The NuBus

1
The NuBus

• Miles Carter
• Ben Stoutenburgh
• Jim Vallerie

2
IntroductionSome NuBus History

• NuBus was originally created in the 1970s at MIT
  as a new parallel computing bus implementation.
• NuBus was a pioneer of 32-bit design that aimed
  to support the predicted market shift towards
  universal, industry wide 32-bit implementations.
• Emphasis was placed on simplicity of install and
  configuration a NuBus hardware card could be
  plugged into any NuBus machine as long as it was
  supplied the correct device driver.
• Like many other components of the same nature,
  NuBus was born under the auspices of the
  Institute of Electrical and Electronic Engeineers
  (ANSI/IEEE STD 1186-1988).
• NuBus was adopted most notably by Apple in their
  Macintosh II personal computer.

3
IntroductionSome More NuBus History

• The entire NuBus system was implemented as a task
  sharing array of devices within the NuBus device
  slots.
• Each NuBus device contained its own independent
  68000 processor, working space in RAM, and basic
  operating system services.
• The taxing load of time consuming/time critical
  tasks can be taken of a NuBus equipped systems
  main logic board and placed in the hands of an
  appropriate independent NuBus processor.
• The success of NuBus was relatively limited.
  Although it saved production costs on the main
  board hardware end, it increased costs for
  expansion device producers.
• NuBuss use in the Macintosh computing platform
  was extinguished when the company adopted the PCI
  bus in the early 1990s.

4
Bus ArbitrationBasic Arbitration Process

• Since more than one processor/device can be
  connected to the same bus, some mechanism is
  needed to deal with simultaneous requests for bus
  usage. Arbitration is how multiple requests are
  recognized and given priority by the system.
• NuBus is of distributed arbitration each device
  on the bus chain (master) monitors the other
  devices (masters) competing for the bus and
  determines whether it should assume control of
  the bus based on its priority in relation to its
  competitors.
• Each slot has its own unique ID code in the bus.
  The ID is fixed and defined for each slot by
  hardwires in the buses backplane. Rather than
  the card defining its ID in the bus, the data
  path of the backplane dictates each slots unique
  ID.
• Overall, NuBus is decentralized, which means that
  no particular slot assumes master control over
  each successive slot in the bus. The NuBus
  system of priority determines which slot in the
  bus assumes control at any given time.

5
Bus ArbitrationThe Details

• NuBus supports multiple bus masters at the same
  time.
• There are four arbitration lines, and one
  prioritization algorithm accessed when multiple
  requestors compete for bus control (usually the
  requestor with the highest ID number is given
  control).
• When control is desired, a requestor asserts
  request for control over its own bus request
  line. It continues this assertion until the bus
  is free and it is awarded control.
• No special master arbiter exists to decide which
  master will get the bus. All priority is handled
  by electrical current flow and multiplexer
  control.

• Simplified circuit for NuBus arbitration
• Lines ID0 to ID3 define the slot location and,
  therefore, the priority level of the master.
• Lines ARB0 to ARB3 are the NuBus arbitration
  lines that run the length of the bus.

6
Bus ArbitrationThe Bus Request Process

• A potential master that wants to use the bus
  places its arbitration level on the 4-bit
  arbitration bus, ID3 to ID0.
• The arbitration bus lines are driven by
  open-collector circuits and can, therefore, be
  driven by more than one card simultaneously
  without creating bus contention.
• An open-collector gate can force the bus actively
  down to an electrically low level or passively
  let it remain in an electrically high level. If
  several open-collector circuits drive the bus in
  different directions, the electrically low state
  always wins. Since NuBus uses negative logic, the
  arbitration number is inverted so that the
  highest level of priority is 0000 and the least
  is 1111.
• The key to NuBus arbitration is simpleif a
  competing master sees a higher level on the bus
  than its own level, it ceases to compete for the
  bus. Each requester simultaneously drives the
  arbitration bus and observes the signal on bus.
  If it detects the presence of a requester at a
  higher level, it backs off.

For further reading visit http//wheelie.tees.ac.
uk/users/a.clements/Buses/Buses.htm
7
Bus ArbitrationAn Example Of Arbitration

• Consider the case in which three masters numbered
  0100 (four), 0101 (five) and 0010 (two)
  simultaneously put the codes 1011, 1010 and 1101,
  respectively, onto the arbitration bus.
• As the arbitration lines are open-collector, any
  output at a 0 level will pull the bus down to 0.
  In this example, the bus will be forced into the
  state 1000. The master at level 2 that puts the
  code 1101 on the arbitration bus will detect that
  ARB2 is being pulled down and will therefore
  leave the arbitrating process and the arbitration
  bus will now have the value 1010. The master with
  the code 1011 will detect that ARB1 is being
  pulled down and will leave the arbitration
  process. The value on the arbitration bus is now
  1010 and the master with that value (five) has
  gained control.
• The signal labeled ARBITRATE permits the master
  to arbitrate for the bus, and the output GRANT is
  asserted if the master wins the arbitration.
• Because NuBus arbitration uses simple
  non-sequential arbitration logic and arbitration
  can take place in parallel with normal bus
  activity, the NuBus is well suited to
  multiprocessing applications.

