ENGS 116 Lecture 11 ENGS 116 / COSC 107 Computer Architecture Introduction Vincent H. Berk September 24 th, 2008 Reading for Friday: Chapter 1.1 – 1.4, Amdahl article Reading for Monday: 1.5 – 1.11

ENGS 116 Lecture 12 Prerequisite Knowledge Assembly language programming Fundamentals of logic design  Combinational and sequential components (e.g., gates, multiplexers, decoders, ROMs, flip-flops, registers, RAMs)‏ Processor Design  Instruction cycle, pipelining, branch prediction, exceptions Memory Hierarchy  Caches (direct-mapped, fully-associative, 2-way set associative), spatial locality, temporal locality, virtual memory, translation lookaside buffer (TLB)‏ Input and Output  Polling, interrupts Multiprocessors

ENGS 116 Lecture 13 What is Computer Architecture? Two viewpoints: Hardware designer’s viewpoint: CPUs, caches, buses, pipelines, physical memory, etc. Programmer’s viewpoint: instruction set – opcodes, addressing modes, registers, virtual memory, etc.  Study of architecture covers both instruction-set architectures and machine implementation organizations.

ENGS 116 Lecture 14 Computer Architecture Is... The attributes of a [computing] system as seen by the programmer, i.e., the conceptual structure and functional behavior, as distinct from the organization of the data flows and controls the logic design, and the physical implementation. Amdahl, Blaauw, and Brooks, 1964

ENGS 116 Lecture 15 Computer Architecture’s Changing Definition 1950s to 1960s: Computer Architecture Course = Computer Arithmetic. 1970s to 1980s: Computer Architecture Course = Instruction Set Design, especially ISA appropriate for compilers. 1990s to 2000s: Computer Architecture Course = Design of CPU, memory system, I/O 2000 to now: Computer Architecture Course = ILP, DLP, TLP, storage

ENGS 116 Lecture 16

7 5 Generations of Electronic Computers (Hwang)‏

ENGS 116 Lecture 18 Computer Tasks Desktop Computing, Lightweight Servers, Laptops  Price-performance (low cost)‏  Communication, Graphics Server Computing, Mainframe Systems  Specific performance, processing power, storage  Availability, Reliability Embedded Computers and DSPs  Power and Memory requirements  Lowest cost for required performance  Real-time or soft-real-time performance

ENGS 116 Lecture 19 Task of Computer Designer Determine which attributes are important for a new machine. Design a machine to meet functional requirements, price, power and performance goals.

ENGS 116 Lecture 110 Basic Computer Organization Control Datapath Memory Input Output Processor

ENGS 116 Lecture 111 Computer Architecture Topics Instruction Set Architecture Pipelining, Hazard Resolution, Superscalar, Reordering, Prediction, Speculation Addressing, Protection, Exception Handling L1 Cache L2/L3 Cache SDRAM Disks, WORM, Tape Multi-Core Coherence, Bandwidth, Latency Emerging Technologies Interleaving Bus protocols RAID VLSI Input/Output and Storage Memory Hierarchy Pipelining, Instruction- Level Parallelism, Thread- Level Parallelism

ENGS 116 Lecture 112 Computer Architecture Topics M Interconnection Network S PMPMPMP ° ° ° Topologies, Routing, Bandwidth, Latency, Reliability Network Interfaces Shared Memory, Message Passing, Data Parallelism Processor-Memory-Switch Multiprocessors Networks and Interconnections

ENGS 116 Lecture 113 Understanding the design techniques, machine structures, technology factors, and evaluation methods that will determine the form of computers in the 21st Century Technology Programming Languages Operating Systems History Applications Interface Design (ISA)‏ Measurement & Evaluation Parallelism Computer Architecture: Instruction Set Design Organization Hardware Course Focus

ENGS 116 Lecture 114 Technology Trends Integrated circuit logic technology  transistor density (feature size)‏  transistor count  cycle speed  multiple cores Semiconductor DRAM  density  latency and bandwidth Magnetic disk technology  density  access time Network technology  bandwidth  latency

ENGS 116 Lecture 115 Scaling in ICs Feature size: minimum size of a single distinguishable/producible item on a chip die  1971 – 10 microns  2001 – 0.18 microns  2003 – 0.06 microns  2006 – 5 nanometers (0.005 microns)‏ Complex relationships:  Transistor density increases quadratically with decrease in feature size  Reduction in feature size requires voltage reduction to maintain correct operation and reasonable reliability Scaling IC wiring:  Signal delay increases with product of resistance and capacitance  Shorter wires can be smaller  Smaller features have higher current leakage

ENGS 116 Lecture 116 Power Consumption of ICs Power requirements per transistor are proportional to load capacitance, frequency of switching and the square of the voltage. Switching frequency and density of transistors increases faster than decrease in capacitance and voltage, leading to increased power consumption == generated heat Pentium 4 consumes 135 Watts of power while the i386 did not even feature a heat-sink Power = ½ x Capacitance x Voltage 2 x Frequency switched

ENGS 116 Lecture 117 Cost and Price Cost of manufacturing decreases over time: learning curve Learning curve is measured as an increase in yield Volume doubling leads to 10% reduction in cost Commodity products tend to decrease cost:  Volume  Competition  Efficiency

ENGS 116 Lecture 118 Difference between Cost and Price

ENGS 116 Lecture 119 Wafers and Dies Chips are produced on round silicon disks Dies are the actual chip, cut out from the wafer Testing occurs before cutting and after packaging

ENGS 116 Lecture 120 Yield and Cost However:  Wafers do not just contain chip-dies, usually a large area, including several chip-dies, is dedicated for test equipment hook-up  Actual yield in mass-production chip-fabs varies between 98% for DRAMS to 1% for new Processors

ENGS 116 Lecture 121 Yield and Cost Switch from 200mm to 300mm wafers:  Although 300mm wafers have lower yield than 200mm wafers, the overhead processing costs per wafer are high enough to make 300mm wafers more cost effective. Redundancy in dies:  Single transistors do fail during production, causing memory cells, pipeline stages, control logic sections to fail  Redundancy is built into the each die by introducing backup-units  After testing, backup units are enabled and failed units can be disabled by LASER  This decreases the chances of small flaws failing an entire die  Few companies give insight into their redundant circuitry numbers

ENGS 116 Lecture 122 Performance Hwang: “The ideal performance of a computer system demands a perfect match between machine capability and program behavior.” Machine capability – enhanced with better hardware technology, innovative architectural features, efficient resource management. Program behavior – affected by algorithm design, data structures, language efficiency, programmer skill, compiler technology. To improve software performance, need to understand how various hardware factors affect overall system performance!

ENGS 116 Lecture 123 Measuring Performance Key measure is time. Response time (execution time): Time between start and completion of a task. Throughput: total amount of work completed in a given time.

ENGS 116 Lecture 124 Comparing Design Alternatives “ X is n times faster than Y” means

ENGS 116 Lecture 125 Benchmarking Real programs; e.g., compilers, photo editing Modified or scripted real programs; e.g., compression algorithms Kernels – small, key pieces from real programs; e.g., Livermore Loops, Linpack. Toy benchmarks – typically 10 to 100 lines of code, useful primarily for intro programming assignments; e.g., quicksort, prime numbers, encryption Synthetic benchmarks – try to match average frequency of operations and operands for a set of programs; e.g., Whetstone, Dhrystone. Benchmark suites – collections of programs; e.g, SPEC CPU2000