100!



#26

Due to some more introductory remarks, the WP 34S Owner's Manual now contains 100 pages for the first time. I didn't expect that amount when we started :-/ Frankly, I doubt I tackled that task if I'd have seen that then.

But now it's slowly moving in the direction of software, where it's said "20% of the code is for what the SW shall do, and 80% is for error prevention and recovery - if the user does something a knowledgeable user shouldn't do". Seems such is life.

FWIW,
Walter


#27

Quote:
...the WP 34S Owner's Manual now contains 100 pages...

100 only? In the header you state 100! ;-)

Nevertheless many thanks for it,

Franz


#28

Well Franz, 100! is just 11000, and the manual has more pages than that... (base 2, of course) :-)

#29

100! -- I started reading this thread thinking that it might be about a sneaky way to compute factorials greater than 69 on an older HP calculator (or the 15C LE or 12C+). But no, it's another 34s topic, sigh.

#30

Quote:
20% of the code is for what the SW shall do, and 80% is for error prevention and recovery - if the user does something a knowledgeable user shouldn't do

I'd have to agree with this. Catching the corner cases and returning the proper results for silly input consumes lots of code space.


- Pauli

#31

*sniff* she's all grown up now!

#32

It is about to get a bit bigger soon once the matrix support settles properly.


And a user guide will be ten times this size ;-)


- Pauli


#33

Quote:
It is about to get a bit bigger soon once the matrix support settles properly.

Since you mention matrix support:

I see you're just working on M.LIN and M.INV - can you say what else is planned to be built-in wrt. matrix functions?

M.RREF would be very nice (could also be used for linear equation systems), and also eigenvalues/vectors would be great. :-)

Franz


#34

I wasn't planning a RREF command. We're really tight on flash and the commonest uses for RREF that I remember are inverting a matrix and solving linear equations. Both of these are available already via the LU decomposition.

I'd like to include eigenvalues and vectors. Again, space is tight and these might have to be done via user code. Another one which would be good to get in would be one of the SVD algorithms. Good for least squares solutions, finding the Moore–Penrose pseudoinverse and many other things. I've often said that any problem in linear algebra starts either with the SVD or the eigenvectors -- not strictly true but not that far off the mark :-)


- Pauli


#35

Quote:
...and the commonest uses for RREF that I remember are inverting a matrix and solving linear equations. Both of these are available already via the LU decomposition.

You're right, but RREF is quite useful if you have more or less equations than variables, or if some equations are linear dependent. What does your M.LIN implementation do in such cases?

#36

It errors out with a dimension mismatch or a singular matrix. I guess this isn't as good :-(

The generalised inverse solves part of these problems, but I've not done that at this stage.


- Pauli


#37

Quote:
It errors out with a dimension mismatch or a singular matrix. I guess this isn't as good :-(

No, not really. ;-)

And that's exactly the reason why I would like such a RREF command, because it solves all these problems.

About your "Matrix?" question: I would be happy also with 5 user pages - builtin matrix functions are always more comfortable than having to code them by yourself (not to forget the problems with only 3-letter labels).

Franz

#38

A question for the folks here.

Implementing the matrix functions will force us to reclaim some of the user flash pages. Currently, matrix support has required three pages and a fourth is very likely.

How many pages of user flash is matrix support worth?


At the moment there are seven program flash pages. One is part of the device backup region which leaves six pages (506 steps each or 3036 steps in total) for whatever the user wants. This will likely drop to five pages (2530 steps).


- Pauli


#39

Is this something that can be implemented as programs that need not be installed?

I personally won't have a large need for matrix operations, but this is certainly a huge addition for some folks.

At the same time, I won't be using all of the user pages either. So for me it is not a problem to use additional pages for matrices.

I just thought it might be possible to make this an option for each person to choose, but I don't know the mechanics around this and if it is possible.

David


#40

Quote:
Is this something that can be implemented as programs that need not be installed?

Not the way it has been done. It would be possible to write user code programs to support matrix operations and they'd be smaller, slower, less accurate and more difficult to use but functional.

Quote:
I just thought it might be possible to make this an option for each person to choose, but I don't know the mechanics around this and if it is possible.

Not without two different firmware images which I'd like to avoid.

Another alternative would be to drop some of the esoteric functionality in favour of matrix support. e.g. zeta and Bernoulli numbers are largish as are some of the complex functions.

- Pauli


#41

Quote:
Another alternative would be to drop some of the esoteric functionality in favour of matrix support. e.g. zeta and Bernoulli numbers are largish as are some of the complex functions.

+1 for this idea! :-)

I'd say matrix functions (e.g. M.RREF or the already existing but commented out row-operations) are more important than such special functions which almost nobody will ever need or use.

Franz


#42

The question is what can be easily redone in user code without sacrificing too much accuracy. User code can go into libraries and thus swapped in and out. We have no provisions of putting native code into user flash regions to make them installable objects. This would require a major rewrite.


#43

Quote:
The question is what can be easily redone in user code without sacrificing too much accuracy.

Yes, that's ONE point - the other is which functions are used much more frequently than others. And I'm sure this point would definitely go to matrix operations in favour of such exotic functions Pauli mentioned (at least I myself have never needed or used any of them in my almost 30 years as mathematician ;-)).

Franz

#44

Argh.

As much as I would love a wonderful set of matrix functions, that is a lot of space. :-(

The RRM programs take about 350 program steps of one user flash area and give results that are VERY respectable and fast as well.

I know these will be even faster and a lot more accurate, but at a trade off.

The AM1 7x7 determinant takes 2 seconds using the RRM program and returns a result of 0.999997 which is awfully close to the real result of 1.

I have seen the result the new functions provide, which rounded to 15-16 digits really is equal to 1. lol

While that's wonderful and I certainly would love such abilities, it would be taking flash from everyone - and I can't make that call.


#45

There is a lot more to matrix support than the five PPC routines and RRM routine :-)

Matrix addition and multiplication?
Transpose?

Also the routines I've coded do things in-situ and without altering the source matrix or doing anything untoward to the stack. i.e. they are much easier to use.


I guess the question comes down to how many programming steps are required on the 34S?


- Pauli


#46

50,000 !

:-)


Seriously, as you said, there is more functionality to be had than the RRM/PPC ROM stuff and to add that is a trade-off to program space.

Still, since pages of flash can be loaded by combining files at upload from a PC, having all 9 flash spaces open is very tempting!

Tough choice. Perhaps you should start a feedback thread with a compelling title to get the feeling here?


#47

It is a tough choice. Given that the 34S is still a moving target firmware wise, the community hasn't had a lot of time to develop complete user apps and get an idea of how much space is needed for user-friendly programs. Also goes the corollary that any user program will fill up all the user space. So, does it really matter what a survey reveals? In the end, no matter how much programming is made available to the user, it will be filled up and we will cry for more.

The HP-41C was built with more memory than was ever seen before in an HP calculator, with a price tag to match. Now with the 41CL we have multiple pages to store complete 41 memory images, including extended memory.

Given the physical hardware limitations of the calculators,I vote for two different firmware images, one with matrix functions and one without. Then I can burn a couple of HP 20B/30Bs with the functionality I want and programming space I need. But that makes firmware development more complex. Now we are experiencing some of the design trade-offs HP has had to make in the past. CRAP!

I WANT IT ALL IN MY 34S and I WANT IT NOW!

Gerry


#48

Sounds like we've got a volunteer to maintain a split distribution :-)


- Pauli


#49

You are a very funny man:). I did sit next to Marcus during the conference but none of his smarts rubbed off on me.

Gerry

#50

"How many pages of user flash is matrix support worth?"

As i have already stated; i think that massive program memory is more important than matrices, those complex functions, or even hyperbolic trig. I seldom need to describe trajectories in flat non-euclidean 4 dimensional spacetime of the sort used in special relativity any more. Maybe it's the bars i go to these days.

