Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!uwm.edu!zaphod.mps.ohio-state.edu!brutus.cs.uiuc.edu!jarthur!spectre.ccsf.caltech.edu!tybalt.caltech.edu!madler
From: madler@tybalt.caltech.edu (Mark Adler)
Newsgroups: comp.sys.handhelds
Subject: Re: HP-48SX: first impressions
Message-ID: <1990Mar11.013428.20139@spectre.ccsf.caltech.edu>
Date: 11 Mar 90 01:34:28 GMT
References: <6750@columbia.edu> <1157@kunivv1.sci.kun.nl> <10069@cbmvax.commodore.com>
Sender: news@spectre.ccsf.caltech.edu
Organization: California Institute of Technology, Pasadena
Lines: 71


Actually, there are two problems with the WALK demo program on page 598.
First, there needs to be a << before ERASE (you get a syntax error because
{ is not a valid local name for ->).  The missing << can actually be found
wandering around the bottom of the page after STEP.  That << should
obviously not be typed in there.

The second problem is that there should be a GXOR before PICT ROT to
display the new image.

Upon fixing these two things, the program can be entered and works as
advertised.  However, the checksum still doesn't match up.  (The number
of bytes does though.)

For those who would like to know, here are the bugs in the various ROM
revisions:

DEFINE rounds numbers in user-defined functions if STD is not the current
display mode---affects only revision A.

Entering a symbolic complex number like '(A,<B)' (where < is really the
funny little angle symbol) meaning in polar form, gives an expression
in A and B that is wrong (the A and B are interchanged so A is the angle
and B is the magnitude)---affects revisions A, B, C, and D.

NXEQ from the Graphics Environment with flag -3 set (SYM off in Modes)
can overwite the current equation with a number.

KGET on a zero length file from another machine will CAUSE THE CALCULATOR
MEMORY TO BE CLEARED!---only in revision A.

If time display mode is set to 24 hours, then minutes are shown
incorrectly in the Alarm Catalog---only in revision A.

INV returns an incorrect result if used on an 8x8 or larger matrix---
affects revisions A, B, and C.

The fixes for most of these are from the Henny Youngman School of Medicine:
"Then don't do that!"  That is, don't DEFINE if not in STD, don't enter
symbolic complex numbers in polar form, don't use NXEQ if in Numerical
Results mode, don't transfer zero length files into your calculator, and
don't use 24 hour mode when looking at the Alarm Catalog.  For INV, there
is a pseudo-fix that divides the matrix into the identity matrix instead
of inverting it.  It also tells you how to remap the 1/x key to call that
program instead of INV (the program uses INV if it's not a matrix).  You
have to be in USER mode for this to work, however.

To the HP guys who pay attention to this newgroup:  it would have been
enourmously clever, and therefore expected of HP, to have a copy of a
function dispatch table copied into RAM, so that just these sort of bugs
can be fixed by supplying new machine code that lives in user ram and
changing the entry in the dispatch table.  So is this the case?

If not, then maybe there are hooks for library ROM's to replace built-in
functions?

In any case, the keyboard trick is not adequate since it does nothing for
programs with INV.  I also wonder whether INV and dividing into an
identity matrix are numerically the same.  I know that with most
scientific subroutine libraries, the inversion routine is much better
at inverting than the matrix divide routine.  (Well, much better for
ill-conditioned matrices anyway.)

Despite the bugs in the ROM and the manual problems, I'm enourmously
impressed with this thing.  I highly recommend it to anyone with the
slightest interest in such beasts and the money to blow on it.  However,
try to get the latest ROM revision (E?) available whenever it is you
buy one.

Mark Adler
madler@hamlet.caltech.edu