This release not only has the big speed and space improvements of V2.5.2 - it also should work almost everywhere! :)
(Plain 2.5.2 had some bugs in setup_dirtree.pl that didn't even provoke a notice from my Perl 5.6, but caused older Perls to abort).
Many small bugfixes & other improvements.
The 2.5.0 release can now be downloaded as a nice complete tar.gz. It seems to work fine with kernel versions slightly different from 2.4.0 (tested it with 2.4.1, 2.4.3 folows tonight). The required total processing time of this version is about 70 minutes on my Celeron 400.
Well, unfortunately I can't upload a release package of 2.5.0 due to a bug in konqueror (and due to netscape not wanting to run on my system), but it's "released" nevertheless (CVS tag release_2_5_0).
Changes from 2.4 to 2.5:
(1) Rewrote most of the shell scripts in Perl (creinig)
(2) Optimized the shell/Perl stuff (processing time reduced by about 87%) (creinig)
(3) Optimized the PS output a bit (file size 60MB instead of 180) (creinig)
(4) Removed some of the dependencies on the kernel coding style (creinig)
(5) Made the code a bit more understandable / extensible (creinig)
(6) Added a0 paper format to posterize (creinig)
Ok, the thing is in a working state again (testing takes simply too long :( ) and it even features some improvements. First it's even faster (3.3 hours) and second (much more importantly) it generates much more compact postscript.
The generated image.ps (originally 180MB) is now only 60MB big - without loss of information. Printers rejoice :)
The readability problem still exists of course (and I don't think there's a solution to that, aside from splitting it up into multiple A0 posters...).
I made some changes to classify_nonstatics.pl, resulting in more than doubled performance. The time spent in this script is down to ~40 minutes (from ~105). That should move the total processing time from 4 hours to about 3 hours (all on my Celeron 400)
Please note that classify_nonstatics.pl may need up to 90MB of RAM at peak time (60MB minimum). Without that the good performance wouldn't be possible.
I don't say it's bug free, but it has generated a huge postscript file which at least partly consists of the proper code. It's only about 160MB instead of the 180 it should be, but that could be because the new code does some slightly stricter filtering.
Well, at least it is significantly faster than the old code. I don't have a hard number as kde crashed in the middle of the timed run, but I think it took about 4 hours on my Celeron 400.
Almost forgot this :)
Ok, this project was created from Rusty Russel's "Linux Graphing Project" (building a 180MB PS file out of the linux 2.4.0 kernel sources). His sources are available in the downloads area.
As to me, well, I did some work on that code (speeding it up quite a bit), and since Rusty has very little time right now he asked me to take care of that stuff for some time. My improved code will appear in CVS soon (no download yet, as there are still some fatal bugs).
I just added Rusty's original "Linux Graphing Project" source tarball to the downloads section. It's specifically tailored to the Linux 2.4.0 kernel sources, and it's quite slow.