Imported Upstream version 4.6.0

[debian/atlc] / ChangeLog

****** Version 4.6.0 Realeased ??? November 2003  
top level directory:

atlc/update 
has been modified so it updates the NEWS.html file. 

atlc/examples directory.
Have removed the following files from the distribution:
atlc/examples/100-Ohm-201b.bmp \
atlc/examples/100-Ohm-401b.bmp \
atlc/examples/200-Ohm-201b.E.bmp \
atlc/examples/200-Ohm-201b.Er.bmp \
atlc/examples/200-Ohm-201b.Ex.bmp \
atlc/examples/200-Ohm-201b.Ey.bmp \
atlc/examples/200-Ohm-201b.U.bmp \
atlc/examples/200-Ohm-201b.V.bmp \
atlc/examples/200-Ohm-201b.bmp \
atlc/examples/200-Ohm-401b.bmp \
atlc/examples/200-Ohm-801b.bmp \
atlc/examples/25-Ohm-201b.bmp \
atlc/examples/25-Ohm-401b.bmp \
atlc/examples/25-Ohm-401h.bmp \
atlc/examples/25-Ohm-801h.bmp \
atlc/examples/50-Ohm-201h.bmp \
atlc/examples/50-Ohm-401h.bmp 

I'm not sure they were that usful, given there are plenty of 
files like 50ohm-201h.bmp etc. 

Remove atlc/examples/13inner_22outer_coax.bmp.txt too, as there was
no 13inner_22outer_coax.bmp to match up with it. 

atlc/docs/html-docs/accuracy.html 
Completed updated all results to those obtained with ver 4.6.0
Put into 4 sections detailing:   
Section 1. Two conductor Transmission Lines with a Uniform Dielectric
Section 2. Two-conductor Transmission Lines with a non-uniform dielectric
Section 3. Accuracy of atlc with coupled lines
Section 4. Conclusions about the accuracy of atlc

Added a completly new section:
2.1 Comparision of atlc and a dual dielectric coaxial cable



atlc/src/non_gui/create_bmp_for_symmetrical_stripline.c
This no longer computes the theretical impedance if the 
structure is too narrow. Only computes if the -v option
is added. 

atlc/src/create_bmp_for_circ_in_circ.c
Will now print sensible error messages if the inner conductor is too big
or if the inner and outer will touch since the offset is too large

atlc/src/set_oddity_value.c
This sets a global array unsigned char **oddity; to a value describing
if the pixel is matallic, has conductors to the right/left etc.
cell_type was not adequate for this, since once you had set one cell,
you did not know the result for those around it. #

atlc/src/non_gui/print_data_for_directional_couplers.c
No longer fails to print intermediate values of impedance if 
verbose is >2 as it did before. 

atlc/src/non_gui/write_fields_for_two_conductor_lines.c
This can now write the oddity values from set_oddity_value()
to a bitmap file, although this has been commented out by default. 

atlc/src/
A huge number of changes were made to this release. Not every
one is fully documented, as to do so I would have spent as long
documenting the code as writing it. 
****** Version 4.5.1 Released 16th October 2003   ****
A problem with multiple dielectrics has been discoved, and aslo with 
multi-threaded operating. As such, both have been disabled. This affects
many files. 

configure.ac
changed version to 4.5.1
No longer accepts the option --with-threads, since I have some
suspicions about threaded code. 

docs/html-docs/examples.html
Removed the example of multi-dielectrics, as that is not working
properly now. 

docs/html-docs/FAQ.html
Changed the answer to 'No' about the qestion of multi-dielectrics

docs/html-docs/multi-processing.html
Mentioned this has been disabled. 

docs/html-docs/accuracy.html
Pointed out multiple dielectrics are not working properly

src/non_gui/definitions.h
Removed two duplicate function definitions for 
free_cmatrix and ustring, whilst removing the #define
around the #ifdef ENABLE_MPI so there is only one  
definition of do_columns now. 