8
Data FormatsGeographic Addressing

• NuBus implements geographic addressing each card
  in a NuBus system is partitioned a unique slice
  of memory from the address space spanned in main
  memory for the entire system.
• The entire NuBus is allocated a 256 MByte address
  space called slot space. This space is
  subdivided into 16 blocks of 16 Mbytes (each slot
  has its own 16 Mbyte slot). NuBus cannot support
  more than 16 slots.
• Designing systems where each bus slot has its own
  space in memory is much less difficult than
  systems that deal with shared bus storage (less
  management of data overlap/collision).

9
Data TransferCommon Standards

• IEEE 1196 standards
• 10 Mhz clock
• Maximum burst throughput of 40Mbps
• Average speeds of 10 to 20Mbps
• NuBus90 standards
• 20 Mhz clock
• Maximum burst throughput of 70Mbps
• Average speed of 30Mbps

10
Battle!PCI vs. NuBus
Feature NuBus PCI bus
Bus clock rate 10 MHz 33 MHz, 66 MHz
Addressing Geographic Dynamic
Signal loading No enforced rules One load per signal
Transaction length determination Determined at start of transaction Determined at end of transaction
Bus termination Resistor network Not required
Bus control arbitration Distributed Centralized
Addressing spaces Memory only Memory, I/O, and configuration
Wait-state generators Slave only Slave and master
Kinds of expansion Cards only Cards and ASIC chips
Timeout 255 bus clocks 5 bus clocks
Burst capability 8, 16, 32, or 64 bytes Any number of bytes
Power allocation 15 W per card 7.5, 15, or 25 W per card
11
A/ROSEWhat is it?

• The Apple Real-Time Operating System Environment
• The Operating System for the Macintosh
  Coprocessor Platform (foundation for connectivity
  products)
• Development started in August 1987
• First operational version in February 1988

12
A/ROSEWhat is it?

• A real-time, multitasking, message-based,
  distributed OS embedded on NuBus cards
• Created to compensate for the lack of processing
  power in the Macintosh Operation System

13
A/ROSEFeatures

• Minimal
• kernel is only 6K
• complete configuration utilizes 23K
• uses 48K buffer
• Multitasking
• pre-emptive multitasking
• round-robin task scheduling
• 32 priority levels

14
A/ROSEFeatures

• Real-Time
• 110 ms context switch time
• 20 ms of latency (guaranteed interrupt response
  time)
• Distributed
• OS is independent on each card
• Message-Based
• tasks communicate to other tasks on same card, or
  other cards across the NuBus
• messages can pass data such as pointers
• thousands of messages can be sent per second

15
A/ROSEUses

• A/ROSE was designed for
• communications and networking
• data acquisition
• signal-processing
• heavy-duty processing
• time consuming or time critical tasks

16
Additional Reading/Interesting Links

• Macintosh NuBus interface code (written in c for
  Linux)
• http//pc1.peanuts.gr.jp/kei/Kernel-Snapshot/lin
  ux/drivers/nubus/nubus.c
• NuBus physical design notes
• http//developer.apple.com/technotes/hw/pdf/hw_17
  .pdf
• NuBus in relation to real-time
• http//developer.apple.com/technotes/hw/pdf/hw_16
  .pdf
• NuBus (IEEE 1196) vs. PCI
• http//developer.apple.com/documentation/Hardware
  /DeviceManagers/pci_srvcs/pci_cards_drivers/PCI_BO
  OK.7.htmlpgfId3317
• DMA Fundamentals on Various PC Platforms
• http//cires.colorado.edu/jjose/DMAFundamentals.
  pdf

17
Sources

• Heid, Jim and Peter Norton. Inside the Apple
  Macintosh. Brady New York. 1989.
• Donald Koscheka, Ernst Young. Exploring NuBus
  Slots. lthttp//www.mactech.com/articles/mactech/
  Vol.06/06.04/NubusSlots/gt.
• Maurer, Joseph. Inside The Macintosh
  Coprocessor Platform And A/ROSE.
  lthttp//www.mactech.com/articles/develop/issue_04/
  coprocessor.htmlgt.
• University of Teesside School of Computing. The
  Bus. lthttp//wheelie.tees.ac.uk/users/a.clements
  /Buses/Buses.htmgt.
• Wikipedia. Nubus. March 2004.
• lthttp//en.wikipedia.org/wiki/NuBusgt.
• Harvey, A.F. DMA Fundamentals on Various PC
  Platforms. lthttp//cires.colorado.edu/jjose/DMA
  Fundamentals.pdfgt.

18
Oh, The Mac II MemoriesOregon Trail!