Core memories

Hash codes of molecules which are already pre-computed are suitable for use in fiill structure searches in database applications. The compression of the code of a chemical structure into only one number also makes it possible to compute in advance the transformation results for a whole catalog. The files can be stored and kept complete in the core memory during execution of the program, so that a search can be accomplished within seconds. 

Figure 7.9. A small portion of a magnetic core memory (after IBM News, 1967).

Today, dynamic random-access memories (DRAMs) are transistor/capacitor-based semiconductor devices, with access times measured in nanoseconds and very low costs. Core memories were made of magnetic rings not less than a millimetre in diameter, so that a megabyte of memory would have occupied square metres, while a corresponding DRAM would occupy a few square millimetres. Another version of a DRAM is the read-only memory (ROM), essential for the operation of any computer, and unalterable from the day it is manufactured. We see that developments in magnetic memories involved dramatic reductions in cost and... 

In this chapter, we also discussed several schemes that allow for the computation of scalar observables without explicit construction and storage of the eigenvectors. This is important not only numerically for minimizing the core memory requirement but also conceptually because such a strategy is reminiscent of the experimental measurement, which almost never measures the wave function explicitly. Both the Lanczos and the Chebyshev recursion-based methods for this purpose have been developed and applied to both bound-state and scattering problems by various groups. 

Figure 14. System for collecting and processing data from the Ge(Li) spec-trometers. The data may he either stored in core memory or collected at random...

Data Reduction. The spectral data can be collected on-line and stored directly on magnetic tape under control of a PDP-8 digital computer 11). Alternatively, they can be collected on a magnetic core memory, then transferred in bulk fashion (Figure 14). The choice of method is dictated by the length of counting required and by the method of treatment of the data (i.e., simple X-Y plot or complete channel information). 

The automation concept implemented at Conroe is called computer production control (CPC).2 CPC systems are driven by a general purpose process-control computer (48-K core memory) centrally located and programmed to monitor production and equipment operations and perform oil and gas accounting. As shown in Fig. 11, the computer is connected to a supervisory control system with a computer interface unit (CIU) to provide remote data acquisition and control function capability. 

The original WOA supervisory control and data acquisition system was configured around the Honeywell HS4400 process computer. Two 4400s in dual redundant processor configuration were installed with automatic failover to the standby should the online unit fail. Each 4400 has 96K of 24-bit word magnetic core memory and a 1.3 million word drum. 

An estimate of the number of tests also is necessary for estimating computer core memory requirements if the data reduction workload is so heavy that computer data storage seems attractive. 

An NMR spectrometer capable of PFT measurements is a combination of a CW circuit such as found in conventional NMR spectrometers [1-5], a computer controllable pulse generator, and a small digital computer of core memory size 8-256 K. Fig. 2.34 outlines this experimental arrangement. 

The PFT NMR spectrum can then be recorded in analog or digital form by a plotter controlled by the computer. If the computer core memory allows, additional subroutines may be stored which automatically compute signal intensities and positions (in ppm values relative to some standard or in Hz) and outputs them to the printer. 

Scattered light collected by the FI.5 lens (f = 30 cm) is relayed to the photomultiplier tube via 1 mm slits, a 1 nm bandwidth interference filter and a polarization filter, to reduce background from flame luminescence. The PDP-11/34 computer instructs the A/D convertor to make a conversion every 100 vsec. The resulting digital data are stored sequentially in core memory. The memory is saturated at 16,000 temperature measurements, at which time the data are transferred to a hard disk memory. The data in this transfer constitute one time... 

Ferrites having a square hysteresis loop (e.g. (Mg, Mn, Zn) Fe204), developed for the now outdated computer core memories, have found applications in the form of a toroidal saturable inductor for regulating the output currents in switched-mode power supplies. 

The spectral region of interest is then detected and processed by means of an optical multichannel analyzer (EG G-PARC model 1215, OMA). The OMA detector SIT (silicon intensified target), EG G-PARC model 1254 has been used. It is operated by the detector controller, EG G-PARC, model 1216, which performs the signal digitization as well. The acquired spectra are displayed in real time on a TV display and on the OMA console. The data storage and processing are also performed by that console which has a 28K of 16 bit core memory and a floppy disk for permanent storage. 

Core memory (the data here are usually "volatile") British books call this the "store" the Germans call it "Speicher." In the old days, ferrite cores were used this memory was not volatile. Nowadays it is usually MOS memory (volatile). 

In this scctioin we describe typical computer requirements for the 0( D) + Ho reaction. The TB, TJ and TK codes use about 2 GBytes of core memory. They have been implemented on a NEC-SXo vector computer (8 Gflops peak performance per processor) at IDRJS/CNRS (Orsay, France). All the codes are efficiently vectorized and use the optimized BLAS and LAPACK linear algebra libraries. 

Rectangular ferrites are utilized in core memory systems in computers... 

At Merck, we are using DEC GT-42 and GT-43 graphics display terminals with 16 K words of core memory. Communication with our host computerll is at 1200 baud using VADIC 3400 series modems. 

A potential difficulty of this approach is that the a vectors (which must also be stored) are not necessarily sparse. Knowles notes81 that even when c is only 1% populated, typically 50% of a will be nonzero. Nevertheless, in order to obtain variationally correct energies, the full a vector must be formed in core memory and its dot product taken with all expansion vectors c. However, once this is done, the only further use of a is in the construction of new subspace vectors hence, Knowles only writes to disk those elements of a greater than some threshold. According to (134)-(135), these neglected elements of a would only contribute to elements of Ac which make very small energy contributions. 

There are two basic families of solution techniques for linear algebraic equations Direct- and iterative methods. A well known example of direct methods is Gaussian elimination. The simultaneous storage of all coefficients of the set of equations in core memory is required. Iterative methods are based on the repeated application of a relatively simple algorithm leading to eventual convergence after a number of repetitions (iterations). Well known examples are the Jacobi and Gauss-Seidel point-by-point iteration methods. 

---