ZeePedia buy college essays online

Advance Computer Architecture

<<< Previous Data Forwarding Hardware, Superscalar, VLIW Architecture Next >>>
Advanced Computer Architecture-CS501
Advanced Computer Architecture
Lecture 21
Reading Material
Vincent P. Heuring&Harry F. Jordan
Chapter 5
Computer Systems Design and Architecture
Data Forwarding Hardware
Instruction Level Parallelism
Difference between Pipelining and Instruction-Level Parallelism
Superscalar Architecture
Superscalar Design
VLIW Architecture
Maximum Distance between two instructions
Read page no. 219 of Computer System Design and Architecture (Vincent
P.Heuring, Harry F. Jordan)
Data forwarding Hardware
The concept of data forwarding was introduced in the previous lecture.
RTL for
Page 220
Last Modified: 01-Nov-06
Advanced Computer Architecture-CS501
forwarding in case of ALU instructions
Stage 3-5
(ra5=rc3)&!imm3: Y Z5);
Stage 3-4
(ra4=rc3)&!imm3: Y Z4);
Instruction-Level Parallelism
Increasing a processor's throughput
There are two ways to increase the number of instructions executed in a given time by a
 By increasing the clock speed
 By increasing the number of instructions that can execute in parallel
Increasing the clock speed
Increasing the clock speed is an IC design issue and depends on the advancements in
chip technology.
The computer architect or logic designer can not thus manipulate clock speeds to
increase the throughput of the processor.
Increasing parallel execution of instructions
The computer architect cannot increase the clock speed of a microprocessor however
he/she can increase the number of instructions processed per unit time. In pipelining we
discussed that a number of instructions are executed in a staggered fashion, i.e. various
instructions are simultaneously executing in different segments of the pipeline. Taking
this concept a step further we have multiple data paths hence multiple pipelines can
execute simultaneously. There are two main categories of these kinds of parallel
instruction processors VLIW (very long instruction word) and superscalar.
The two approaches to achieve instruction-level parallelism are
­  Superscalar Architecture
A scalar processor that can issue multiple instructions simultaneously is said to be
­  VLIW Architecture
A VLIW processor is based on a very long instruction word. VLIW relies on
instruction scheduling by the compiler. The compiler forms instruction packets which can
run in parallel without dependencies.
Page 221
Last Modified: 01-Nov-06
Advanced Computer Architecture-CS501
Difference between Pipelining and Instruction-Level Parallelism
Instruction-Level Parallelism
Single functional unit
Multiple functional units
Instructions are issued sequentially
Instructions are issued in parallel
Throughput increased by overlapping the Instructions  are  not  overlapped  but
instruction execution
executed in parallel in multiple functional
Very little extra hardware required to Multiple functional units within the CPU
implement pipelining
are required
Superscalar Architecture
A superscalar machine has following typical features
 It has one or more IUs (integer units) , FPUs (floating point units), and BPUs (branch
prediction units)
 It divides instructions into three classes
o Integer
o Floating point
o Branch prediction
The general operation of a superscalar processor is as follows
 Fetch multiple instructions
 Decode some of their portion to determine the class
 Dispatch them to the corresponding functional unit
As stated earlier the superscalar design uses multiple pipelines to implement instruction
level parallelism.
Operation of Branch Prediction Unit
BPU calculates the branch target address ahead of time to save CPU cycles
Branch instructions are routed from the queue to the BPU where target address is
calculated and supplied when required without any stalls
BPU also starts executing branch instructions by speculating and discards the results
if the prediction turns out to be wrong
Superscalar Design
The philosophy behind a superscalar design is
 to prefetch and decode as many instructions as possible before execution
Page 222
Last Modified: 01-Nov-06
Advanced Computer Architecture-CS501
and to start several branch instruction streams speculatively on the basis of this
and finally, discarding all but the correct stream of execution
The superscalar architecture uses multiple instruction issues and uses techniques such as
branch prediction and speculative instruction execution, i.e. it speculates on whether a
particular branch will be taken or not and then continues to execute it and the following
instructions. The results are not written back to the registers until the branch decision is
confirmed. Most superscalar architectures contain a reorder buffer. The reorder buffer
acts like an intermediary between the processor and the register file. All results are
written onto the reorder buffer and when the speculated course of action is confirmed, the
reorder buffer is committed to the register file.