Added the variable 'size' to several functions - 
do_fd_calculation.c, write_fields_for_directional_couplers.c,
write_fields_for_two_conductor_lines.c


src/non_gui/do_fd_calculation.c
Added variable 'size' as a command line argument to stop
it needing to be declared global. 

src/non_gui/do_fd_calculation.c
Added variable 'size' as a command line argument to stop
it needing to be declared global. 

src/non_gui/write_fields_for_two_conductor_lines.c
Added variable 'size' as a command line argument to stop
it needing to be declared global. 

src/non_gui/atlc.c
size_t size is no longer global but local to main. This would at first
checking appear to stop the issue with all (which was one or two) 
test failures on AIX 5.2. Later I found this is not so and there
is a problem in the multi-threaded code for atlc.

tests/WARNING--The-next-test+benchmark-takes-a-long-while.test
return 0 (pass), as previously it did multi-dielectrics. 

tests/Makefile.am
Removed 2[b-h]* from the tests. 2a-create_bmp_for_rect_in_rect.bmp.test
is still used. 


****** Version 4.5.0 Released 13th September 2003 ****

The ChangeLog shows version 4.5.0 being released on
the 13th September, when it fact it was the 11th of 
October. 

WRONG XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
****** Version 4.5.0 Released 13th September 2003 ****
WRONG XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

NEWS
mention lattest news - mainly correction of a few errors
and the fact its no longer necessary to link in gsl. 

remoterun 
changed, as the old version was accidently deleted and 
I could not be bothered to read from tape. Might not be
100% okay, but it is not needed for use of atlc - only
during testing. 

'update' added to CVS tree
Added a file called 'update' that generates docs/html-docs/*.1.html
and docs/html-docs/BUGS.html This uses header and footer for the 
generation of docs/html-docs/BUGS.html

'header' and 'footer'' added to CVS tree
used with 'update' to generate docs/html-docs/BUGS.html

README.checks
Removed, as information is out of date. 

tests/find-zero-length.test
Added a file to check for the existance on zero length
files in the source distribution. 

configure.ac
Incremented version to 4.5.0
Removed tests for gsl and all that involved. 
Removed tets for the sizeof short, int and long
as they are not needed. 

man/man1/create_bmp_for_stripline_coupler.1
changed email address to make it less easy for spammers to grab. 

man/man1/man-pages.html 
Added a new man page for design_coupler.1.html

src/calculate_Zodd_and_Zeven.c:
Added #include gsl_types.h and #include "gsl_definitions.h"

src/non_gui/usage_create_bmp_for_symmetrical_stripline
made it print the fact the -v option computes the 
theoeretical impedance. Before the binary did it, but
the usage message did not say so. 

docs/html-docs/jpgs/home-email.jpg 
Added new file with my email address, that is not easy to 
read automatically. 

Removed all occurance of my email address from the html
files, so that those published on the web are less likely
to attract spam. 

src/ 
All the following had my email address removed
usage_create_bmp_for_stripline_coupler.c 
usage_create_bmp_for_rect_in_rect.c 
usage_create_bmp_for_rect_in_circ.c
usage_create_bmp_for_rect_cen_in_rect_coupler.c 
usage_create_bmp_for_rect_cen_in_rect.c 
usage_create_bmp_for_microstrip_coupler.c
usage_create_bmp_for_circ_in_rect.c 
usage_create_bmp_for_circ_in_circ.c 
usage_atlc.c swap_conductor_voltages.c setup_arrays.c
read_bitmap_file_headers.c
print_data_for_two_conductor_lines.c
print_data_for_directional_couplers.c
print_copyright.c
memory.c
get_options.c
get_file_pointer_with_right_filename.c
get_Er1_and_Er2_colours.c
finite_difference_single_threaded.c 
finite_difference_multi_threaded.c
finite_difference_mpi.c
find_maximum_values.c
find_energy_per_metre.c
find_electric_fields.c
fill_rect_in_rect.c
fill_rect_in_circ.c
fill_circ_in_circ.c
exit_with_msg_and_exit_code.c
do_fd_calculation.c
create_bmp_for_rect_in_rect.c
create_bmp_for_rect_in_circ.c
create_bmp_for_rect_cen_in_rect_coupler.c

tests/8a-design_coupler.bmp.test
Removed the test for gsl library and so to exit with 77 if not present. 

