250 likes | 270 Views
Cosc 3P92. Week 1 Lecture slides. Psychiatrist to patient "You have nothing to worry about - anyone who can pay my bill is certainly not a failure.". Class Room. Ground Rules Cell Phones My contact times Open Door Policy Email bockusd@brocku.ca. Course Overview. Basics.
E N D
Cosc 3P92 Week 1 Lecture slides Psychiatrist to patient "You have nothing to worry about - anyone who can pay my bill is certainly not a failure."
Class Room • Ground Rules • Cell Phones • My contact times • Open Door Policy • Email bockusd@brocku.ca
Course Overview Basics • introduction, overview of basic architecture • circuits, microprocessor chips • microprogramming - CPU, ALU • memory - chips, cache, virtual • input/output • instruction sets Advanced • RISC vs CISC • parallel computers • special applications
Introduction • computer: mechanism which processes information based on a predefined sequence of instructions • digital computer: based on counting (base 2) analog computer: based on measurements (voltage thresholds) • see text sections 1.3, 1.5 for history • first stored-program computer proposed by John von Neumann (1945) - instructions interpreted sequentially - binary representation of data - typical von Neumann architecture
1. CPU organization registers A ALU input registers B ALU A + B ALU output register
CPU organization • ALU - arithmetic logic unit - add, subtract, AND, ... - often supplemented by floating-pt processors • instructions: 1 - register-register 2 - register-memory 3 - memory-memory
2. Memory • organization: bits (0,1) byte (8 bits) word (1,2,4,... bytes) "boards" (1 megabyte) • addressing: normally by byte address (convention) - “address alignment” • word size is a measure of CPU power, but address and data path size between CPU and memory is a crucial factor • main memory: hardware memory, usually chips secondary memory: external cheap storage (hard discs, floppies,...)
3. Input Output device • secondary storage (hard discs, floppies, tapes,...) • interactive devices (terminals, mice, ...) • paper output (printers, plotters, FAX machines...) • modems cable modems • CD ROMS
Fetch Decode (CU) Execute (CU+ALU+Reg+Mem+IO) An instruction cycle An execution cycle • Instruction cycle steps: 1. Fetch–the control unit reads an instruction from memory; 2. Decode–control unit decodes the instruction based on its internal logic; 3. Execute–depending the operation code (opcode) of the instruction, it enables the control signals to the ALU and/or the Input/Output devices to perform the required operation. Control Unit (CU+Mem+Reg) Arithmetic Logic Unit Registers CPU (M.Cheng)
Execution cycle • • The control unit interprets instructions, and is either • hardwired or microprogrammed. • • registers include: program counter, instruction register, effective address register, accumulator and/or set of general-purpose registers. • Arithmetic Logic Unit (ALU) performs basic binary addition, shift, and comparison operations. More advanced ones do multiplication, division, and floating-point operations. • When an instruction is being decoded, operand(s) (the data) may have to be prefetched as well. • Data path cycle: running (2) operands through the ALU and storing the result • of central importance to CPU design
Basics • A CPU is an instruction sequencer and executor. - different CPU’s have different instruction sets • A set of high-speed registers is incorporated into the CPU to assist instruction execution. LOOP EAR := EffectiveAddressOf( PC ); IR := MBR; PC := PC + 1; EAR := EffectiveAddressOf( Operand( IR ) ); Execute( OpCode( IR ), MBR ); END PC (program counter) IR (instruction register) EAR (effective address register) MBR (memory buffer register) • A simple (one-address) instruction executor can be defined as follows.
Basics • • The Program Counter holds the address of the next instruction to be executed. • • The Instruction Register holds the current instruction. • • On many modern processors, there is a distinction between a logical and a physical (or effective) address. Once the physical address of a memory word is calculated and loaded into the Effective Address Register, the content of the word is retrieved from memory to the Memory Buffer Register. • Note that the execution processor can be implemented in 2 ways: • 1. Hardware: direct instn execution • 2. Interpreted: indirect instn execution • also, hybrid methods that use both h/w and interpretation
Conceptual Layers of a Modern Computer High-level language layer Application layer compiled Operating System Assembly language layer Hardware assembled Machine language layer interpreted Micro-instruction layer directly executed Hardware layer (M.Cheng)
instruction stream data stream Control Unit ALU Single-Instruction-Single-Data stream (SISD) data stream 1 ALU instruction stream Control Unit data stream N ALU Single-Instruction-Multiple-Data stream (SIMD) Classes of Computers
instruction stream 1 Control Unit ALU data stream instruction stream N Control Unit ALU Multiple-Instruction-Single-Data stream (MISD) instruction stream 1 data stream 1 Control Unit ALU instruction stream N data stream N Control Unit ALU Multiple-Instruction-Multiple-Data stream (MIMD) (M.Cheng) Classes of Computers.
