Comments by "GH1618" (@GH-oi2jf) on "Computerphile" channel.

  1. He gets off to a bad start. The alternative to “floating point” is not “integer” but “fixed point.” It would be better if he got that right at the beginning.
    3
  2. You should mention that software implementations can be accelerated by generating multiple bits in one iteration using a lookup table. For example, if you need pseudorandom 8-bit values, you might use a 32-bit generator, but shift out 8 bits and look up the next value for the xor in a 256-word table.
    3
  3. JJ K — He is about as famous as one can be.
    3
  4. The easiest way is to get a table of primitive polynomials that is published in an academic journal. That’s what I did, long ago, but I’ve forgotten where it came from. I might have a reference to it somewhere in my poorly organized files.
    2
  5. In general, an n-bit generator of this type will have a period of 2^n-1, with all zero being unreachable. Getting the maximum period requires using the correct polynomial.
    2
  6. Sure, but there are many uses for a pseudorandom bit generator.
    2
  7. Any way you like. If you want a different sequence every time you use it, you can just put a clock value in it at the start.
    2
  8. Correct. But if you use, say, a 32-bit generator, the period will be 2^32-1 (if properly implemented). That is probably much longer than you need, so you just have to randomize your starting seed.
    2
  9. “C” was influential beyond all reason. It has been a bad influence, in my opinion, despite its great success.
    2
  10. The purpose is to generate long strings of pseudorandom bits, which have many applications.
    2
  11. I’m pleased to see the ABC mentioned here, along with other early computers. One thing I like about Atanasoff was his view on priority. He said "I have always taken the position that there is enough credit for everyone in the invention and development of the electronic computer." That’s the right spirit. Remember people for their individual contributions without arguing about who should be considered the inventor.
    2
  12.  @Sh-hg8kf  — Computers are implemented with binary logic, so we use binary polynomials for this algorithm. The coefficients are always zero or one, representing the absence or presence of that power of x.
    2
  13. kasuha — True for ASCII, but nit Baudot-Murray.
    1
  14. Any number can be constructed from bits.
    1
  15. By the way, they are “Hollerith cards.” Perhaps they don’t use that term in the UK because Herman Hollerith was Usanian. The Hollerith card predates the ASCII character set. Most computers which used card input predominately did not use ASCII characters.
    1
  16. Pseudorandom numbers are useful.
    1
  17. methanbreather — “von Neumann architecture” just means that one memory device contains both instructions and data. That principle originates with the draft report on the EDVAC.
    1
  18. Edison (poor choice of name, however)
    1
  19. True randomness isn’t that important.
    1
  20. Gego/XAREN ゲゴザレン — “Based on” is too strong. The designer of ENIAC did meet Atanasoff and got some ideas from him, but it is not an elaboration of ABC.
    1
  21. You need a primitive irreducible polynomial of the right order for the number of bits.
    1
  22. Philip Stuckey — Von Neumann created game theory. It is not about cheating at poker.
    1
  23. This technique is based on division of primitive polynomials. A primitive polynomial is comparable to a prime number.
    1
  24. About a third of my computer career was spent working with one’s complement machines. They worked well. The extra zero was not a big deal. The hardware took care of it.
    1
  25. As you shift, you take pseudorandom bits off the upper end of the shift register. One iteration gives you one bit.
    1
  26. You will tend to get lots of strings of zero bits at the beginning, but it won’t take long to get past that. You can use a random seed to start at soewhere in the pattern beyond the beginning, then you won’t see much of that.
    1