Supercomputers availability

The CRAY-1 vector processing computer at the Science Research Council s (S.E.R.C) Daresbury Laboratory, is at the centre of a network providing large scale computational facilities for Universities in the United Kingdom. This is the only supercomputer available at present to Quantum Chemists in the U.K., and this article will therefore be restricted to experience gained on the CRAY-1, although this experience will undoubtedly be relevant to future applications on machines such as the ICL Distributed Array Processor (DAP) (see reference (2) for a detailed description) and the CDC Cyber 203/205. 

Most interesting polymer properties are observed at frequency scales of 10 Hz and lower This means that, in order to compare atomistic simulation results to experiment, one should be able to reproduce the model behavior at time scales spanning more than 10 orders of magnitude This task would be daunting even for a computer 10,000 times faster than the most powerful supercomputers available today. 

Even the most powerful supercomputers available today can only handle up to about a million atoms. If we are interested in the properties of single molecules or small drops or clusters of diameter 100 A or less, this is not a problem. Most of the time, however, one is interested in the properties of bulk materials. 

It is a truism (cliche ) that microcomputers have beeome more powerful on an almost exponential eurve sinee their advent more than 30 years ago. Molecular orbital calculations that I ran on a supercomputer a decade ago now run on a fast desktop mierocomputer available at a modest priee in any popular eleetionies store or by mail order eatalog. With this has eome a eomparable inerease in software sophistication. 

Another important consideration is the amount of labor necessary on the part of the user. One major difference between different software packages is the developer s choices between ease of use and efficiency of operation. For example, the Spartan program is extremely easy to use, but the price for this is that the algorithms are not always the most efficient available. Many chemistry users begin with software that is very simple, but when more sophisticated problems need to be solved, it is often easier to learn to use more complicated software than to purchase a supercomputer to solve a problem that could be done by a workstation with different software. 

Supercomputers, such as the CRAY X-MP, CRAY Y-MP, and CRAY-2, are partially available and used for flow-sheet and optimization studies (7—10). Optimization modules usiag linear and nonlinear programming (LINPRO and UNLPl, based on a revised simplex, and Davidson-Eletcher-PoweU and Broyden methods, respectively) are available ia MicroMENTOR (11). 

Supercomputers from vendors such as Cray, NEC, and Eujitsu typically consist of between one and eight processors in a shared memory architecture. Peak vector speeds of over 1 GELOP (1000 MELOPS) per processor are now available. Main memories of 1 gigabyte (1000 megabytes) and more are also available. If multiple processors can be tied together to simultaneously work on one problem, substantially greater peak speeds are available. This situation will be further examined in the section on parallel computers. 

Many institutions have hundreds, or even thousands, of powerful work stations that are idle for much of the day. There is often vastiy more power available in these machines than in any supercomputer center, the only problem being how to harness the power already available. There are network load-distribution tools that allocate individual jobs to unused computers on a network, but this is different from having many computers simultaneously cooperating on the solution of a single problem. 

MIMD Multicomputers. Probably the most widely available parallel computers are the shared-memory multiprocessor MIMD machines. Examples include the multiprocessor vector supercomputers, IBM mainframes, VAX minicomputers. Convex and AUiant rninisupercomputers, and SiUcon... 

TABLE 8.1 Actual and Estimated Supercomputing Resources Available to Researchers in the United States, 1980-1990 Year Number of Cray I-Class Supercomputers ... 

Before the advent of modem computer-aided mathematics, most mathematical models of real chemical processes were so idealized that they had severely limited utility— being reduced to one dimerrsion and a few variables, or Unearized, or limited to simplified variability of parameters. The increased availability of supercomputers along with progress in computational mathematics and numerical functional analysis is revolutionizing the way in which chemical engineers approach the theory and engineering of chemical processes. The means are at hand to model process physics and chenustry from the... 

The supercomputer power that is becoming available will provide the opportunity to combine... 

The work described in this paper is an illustration of the potential to be derived from the availability of supercomputers for research in chemistry. The domain of application is the area of new materials which are expected to play a critical role in the future development of molecular electronic and optical devices for information storage and communication. Theoretical simulations of the type presented here lead to detailed understanding of the electronic structure and properties of these systems, information which at times is hard to extract from experimental data or from more approximate theoretical methods. It is clear that the methods of quantum chemistry have reached a point where they constitute tools of semi-quantitative accuracy and have predictive value. Further developments for quantitative accuracy are needed. They involve the application of methods describing electron correlation effects to large molecular systems. The need for supercomputer power to achieve this goal is even more acute. 

Calculation of thermochemical data using high-quality ab initio electronic structure calculations has been a long-standing goal. However, the availability of supercomputers and new theoretical techniques now allow the calculation of thermochemical properties with chemical accuracy, i.e., AHf to within... 

How does a rigorously calculated electrostatic potential depend upon the computational level at which was obtained p(r) Most ab initio calculations of V(r) for reasonably sized molecules are based on self-consistent field (SCF) or near Hartree-Fock wavefunctions and therefore do not reflect electron correlation in the computation of p(r). It is true that the availability of supercomputers and high-powered work stations has made post-Hartree-Fock calculations of V(r) (which include electron correlation) a realistic possibility even for molecules with 5 to 10 first-row atoms however, there is reason to believe that such computational levels are usually not necessary and not warranted. The Mpller-Plesset theorem states that properties computed from Hartree-Fock wave functions using one-electron operators, as is T(r), are correct through first order (Mpller and Plesset 1934) any errors are no more than second-order effects. 

How well has Dill s prediction held up In 2000, the first ever microsecond-long molecular dynamics simulation of protein folding was reported. It required 750,000 node hours (equal to the product of the number of hours times the number of processors) of computer time on a Cray T3 supercomputer. According to Dill s prediction, this length of simulation was not to be expected until around 2010. However, as noted above, Dill s analysis does not take into account large-scale parallelization—which, unless the computation is communications-limited, will effectively increase the speed of a computation in proportion to the number of processors available. 

---