Superscalar Processors
Examples of superscalar processors
o PowerPC 601
o Intel P6
o DEC Alpha 21164
VLIW Architecture
VLIW stands for "Very Long Instruction Word" typically 64 or 128 bits wide. The longer
instruction word carries information to route data to register files and execution units.
The execution-order decisions are made at the compile time unlike the superscalar design
where decisions are made at run time. Branch instructions are not handled very efficiently
in this architecture. VLIW compiler makes use of techniques such as loop unrolling and
code reordering to minimize dependencies and the occurrence of branch instructions.
Page 223
Last Modified: 01-Nov-06
Table of Contents:
  1. Computer Architecture, Organization and Design
  2. Foundations of Computer Architecture, RISC and CISC
  3. Measures of Performance SRC Features and Instruction Formats
  4. ISA, Instruction Formats, Coding and Hand Assembly
  5. Reverse Assembly, SRC in the form of RTL
  6. RTL to Describe the SRC, Register Transfer using Digital Logic Circuits
  7. Thinking Process for ISA Design
  8. Introduction to the ISA of the FALCON-A and Examples
  9. Behavioral Register Transfer Language for FALCON-A, The EAGLE
  10. The FALCON-E, Instruction Set Architecture Comparison
  11. CISC microprocessor:The Motorola MC68000, RISC Architecture:The SPARC
  12. Design Process, Uni-Bus implementation for the SRC, Structural RTL for the SRC instructions
  13. Structural RTL Description of the SRC and FALCON-A
  14. External FALCON-A CPU Interface
  15. Logic Design for the Uni-bus SRC, Control Signals Generation in SRC
  16. Control Unit, 2-Bus Implementation of the SRC Data Path
  17. 3-bus implementation for the SRC, Machine Exceptions, Reset
  18. SRC Exception Processing Mechanism, Pipelining, Pipeline Design
  19. Adapting SRC instructions for Pipelined, Control Signals
  20. SRC, RTL, Data Dependence Distance, Forwarding, Compiler Solution to Hazards
  21. Data Forwarding Hardware, Superscalar, VLIW Architecture
  22. Microprogramming, General Microcoded Controller, Horizontal and Vertical Schemes
  23. I/O Subsystems, Components, Memory Mapped vs Isolated, Serial and Parallel Transfers
  24. Designing Parallel Input Output Ports, SAD, NUXI, Address Decoder , Delay Interval
  25. Designing a Parallel Input Port, Memory Mapped Input Output Ports, wrap around, Data Bus Multiplexing
  26. Programmed Input Output for FALCON-A and SRC
  27. Programmed Input Output Driver for SRC, Input Output
  28. Comparison of Interrupt driven Input Output and Polling
  29. Preparing source files for FALSIM, FALCON-A assembly language techniques
  30. Nested Interrupts, Interrupt Mask, DMA
  31. Direct Memory Access - DMA
  32. Semiconductor Memory vs Hard Disk, Mechanical Delays and Flash Memory
  33. Hard Drive Technologies
  34. Arithmetic Logic Shift Unit - ALSU, Radix Conversion, Fixed Point Numbers
  35. Overflow, Implementations of the adder, Unsigned and Signed Multiplication
  36. NxN Crossbar Design for Barrel Rotator, IEEE Floating-Point, Addition, Subtraction, Multiplication, Division
  37. CPU to Memory Interface, Static RAM, One two Dimensional Memory Cells, Matrix and Tree Decoders
  38. Memory Modules, Read Only Memory, ROM, Cache
  39. Cache Organization and Functions, Cache Controller Logic, Cache Strategies
  40. Virtual Memory Organization
  41. DRAM, Pipelining, Pre-charging and Parallelism, Hit Rate and Miss Rate, Access Time, Cache
  42. Performance of I/O Subsystems, Server Utilization, Asynchronous I/O and operating system
  43. Difference between distributed computing and computer networks
  44. Physical Media, Shared Medium, Switched Medium, Network Topologies, Seven-layer OSI Model