go to Xputer pages

homepage | impressum |  survey | 

The E.I.S. page
(summary only)

TU Kaiserslautern

 Karlsruhe Institute of Technology (KIT) homepageInstitut für Technik der Informationsverarbeitung (ITIV) des Karlsruher Institut für Technologie (KIT)

Prof. Dr.-Ing. Reiner Hartenstein

Homepage | Unversity page | CV |     Offices: Kaiserslautern | Karlsruhe      |  Impressum

IEEE fellow
SDPS fellow
FPL fellow      

Reiner Hartenstein

 

Founder of the "E.I.S.-Projekt", the multi-university introduction of the Mead-&-Conway revolution: first on the entire continent (Europe & Asia) - incubator of the EU-funded EUROCHIP action still running to-day worldwide. "E.I.S." = "Entwurf Integrierter Schaltungen" (design of integrated circuits).

 

Wikipedia on the E.I.S. Projekt (in German): click here.   

Reinventing Computer Engineering: Zur Geschichte (German language) des E.I.S.-Projekt  - in English language: E.I.S. project:    http://www.fpl.uni-kl.de/staff/hartenstein/eishistory_en.html      

Early EDA innovations by Lynn Conway's Lmbda Notation: http://hartenstein.de/Hartenstein-EDA-Innov-Europe-3.pdf.       the KARL language

Other Projects: PATMOS project, CVT projectCVS project      E-I.S. Projekt

Again Reinvent Computer Engineering  |  European EDA industry killed

next FPL conference:

http://fpl.org

next PATMOS conference

The Mead-&-Conway Revolution: Reinventing Computer Science

Throughout the 70ies (and earlier) each microchip fab line had different layout design rules, mask polarities, alignment marks, process test patterns, scribe lines and more, and, it was nearly impossible for outsiders to access such prototyping [Lynn Conway]. The local design rules have been books of about 50 pages or more. All product designs could be implemented only by experts of the fab line company. It was clear that commercially viable chips would contain several million transistors by the early 1990s. Here Carver Mead coined the term "Moore's Law". In 1976 a major DARPA study of the basic limitations of microelectronics fabrication urgently recommended research into the system design implications of “very-large-scale integrated circuits” – pointing out that no methods existed for coping with such complexity [1].

In the early 80ies the Mead-&-Conway revolution has changed this. Making use of the new self-aligned silicon-gate technology coming along with MOS replacing bipolar, Lynn Conway, joining PARC in 1973, invented scalable and massively simplified MOS design rules on only 3-4 pages, replacing a book with more than 40 pages. Suddenly a clean separation between chip design and fabrication was possible, and Carver Mead coined the term “silicon foundry” for manufacturing-only firms which ‘print’ externally generated designs, meanwhile a market of more than 30 billion US-dollars. With the MPC79 action Lynn Conway pioneered the cooperation between prototyping designers and the silicon foundry by merging 19 designs into a single multi-project chip (MPC), by solving all the mask making data format translation problems, and interfacing all designers to this service via the internet (at that time the ARPANET).

The Mead-&-Conway revolution triggered reinventing computer science. Lynn Conway wrote, that a huge and previously-unknown territory for creative architectural innovation had opened up [2]. The new approach enabled highly simplified methods for silicon chip design enabling architects to design systems from top-to-bottom throughout the hierarchy of levels. Carver Mead coined the phrase “Tall Thin Man” (my phrase “Tall Thin Designer” should also include the “Tall Thin Woman”). This phrase illustrates, that now each person can work successfully through several abstraction levels. This was the birth of the microchip design as its own

 discipline, independent of the fabrication technology discipline. Within about 2 years about 120 universities worldwide started reinventing their CS and EECS curricula [3] by establishing Mead-&-Conway courses and starting many new research areas [4] such as computing systems research and EDA research, a climate also for the birth and healthy growth of the EDA industry. The multi-university “E.I.S.-Projekt” [5] was the very early German contribution, the first on the “continent” (Europe and Asia).

To Reinvent Computing again? Now, after 30 years? Why ?

Because of the Mead-&-Conway revolution based on Moore's Law software engineering got more performance by just waiting for the next faster microprocessor. So there was not much motivation to reinvent software engineering, since there was also no paradigm shift away from the dominance of the von Neumann model. This has changed in 2004 by a disruptive strategy shift of the microprocessor industry from ever faster single processors to manycore chips. Industry dumped its power problem on the software: impossible to be solved in a reasonable time. Most code is “stiff ware” (legacy code) so that extensive changes are extremely difficult. After reinventing hardware design we now we are forced to reinvent software engineering. In contrast to the Mead-&-Conway times, they do not like to hear this message. Now not only supercomputer programmers, but they all have to learn, how to master parallelism in programming and system architecture.

We need to bridge the hardware / software chasm

In addition to Moore's Law we now have to watch "Coore's Law" Because manycore will remove only a smaller part of the power wall we are also forced to move over to heterogeneous systems also including accelerators, such as for instance,  FPGAs. We again have to reinvent computing, this time also for bridging the hardware / software chasm [7]. Also see keynotes about this topic area [9].

Reinventing Computing several times [8]

The first electrical computer produced in series was ready in 1884 - reinvented after historic mechanical computers. The ENIAC, ready around 1945, again was a reinvented computer: a paradigm shift from data-driven to instruction-driven (from anti-machine to von Neumann paradigm). The introduction of Reconfigurable Computing also means reinventing - back to data-driven, but implemented by microelectronics.

Literature

[1] I. Sutherland, C. Mead and T. E. Everhart, “Basic Limitations in Microcircuit Fabrication Technology”, R-1956-ARPA, Nov. 1976. http://www.dtic.mil/cgi-bin/GetTRDoc?AD=ADA035149 (mirror)

[2] Lynn Conway: The VLSI Archive: An online archive of documents and artifacts from the Mead-Conway VLSI design revolution; http://ai.eecs.umich.edu/people/conway/VLSI/VLSIarchive.html

[3] I. Sutherland and C. Mead, “Microelectronics and Computer Science”, Scientific American, Nov. 1977 http://ai.eecs.umich.edu/people/conway/VLSI/BackgroundContext/Sciam/SM.SciAm77.pdf (mirror)

[4] Committee on Innovations in Computing and Communications, National Research Council (editors): Funding a Revolution (Lessons from History); National Academy Press in 1999, http://ai.eecs.umich.edu/people/conway/Impact/FundingaRevolution.html (mirror)

E-I.S. Projekt  [5] Reiner Hartenstein: Early EDA Innovations in Europe driven by Lynn Conway’s Lambda Notation; http://hartenstein.de/EIS/EDA-Innovation-Europe.pdf 

[6] Paul McLellan: On-chip supercomputers, AMBA 4, Coore's law; 07-11-2011, Semiwiki http://www.semiwiki.com/forum/content/624-chip-supercomputers-amba-4.html (mirror)

[7] Burton Smith: Reinventing Computing; 7th Symposium of the Los Alamos Computer Science Institute, October 17-19, 2006, Santa Fe, New Mexico. http://www.cct.lsu.edu/~estrabd/LACSI2006/Smith.pdf (mirror)

[8] Reiner Hartenstein: The Reinvent computing page; http://hartenstein.de/EIS2/

[9] Reiner Hartenstein: keynotes:  http://hartenstein.de/keynotes.htm

[10] Xputer-related Literature  http://xputer.de/Hartenstein-Xputer-related-Literature_2.pdf