Threaded Mode | Linear Mode

wojtek · 04-19-2016, 02:49 PM

How would you determine the number of data registers set in the past (and forgotten) in the HP41C?
There is no SIZE? command in this calculator.

Jake Schwartz · 04-19-2016, 03:21 PM

(04-19-2016 02:49 PM)wojtek Wrote: How would you determine the number of data registers set in the past (and forgotten) in the HP41C?
There is no SIZE? command in this calculator.

This is a question people had been asking since the machine was introduced. It was solved relatively early on in the PPC ROM with the "S?" (Size Finder) routine, which contains a call to another routine within the module. If you can't get your hands on a PPC ROM module, perhaps others here can chime in with a reasonable alternative.

Jake

RMollov · 04-19-2016, 03:42 PM

(04-19-2016 02:49 PM)wojtek Wrote: How would you determine the number of data registers set in the past (and forgotten) in the HP41C?
There is no SIZE? command in this calculator.

Very simple. Start RCL-ing from maximum possible register number down. Depending on how much memory is available will take you less than a minute to figure it out. Wink

wojtek · (This post was last modified: 04-19-2016 04:13 PM by wojtek.)

(04-19-2016 03:42 PM)RMollov Wrote:
(04-19-2016 02:49 PM)wojtek Wrote: How would you determine the number of data registers set in the past (and forgotten) in the HP41C?
There is no SIZE? command in this calculator.
Very simple. Start RCL-ing from maximum possible register number down. Depending on how much memory is available will take you less than a minute to figure it out.

:-) :-) good joke, thank you, I know that way of course.
Less than a minute? Even if you have quad memory installed?

BobVA · 04-19-2016, 07:12 PM

(04-19-2016 04:11 PM)wojtek Wrote: ...Less than a minute? Even if you have quad memory installed?

Bisection search rather than sequential?

wojtek · 04-19-2016, 07:39 PM

(04-19-2016 07:12 PM)BobVA Wrote:
(04-19-2016 04:11 PM)wojtek Wrote: ...Less than a minute? Even if you have quad memory installed?

Bisection search rather than sequential?

Maybe. Allowing for the worst case - 317 data registers it would take 8 steps:
318/159/240/270/295/307/313/316=HIT (or something like this)
I would rather think about a procedure based on bisection search and hit or miss rule and using the flag 25 - error ignore.

Dieter · (This post was last modified: 04-19-2016 08:41 PM by Dieter.)

(04-19-2016 02:49 PM)wojtek Wrote: How would you determine the number of data registers set in the past (and forgotten) in the HP41C?
There is no SIZE? command in this calculator.

In the early HP41 days, SIZE finders were among the most popular routines. They could be found in several issues of HP key notes and most probably many similar publications. With the advent of Synthetic Programming, faster synthetic versions became available. And of course the PPC ROM includes such a function. You may take a look at the books by Bill Wickes and/or the documentation of the mentioned ROM.

Eventually the X-Functions module became available, and so the problem was solved by the included SIZE? command. Since it offers many other useful functions, I would recommend getting such a module.

Dieter

Dieter · 04-19-2016, 08:55 PM

(04-19-2016 04:11 PM)wojtek Wrote: :-) :-) good joke, thank you, I know that way of course.
Less than a minute? Even if you have quad memory installed?

Hmmm... I tried this very (!) simple program...

Code:

LBL"SIZE??"

,4

SIGN

SF 25

LBL 01

RCL IND L

FC? 25

GTO 02

ISG L

GTO 01

LBL 02

LastX

INT

END

... and my physical '41 required 9 seconds for SIZE 050. So even in the worst case the result indeed should appear within about a minute. ;-)

Dieter

RMollov · (This post was last modified: 04-19-2016 11:08 PM by RMollov.)

(04-19-2016 08:55 PM)Dieter Wrote: Hmmm... I tried this very (!) simple program...

Code:

LBL"SIZE??" ,4 SIGN SF 25 LBL 01 RCL IND L FC? 25 GTO 02 ISG L GTO 01 LBL 02 LastX INT END

... and my physical '41 required 9 seconds for SIZE 050. So even in the worst case the result indeed should appear within about a minute. ;-)

Dieter

It took 17 secs to find out there were 100 registers allocated on my beast (which I already knew anyway). Nice program BTW, I love watching the flying goose.
Cheers,

Sylvain Cote · (This post was last modified: 04-20-2016 05:10 AM by Sylvain Cote.)

The following program (64 bytes) will find the number of registers in nine lookup. (160, 80, 40, 20, 10, 5, 3, 2, 1)

Code:

01 LBL "SIZE?"    // Entry Point

02 SF 25          // Start with substraction

03 0              // Register index

04 320            // Register offset (startup value, max number of registers)

05 LBL 00         // Loop

06 2              //

07 /              // Divide register offset by 2

08 .5             //

09 +              // Add .5 for odd divisions

10 INT            // Get interger part

11 FS? 25         // If last memory access was valid

12 +              // Add register offset to register index

13 FC? 25         // If last memory access was invalid

14 -              // Substract register offset from register index

15 LASTX          // Get back register offset

16 SF 25          // Ignore next error

17 RCL IND Y      // Memory access request

18 FS? 25         // If successful

19 RDN            // Drop unneeded recalled value

20 1              // Exit loop condition

21 X=Y?           // 

22 GTO 01         // leave loop if the offset is 1

23 RDN            // Drop exit loop test value

24 GTO 00         // Loop for next iteration

25 LBL 01         // Almost done

26 RDN            // Drop exit loop test value

27 RDN            // Drop register offset

28 SF 25          // Ignore next error

29 RCL IND X      // Memory access request

30 FC? 25         // If last memory access was invalid

31 GTO 02         // Odd number of registers allocated, no compensation needed

32 RDN            // Drop unneeded recalled value

33 1              // Even number of registers allocated, compensation needed

34 +              // Compensate the value found by 1

35 LBL 02         // Compensation terminated

36 X<0?           // If the number is negative, thene there was no data register allocated

37 0              // 0 for no data registers

38 END            // The number of registers has been found

On a unmodified HP-41CX Halfnut, the program consistently runs in 5.7 seconds.

Measured with the stopwatch
Insert after LBL "SIZE?" -> 0, SETSW, RUNSW
Insert before END -> STOPSW, RCLSW, X<>Y

Have fun!

Sylvain

Edit: add run time

Werner · (This post was last modified: 04-21-2016 09:11 AM by Werner.)

Here's my non-synthetic version, 36 bytes without the END

Code:

>LBL"SIZE?"

 255

 ENTER

 ISG X

>LBL 01

 SF 25

 VIEW IND Y

 CLD

 FC?C 25

 CHS

 2

 /

 ST+ Y

 ABS

 INT

 X>0?

 GTO 01

 RDN

 LASTX

 +

 END

Cheers, Werner
*Edited a few times. Now, after 25 years or so, I seem to be able to find improvements I couldn't find before..

Dieter · 04-20-2016, 12:03 PM

(04-19-2016 11:06 PM)RMollov Wrote: It took 17 secs to find out there were 100 registers allocated on my beast

Twice the SIZE, twice the time. ;-)
But...

(04-19-2016 11:06 PM)RMollov Wrote: Nice program BTW, I love watching the flying goose.

...OK, it's a nice program for goose watchers, but otherwise it's simply stupid. #-)
I also tried a bisection method, and this runs in about 5 seconds for any SIZE.

Dieter

RMollov · 04-21-2016, 11:14 AM

(04-20-2016 08:40 AM)Werner Wrote: Here's my non-synthetic version, 36 bytes without the END
...
*Edited a few times. Now, after 25 years or so, I seem to be able to find improvements I couldn't find before..

I find that fenomenon too myself. Kudos btw, as usual.
Cheers

RMollov · 04-21-2016, 11:27 AM

Some 30+ years ago, a structural engineer I worked for had a 41cv program to calculate if pile depth is enough to support future structure or something; I only did surveying work for him. The program used to run for more than 5 hours, sometimes 6 or 7. We used to sit in the pub nearby watching the flying goose while sipping gin and tonic and other similar stuff. That was great! That's how I fell in love with the goose.

Dieter · (This post was last modified: 04-21-2016 05:49 PM by Dieter.)

(04-20-2016 08:40 AM)Werner Wrote: *Edited a few times. Now, after 25 years or so, I seem to be able to find improvements I couldn't find before..

I prefer the original version – the new one is slightly shorter (while I doubt it's faster), but the display looks a bit weird because of the VIEW / CLD sequence.
Here's another option that does not require a second flag test:

Code:

>LBL"SIZE?"

 255

 ENTER

 ISG X

>LBL 01

 SF 25

 ENTER

 CLX

 RCL IND Z

 RDN

 FC?C 25

 CHS

 2

 /

 ST+ Y

 ABS

 INT

 X>0?

 GTO 01

 X<> L

 +

 END

The CLX disables stack lift so that the RCL overwrites X if the register is available, otherwise it leaves a zero.

Otherwise I really like your code. It's short, fast and effective.

Dieter

**Gene** · 04-21-2016, 05:49 PM

Couple of items from PPC Journal issues, just for the fun of it.

HP-41C Shortest Size Finder by Rolf R. Schmitt - 21 bytes
PPC Calculator Journal V7 N8 P26 October 1980

Code:

01 LBL "S" ;Shortest Size Finder Routine

 02 0

 03 SF 25

 04 LBL 01

 05 VIEW IND X

 06 FS? 25

 07 ISG X

 08 STOP

 09 GTO 01

 10 END

HP-41C Short Size Finder Routine by Geoff Smith - 25 bytes
PPC Calculator Journal V7 N6 Pg 36 Jul/Aug 1980

Code:

01 LBL "S" ;Short Size Finder Routine

 02 .5

 03 SF 25

 04 LBL 01

 05 VIEW IND X

 06 ISG X

 07 FS? 25

 08 GTO 01

 09 1.5

 10 -

 11 END

HP-41C Size Finder Routine by Justus A. Villa - 31 bytes[/php]
PPC Calculator Journal V7 N6 P36 Jul/Aug 1980

Code:

 01 LBL "S" ;Size Finder Routine

 02 257

 03 ENTER

 04 LBL 00

 05 2

 06 /

 07 SF 25

 08 ARCL IND Y

 09 FC? 25

 10 CHS

 11 +

 12 LASTX

 13 ABS

 14 INT

 15 X#0?

 16 GTO 00

 17 +

 18 RTN

 19 END

HP-41C Small Size Finder Routine by Ron Knapp - 34 bytes
PPC Calculator Journal V7 N2 P38 Feb-Mar 1980

Code:

01 LBL "S" ;Small Size Finder Routine

 02 9

 03 -1

 04 512

 05 LBL 00

 06 2

 07 /

 08 SF 25

 09 ST+ Y

 10 ABS

 11 ARCL IND Y

 12 FC?C 25

 13 CHS

 14 DSE Z

 15 GTO 00

 16 X<0?

 17 CLX

 18 +

19 RTN

Werner · (This post was last modified: 04-22-2016 05:31 AM by Werner.)

(04-21-2016 05:24 PM)Dieter Wrote:

Code:

>LBL"SIZE?" 255 ENTER ISG X >LBL 01 SF 25 ENTER CLX RCL IND Z RDN FC?C 25 CHS 2 / ST+ Y ABS INT X>0? GTO 01 X<> L + END

The CLX disables stack lift so that the RCL overwrites X if the register is available, otherwise it leaves a zero.

Otherwise I really like your code. It's short, fast and effective.

Dieter

The previous version at 37 bytes used almost exactly that trick. Almost, since ENTER itself also disables stack lift and so the CLX is not needed ;-))

And I tried out the VIEW/CLD on a real 41, the VIEWs do not show, that's why I thought it was acceptable.

Cheers, Werner
And now I'm going to read Gene's post ;-)

Dieter · 04-21-2016, 07:38 PM

(04-21-2016 07:19 PM)Werner Wrote: The previous version at 37 bytes used almost exactly that trick. Almost, since ENTER itself also disables stack lift and so the CLX is not needed ;-))

Arrrrgghh... *facepalm*
OK, I was really blind here. #-)

(04-21-2016 07:19 PM)Werner Wrote: And I tried out the VIEW/CLD on a real 41, the VIEWs do not show, that's why I thought it was acceptable.

I do not like the goose jumping back on every CLD.
Among the versions posted by Gene some use ARCL instead of RCL so that the stack is not affected. I wonder if this may slow down the program.

Dieter

RMollov · 04-22-2016, 04:16 AM

(04-20-2016 08:40 AM)Werner Wrote: Here's my non-synthetic version, 36 bytes without the END

Code:

>LBL"SIZE?" 255 ENTER ISG X >LBL 01 SF 25 VIEW IND Y CLD FC?C 25 CHS 2 / ST+ Y ABS INT X>0? GTO 01 RDN LASTX + END

Cheers, Werner
*Edited a few times. Now, after 25 years or so, I seem to be able to find improvements I couldn't find before..

I got rid of CLD and replaced VIEW IND Y with ARCL IND Y and it's now slightly quicker plus the goose flies smoooothly Wink

Cheers

Werner · (This post was last modified: 04-22-2016 05:41 AM by Werner.)

(04-21-2016 07:38 PM)Dieter Wrote: I do not like the goose jumping back on every CLD.
Among the versions posted by Gene some use ARCL instead of RCL so that the stack is not affected. I wonder if this may slow down the program.

Dieter

The goose jumping backward doesn't bother me - the MATRIX program in the standard module used CLD as a NOP after ISG, so it showed that behaviour, too.

I didn't want to use ARCL, I always try to use as few resources as possible.

Also, again I must praise Free42: did you know that (for any x)

SF 25
SIZE x
ENTER
RCL x

still has stack lift disabled? I didn't. But that's how it is in the 41 and 42. And Free42.

Werner