🚀 go-pugleaf

"Charlie Gibbs" <cgibbs@nowhere.in.particular> writes:

>How would this compare with the infamous SLT (search list) instruction,
>which was available on certain 360s by RPQ.  If anyone is interested,
>I might be able to dig up a couple of write-ups, one serious and one
>tongue-in-cheek.  SLT's many implied operands gave the latter write-up
>its subtitle of "And you thought BXLE was bad?"

Implied operands are (to put it politely) no fun whatever.  In at least
one case they bit IBM in the pocketbook: the 360/91 was a non-microcoded
system (everything hardwired, which is probably why it didn't have the
decimal arithmetic instructions).  Several years after the last unit
rolled off the assembly line and the designers were dispesed to other
projects, the /91 at Oak Ridge National Labs was found to have a design
defect in the TRT instruction (which has an implied destination operand
of GPR1) if the explicit source operand used GPR1 as a base register.

This usage was consistent with the PoO, and the /91 was on maintenance.
I'm told that there was lots of unhappiness when the hardware designers
had to be brought back to the /91 project to redesign that part of the
logic and verify that the fix didn't screw up something else.

Joe Morris