On 10/7/22 17:10, Paul Koning via cctalk wrote: > Then of course some card devices feed the narrow edge (1 column) first rather > than the top or bottom row. CDC had a reader that worked column-wise and a > punch that worked row-wise, so the interface logic for the punch needed a > transpose operation while the reader doesn't -- given that both would > transfer card data as a word per column. But Electrologica used the same > hardware without the transpose logic in the controller, so the software would > see columns from the reader but would have to construct rows to send to the > punch.
CDC 405 reader and 415 punch. Both pretty noisy, but I found the 415 more aurally irritating--and easily overheated if you ran a big punch job. Early IBM gear read only the first 72 columns of an 80 column card; hence the restriction for early FORTRAN to 72 columns. Made sense reading row-binary into 2 36-bit words. Generally, the 405 was very fast and trouble free--until a card was fed that got stuck in the works, resulting in lots of "accordion pleated" cards in the stacker. It was a marvel of engineering, with its machine-gun rattle of 1200 cards/minute. You could load two boxes (4000 cards) in the hopper. --Chuck