Micro-architecture

The micro-architecture (Britain: It is called uarch and µarch by Microarchitecture and the shortened form) is chiefly a computer architecture in the microprocessor at first in the microprogram method now.

The design and mounting the electric circuit that can completely describe the movement of the hardware such as CPU and DSP are concretely indicated. The term of micro-architecture is often used in the processor industry though there is “Computer configuration” (computer organization) in an academic term, too.

Both designs of the micro-architecture and the instruction code (instruction set architecture and ISA) are parts of the computer-aided design. The instruction set is chiefly called for Intel and the design of “Architecture” and each microprocessor (mount) is called “Micro-architecture” (The example: IA-32 Intel architecture and Core micro-architecture).

History

A lot of computers made from the 1950′s to the 1970′s achieved the instruction row of the computer, and it did deciphering and the control logic to execute it had been achieved by the microprogram method. The bit in the word of the microprogram was an electric signal that controlled the unit that actually calculated. The term of micro-architecture was used to describe the unit controlled in the word of the microprogram in this age.

Relation to instruction set architecture

The instruction set architecture of the micro-architecture and the computer is clearly distinguished. The instruction set architecture is an abstract image of the computer system seen from machine language (assembly language) programmer, and the one including the format of the mode of the instruction set and the memory address specification, registers, the addresses, and data etc.Computer configuration (computer organization) is the one that the system was concretely described from ISA at a lower level.

The composition of the computer is very connected them to achieve the specification of the element and architecture that forms the system, and shows to be interactive.

It might have the instruction set architecture as which a different machine is the same and it has the same program, and it might have a different micro-architecture though it is executable. Such a different micro-architecture (With the advancement of the semiconductor processing technology) enables the processor at the new generation to demonstrate a high performance level compared with the generation before. Theoretically, one micro-architecture (Especially, if it contains the microcode) can be used to achieve two different instruction sets by programming two different storage controls (control store).

The micro-architecture of the machine is usually expressed in the block chart where the register of the machine, the bus, and the connection of the functional block are shown. For instance, the support of the design of the structure of the two or more data operation (SIMD etc.) by a floating point, an integer, a branch prediction, and a single instruction and pipelines (For instance, stages of the instruction fetch, the decipherment, the allocation, and execution, etc. might be included in the easy one) and cache memories (level 1 and level 2 interface) and peripherals etc. ..kind (.. are included in figure about many execution unit execution units.

It is structural of the layout of a physical actual circuit and hardware, packages, and, additionally, physical details are called mount of the micro-architecture. There is a possibility that is a different hardware implementation though two machines are the possibility with the same micro-architecture and are expressed in the same block chart.

Modern micro-architecture

The pipelining is a method of data processing used most in modern micro-architectures. This technique is used for most modern microprocessors, the micro controllers, and DSP. The pipeline design can do two or more overlapping execution processing, and behaves as if the assembly line of the factory. Four that there are some basic stages in pipelining in micro-architecture design. It fetches in the stage, there are a decipherment, execution, and writing, etc. , and, for instance, there is architecture including the stage such as the memory accesses about the rest, too. It is one of the center work. doing by the [maikuroa-kutekucha] design the design of the pipeline

The execution unit is indispensable for [maikuroa-kutekucha]. It consists of SIMD that processes store unit, divergence expectation unit, and two or more arithmetic and logic unit (ALU), floating point unit (FPU), and loading/instructions for the execution processing unit. These units control of the processor and operate it. The selection of the number of execution units, those processing time (latency), and the processing performance (throughput) are the one the center among the work of the micro-architecture design. The connection of the size, latency, throughput, and the system to the memory is the decisions in the micro-architecture design.

It is thought to be a part of micro-architecture design process whether to contain the peripheral circuitry of the memory controller etc. for instance among the design matters of the system level. The selection of the performance and the connection method is included in this.

When differing from the architecture design, and aiming at a specific performance level, attention is elaborately paid by the limiting condition in the micro-architecture design. Because the decisions in the [maikuroa-kutekucha] design influence the system connecting directly, it is necessary to take account to the following problems.