1. Forum
  2. Usenet
  3. COMP.ARCH

  • Vanished Post (Cray-I Basic Gates)

    From Quadibloc@21:1/5 to All on Fri Feb 23 17:55:39 2024
    I saw this post by Thomas Koening:

    Reading literature on the Cray-I, one finds that their logic was
    implemented in ECL using 5/4 NAND gates.

    Wikipedia claims that these were, in fact, XOR gates, but the only
    source to that claim is a link to a datasheet, without a source
    stating that this chip was indeed used in the Cray-1, so I am
    disinclined to believe that (and I may just remove it as being
    unsourced, and in contradiction to published literature).

    The pictures of circuit boards that I found that were big enough
    to read the markings only had the flip-flops and the static ROMs
    on them.

    So, does anybody have an idea of what they actually used (and
    maybe a source)?


    But when I tried to reply to it, my newsreader crashed, and this
    thread did not reload. (It got into Google Groups under the wire,
    so this is where I found the text.)

    I was going to reply - that his skepticism was quite right
    about Wikipedia's claim that it was XOR gates. You can build
    anything from enough NAND gates, including an XOR gate. But
    XOR gates can't be used to make AND and OR, so they can't be
    used as a starting building block.

    John Savard

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Lawrence D'Oliveiro@21:1/5 to Quadibloc on Fri Feb 23 23:06:45 2024
    On Fri, 23 Feb 2024 17:55:39 -0000 (UTC), Quadibloc wrote:

    I was going to reply - that his skepticism was quite right about
    Wikipedia's claim that it was XOR gates. You can build anything from
    enough NAND gates, including an XOR gate. But XOR gates can't be used to
    make AND and OR, so they can't be used as a starting building block.

    Or in short: you can build symmetry out of asymmetry, you can’t build asymmetry out of symmetry.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From David Schultz@21:1/5 to Quadibloc on Fri Feb 23 17:41:17 2024
    On 2/23/24 11:55 AM, Quadibloc wrote:

    I was going to reply - that his skepticism was quite right
    about Wikipedia's claim that it was XOR gates. You can build
    anything from enough NAND gates, including an XOR gate. But
    XOR gates can't be used to make AND and OR, so they can't be
    used as a starting building block.

    You can also build anything from NOR gates. Since the NOR is faster than
    NAND, I find it curious that NAND is used.

    Unless that emitter follower output and pull down/termination resistor
    on inputs is being used for wired-AND sorts of games. Probably as that
    would save a bunch of gates.

    I can't find anything resembling a schematic but I did find this:

    https://www.computerhistory.org/revolution/supercomputers/10/7/915

    A page full of sum of products.

    --
    http://davesrocketworks.com
    David Schultz

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Lawrence D'Oliveiro@21:1/5 to David Schultz on Fri Feb 23 23:47:31 2024
    On Fri, 23 Feb 2024 17:41:17 -0600, David Schultz wrote:

    You can also build anything from NOR gates. Since the NOR is faster than NAND, I find it curious that NAND is used.

    Could the bias be coming just from the notation used? The fact that,
    because of the adaptation of traditional mathematical operator precedences
    to Boolean algebra, it is easier to write a sum of products than a product
    of sums, even though the two formulations are exact duals?

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From MitchAlsup1@21:1/5 to Lawrence D'Oliveiro on Sat Feb 24 01:08:29 2024
    Lawrence D'Oliveiro wrote:

    On Fri, 23 Feb 2024 17:41:17 -0600, David Schultz wrote:

    You can also build anything from NOR gates. Since the NOR is faster than
    NAND, I find it curious that NAND is used.

    First of all we must re-visit deMorgans logic law:: although there are several the most useful one is: A NAND gate operating on true signals is a NOR gate operating on complement signals.

    In true signal logic (high = 1 low = 0)
    A NAND followed by a NAND is a large fan-in AND.
    A NAND followed by a NOR is a multiplexer.

    In complement logic (high = 0 low = 1)
    A NOR followed by NAND is a multiplexer
    A NOR followed by a NOR is a large fan-in OR

    So, unless you state the logic voltage (or current) levels we cannot tell
    if a NOR is faster than a NAND. In CMOS NANDs are faster, in ECL NORs are faster both under the higher voltage = 1 criterion.

    Could the bias be coming just from the notation used? The fact that,
    because of the adaptation of traditional mathematical operator precedences
    to Boolean algebra, it is easier to write a sum of products than a product
    of sums, even though the two formulations are exact duals?

    Yes, it is called deMorgans Laws.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From John Savard@21:1/5 to [email protected] on Fri Feb 23 23:22:20 2024
    On Fri, 23 Feb 2024 17:55:39 -0000 (UTC), Quadibloc <[email protected]d> wrote:

    But when I tried to reply to it, my newsreader crashed, and this
    thread did not reload. (It got into Google Groups under the wire,
    so this is where I found the text.)

    While the PAN newsreader eventually did load the message again,
    I am quite annoyed that it seems to now frequently crash.

    I had used it before, and it had been stable.

    So I am returning to what I *know* works. Forte Free Agent 3.3.

    John Savard

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Brett@21:1/5 to John Savard on Sat Feb 24 23:56:23 2024
    John Savard <[email protected]d> wrote:
    On Fri, 23 Feb 2024 17:55:39 -0000 (UTC), Quadibloc <[email protected]d> wrote:

    But when I tried to reply to it, my newsreader crashed, and this
    thread did not reload. (It got into Google Groups under the wire,
    so this is where I found the text.)

    While the PAN newsreader eventually did load the message again,
    I am quite annoyed that it seems to now frequently crash.

    I had used it before, and it had been stable.

    So I am returning to what I *know* works. Forte Free Agent 3.3.

    John Savard


    Had the same problem with NewsTap after a decade of use, database got corrupted.
    Just unsubscribe to everything and resubscribe.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)