I see why you and the rest of the trinity don't want to have two or more final versions of the 34S floating around out there. I can also see how Eric Rechlin wouldn't like that either. You might give it another thought though.

Whatever you do, thanks.


#51

Where were you at HHC?


#52

Thanks for asking Gene. If i had been, Marcus would have drank even more beer. Did i miss any interesting "deciduous trees"?

Too much work, too much indecision on the part of my employer, too much of my own house's roof open to the sky.


#53

Hey, that's three too's.

Hope you can come next year.

#54

Quote:
I seldom need to describe trajectories in flat non-euclidean 4 dimensional spacetime of the sort used in special relativity any more. Maybe it's the bars i go to these days.

LOL!

#55

100 pages barely scratches the surface. It deserves to be, and users deserve, several times that much, not so much to say "mine is bigger than yours" but because clear explanations and examples take that much. The PPC ROM manual is a shining example of good documentation.

It's that much more reason to take advantage of Gene and Jake's generous offer to fully document the WP 34S, especially including beginner info. Doing so will free you to do other things. Thank you for everything you have done thus far.


#56

I agree with Brian. The original 15C Owners Handbook was 300 pages. Given the huge addition of functionality, I think a 34S manual with similar detail would be 800-1,000 pages.


#57

Quote:
I think a 34S manual with similar detail would be 800-1,000 pages.


...and since the 34S computes 100! to be 9.332621544394415 E157, we have our work cut out for us :-)

Jake

#58

The HP 48G/GX had its User's Guide (which underwent at least eight editions) and its Advanced User's Reference Manual (three). They're exactly what one would call the FM: read it and you'll find everything, or nearly so.

HP left it at that.

Then others wrote books on using the 48G/GX. Grapevine Publications put out four books and a Pocket Guide, and Thomas Adams wrote or cowrote two more. Some books, written for the 48S/SX, are pertinent to the 48G/GX as well. There may be others of which I am not aware. They all have their place alongside the FM.

I would think that 100! pages by a principal in the 34S project would qualify, after minor fixes, as the FM. Rather than setting up a wiki within which to make the FM satisfy everyone -- an impossible project -- might it not be better for anyone with a Grapevine-like idea just to do it?

By all means read and proofread the FM, and suggest fixes to the author. Then let the FM be, and let a thousand flowers bloom.


Possibly Related Threads…
Thread Author Replies Views Last Post
  HP-100, 200, 300LX & others + Macintosh Matt Agajanian 12 3,801 07-16-2012, 01:38 PM
Last Post: Marcus von Cube, Germany
  41CX was 95% working - now 100% working Russell Dobson 5 1,892 03-27-2012, 05:04 AM
Last Post: Russell Dobson
  OT: Commodore PR-100 manual Bart (UK) 4 1,790 03-12-2012, 12:08 PM
Last Post: Paul Berger (Canada)
  HP-65 makes All-TIME 100 Gadgets list. Alex L 32 7,472 10-27-2010, 10:31 AM
Last Post: Jim L (OHIO)
  Commodore PR 100 Nigel Bamber 1 1,015 05-22-2010, 04:21 PM
Last Post: Thomas Radtke
  HP 42S sells for < $100 H**l freezes Michael de Estrada 25 6,343 10-20-2009, 11:43 PM
Last Post: DavidShenk
  HP OmniGo 100 Question Maximilian Hohmann 4 1,569 05-27-2007, 09:41 PM
Last Post: Raymond Del Tondo
  100-99.99-0.01 in a spreadsheet Karl Schneider 17 4,268 02-10-2007, 09:49 PM
Last Post: Rodger Rosenbaum
  enter 100 on the 50g hugh steers 4 1,484 01-05-2007, 02:28 AM
Last Post: vincent
  HP 15C new for $100 !??? Chris Woodhouse 16 3,725 08-04-2006, 11:21 AM
Last Post: Ed Look

Forum Jump: