🚀 go-pugleaf

> Why would you wish that IBM mainframes were the brain-dead architecture
> that the PC uses?  You need the '386 before you can truly multitask safely,
> and even if a 386 *can* go at 15MIPS, it's still useless because it
> spends 12 of them doing loads and stores out of the 4 registers...

Yep.  That's the problem with people who like to claim that a Sun
4/490 is 23 mips -- so what?  If it wastes 90% of those cheap cycles
running grossly inefficent C code, the true speed is 2.3 mips.  If it
takes you 10 instruction to do what can be done in one instruction on
a 370, it's absolutely unreasonable to even talk about mips.  That's
why IBM resists giving mips ratings.

> At the risk of starting a holy war, MS-DOS and the Intel 8x86 architecture
> are the single worst thing to happen to computing.

Nah, *ix is.  At least not many people claim that MS-DOS is a
reasonable system.  But there are large numbers of people who think
that *ix is the salvation of computing.

Flames cheerfully ignored.


> Does anybody out there have a war story of a "undocumented" opcode or
> similar feature on an IBM 360/370/etc mainframe?

How about an instruction that under certain conditions gives the wrong
answer, guaranteed?  Someone sprung that one on me.  On a 360/91,
there is a bug in the TRT instruction if you use R2 for the translate
table base register.  I also was told a story about that same 360/91.
When it was first installed, apparently a STM that started in your key
and crossed a page boundary into another key FINISHED and THEN gave a
protection exception.  IBM wouldn't believe it until a systems
programmer toggled a demonstration program into the machine via the
front panel switches...  (First they claimed it was an O.S. bug.)

/Leonard