First Draft of a Report on the EDVAC

John von Neumann's 1945

Below is a copy of John von Neumann's EDVAC Report, in 150dpi image form. For some reason I was unable to find a copy of this famous paper anywhere on the net, which is both a shame and a shock, considering it's importance. This copy was taken from a book owned by the Computer History Museum, who kindly let me photocopy it. OCR sounds reasonable enough, but all I have are photocopies, and the paper is filled with un-OCR-able symbols, and right now I have no time to deal with all this. Volunteers accepted.

It may not be an easy read, if you can't shed modern pre- (or is it post-) conceptions. Nearly the only "modern" concept taken for granted in this 1945 paper is that orders (instructions) and data would be stored in the same memory -- and even that wasn't quite certain; in this report JvN made it clear that orders and data would be distinguished by a special digit (bit), presumably so that the hardware could tell the difference. (This was abandoned by the time any machines were built.) The paper also fully embraces the graphical terminology of the McCulloch-Pitts brain-as-assembly-of-logical-neurons theory, part of the Cybernetic fad. McCulloch-Pitts was all the rage in electronic-computer circles of the time (and in fact mutated into the AND and OR notation used today); many things from this historical period is steeped in it.

As Martin Davis writes in THE UNIVERSAL COMPUTER, JvN describes his computer in terms of it's logical design -- as opposed to the metal-head approach of Mauchly and Eckert. Even the most rudimentary glance at the design of ENIAC makes it clear that it was, even in 1946, an utter piece of shit, about 10 times too many tubes for the job, a truly terrible design even by then-current standards. Don't get me started. (For the record, ENIAC was not a stored-program computer, it was a programmable calculator. It was a phenomenonally important machine, but it was NOT a stored-program computer, thank you, regardless of von Neumann's incredibly clever hack to allow it to programmatically modify program, and in spite of H.H. Goldstine's disingenous writings (for example, in his 1972 THE COMPUTER FROM PASCAL TO VON NEUMANN he relegates the entire British history of computing to an appendix! when in fact the first TWO actually-operating stored program digital computers were british, not american (Manchester Baby, aka SSEM, a test hack; and EDSAC). I guess I got myself started. I'll shut up now.

Anyways, the EDVAC Report is a brilliant paper. It was published a few months before Turing's ACE Report, and is therefore taken as gospel that Turing got all his ideas from JvN. In fact the concept of "stored program digital computer" was organically known by nearly everyone involved in automatic computing, and further, during the war there was much direct contact, and JvN acknowledged he had read Turing's 1936 COMPUTABLE NUMBERS paper and knew about the work at Bletchley Park, etc. In other words, they all knew each other's business.

Turing's ACE Report is even more amazing, in fact (and at the same time it illuminates Turing's worst follies). Though it wasn't publically known until the 1970's when Brian Randell seems to have got the British to declassify some of the Bletchley Park work, Turing had a big head start on the design of a practical computer. In fact by 1946 he was writing software libraries, floating point code, etc. (Though JvN is considered to have written "the first [explicit] computer program" (hmm, ask Zuse; but even 19th century jaquard loom programmers were thinking sequentially and algorithmically, so what the hell does this mean anyways, (for that personally I would ask Thomas Kuhn, who in fact has a lot to say about this sort of thing (and these parentheticals are bad enough to put my sanity in doubt, no? and what sort of person reads them, hmm?))), a sort routine that was supposed to have accompanied the EDVAC Report (in which he contradictorally dismisses non-mathematical uses of the computer, and in later years would dismiss automatic programming (aka compilers) as a waste of time); alas, it appears to be the first instance of vaporware, as there seems to be no record of it, though Knuth says he saw it, which is probably good enough for anyone.) Turing was also his own worst enemy when it came to getting his projects taken seriously, and the ACE project as he envisioned it never came to be. ACE contained critical features that got excised by mindless bureaucrats utterly lacking in vision (and no tolerance for Turing's unfortunate lack of finesse), such features not to appear generally until a decade later (and now utterly unthinkable without).

>inhale< But that's another story for another time.

Tom Jennings, March 2001




Page 177

Page 178

Page 179

Page 180

Page 181

Page 182

Page 183

Page 184

Page 185

Page 186

Page 187

Page 188

Page 189

Page 190

Page 191

Page 192

Page 193

Page 194

Page 195

Page 196

Page 197

Page 198

Page 199

Page 200

Page 201

Page 202

Page 203

Page 204

Page 205

Page 206

Page 207

Page 208

Page 209

Page 210

Page 211

Page 212

Page 213

Page 214

Page 215

Page 216

Page 217

Page 218

Page 219

Page 220

Page 221

Page 222

Page 223

Page 224

Page 225

Page 226

Page 227

Page 228

Page 229

Page 230

Page 231

Page 232

Page 233

Page 234

Page 235

Page 236

Page 237

Page 238

Page 239

Page 240

Page 241

Page 242

Page 243

Page 244

Page 245

Page 246