Delete comment from: Ken Shirriff's blog
Thinking of the core rope memory, from your description a 'value of '1' requires looping through a core, while a '0' bypasses a core. It seems like therefore a '1' weighs more than a zero, in terms of flying weight of the core rope module. If that is true, I wonder how much more a '1' in core rope weighs? I read here that a single core can be used for multiple bits, so the extra weight of a core is shared by the number of wires running through it representing a '1' bit. I'm sure that statistically the overall percentages of '1's and '0's is close to 50/50 looking at instruction encoding and the SW in ROM. If there were a large difference it could imply an inefficiency. For example if the instruction encoding was such that many instructions were mostly '0's, perhaps the instruction word should use less bits. Knowing there may have been a few versions of the rope core, as the SW matured, I wonder if the weight of the versions differed due to the number of cores required to implement the binary image in each version? PS - thanks for these interesting posts about the AGC. What a unique and significant piece of computing hardware!
Sep 30, 2019, 2:27:16 PM
Posted to A computer built from NOR gates: inside the Apollo Guidance Computer