In sample tests the complex problem mtest6 was fixed as mtest6g. see:
RubyApfp07 has truly arbitrary precision logarithms - finally. I had a hard time figuring this out for some reason. RubyApfp07 is used by RubySode07. Also improved norm & naturally sqrt & "**".
NOTE: this is still an alpha release. I have fixed some problems that I haven't pulled together yet & still need to fix a few more. Most of the problems are in APFP (but not ureal). They mainly affect performance & memory usage.
Sample problems & solutions using usode are now on the web page http://sode.sf.net/ They are in the download as well.
There is a new version of sode -- usode. It uses APFP (a sourceforge project) for floating point.
You can uses either apfp or ureal found there. It is still very new. The latest version of usode is found in cvs. I have decided to make more releases yet. There is a release for APFP as it is more stable. ( I think)
I intend to add a "machine" section to isolate machine dependences -- such as constants to avoid under/over flow. Any suggestions?
You can do the same thing with atomall.
You can create batch files for your OS to run the Icon or Unicon compilers (redirecting the stdout & stderr to files) This make the easer to look at & see them. The only drawback is that you get no prompt for the problem file name (minus the suffix).
Sode33 was released yeaeterday. Fixing problems having to do with H and input checking (mostly).
sode32, now released, does basic validation on the input data. This occurs when the generated program is run.
Sode30 is released with much better error control. And a fix for diff's on the RHS.
It turned out that sode28 had a severe bug introduced, and some others not yet fixed. THE bug has been taken care of, & some of the others. These have been checked into cvs. A release 29 will be made soon.
SODE28 with improved singularity handling and a little better step size conrtrol. Also there is a new variable which may be used in the problem file -- look_poles . If set to 1, poles will be looked for. If 0 , the defaullt, they won't be.
I'm still working on step size control, mostly with the STIFF hires diffeq. I get overflow or underflow after many iterations. I hope to have this fixed soon.
The web pages at http://sode.sourceforge.net have been revised.
SORRY -- I'm still working on this.
Mostly improving step size control. Also making corrections on .ide files.
SODE25 will mainly improve the increment handling.
Consolidating SODE23 & improving efficiency of atomall.
cvs containes an improved version of SODE.icn for atomall. Also The corrected version of sats.txt is in cvs.
Sats23.zip & sats23.tar.gz contain fixes for sats.txt & sats.icn in sode23.zip & sode23.tar.gz.
You will also want to make the change the change to sats.txt in the prior news message.
in wrk_prod, the "&" should be "and".
sode23 seemd to fix the trouble if a small enough increment is taken. (adjust_h := 0; if needed). The code the adjust the increment needs examination -- but I had to get the rest right first.