Interesting patterns for HP-42S

[Tom Grydeland]

There are doubly-periodic pixellated patterns in which the frequencies of lit to unlit pixels are equal across all rows, columns and diagonals (left and right) of the pattern. When the basic pattern is of size 8, the frequencies can be equal. Looking for particular symmetries can lead the way to some strikingly beautiful patterns. I was curious to find out how compactly I could represent these in the HP-42S, and came up with the program listed below. (Rv is roll down, * is multiply).

Enter a number from 0 to 10 in ST X and XEQ "DPP". (only 0, 1, and 5 shown below, to save a bit of typing)

The underlying mathematics for the pattern search was published by prof. Ben Johnsen in the norwegian mathematical journal Normat in 2001. The paper (in Norwegian) is not available on-line, but I can provide copies upon request.

A small selection of patterns with other sizes and symmetries is shown in this album

00 { 46-Byte Prgm }

01 LBL "PSHOW"

02 CLLCD

03 ALENG

04 1e-5

05 *

06 1.131

07 +

08 STO 00

09 LBL 01

10 1

11 RCL 00

12 AGRAPH

13 X<>Y

14 8

15 +

16 X<>Y

17 AGRAPH

18 +

19 CLX

20 ISG 00

21 GTO 01

22 END

00 { 344-Byte Prgm }

01 LBL "DPP"

02 GTO IND ST X

03 LBL 00

04 3903507479

05 3895678231

06 GTO 96

07 LBL 01

08 4036562913

09 1013367582

10 GTO 96

...

23 LBL 05

24 4039905177

25 4029878118

26 GTO 96

.... (add more patterns here, if so desired)

46 LBL 96

47 CLA

48 XEQ 97

49 Rv

50 XEQ 97

51 XEQ "PSHOW"

52 RTN

53 LBL 97

54 4

55 STO 00

56 Rv

57 LBL 98

58 XEQ 99

59 DSE 00

60 GTO 98

61 RTN

62 LBL 99

63 ENTER

64 ENTER  (if anyone know a 1-instruction DUP which does not disable stack lift, tell me)

65 255

66 AND

67 XTOA

68 Rv

69 8

70 ROTXY

71 END

[C.Ret]

Quote:

---

[...]

62 LBL 99

63 ENTER

64 ENTER  (if anyone know a 1-instruction DUP which does not disable stack lift, tell me)

[...]

---

May I suggest RCL ST X ?

[Tom Grydeland]

yes, that did occur to me after posting. It works, but uses the same # of bytes.

[C.Ret]

Exact. No byte are spared here!

Your code triggers me enough to let me try to observe your curious and highly symmetrical pixel patterns.

I have adapt your code to my HP-28S to take advantage of the larger LCD screen (double size in row compare to HP-42S).

« 64 STWD                  @  Use full size binary words

   { #E8AACC17E8335515h    

     #F0990FE13C66C31Eh    

     ...                   @ Only pattern 0, 1 and 5 in list 

     #F0CC0F99F0330F66h

     ...                   @         to save keystrokes

   }

   SWAP 1 + GET            @ get pattern from code list

   ""                      @ initiate graphic string

   1 8 START

         SWAP RRB          @ build 8x8 pixels pattern

         DUP #FFh AND      @ RRB is rotate right one byte

         B->R CHR ROT +    @ from code last byte into graphic string

   NEXT

   DUP OVER +              @ grow gr. String up to 16 pixels

   DUP + DUP + DUP +       @ 32 p. -> 64p.-> 128p.

   + 0 CHR +               @ adjust length to exactly 137 pixels /1 row on HP28C/) 

   DUP + DUP +             @ repeat for each of the 4 rows

   ->LCD                   @ display graphic string 

 »

‘DPP’ STO

Usage, enter 0, 1 or 5 and press [ DPP ] menu key to see corresponding doubly-periodic pixellated patterns full screen (press [ ON ]+[ L ] to online-print screen).

I am now one the way to code other high symmetrical 8x8 pixel patterns.

Edited: 6 Mar 2012, 5:36 p.m.

[Tom Grydeland]

Very nice! I don't have a 28S, so I cannot try your programs, but the output looks exactly correct. The x48 emulator works for building the string, but it doesn't like the string argument to "->LCD" -- I haven't tried particularly hard, so I might have missed some subtlety.

The larger screen of the 28 could also be used to tile up the pattern, ie use 2x2 screen pixels as the basis for the pattern. Not exactly sure what kind of bit twiddling would turn the nybble "abcd" into "aabbccdd", perhaps a look-up table would be best?

I have some 2537 patterns of this weight and size (i.e. 4 of every 8 pixels are black, starting anywhere and counting horizontally, vertically or diagonally) of which 919 have more than one symmetry. They are not coded up as hexadecimal strings, however (searches started out with patterns of other sizes), but I should be able to recode to hex-strings without too much trouble.

[Marcus von Cube, Germany]

Tom, the 48 series has a more evolved graphic object concept. Hunt for GROB in the documentation!

[Tom Grydeland]

Here are some more interesting patterns for you to try, in hex form:

#17A33AD8C68D6C71h

#271BD8B18D4E72E4h

#17A3725CC68DE839h

#17C56CD88EA33972h

#27A96A5695E41BD8h

#27B16A568DE4599Ah

#1B63D436CA4DB1ACh

#2B53ACB24DCA35D4h

#2739CAB28D6C53D4h

#2739C6B2AC4D53D8h

#271B72B18D4ED8E4h

#2739D493AC4D72CAh

#0F69690FF09696F0h

#1763E84E9CC5B13Ah

#1763E45C8EC9B13Ah

#1771A33A4EC9E49Ch

#17A33AE88EC55C71h