Added a second checksum:
aebd9a6d1c1db548d39723edec454640
that is acceptable, as this will be produced on 
Windoze machines.

tests/6e-create_50ohm_401Pixel_high_symmetrical_stripline.test

tests/7n-create_bmp_for_stripline_coupler.bmp.pre-atlc.txt.test
Removed the test for gsl library and so to exit with 77 if not present. 

tests/10b-create-lots-of-examples-check-last-one.test
Removed the test for gsl library and so to exit with 77 if not present. 

tests/10a-create-lots-of-examples-check-last-one.test
Removed the test for gsl library and so to exit with 77 if not present. 

src/non_gui/gsl_sf_ellint_Kcomp.c
Added new file that basically has the gsl_sf_ellint_Kcomp
function for computing the elliptic integral. This requires
several functions for it to opperate, as it calls several 
functions. These are all in the same file gsl_sf_ellint_Kcomp.c


src/non_gui/definitions.h
Added quite a few function definitions, #defines etc to allow
the gsl_sf_ellint_Kcomp function to work. It needed several
other functions added. 


****** Version 4.4.4 Released 13th September 2003 ****
configure.ac
Version incremented to 4.4.4

NEWS
mention made of the fact the sources have been
changed primarily so it should build under Windoze.

docs/html-docs/jpgs/Makefile.am
added bluefish.jpg properly to the file - it was 
missing partically before, so bluefish.jpg was
not included in the distribution. 

docs/html-docs/create_bmp_for_symmetrical_stripline.1.html
Autogenerated, reflecting changes in man page, which
was a change in the warning information about the 
inability of man pages to fully describe the atlc
project. 

docs/html-docs/create_bmp_for_symmetrical_stripline.1.html
Autogenerated, reflecting changes in man page, which
was a change in the warning information about the 
inability of man pages to fully describe the atlc
project. 

man/man1/sysdata.1
Removed text saying man page is not a full set of 
docs, since in the case of the simple sysdata.1, the
man page adequately describes the function. This is 
no so really with some of the other man pages, because
your can't really describe fully how to use such   
programs with man pages. 

src/non_gui/create_bmp_for_rect_in_rect.c
Changed fopen to open in binary mode, to make it work 
under Windoze. 

src/non_gui/find_optimal_dimensions_for_microstrip_coupler.c
Changed fopen to open in binary mode, to make it work 
under Windoze. 

src/non_gui/create_bmp_for_symmetrical_stripline.c
Changed fopen to open in binary mode, to make it work 
under Windoze. 

src/non_gui/create_bmp_for_rect_in_circ.c
Changed fopen to open in binary mode, to make it work 
under Windoze. 

src/non_gui/create_bmp_for_rect_cen_in_rect_coupler.c
Changed fopen to open in binary mode, to make it work 
under Windoze. 

src/non_gui/create_bmp_for_rect_cen_in_rect.c
Changed fopen to open in binary mode, to make it work 
under Windoze. 

src/non_gui/create_bmp_for_microstrip_coupler.c
Changed fopen to open in binary mode, to make it work 
under Windoze. 

src/non_gui/create_bmp_for_circ_in_rect.c
Changed fopen to open in binary mode, to make it work 
under Windoze. 

tests/8a-design_coupler.bmp.test
Changed so the temporary storage is no longer in /tmp
but instead in $top_builddir/tmp/results. This should be 
compatible with Windoze, but I don't know for sure if it is
yet. I don't build Windoze binaries. 

tools/src/mymd5sum.c
Changed fopen to open in binary mode, to make it work 
under Windoze. 

tools/src/myfilelength.c
Changed fopen to open in binary mode, to make it work 
under Windoze. 

tools/src/locatediff.c
Changed fopen to open in binary mode, to make it work 
under Windoze. 

tests/benchmark.c
Changed fopen to open in binary mode, to make it work 
under Windoze. 


