Hello,

I was doing some tests with the numeric integration and noticed that it was not as accurate as other calculators. Then, I found this discussion about the subject, so after reading it, I have some questions:

For Team 34s:

- Is there any plan to improve the built in integrator?

- Pauli talks about a wrapper that would use the built in integrator. Do you have any example of that, or, is there any basic way to do it so one can adapt it easily to any function?

For Gene:

- In the same thread, you talk about a “matrix series modeled after the M1 - M5 routines from the PPC ROM”. Is this already available somewhere?

Please excuse me for these evidently badly informed questions

Thanks,

Miguel

*Edited: 12 Sept 2011, 1:54 p.m. after one or more responses were posted*

Buenas tardes Miguel,

It would significantly help answering your questions if you would include the link to the discussion you reference. And, BTW, please omit the "pre" to allow for free text flow in your messages. Thanks in advance,

Walter

Terribly sorry about that. I edited the entry.

Gracias,

Miguel

The idea is to subdivide the interval in several subintervals and integrate with the built in function in each subinterval. Repeat the step with an increased number of subintervals and compare the results. If they are close enough together, you're done.

Ok. Understood.

Thank you very much Marcus,

Miguel

Here is a newer post about the matrix math:

http://www.hpmuseum.org/cgi-sys/cgiwrap/hpmuseum/archv020.cgi?read=190327

Re: Talking about the WP 34s at the internal HP calculator forum

Message #3 Posted by gene wright on 11 Aug 2011, 4:03 p.m.,

in response to message #2 by Walter B

Good! Did not know you could read it without registration.

Yes, it has quieted down, but I wanted to again point people in the direction to find out more info.

On another note...I have the first draft of the matrix programs for the 34s keyed in. Debugging now. 197 lines is all it takes. Could do a 6x6 system or inverse, 8x8 determinant. That's a limit based on 100 registers available.

----

It looks like things are moving forward.

I hope to present the RRM routines at HHC in a couple of weeks, after which they will be public.

Looking forward to. Will have to use, however, the global edition of extendable ears.

I'll be watching for those extendable ears under the door.

For any non-NDA sessions, I would expect video to be up fairly shortly thereafter.

Quote:

- Is there any plan to improve the built in integrator?

No.

Quote:

- Pauli talks about a wrapper that would use the built in integrator. Do you have any example of that, or, is there any basic way to do it so one can adapt it easily to any function?

I'm not aware of any example code to do this.

To do much better then integrator needs to be adaptive. The current integrator calculates two different estimates of the integral. With a bit of programming it should be possible to integrate over an interval and if the two estimates disagree by too much, to subdivide the interval and try again.

Quote:

In the same thread, you talk about a “matrix series modeled after the M1 - M5 routines from the PPC ROM”. Is this already available somewhere?

I've also done a bit of work on some matrix code.

The M1 - M5 routines are the basics commands necessary to row reduce a matrix, they are not a full matrix package by themselves.

These do not include enough functionality to do a matrix multiplication or transpose e.g. They can, however, be used to calculate an inverse or a determinant very easily.

They are a great basis for further work & definitely worthwhile having.

- Pauli