On Wed, May 25, 2016 at 6:50 AM, Paul Koning <paulkon...@comcast.net> wrote:
> > The emulator was loaded, on those rare occasions where the memory got > wiped, using the "Emulator IPL" button, from a binary card deck. That deck > was pretty slick: it was a channel program loop. No CPU code involved at > all; the first card was a 4 (?) entry channel program that would read the > remaining cards, which were a standard assembler output (object deck). > Self modifying channel code: since each object card contained the address > and length for its data, the channel program would pick up those two fields > and drop them into the third channel command, which would transfer that > number of bytes to that address. > > The DPS8 bootload was similar -- channel code to read a record from tape into memory. The record was placed in memory such that it overwrote the interrupt vector (actually fault pair), so that when the read completed and the tape drive raised the interrupt, the CPU jumped into the data just read. -- Charles