This (8/26/2016) is the README for rev0062 of omnisode.
I added support for Maple 2016 (under Ubuntu Linux 14.04).
I no longer have Maple for Windows.
Also I had to make small change to tomni.rb for difference in command line order for gcc and g++. And commented out some dead code and made small corrections
to format statements noticed by newer c & c++.
This (4/24/2015) is the README for rev0061 of omnisode.
Slight change for APFP.
This (4/24/2015) is the README for rev0060 of omnisode.
Slight changes for Maple.
This (4/24/2015) is the README for rev0059 of omnisode.
Fixed slight problem with release script.
This (4/24/2015) is the README for rev0058 of omnisode.
I have added support for my arbirarary precision floating point (APFP)
Ruby code. It is too slow to be of much practical use, but the interval
arithmetic may be of some interest. See http://apfp.sourceforge.net/
for more on it. Also I added a neg function for monadic minus. It was simpler
than an actual monadic minus and increases efficiency just as much.
Also, when using APFP, the floats in the ode file must be passed through the
convrsion function "c". This doesn't hurt anything for the other languages.
The optimization of H is turned off for now. Also looking for poles can
be turned on or off. I had tested quite a bit for c, c++, Ruby and APFP.
Now I have made quite a few changes for Maxima and it seems to work with
limited testing. I am making this "release" in order to move code to another
computer where I can work on Maple. Also to test revised release script.
This (9/23/2014) is the README for rev0057 of omnisode.
Rev0056 was a mistake - the code was unchanged. But I did
discover I could use a much smaller h in many cases.
I am including updated ode and odes test files.
Otherwise the code in 0056 and 0057 matched 0055.
This (9/22/2014) is the README for rev0056 of omnisode.
I was comparing my results with Dr Y. F. Chang & Dr George Corliss's
ATOMFT. I was reminded more accutely of something I was aware of.
Namely I waa having to use a smaller h (increment) than I should.
After thinking where such an error could occur I examined the generated
code for the "jump" logic carefully, which waa where I thought such an
error most likely. I discovered an "/" that should have been an "*".
I have rerun many of the tests, and seem able to use a much larger increment
than before. In partucular I used an increment of 1.0 in "sin_lg_h.ode"
and still got 16 place accuracy with gcc doubles. Two cases still do not
give good accuracy and both seem insensitive to smaller h - mamely
mtest2.ode and mtest6.ode. They need more attention. Also I made the
display interval default to 0.1 - which solved a problem I discovered.
This (8/28/2014) is the README file for rev0055 of omnisode.
Added evalf in Maple for estimation of error
Changed "analytic" to "closed_form"
Used glob_prec instead of literals in check in error calculations.
Simplified iteration count - eliminating unneeded variables.
Removd some commented out code.
This (8/27/2014) is the README file for rev0051 of omnisode.
Fixed problem in preodein.rb regarding absolute value.
Fixed problem with indentation in preindent.rb.
This (5/17/2014) is the README file for rev0050 of omnisode.
The only significant change since rev0049 is a correction for estiimating
the error for Maple -- using the value of Digits correctly in estimating it.
It seems to be functioning well, although I have plans for more functionality.
The focus since rev0046 has been on the user interface. The shell scripts
have been completely replaced by a ruby program (tomni.rb) which runs omnisode
for linux, cygwin or windows. It also works for any of c, c++, ruby, maxima
or maple. It reads a file of names of ode files (---.odes). You now simple
type "ruby tomni.rb" at a prompt (it trys to reecognize the os, and prompts
if it doesn't). Then it asks for the name of the odes file containing the
names of the ode files (without the ode suffix) and optionally followed
by comments in in quotes. If the directory this would use (using the language,
odes file name and os) already exists you may select a different part of the
name for the odes part. Then you are prompted for which lanhuages you want
omnisode to be executed for. Then it runs omnisode for all the files in the odes
file for each language selected. The results are placed in the
html/omniresults/language_odes_os directory. where language is replaced by the
language, odes is replaced by the odes name (or its replacemeny) and os by the
name of the os. A web browser can then be pointed at table.html in each
directory so created.
tar -xzvf omnisode0050.tar.gz
to extract the files.
The contents will be placed in a subdirectory which will be created.
Documentation is on this web site http://sode.sourceforge.net/ at:
(on input ode file - defines diffeq's to be solved)
(on the shell scripts to run omnisode)
The documentation needs some updating (which I plan to do soon) but is
I have extensively edited this README file when masking rev0049 to make it
I plan to edit the other documentation soon, it is on the web site, but
not in the archive.
The main thing missing is that the size of the increment is not determined
well. It seems that glob_max_h in effect controls glob_h - the increment.
I have been busy with determining how close any singularity is, and now
have that working most of the time - there are test results of this on the
web site. These are prerequisite to selecting a glob_h. Also it would be
good it a different max_terms could be recommended if the one in the
input ode file was not a good one - it cannot be changed dynamically.
The program can generate code for c, c++, ruby, maxima and maple.
I may be reached by email at firstname.lastname@example.org
The omnisode web site is
and Dennis Darland's personal web site is: