home *** CD-ROM | disk | FTP | other *** search
-
-
- @closing(5-SEP-85)
-
-
- This document outlines the usage of my version of the Bose-Nelson
- sorting algorithm presented in the September 1985 issue of Dr.
- Dobbs. As far as I could tell, both programs that were presented
- (in good faith I presume), were not functional as written.
- Rather than try to correct the bugs, I wrote my own version of
- the recursive program.
-
- The version of "BOSE.COM" that is contained in this library will
- run on Z80 CP/M only. You will have to recompile "BOSE.C" to
- suit your machine and C compiler, if this is not the case.
-
- The "BOSE.COM" program will generate a program containing the
- swap pairs for a known, fixed set of elements. To use the
- program, type "BOSE <#> <filespec.ext>", where "#" is the number
- of elements to be sorted and "filespec.ext" is any legal file
- name that you wish. I use ".c" as my extension.
-
- A few notes about the efficientcy of both the speed of
- generating the swap pairs and the object code size. The
- Bose-Nelson algorithm begins to take excessive amouts of time
- when the number of elements gets above 200-300. For this many
- elements, there are about 12,000 ( right! 12K) swaps. Thats
- about 20k in swap code alone for my rather inefficient (remaining
- nameless) C compiler. However, the speed is quite excellent.
- For the example ("STEST.C") sort of 50 random integers, it puts
- quicksort to shame. Try it!
-
- Anyway, enjoy the program. I'm sure that you will be pleased to
- have a working version of the algorithm. Several improvements
- are in order, but the first that comes to mind is to reduce that
- number of swaps involved for large amounts of elements. It was
- mentioned in Dr. D's that this routine did not generate optimal
- sorts. They weren't kidding.
-
- @closing(Mark D. Lougheed
- 6704 Sierra Drive S.E.
- Lacey Wa
- 98503)
- legal file
- name that you wish. I use ".c" as my extension.
-
- A few notes about the ef