Harvard architecture

E87766

Harvard architecture is a computer architecture design that uses separate memory and signal pathways for instructions and data, enabling simultaneous access and often improved performance over unified-memory designs.


Statements (48)
Predicate Object
instanceOf computer architecture
instruction set architecture concept
aimsTo improve performance over unified-memory designs
canBeImplementedAs modified Harvard architecture
strict Harvard architecture
canImprove throughput of pipelined processors
canReduce instruction fetch bottlenecks
contrastsWith Princeton architecture
von Neumann architecture
differsFrom von Neumann architecture in bus structure
von Neumann architecture in memory organization
emergedIn 1940s
enables simultaneous access to instructions and data
hasAlternativeName Harvard computer architecture
hasLimitation more complex memory design than unified-memory systems
more complex programming model for self-modifying code
requires explicit mechanisms to write to instruction memory
hasProperty can allow different access times for instructions and data
can allow different memory technologies for code and data
can allow different word widths for instructions and data
can fetch instructions and data in parallel
can simplify pipeline design
physically separate instruction and data memories
reduces contention between instruction fetch and data access
separate instruction and data buses
historicallyAssociatedWith Harvard Mark I computer
influenced modified Harvard architecture
isBasisFor modified Harvard architecture
isCharacterizedBy no single shared bus for instructions and data
no single shared memory for instructions and data
isCommonIn AVR microcontrollers
PIC microcontrollers
many DSP chips
isDiscussedIn computer architecture textbooks
isNamedAfter Harvard University NERFINISHED
isTaughtIn computer engineering curricula
computer science curricula
isUsedIn digital signal processors
embedded systems
microcontrollers
some DSP-oriented CPUs
some RISC microcontrollers
originatedAt Harvard University NERFINISHED
supports parallel instruction fetch and data access
uses separate memory for data
separate memory for instructions
separate signal pathways for data
separate signal pathways for instructions

Referenced by (3)
Subject (surface form when different) Predicate
Computer Command System ("non–von Neumann architecture")
architecture
von Neumann architecture
contrastedWith
Harvard architecture ("Harvard computer architecture")
hasAlternativeName

Please wait…