This program takes input in rows and columns as decimals 0..9 and organises the data so that most typical values gather at the top-left and most untypical values bottom-right. The numbers in the headers of the resulting rows/columns are their original position in the data and the numbers in the footer are the conformism values at the time of being "thrown out" of the calculations.

Detailed usage:
You can write your input text in the textbox or into a file Alt+I or "Intput" checkbox toggles that. Filenames can be relative or absolute
You can see the output in the other textbox or put it into a file with Alt+O toggling the choice.
Alt+S will start the program (performance is poor about O(n²). Alt+E will exit. Processing time shows the time it took in milliseconds.
Examples from performance on Core 2 T7200 @ 2GHz:

This isn't exactly nxn, but a bit less. When increasing data size 1.5 times, it will take 2 times longer or something O(n^1,5) maybe

Don't know if it has a better name in English, but that's what it was called in school: "Miinustehnika"

Implementation in ASM
64-bit source+executable
32-bit source+executable
Test data generator in FASM macrolanguage
(64 one is about 10% faster than 32)

just wondering: what's it good for?

Preparing for Huffman encoding?

exercise for the reader, perhaps?

heh, I don't know if NP-hard algorithm should be used, but maybe...

Well, I don't know what is it used for exactly, but the lecturer had some examples like perhaps statistics. You have persons that have a car and others that don't. Then you have people with good healt and others with bad
Now you can put them in a table as 0-bad, 1-good, 0-no car, 1-w/ car like this:

After arranging you can come to some conclusions like "people with cars have commonly healthproblems" or something like that...

just running exe i supposed there buttons should be somewhere below my screen
(i use 800x600 resolution - i hope it is valid )

Damn, I'm really sorry - I was just thinking that my 1280x800 res laptop is the lowest you can get and thought that it would run no prob. on 1024x768 too. There is a really simple fix for this:
Around lines 760-766 (depending on source 32/64) change the upper-left to some other point on screen:
dialog main_dlg,"Subtraction technique",7,7,390,375,WS_CAPTION+WS_POPUP+WS_SYSMENU+DS_MODALFRAME
I put 7,7 there because I don't like it at the very corner

thnx there is nothing esspecial here in real: more and more of new application can not reduce their windows to fit into less then 1024x768, forcing users sell their aquariums and buy TFT, but i prefer low resolution even on TFT as more soft for eyes (usage of high resolution with large fonts brings other kind of problems with some software )
----
in real you change left upper corner position, but i need smaller size
i had try to modify dialog, but failed to compile: fasm says "Out of memory" (even fasmw on original M..Eng32.asm) (maybe because of national labels ??? )

here my variant of dialog (modify labels and "01010" text to get it proper if you wish to try)

Well its not about labels, but should be called "BAD PROGRAMMING" I've allocated rb 1 shl buffersize so when buffersize is 27 then about 131MB will be allocated and FASM needs all that space at the compile-time. In case there is other stuff at the end...
My compiler setup says 524288 in the memory inputbox

I've done some very resolution-independent apps in MenuetOS, but in Windows they're making it so hard to do :S I'll have to catch the resize event and then calculate everything based on DPI ^o) or sth. The project is in "good enough" status.

now clear, thanx

over half of gigabyte?!? wow, man, there surely is a better way