IBM 610 workstation computer 3389

David Dyer-Bennet

In addition, the A and B bits were used in the addressing scheme. The address field was three decimal (BCD) characters, so max address was 999). The A&B bits of the 1st character were a binary number (0-3) indicating the number of thousands to be added to the address, and those on the 3rd character were the number of 4000s to be added. This gave a max address of 3*4000 + 3*1000 + 999 = 15999. The A&B bits on the 2nd character were the index register to be used (0=none).

I once got caught by this. There was a dirty trick in common use of using the word mark to change the value of a switch. A NOP (N) followed by a branch (B xyz) did the jump, but clear the word mark on the B, and the whole sequence became a NOP. Unfortunately, the processor evaluated the addresses even in a NOP instruction, and the apparent address was taken from the next three characters after the N opcode, so the 2nd character of the branch address became the 3rd character of the evaluated address for the NOP. if the branch address was indexed (typically used for a jump table), this could result in an adress beyond the installed physical memory, and a machine fault. In my case, a program that ran perfectly well on a 16k machine failed on a 12k one, because the branch address used index register 3.

So much for self-modifying code.

--brian

-- Wellington, New Zealand

"What's life? Life's easy. A quirk of matter. Nature's way of keeping meat fresh."