****** Version 4.4.3 Released 11th September 2003 ****

Replace 'kb' and 'Mb' with 'kB and MB, as I was advised the lower
case is used for bit and the upper case for byte. These required 
changes in tests/benchmarks.test

man/man1/sysdata.1
Added a man page for the program sysdata in man/man1/sysdata.1
This has about 11 examples, which are there more for my own use
that anyones elses. When possible I'll try to expand the files
that gather data about the hardware.

man/man1/*.1
Corrected many typo, spelling errors etc in all the man pages.
more significantly, the 


docs/html-docs/accuracy.html
Corrected a few typos and made the page a bit clearer
I hope. Any HTML that did not strictly follow the stanard has
been corrected, so it should pass the test at 
http://validator.w3.org/check/referer
hence an W3C logo has been added at the bottom of the page.

docs/html-docs/accuracy.html
A bluefish logo has been added too.

The spamtrap on the page has been changed a bit, which is a bit
annoying as it is now visable. But the code was not strict 
HTML before, so there were problems in making it valid 
otherwise. 

tests/Makefile.am
References to the gsl library have been removed, so the 
thread tests will pass if the gsl library is not present.
This was changed just in case someone runs a binary 
in pkgadd format on a Sun without having the gsl library
installed. Some binaries will break, but Test_threads_a
and Test_threads_b should now still pass, even if 
the gsl library is not present. No functions from gsl
are needed on these two files, so it was sensless linking
them against the library. 

tests/try_portable.c
The code has been changed a bit so it should now (I hope)
compile under Windoze too. Someone has done a Windoze port
but found a bug in the code, which has been corrected. 

tests/try_solaris.c
Added code to make it cleaner and so no variables 
get declared if not needed.

****** Version 4.4.2 Released September 6th 2003 ****

src/non_gui/readbin.c
swapped the second (size) and third (number) arguments around
on an fread, as were the wrong wary around - not that I think
it will make any difference. 

src/non_gui/atlc.c
swapped the second (size) and third (number) arguments around
on an fread, as were the wrong way around - not that I think
it will make any difference. 

src/non_gui/check_for_boundaries.c
put a check in for j=0 which could cause a progrom on the 
Digital UNIX (spe145.testdrive.compaq.com) system on 
the HP testdrive site. A real bug found at least!!!!

tests/try_bsd.c
put some checks for NGROUPS, including a couple more 
header files if NGROUPS is not defined. This causes a 
problems on OpenBSD 3.2 for SPARC (my SS20), so was 
fixed.


tests/try_linux.c
Added a call to get the number of CPUs. Appears not to be
documented, so a bit risky and gives the number of processors
configured, rather than online, but its is better than nothing. 

****** Version 4.4.1 Released September 4th 2003 ****

configure.ac
Changed version to 4.4.1 
Added sys/sysctl.h and sysctl as these are used on BSD try_bsd.c
to gather information on a BSD UNIX system. 

docs/html-docs/index2.html
Changed a typo or two. Added an entry for Unixware, as someone emailed
me to say iversion 4.0.0 works on UnixWare. 

tools/src/locatediff.c
Corrected an error so this now works. A change recently caused this to
break. It now reads both files. 

tests/try_irix.c
Much improved, finds information about cache and main ram on IRIX now
Also speed of CPU, and type of CPU and FPU, although the latter are not
decoded into a sensible number like R10000. 

tests/benchmark.test  
information now prints the fact the IRIX information is good, not 
poor as in prior versions. 

THANKS 
Added new testers and thanks the cray site's administrators. 

Now informs the amount of RAM on Linux and BSD systems is a little
larger than the value reported.

tests/Makefile.am
Added a new binary sysdata, which prints data in this format. 