Classes of Computers • SISD: conventional serial computer. • SIMD: vector processor - each instruction operates on a data vector rather than a single operand, e.g. CRAY • MISD computer in general does not exist. But for reliability reason, there are computer systems with redundant processors which executes the same input by several independent processors. • MIMD computer is actually called a multiprocessor, e.g, C.mmp at CMU.
(To Memory or I/O Devices) Data Bus Registers MBR PC CU ALU IR EAR Address Bus (To Memory or I/O Devices) Control Bus (M.Cheng) Basics • Three different types of bus used to interconnect the CPU, memory and I/O devices. Control bus (bidirectional)–carries the necessary commands and control signals to the various parts of the system. Address bus (unidirectional)–specifies the address of the memory word or I/O device that the CPU wants to communicate. Data bus (bidirectional)–carries the data transmitted between CPU, memory and I/O devices.
Basics. • If a computer has an N-bit address bus, then it can address up to 2**N unique locations. • Every addressable unit may be of different sizes from one computer to another. It may be bit-addressable (e.g., Burroughs B1700), byte-addressable (e.g., Intel 8086), or word-addressable. (Note: There is no general agreement what a word size should be.) • When we speak of a computer of size N-bit, we usually refer to the size of its data bus. It is possible to have an 8-bit computer with word size 16-bit (e.g., Intel 8088, Motorola 6800).
- e.g., Motorola MC68000, DEC VAX-11, IBM 360 - The CPU contains a set of general purpose (scratch-pad) registers. - Typically, this type of processors supports two-address or three-address instructions. Computer Structures • General register machine - A two-address instruction: ADD X , Y % Y := Y + X; where X,Y and Z refer to a memory location or a register. - A three-address instruction: ADD X , Y, Z % Z := X + Y;
• Accumulator machine Computer Structures - e.g., Rockwell 6502, Motorola 6809 or 6800 series - There are one (or more) accumulators: general scratch-pad registers - all data processing instructions refer to at least one accumulator, hence one-address instructions of the form: LDA X % ACC := X; ADD X % ACC := ACC + X; where X refers to a memory location.
• Stack machine Computer Structures - e.g., Hewlett-Packard HP3000, Burroughs B1700 - All instructions are assumed to operate on the top of a LIFO stack data structure - A dedicated register, called stack pointer (SP), refers to the top of the stack. - All instructions must refer to the stack pointer implicitly, hence zero-address instructions. - Different instructions result in different amount to be added to or subtracted from the SP automatically. - A zero-address instruction has the following form: PUSH X % ++SP; *SP := X; POP X % X := *SP; - - SP; ADD % - - SP; *SP := *(SP+1) + *SP; SUB % - - SP; *SP := *(SP+1) - *SP; where X refers to a memory location.
Supercomputer Types of Computers - e.g. CRAY - usually SIMD • Maxicomputer - e.g., CDC7600, IBM 370 - “ mainframe computer” • Midicomputer - e.g., DEC VAX 8600, IBM 4381, high-end SGI’s (?) - “supermini computer” • Minicomputer fuzzy area • e.g.,higher-end Sun’s & SGI’s, DEC PDP-11 (16-bit), Data General NOVA • classification is obsolete • Microcomputer e.g., IBM PC’s, Apple Macintosh, Amiga, Atari, SGI, Sun Cell phones, Pads, iPod etc. are the next generation of micro computers.
Design goals for computer architectures • performance - raw speed eg. MIPS, MFLOPS - speed with respect to a particular application, programming language or environment e.g., digital signal processing, FORTRAN, ADA, Prolog, Graphics, real-time systems Compatibility - with previous versions; other hardware & standards - with future versions
Design goals for computer architectures. • ease of use - easy to program e.g., instruction set, interaction with external devices - easy to add peripherals e.g., high-speed optical disks, a communication switching network