Regarding 24 bit addressing with 32 bit arithmetic. ?When addressing memory in 24 bit mode, only the low 24 bits of the 32 bit generated address are used.
I worked in Poughkeepsie in the early '70s adding MP support to MVS. ?I can remember one incident when Kingston assembled the first fully configured 168 MP that had the maximum memory of 8 Meg on each side for a total of 16 Meg in MP mode. ?Kingston notified MVS development that they had some available machine time on the machine since it had successfully executed all the initial diagnostics. ?Some of the MVS development guys took their disk packs and drove over the river up to Kingston and loaded the disk drives and hit the load button.
The lights blinked for several seconds - then went dead. ?Second attempt resulted in same thing. ?They then took a memory dump, and found nothing in memory. ?They took their disk packs and went back home. ?They discovered that the IPL bootstrap program would determine the memory size of the machine by going into a loop zeroing out 256 bytes of memory at a time expecting to get an addressing exception when it found the end of memory. ?Problem was that with 16 Meg the addressing exception never occurred. ?The 32 bit address went above 16 Meg, but only the low 24 bits were used so the loop clearing memory ended up zeroing out low core and wiped itself out coming from the other direction. ?Very amusing at the time.
