MIMD). SIMD computers typically have many, usually low power, computing... 

The Thinking Machines Corporation (TMC) CM5 MPP system was announced in November 1991. This processor is a distributed-memory MIMD machine that supports both the SIMD and MIMD programming models. The hardware is composed of a fat tree of compute processors that are used and administered from control processors. The compute processing nodes consist... 

During the early part of this decade, there was a vivid debate between proponents of MIMD (multiple instruction stream, multiple data stream) and SIMD (single instruction stream, multiple data stream) type parallel computers. This taxonomy, which stem from Flynn s classification [25,26] of possible computer architectures, can now be said to be mostly of historical and academic interest. 

While SIMD computers have been shown to be applicable to a much wider range of applications than was first thought possible they have not been commercially successful. Companies which sell these types of machines have had to leave the arena of general purpose computing and can now only be found in a few special areas. In short, SIMD type machines have not been able to cope with the tougher, post cold-war, commercial market and variations of MIMD parallel computers are the only types left on this market. 

Another common programming model is the data parallel model. In this model the parallelism is extracted from the parallel operations that can be performed on arrays of data. Even though this programming model was first introduced on SIMD machines it is a misconception to believe it to be tied to these types of machines. The data parallel model is directly supported by HPF [29]. The model is clearly of great generality and very elegant, but not all parallel algorithms can easily be expressed in HPF and also compiled into efficient code. However, as compilers continue to improve this situation is sure to improve. 

Joel F. Liebman, Jose Artur Martinho Simdes and Suzanne W. Slayden... 

Simde, R. A., and Evenson, J. K. (1987). Serine incorporation into the selenocysteine moiety of glutathione peroxidase. J. Biol. Chem. 262,933-937. 

---