Tomasulo

1
Tomasulos Algorithm and IBM 360

• Srivathsan Soundararajan

2
What we have seen till now!!!

• Single-cycle datapath.
• Multi-cycle datapath.
• N-stage Pipelined datapath.

3
Tomasulo's algorithm

• A hardware algorithm for controlling the
  execution of multiple functional units with
  varying latencies in a pipelined CPU
  micro-architecture.
• A general mechanism for register forwarding and
  hazard detection.
• The key idea is to virtually execute each
  instruction in a single cycle.
• Out-of-order execution of instructions.

4
So what is so special???

• Let instructions behind stall proceed.
• Decode instructions and check for structural
  hazard.
• Wait until no data hazard and then read operands.

5
Three Stages

• Issue if reservation station free (i.e. no
  structural hazard), control issues instruction
  and sends operands (renames registers).
• Execution if both operands ready, then execute.
  If not, watch common data bus for result.
• Write result if CDB available, write on common
  data bus to all awaiting units mark reservation
  status available

6
Virtual result a promissory note

• All registers are modified so that they can
  either hold a true result or a virtual result.
• When an instruction is issued, a virtual result
  is placed in the instruction's destination
  register.
• A functional unit is assigned to compute the real
  result.
• The virtual result is replaced by the real result
  when the functional unit has completed its
  computation.

7
Tomasulo Organization
8
The concept

• Each instruction, as it arrives, fetches its
  operands from a special register file.
• Each register in this file holds either an actual
  value, or a tag indicating the reservation
  station that will produce the register value when
  it completes.
• The instruction and its operands (either values
  or tags) are stored in a reservation station
  (RS).
• The RS watches the results returning from the
  execution pipelines, and when a result's tag
  matches one of its operands, it records the value
  in place of the tag

9
A good example

• http//www.cs.umd.edu/users/saltz/cmsc411-s97/toma
  sulo.htm

10
Why Tomasulos Algorithm

• Hazard detection.
• Dynamic Scheduling (i.e. Hardware reorganizes
  instructions)

11
IBM 360

• The IBM 360 introduced many new concepts,
  including dynamic detection of memory hazards,
  generalized forwarding, and reservation stations.
• The approach is normally named Tomasulos
  algorithm

12
Installation of the IBM 360/91 in the Columbia
Computer Center machine room in February or March
1969
Photo AIS archive.
13
IBM 360

• Was introduced by the team led by Michael Flynn
  in 1966.
• The internal organization of the 360/91 shares
  many features with the Pentium III and Pentium 4,
  as well as several other microprocessors.
• One major difference was that there was no branch
  prediction in the 360/91 and hence no
  speculation.
• Another major difference was that there was no
  commit unit, so once the instructions finished
  execution, they updated the registers.
  Out-of-order instruction commit led to imprecise
  interrupts, which proved to be unpopular and led
  to the commit units in dynamically scheduled
  pipelined processors since that time.

14
IBM 360

• Although the 360/91 was not a success, the key
  ideas were resurrected later and exist in some
  form in the majority of microprocessors like
  Pentium II, Power PC 604 etc...
• It ran under Operating System/360 -- a powerful
  programming package of approximately 1.5 million
  instructions that enabled the system to operate
  with virtually no manual intervention.

15
IBM 360

• Within the central processing unit (CPU), there
  were five highly autonomous execution units which
  allowed the machine to overlap operations and
  process many instructions simultaneously.
• The five units were processor storage, storage
  bus control, instruction processor, fixed-point
  processor and floating-point processor. Not only
  could these units operate concurrently, they
  could also perform several functions at the same
  time.

16
Some uses

• The IBM-360 family of computers ranged from the
  model 20 minicomputer (which typically had 24 KB
  of memory) to the model 91 supercomputer which
  was built for the North American missile defense
  system.

17
Quite huge isnt it???
18
References

• http//www.d.umn.edu/gshute/cs2521/arch/tomasulo.
  html
• http//euler.ecs.umass.edu/arch/parts/Part6-tomasu
  lo.pdf
• http//www.cs.indiana.edu/classes/p415-sjoh/readin
  gs/smv/CadenceSMV-docs/smv/tutorial/node36.htmlto
  masulo1
• http//www.beagle-ears.com/lars/engineer/comphist/
  ibm360.htm
• http//www.eecs.ucf.edu/lboloni/Teaching/EEL5708_
  2006/slides/Tomasulo.ppt268,9,Three Stages of
  Tomasulo Algorithm
• http//www.cs.unc.edu/montek/teaching/fall-03/lec
  tures/lecture-11.ppt723,4,Tomasulo Organization
• http//www.ece.cmu.edu/jhoe/distribution/2005/741
  /proj3.pdf
• http//www.columbia.edu/acis/history/36091.html