Evaluation of Systems Architecture - Essay Example

School Evaluate Systems Architecture CPU A CPU can be described using various characteristics. These include the memory addressing techniques, type of registers and their number, the design of its instruction set and others. Over the evolution of computers, many architectures have come up. The most important and the basis of modern computers is the Von Neumann architecture. Some of the household names in CPU architecture include Intel, IBM and Sun Microsystems. The Von Neumann architecture identified the CPU to be made of three main components. These are the control unit, arithmetic & logic unit and the memory. The kind of memory used on CPUs has to be fast speed and is limited in size. These are the registers. Below is the Von Neumann architecture on which all other architectures are based; Traditionally, CPUs did not have many registers and comprised numerous memory addressing techniques. However, after intensive research, it was found that these features made the computers less efficient. Major issues included rare use of specialized instructions, increased the frequency of memory access due to few general-purpose registers. This feature also reduced the amount of addressable memory space for computers. The time required for instruction execution was also unnecessarily lengthened by lack of registers. An example of such processors is the intel x86 processor. VLIW (very long instruction word) and EPIC (Explicitly parallel instruction computer) These came up as a solution to increase the efficiency of computers. They use a technique called parallelism, which has the effect of increasing the overall speed of execution of computers. Through data dependency where the output of one operation is the input of the next, concurrency can be achieved. CPUs using VLIW architecture have 64 general purpose registers at their disposal. Their instruction words are 128 bits long which is then divided into 4 atoms each 32 bits long. Below is a graphical depiction of VLIW architecture. Intel was the originator of the EPIC architecture and used it on their IA-64 processor. It clusters instructions before execution. Its instruction word size is 41 bits with 256 registers. The clusters are 128 bits wide. Execution of instructions occurs in cycles. Instructions and data are stored in the computer’s hard disk. For execution, they are loaded in main memory from where the CPU can reference them. The fetch execute cycle then begins. The CPU speed is much faster than memory. The time it takes to fetch an instruction from main memory into the CPU is enough for execution of an instruction. Therefore, CPUs are equipped with an onboard fast memory chip called the cache. This is where most frequently executed instructions are kept. The CPU communicates with main memory over a dedicated memory bus. Through some designed algorithms, it is able to predict the next instruction to be referenced, fetches it from main memory, and stores it in the cache. The program counter assists the ALU to determine the next instruction that will be referenced. Main memory The main memory also called RAM is where programs are loaded as they are being executed. Different memory management techniques are used to divide the memory so that multiple programs can be executed on the processor. These include paging and segmentation. The CPU fetches instructions % data and stores output values via the memory bus. It is dedicated and used specifically for memory access. This makes it fast and matches it up with the speed of the processor. The speed of processors has been on a steady increase for the past few years. New specifications come up every few months. Technology in memory speeds has not matched up with these innovations in processor speed. The fetch cycle would waste CPU capability if measures are not taken to compensate for this. Therefore, faster but less capacity and more expensive memory techniques have come up. These are the registers and cache memories. They are on the CPU chip therefore fetching instructions, data and storing results is several magnitudes faster (Englander, 1996). Registers A CPU is equipped with several registers. Their storage capacity is limited but they are very fast. The main ones include: Program counter It holds the executing instruction’s address. With several algorithms and the ALU, the program counter makes predictions of the next instruction to be executed and its reference. Instruction register The PC above holds the address of the instruction being executed. However, the real instruction is held in the instruction register. It serves the same function as the human brain in executing instructions (Carpinelli, 2001). Memory address register and memory data register The memory address register stores the location of a memory address. Memory management algorithms mainly use it. The memory data register is a buffer that is used to store data being read from any memory location. During execution of instructions, data moves between the control unit, the arithmetic and logic unit, the registers and the main memory. Possible actions include loading values from one register into another. The new value overwrites the existing one in that register. This is called the overwrite operation. It is also possible to combine two values stored in different registers in one register. This is called the append operation. It is common in arithmetic operations and memory accesses. One could also issue several commands to test values in a register. However, these instructions are low level and not visible in modern computers. The abstraction of operating systems conducts them automatically when needed (Stalings, 2000). Multiprocessing To make efficient use of the expensive computer resources, several processes can be executed simultaneously. Resources such as the memory are subdivided among the processes being executed. With the current technology in processors, there is an unlimited degree of multiprocessing. The CPU time is subdivided into several time slots and the CPU switched back and forth among the processes being executed. Manufacturers of operating systems develop various algorithms for concurrent execution of processes. There has to be communication between the processor and a scheduler with a timing mechanism to manage execution of the multiple processes. A dedicated bus is usually present on multiprocessing computers for synchronization of executing processes. References Carpinelli, J. D. (2001). Computer systems organization & architecture. Boston, MA: Addison- Wesley. Englander, I. (1996). The architecture of computer hardware and systems software: An information technology approach. New York: Wiley. Stallings, W. (2000). Computer organization and architecture: Designing for performance. Upper Saddle River, N.J: Prentice Hall. Read More