Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!think.com!zaphod.mps.ohio-state.edu!wuarchive!psuvax1!psuvm!wvnvm!bryan
Organization: West Virginia Network
Date: Wednesday, 29 May 1991 11:59:27 EDT
From: Jerry Bryan <BRYAN@wvnvm.wvnet.edu>
Message-ID: <91149.115927BRYAN@wvnvm.wvnet.edu>
Newsgroups: comp.lang.asm370
Subject: Re: disabling CTRL-BREAK (dos)
Distribution: inet
References:  <9105250259.AA11782@ucbvax.Berkeley.EDU>


>
>Does anybody out there have a war story of a "undocumented" opcode or
>similar feature on an IBM 360/370/etc mainframe?  My personal favorite
>was when I was doing MUSIC hacking, and a kernel routine of mine took a
>branch to East Fishkill on an IBM 4341, hit a data area, and tried to
>execute a SIGP instruction.  Oddly enough, it died on a privop
>exception, not the regular operation exception you'd expect.  I never
>got brave enough to explore whether the 4341 *really* had a SIGP
>instruction, or whether CP's instruction simulation was doing something
>to me....
>

The old Waterloo WATBOL compiler used invalid opcodes on purpose (trapping
them with SPIE macro) to effect paragraph tracing.  One of the opcodes
they chose invoked the APL microcode assist on a 370/148 rather than
generating on operation exception.  We therefore
found the problem of why tracing wouldn't work by reading the 148's
microcode.

(By the way, a 148 would run APL faster than a 168 because of the
148's microcode assist.)