Chapter 8 CPU and Memory: Design, Implementation, and Enhancement The Architecture of Computer Hardware and Systems Software: An Information Technology Approach 3rd Edition, Irv Englander John Wiley and Sons  2003

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-2 CPU Architecture Overview  CISC – Complex Instruction Set Computer  RISC – Reduced Instruction Set Computer  CISC vs. RISC Comparisons  VLIW – Very Long Instruction Word  EPIC – Explicitly Parallel Instruction Computer

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-3 CISC Architecture  Examples  Intel x86, IBM Z-Series Mainframes, older CPU architectures  Characteristics  Few general purpose registers  Many addressing modes  Large number of specialized, complex instructions  Instructions are of varying sizes

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-4 Limitations of CISC Architecture  Complex instructions are infrequently used by programmers and compilers  Memory references, loads and stores, are slow and account for a significant fraction of all instructions  Procedure and function calls are a major bottleneck  Passing arguments  Storing and retrieving values in registers

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-5 RISC Features  Examples  Power PC, Sun Sparc, Motorola  Limited and simple instruction set  Fixed length, fixed format instruction words  Enable pipelining, parallel fetches and executions  Limited addressing modes  Reduce complicated hardware  Register-oriented instruction set  Reduce memory accesses  Large bank of registers  Reduce memory accesses  Efficient procedure calls

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-6 CISC vs. RISC Processing

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-7 Circular Register Buffer

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-8 Circular Register Buffer - After Procedure Call

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-9 CISC vs. RISC Performance Comparison  RISC  Simpler instructions  more instructions  more memory accesses  RISC  more bus traffic and increased cache memory misses  More registers would improve CISC performance but no space available for them  Modern CISC and RISC architectures are becoming similar

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-10 VLIW Architecture  Transmeta Crusoe CPU  128-bit instruction bundle = molecule  4 32-bit atoms (atom = instruction)  Parallel processing of 4 instructions  64 general purpose registers  Code morphing layer  Translates instructions written for other CPUs into molecules  Instructions are not written directly for the Crusoe CPU

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-11 EPIC Architecture  Intel Itanium CPU  128-bit instruction bundle  3 41-bit instructions  5 bits to identify type of instructions in bundle  bit general purpose registers  bit floating point registers  Intel X86 instruction set included  Programmers and compilers follow guidelines to ensure parallel execution of instructions

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-12 Paging  Managed by the operating system  Built into the hardware  Independent of application

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-13 Logical vs. Physical Addresses  Logical addresses are relative locations of data, instructions and branch target and are separate from physical addresses  Logical addresses mapped to physical addresses  Physical addresses do not need to be consecutive

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-14 Logical vs. Physical Address

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-15 Page Address Layout

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-16 Page Translation Process

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-17 Memory Enhancements  Memory is slow compared to CPU processing speeds!  2Ghz CPU = 1 cycle in ½ of a billionth of a second  70ns DRAM = 1 access in 70 millionth of a second  Methods to improvement memory accesses  Wide Path Memory Access  Retrieve multiple bytes instead of 1 byte at a time  Memory Interleaving  Partition memory into subsections, each with its own address register and data register  Cache Memory

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-18 Memory Interleaving

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-19 Why Cache?  Even the fastest hard disk has an access time of about 10 milliseconds  2Ghz CPU waiting 10 milliseconds wastes 20 million clock cycles!

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-20 Cache Memory  Blocks: 8 or 16 bytes  Tags: location in main memory  Cache controller  hardware that checks tags  Cache Line  Unit of transfer between storage and cache memory  Hit Ratio: ratio of hits out of total requests  Synchronizing cache and memory  Write through  Write back

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-21 Step-by-Step Use of Cache

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-22 Step-by-Step Use of Cache

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-23 Performance Advantages  Hit ratios of 90% common  50%+ improved execution speed  Locality of reference is why caching works  Most memory references confined to small region of memory at any given time  Well-written program in small loop, procedure or function  Data likely in array  Variables stored together

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-24 Two-level Caches  Why do the sizes of the caches have to be different?

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-25 Cache vs. Virtual Memory  Cache speeds up memory access  Virtual memory increases amount of perceived storage  independence from the configuration and capacity of the memory system  low cost per bit

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-26 Modern CPU Processing Methods  Timing Issues  Separate Fetch/Execute Units  Pipelining  Scalar Processing  Superscalar Processing

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-27 Timing Issues  Computer clock used for timing purposes  MHz – million steps per second  GHz – billion steps per second  Instructions can (and often) take more than one step  Data word width can require multiple steps

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-28 Separate Fetch-Execute Units  Fetch Unit  Instruction fetch unit  Instruction decode unit  Determine opcode  Identify type of instruction and operands  Several instructions are fetched in parallel and held in a buffer until decoded and executed  IP – Instruction Pointer register  Execute Unit  Receives instructions from the decode unit  Appropriate execution unit services the instruction

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-29 Alternative CPU Organization

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-30 Instruction Pipelining  Assembly-line technique to allow overlapping between fetch-execute cycles of sequences of instructions  Only one instruction is being executed to completion at a time  Scalar processing  Average instruction execution is approximately equal to the clock speed of the CPU  Problems from stalling  Instructions have different numbers of steps  Problems from branching

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-31 Branch Problem Solutions  Separate pipelines for both possibilities  Probabilistic approach  Requiring the following instruction to not be dependent on the branch  Instruction Reordering (superscalar processing)

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-32 Pipelining Example

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-33 Superscalar Processing  Process more than one instruction per clock cycle  Separate fetch and execute cycles as much as possible  Buffers for fetch and decode phases  Parallel execution units

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-34 Superscalar CPU Block Diagram

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-35 Scalar vs. Superscalar Processing

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-36 Superscalar Issues  Out-of-order processing – dependencies (hazards)  Data dependencies  Branch (flow) dependencies and speculative execution  Parallel speculative execution or branch prediction  Branch History Table  Register access conflicts  Logical registers

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-37 Hardware Implementation  Hardware – operations are implemented by logic gates  Advantages  Speed  RISC designs are simple and typically implemented in hardware

Chapter 8: CPU and Memory: Design, Implementation, and Enhancement 8-38 Microprogrammed Implementation  Microcode are tiny programs stored in ROM that replace CPU instructions  Advantages  More flexible  Easier to implement complex instructions  Can emulate other CPUs  Disadvantage  Requires more clock cycles