Text preview for : A_Response_to_the_Star_LSI_Workstation_Goals_Ver_1.0_Nov78.pdf part of xerox A Response to the Star LSI Workstation Goals Ver 1.0 Nov78 xerox sdd memos_1978 A_Response_to_the_Star_LSI_Workstation_Goals_Ver_1.0_Nov78.pdf
Back to : A_Response_to_the_Star_LS | Home
Xerox
Private
Data
A Response to the
Star LSI Workstation Goals
Version 1.0
November 1978
Draft of 1:00 P.M. November 3,1978
The Star LSI Workstation Goals document sets guidelines for the implementation of a low
cost workstation (terminal, processor, and peripherals) for the OIS Star product. This
document responds to those goals by the:
1. Enumeration and brief analysis of several alternative technical approaches.
2. Elaboration and analysis of the Wildflower approach.
Although Wildflower, as described in the documents currently available, will not be adequate
for the workstation design; we find that a the use of a bipolar microprocessor which is
carefully synchronized to the memory and I/O controllers does offer the best approach to
the workstation implementation. .
Version 1.0 represents current thinking concerning the design of the Star LSI Workstation.
The reader is encouraged to send comments to Bob Belleville (Belleville @ ParcMaxc or
8*923*4520); This file is stored on [IRIS] LsiWS 1.0 > LsiWS-Part1.press
and Part2.press. .
. Prepared by: Robert L BellevnIe
Ronald C. Crane
Robert B. Garner
J. Pitts Jarvis
E. A. Miller
RoyC.Ogus
Stephen c. Purcell
C. Richard Snow
Approved by Robert M. Metcalfe
SDD System Architecture
Approved by David E. Liddle
SDD
XEROX
Systems Dcvelopmcnt Department
3408 Hillview Avenue
Palo Alto. Califomia 94304
This document is Xerox Private Data.
Xerox
Private
Data
'8'
69. X~rox
~'lf~n !? to
.
the Star LSI Workstation Goals -. November 1978
Table of Contents
1. Introduction, Conclusions, and Rccomendations 1
2. Mandatory References 3
3. Informational References 3
4. Assumptions 4
5. The Alternatives 5
6. A Synchronous Multi-Tasking Microprocessor 10
6.1 Introduction 10
6.2 The Mesa Emulator 15
6.3 Full Page Display 17
6.4 SAlOOO/ SA4000 Rigid Disk 18
6.5 Xerox Wire 20
6.6 Low Speed Peripheral Bus 26
6.7 Performance Estimates 32
6.8 Packaging/EME/Environmental 34
7. Development Tools 35
8. Field and Factory Diagnostics 39
9. 01S Architectural Impact and Risks 41
10. Development Schedule 43
11. Cost Estimate 44 .
Appendix
A. Microcode 45
B. Propogation Delay Estimates 51
69. Xerox
"oil- Private
"""'fjy, Data
69. Xerox
A Response to the Star LSI Workstation Goals - November 1978 "0111Private
""I/fj" Data
A Response to tile
Star LSI Workstation Goals
1. Intro(luction, Conclusions, and Recomendations
The Star LSI Workstation Goals document [reference 1] sets guidelines for the implementation of a
low cost workstation (terminal. processor. and peripherals) for the OIS Star product. This document
responds to those goals by the:
1. Enumeration and brief analysis of several ~ternative technical approaches.
2. Elaboration and analysis of the Wildflower approach.
This document was written during three weeks of intensive study by a team of 8 engineers. It
should not be represented as an exhaustive examination of all the available alternatives, but rather,
a careful look at the most promising ones.
1.1 Introduction
Wildflower [reference 2] is a high performance microprocessor implemented with the 2901. bipolar.
4-bit cpu slice. Like the Alto and the DO. Wildflower shares this processor among the peripheral
controllers to realize simple and compact controllers. Wildflower simplifies the approach taken in
the Alto and DO in two crucial ways:
1. Wildflower places a strict limit on the number and kind of peripherals it can support.
Specifically, only four are allowed - full page display. rigid disk. Xerox Wire, and low
speed microprocessor bus which handles serial communication, auxiliary media.
printers, and floppy disk.
2. The Wildflower processor is shared among these peripheral controllers in a strict time
synchronous way. not simply on demand as in the Alto and DO. l11is allows each
controller to be designed with a known 110 latency. As we shall see, this results in a
dramatic simplification of the entire system over the Alto or the DO; however. the
performance is comparable to the DO in the execution of Mesa bytecodes.
