Das Video kommt von YouTube: erst beim Abspielen verbindet sich die Seite mit YouTube (Google).
Von Neumann and Harvard CPU Architectures
Das Wichtigste aus dem Video
Tipp auf eine Zeit – das Video springt genau dorthin.
Transkriptautomatisch erstellt · 36 Zeilen
- hello we're going to look at the differences between the Von neyman and the Harvard CPU architectures word architecture is very vague when we talk
- about architectes we're thinking more about the overall design of a system rather than specifics what are the design principles of a system back when
- computers look like this in the 1940s in the early days of computing different designs were put forward and suggested and evaluated and two of the big ones
- were Von neyman and Harvard both of which have been discussed and thought about and used for now 70 years St the obvious computers nowadays are very very
- different to the 1940s although aspects of their design are exactly the same so Von neyman is arguably the one which has been the most widely used this person
- here is Von neyman he was at Princeton University so is often also called for Princeton architecture and really he put forward or at least his team put forward
- things which are used in modern CPU so his suggestion was that a computer should have one control unit one arithmetic logic unit the control unit
- overseas the process the AL carries out maths and logic operations he also suggested that CPUs should have a small bit of memory inside the CPU which are
- registers and he also said there needs to be a larger memory somewhere else in the computer that needs to hold both instructions and data and that last
- point is the most critical one when we are comparing Von Norman architecture to Harvard architecture so in Harvard architecture here's a picture of Harvard
- for comparison this architecture is basically the same idea as Von nyman except people at Harvard proposed that instructions and data should be held in
- separate memory units and what that means in reality is we've got one stick of RAM for instructions one stick of RAM for data and if you know about computers
- you'll know that we don't have different sticks of RAM for different things like this so therefore Von noyman is what is used in most modern computers but
- Harvard suggested we need to have two different memory units and consequently this means if we were to implement Harvard architecture we need to also
- have two sets of buses whereas in vman we've got one set of buses Harvard should have two so here's how Bon noyman architecture looks like we've got our
- CPU we've got memory which in essence is normally Ram but could be ROM and we've got one set of buses address data and control now the difference between
- instructions and data can be confusing because the word data is such a vague word anyway instructions are things like input and subtract these tell the
- computer to do something a lot of the time these instructions work on some data so sub five probably doesn't mean subtract five probably means subtract
- the value at location five in memory which could be a value like 17 or 105 those are examples of data which are used by instructions an instruction like
- imp doesn't need any data but but these ones do and the whole idea of vman is they're held in the same memory unit they might be separate but they're held
- together in this stick of ram whereas in the Haren model it's a little bit more complicated because we've now got two separate components one of these
- components is for holding the instructions and one of these components is for holding the data and because they are separate we need to connect up to
- the CPU via buses that's why we've got two sets of buses in Harvard compared to one in v and because one of these memories is for
- instructions we might rename the data bus to the instruction bus when talking about Harvard although it's the same thing they're both bir directional they
- both travel to and from the CPU now this diagram might give you a clue as to why one noyman architecture has been adopted and is used essentially one noyman has
- become the dominant architecture because it's less complex and therefore less costly to manufacture because you need to have less Hardware we need to have
- two separate memory components in Harvard and two separate buses that is just a bit more work a bit more money to be spent on that system now it's not a
- no-brainer because Harvard does have benefits in Von noyman a key problem is you cannot either read or write and instruction data at the same time that's
- because memory can only do one thing at a time you cannot simultaneously ask it to read an instruction while you will also asking it to read some data which
- might be required if you are pipelining for example we just can't do do that simultaneously with only one memory now this issue is called the Von Norman
- bottleneck a bottleneck is a holdup where you're waiting for something else to finish before your task can complete and we just don't have this bottleneck
- in Harvard because in Harvard we've got separate buses separate Memories One of These Memories can be fetching an instruction while the other memory is
- fetching some data so they can happen simultaneously which means it can be a little bit faster and can be a little bit more efficient so despite being more
- complicated Harvard is often used for what called realtime systems a realtime system will respond immediately there will not be a delay because for its
- purpose that is not acceptable so Road time systems are often used in things like airplanes or medical situations but we can't be waiting around we need to be
- processing stuff immediately even if it means the hardware is a little bit more complicated to manufacture
Zum Nachlesen
Von-Neumann-ArchitekturDie Von-Neumann-Architektur (VNA) ist ein Referenzmodell für Computer, wonach ein gemeinsamer Speicher sowohl Computerprogrammbefehle als auch Daten hält.
Harvard-ArchitekturDie Harvard-Architektur bezeichnet in der Informatik ein Schaltungskonzept, bei dem der Befehlsspeicher logisch und physisch vom Datenspeicher getrennt ist.
ProzessorAufbau und Funktionale Einheiten · Hauptprozessor (CPU) und Mehrprozessorkerne · Steuer- bzw. Leitwerk · Rechenwerk und Register · Datenleitungen · Caches und MMU.
SpeicherwerkIm Gegensatz zur Harvard-Architektur befinden sich in der Von-Neumann-Architektur Daten und Befehle im gleichen Speicher. In modernen Mikrocomputern besteht …