[Librecores Discussion] Discussing EDSAC museum on FPGA for GSoC project with FOSSi

Jeremy Bennett jeremy.bennett at embecosm.com
Fri Mar 10 20:51:39 CET 2017


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 09/03/17 18:28, Hatim Kanchwala wrote:
> Hello!
> 
> I hope this finds you well. I am Hatim, an undergraduate in my
> pre-final year pursuing Bachelors in Electrical Engineering. My
> academic interests lie in the computer organisation and
> architecture realm, and I am looking forward to taking up higher
> education in the said domain. I am keenly interested in the project
> titled "First-generation Museum on FPGA" with FOSSi under GSoC. I
> am interested in replicating the EDSAC on an FPGA.

Hi Hatim,

Good to hear. I am one of the mentors on this project.

> Over the other first-generation computers, I prefer to work on
> EDSAC mainly because replicating it seems rather challenging as
> exhaustive documentation on it is unavailable. It is widely known
> that the original team working on the EDSAC tweaked the designs
> multiple times. It is this challenge at the heart of this GSoC
> project that excites me!

This wasn't explicit in the project summary, but it is certainly an
area we are interested in.

> Here are a couple of my thoughts and ideas on the project - 1.
> Going with a top-down approach, here is a preliminary breakdown of 
> work - the project will mainly involve studying the EDSAC from
> existing documentation and breaking down the architecture into
> independent logical subdivisions (just like the EDSAC itself is
> divided up into 120 panels performing specific functions). We will
> then need to identify the data flow and circuits to build
> abstractions through Verilog modules (for instance, identifying
> that a circuit behaves like a shift register or a counter, etc.).
> Once the known circuitry is mapped onto proper modules, we will
> have to fill in the gaps arising out of missing documentation. This
> will be an iterative process and after each cycle, we will move
> closer and closer to replicating the EDSAC in its entirety. A more
> long-lasting task would be to iron out the anomalies arising out of
> corner cases (which, IMHO, are notorious for making themselves 
> visible only in tests). Please let me know your thoughts/feedback
> on the above approach.

Certainly a good approach. One issue we have discussed on this list is
what to use as proxies for external interfaces, such as the mercury
delay lines, tape reader/writer and switches and lights. One can
certainly just emulate them in FPGA and/or software, but perhaps a
low-cost physical approach would help bring the project to life.
Possibly a low cost thermal printer and business card reader could
make a proxy for a paper tape writer/reader.

The suggested low cost FPGA board (MyStorm) has plenty of expansion
capability, making this feasible.

> 2. This project aims to help hobbyists and programmers learn about
> the inner workings of computers, trace our computer heritage, and
> to "tangibly" see how far we have come from vacuum tube computing
> era to the current supercomputing era. To this end, one of the main
> things we will have to focus on is - documentation. The project
> will be a complete learning resource only when we have both the
> code and the supporting documentation.

Absolutely - this is the key reason for this project. To make
important historical computers accessible for the current generation
of computer engineers.

> Over the coming days, I will continue to work on evolving (1) and
> (2) into a more concrete proposal.

We look forward to it.

> 3. The National Museum of Computing in Bletchley Park (UK) is
> currently working on reconstructing the EDSAC as it was in 1951,
> and should have it operational by late 2017. TNMOC also harbours
> the Harwell Dekatron/WITCH, which was rebooted in 2012. A senior of
> mine was working on building a WITCH emulator written in C. Back
> then I got a chance to study the WITCH and I even made some
> contributions to his emulator work. Those of us (like me) who have
> an academic interest in computers can enjoy playing around in the
> FPGA museum in lieu of visiting Bletchley Park!

We are very familiar with this, and the UK Computer Conservation
Society are aware of what we are proposing here. The rebuild of EDSAC
is a faithful reconstruction of the physical machine. There are also
pure software emulations, one by Martin Campbell-Kelly at the
University of Warwick and one to celebrate EDSAC's 60th anniversary by
Martin Richards at the University of Cambridge Computer Lab.

We are aiming for something in between, a "re-imagining" of EDSAC
using modern hardware technology, to bring history alive for a new
generation.

> 4. This is an idea - how about instead of using Verilog, we use
> Chisel? I like the idea of having an open-source hardware
> construction language powering an open-source effort! I am
> experienced with Chisel only at a basic level, but will be picking
> it up over the next 2-3 weeks. What do you think?

Chisel, while open source, is still relatively new. Outside Cambridge
and Berkeley it is not yet widely used compared to Verilog and VHDL.
It consequently is still relatively thin on good open source tools.
By using Verilog the project achieves the widest usage.

But this is not a hard rule - we are looking for proposals that have
thought through these issues. What is going to be the best way to make
this work accessible to the widest community around the world.

> A little background about myself - I am experienced with FPGAs and
> Verilog HDL. Under the guidance of my VLSI Design course
> instructor, I am currently working on a project to prototype the
> Intel 8085 microprocessor on a Xilinx Spartan-3E FPGA, using
> Verilog HDL. I'll also be adding basic pipelining to the prototype 
> in a later stage. My objective is to investigate the performance
> delta due to the addition of pipelining. I will be open-sourcing
> all of the code generated as part of this project once I pass
> evaluation at my University. In an earlier course, I gained
> exposure to developing full-custom designs and simulating them
> (couldn't fabricate any design as my University does not have the
> requisite equipment). If you wish to peruse my profile further,
> please have a look at my LinkedIn profile [1], at my CV [2], or at
> my GitHub profile [3]. I'd be glad to answer any questions you
> have.

Sounds good. I look forward to seeing your completed proposal.
> 
> Looking forward to hearing from you and discussing things further.
> Will let you know of any updates I receive from TNMOC. Thank you
> for your time.
> 
> Have a good day!

You too!

Best wishes,


Jeremy
> 
> Hatim Kanchwala Undergraduate, Electrical Engineering Indian
> Institute of Technology Patna hatim at hatimak.me
> <mailto:hatim at hatimak.me>, hatim.ee14 at iitp.ac.in 
> <mailto:hatim.ee14 at iitp.ac.in>
> 
> References - [1] https://www.linkedin.com/in/hatimak 
> <http://s.bl-1.com/h/CWhFS5s?url=https://www.linkedin.com/in/hatimak>
>
> 
[2] http://hatimak.me/CV.pdf
> <http://s.bl-1.com/h/CWhFXVv?url=http://hatimak.me/CV.pdf> [3]
> http://github.com/hatimak 
> <http://s.bl-1.com/h/CWhGdtx?url=http://github.com/hatimak>
> 


- -- 
Tel:     +44 (1590) 610184
Cell:    +44 (7970) 676050
SkypeID: jeremybennett
Twitter: @jeremypbennett
Email:   jeremy.bennett at embecosm.com
Web:     www.embecosm.com
PGP key: 1024D/BEF58172FB4754E1 2009-03-20
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2

iEYEARECAAYFAljDA8sACgkQvvWBcvtHVOGwogCgjkeNLEl8rNbeWaYc7bDRoHFj
mpkAn1KL3DUPTcO3js7AzGneNeYmVYXl
=2i5v
-----END PGP SIGNATURE-----


More information about the Discussion mailing list