Раздел 5. Первое занятие

195

advantages over the other types if it was intended for use in exacting or inaccessible environments.

Further developments improved the logic density of CMOS. The only cloud on the CMOS horizon comes from a new development of the normal bipolar circuit. A new semiconductor configuration, called integrated injection logic (IIL), has been devised²¹ which eliminates²² the need for any resistors²³, capacitors²⁴ or transistor isolation. This enables an extremely compact logic circuit to be formed which has a low power consumption while maintaining the normal speed of tran- sistor-transistor logic (TTL).

Nowadays, the CPU must perform a number of functions simul­taneously, for example, fuel-mixture calculations and ignition advance control in a car. Also, it must be simple enough to be built economi­cally. Software development cycles must be short, all the computation proceeds concurrent.

2. The versatility of the microprocessor has altered the entire architecture of modern computer systems. No longer²⁵ is the process­ing of information carried out only in the computer’s central process­ing unit. Today, there is a trend towards distributing more processing capability throughout a computer system, with various areas. For ex­ample, an input-output port may have a controller to regulate the flow of information through it. At times, the controller may accept com­mands from the CPU and send signals back in order to coordinate its operations with those of the rest²⁶ of the system; at other times, the controller may operate independently of the CPU.

Provisions can and are being made in the architecture. Thus, the architectural improvement is parallel and distributing computation.

At present, scientists announced dramatic new breakthroughs²⁷ in molecular electronics: they fabricated the circuits using an advanced system of manufacturing called nano-imprint²⁸ lithography. Capacity and performance could be extended enormously²⁹ by layering molecu- lar-switch devices on conventional silicon. This will push advances in future computer technology far beyond³⁰ the limits of silicon.

An ordinary computer simulates³¹ a neural network. Actually, it is an artificial neural networks. Neural networks are known to be mas­sively parallel distributed processing systems.

13*

Микроэлектроника настоящее и будущее

Nowadays, many large mainframes may have more than one CPU (they may include even thousands of processors). Machines with new architecture possess highly parallel structures but each of these executes instruction streams³² that are unrelated. They act on different data spac­es ³³. SIMD machines often have a large number of processing units that all may execute the same instruction on different data in lock-step³⁴. So a single instruction manipulates many data items³⁵ in parallel.

Another subclass of the SIMD systems are the vector processors. Vector processors act on arrays³⁶ of similar data rather than on single data items using specially structured CPUs. When data can be manipu­lated by these vector units, results can be delivered³⁷ with a rate of one, two and — in special cases — of three per clock cycle (a clock cycle being defined as the basic internal unit of time for the system). So vector pro­cessors execute on their data in an almost parallel way but only when executing in vector mode. In this case they are several times faster than executing in conventional scalar mode³⁸. For practical purposes vector processors are therefore mostly regarded as SIMD machines.

Distributing microprocessing is a technique in which the main microprocessor of the PC directs other microprocessors throughout the PC system to perform specific functions for it and report their sta­tus (состояние).

New forms of I/O are also acquiring³⁹ sophisticated capabilities with distributed microprocessing. These “intelligent” I/O modules perform some of the calculations formerly done by the main micro­processor, store information temporarily⁴⁰, and do other functions under the direction of the main microprocessor.

Some remote⁴¹ I/O modules have microprocessors resident (зд. находящийся) in the modules. Remote I/O modules use the resident microprocessors to shorten the effective scan⁴² time. However, with independent intelligence⁴³ in the I/O, if some thing happens to the PC, the I/O module might already have acted on misinformation. Hence, I/O modules with a resident microprocessor should include appropri­ate⁴⁴ instructions for fail-safe shutdown⁴⁵ should the PC develop a fault⁴⁶.

A trend that is beginning to emerge in microprocessor design is the incorporation of trouble-shooting aids⁴⁷, heretofore (до сих пор) available only on larger computers.