It is likely that a synchronous processor like Wildflower would result in the lowest cost workstation;
however. this type of design leaves little flexibility for either configuration changes or for variations
in the specifications or availability of integrated circuit components. This increases the risk in
conjunction with a tight development schedule.
A more traditional bus organized microprocessor design is presented in chapter 5 as the primary
alternative to a Wildflower-like approach. The higher estimated cost of this apprexlch is offset by
increased flexibility in configuration and in the fact that individual components can be defined and
developed in parallel during the development cyele.
1.2 Conclusions
Chapter 5 introduces five workstation alternatives:
o Cost-reduced Alto
o Cost-reduced DO
o Custom LSI
o Bus organized microprocessor
o Synchronous multi-tasking microprocessor
XerOl(
Private
Data
6~ Xerox
11011 Private .
./J9 ..et/t4SPOf e to the Star LSI Workstation Goals ~ November 1978 2
The number of modifications needed to efficiently implement the Mesa/Pilot environment on a
cost-reduced Alto eliminate this as a viable approach.
There is little that can be done to dramatically reduce the cost of the DO because of its complex
controllers, processor. and memory system.
Custom LSI, built within Xerox or outside. requires a fully tested architecture consistent with the
LSI technology. which we do not now have, and more lead time than we have for the LSI
workstation IMO. ED's Standard Silicon Bus looks very attractive; however, we have no experience
with the use of this technology for the construction of an OIS processor.
The cost estimates in section 11 and the cost figures in the Star LSI Workstation Goals show that
only about $1100 can be used for the processor and the peripheral controllers. This includes the
cost of the PWB, connectors, and and memory. These costs are essentially the same in either a MSI
or LSI implementation. At most LSI implementation can save a few hundred dollars in the total
workstaion UMC.
The bus-organized microprocessor is attractive. However. the use of even the most advanced 16-bit
monolithic microprocessor would require extensive changes to the Mesa development environment
which would be both time-consuming to develop and difficult to maintain.
The use of a custom-designed bipolar .microprocessor is essential to the efficient use of the
Mesa/Pilot environment. Careful examination of the use of this class of processor in the bus
organized approach results in the observation that it is the synchronous relationship of the memory.
processor, and I/O that gives the most attractive design.
The multi-tasking synchronous processor SUMIT divides the use of- the high performance
microprocessor amoung the four I/O devices (disk. display, X-wire. and low speed bus) to realize
simple. low chip count controllers. The type of processor is also well suited to the execution of the
Mesa byte stream.
1.3 RecomendatioDS
So, we conclude that a bipolar microprocessor which is relatively tightly coupled to the the I/O
controllers is the only viable design alternative for the implementation of the Star Workstation. A
processor like SUMIT is the most likely to meet both the cost and schedule constraints.
Furthermore, we recommend that the design of such a processor begin within SDD as soon as
possible so that the proposed schedules can be met.
,
With respect to the implementation of the workstation with LSI technology, we conclude that
because longer schedules will be needed we should start now to develop such a machine. However,
since the cost of the workstation is dominated by the peripherals and the cabinetry, the UMC cost
reduction of such a station will not be very dramatic.
Xerox
Private
Data
69. )(erox
A Response to the Star LSI Workstation Goals -. November 1978 "Oll~rivate
"fI/Dy, Data
2. Mandatory References
[1] Ruebel, R. et. at, Star LSI Workstation Goals, Xerox Internal Memo, 11 September 1977.
[2] Lampson, B. W., Wildflower Manual (Draft), Xerox Palo Alto Research Center, 18 August
1978. Filed on [IVY] WFManua1.Press and [IVY] WFDrawings.Press.
3. Informational References
These background references provides more depth on many of the subjects covered in this
document.
[31 The OIS Processor Principles of Operation, Systems Development Department, April 9, 1977.
[4] ALTO: A Personal Computer System Hardware Manual, Xerox Palo Alto Research Center,
February 1978.
[5] DO Processor Functional Specificaton, Electronic Operations Department, January 16, 1978.
[6] Ogus, Roy C., Xerox Wire Specification, Systems Development Department, April 5, 1978.
[7] Microprocessor Data Manual II, Electronic Design 21, October 11, 1978.
[8] Crane, R. C., Star Grounding, Systems Development Department. September 13, 1978.
[9] Wick, J., DO Timing Simulation, Systems Development Department, October 18, 1978.
69. Xerox
"Oil
"fI/Dy,
Private
Data
6W. Xerox
.