Hardware provider:    Sun_Microsystems
Hardware platform:    SUNW,Ultra-80
Machine:              sun4u
Sysname:              SunOS
Release:              5.9
Version:              Generic_112233-06
Nodename:             sparrow
#CPUs supported:      4
#CPUs online:         4
CPU type:             sparcv9
FPU type:             sparcv9
Speed:                450 MHz
RAM:                  4096 Mb
L1 data cache         unknown kb
L1 instruction cache: unknown kb
L2 cache:             unknown kb

src/non_gui/finite_difference_multi_threaded.c 
removed the argument 'number_of_iterations' as not used

src/non_gui/write_bitmap_out.c
added checks for the return value of fwrite

src/non_gui/write_fields_for_two_conductor_lines.c
added checks for the return value of fwrite

src/non_gui/memory.c
removed all unneeded functions from this file

src/non_gui/convert_create_bmp_for_rect_in_rect_dimensions_to_integers.c
removed extern variables Ers, colours and names, as these are defined
in the header files definitions.h, Ers.h and Erdata.h

src/non_gui/man_functions...
Tidied up, to keep Sun's lint happier

BUGS
Removed bug about linking with old versions of gsl. 
Corrected the fact hardware data is gathered on IRIX
okay and its only Linux and BSD it fails on. 


****** Version 4.4.0 Released August 30th 2003 ****
BUGS 
Added  0 - those I don't know about, 
Added  3 - tools/src/locatediff don't work on AIX. 
Added 11 - documentation gets out of date
Added 12 - CVS sometimes get out of date  

configure.ac
Updated the version to 4.4.0, as there are some major(ish)
changes to allow compilation on the Cray Y-MP EL, which
previously atlc would not work on. 

Added a test for the size of a short, since on a Cray Y-MP, an int is 8
bytes, so this screwed up things in src/non_gui/definitons.h, as 
it needed to define int32 as something. 

Changed the default from not enabling hardware information to enabling it.
It can be changed with --disable-hardware-info

Added a test for _SC_CRAY_NCPU and _SC_CRAY_CPCYCLE, which will help
debugging any problems. 

Added check for sys/vid.h, which I'm hoping is just on IRIX. 

NEWS
mention atlc now runs on toys and supercomuters and the fact an option has 
nw become the default. 

README.hardware-info
new file added, listing information about what hardware data is 
gathered on what systems. 

README.--enable-hardware-info
removed, as the option no longer exists - it's the default. 

INSTALL
Changed install insturctions, as --enable-hardware-info is now the 
default. 

Makefile.am
Added README.hardware-info removed README.--enable-hardware-info

tests/Makefile.am
Added test 5c-check_numeric_data_from_atlc-2conductor.test
as well as try_irix.c and try_unicos.c
tests/try_unicos.c
Added a function try_unicos() to obtain data about the hardware of 
a Cray supercomputer. 

tests/try_linux.c
Removed some variables that were now now longer needed and so generated
a warning from the compiler. Variables were operating_systems, i, ret
and one other I think. 

tests/try_solaris.c
Removed some variables that were now now longer needed and so generated
a warning from the compiler. Variables were operating_systems, i, ret
and one other I think. 

tests/try_irix.c
Added a file to find out information on the SGI running IRIX. 
Currently little information is found. 

tests/try_unicos.c
Added a file to find out information on the Cray. 

tests/Makefile.am
Added entries for try_irix.c and try_unicos.c

tests/benchmark.test
Changed the wording about what data is collected on each system, as
previously it stated only Solaris was reasonably well supported, but
that is not true now, so it has been changed. 

tests/1g-create_bmp_for_symmetrical_stripline.U.bmp.test
Added the md5 checksum 2a9644a0206c2314006e298f07aae233 as being okay,
since this is generated on the Cray Y-MP and differs by only a single
pixel and a single gray-level, from that on the Sun, suggesting 
its just rounding errors. (I don't think the Cray implements IEEE maths
anyway). 
 
tests/7g-create_bmp_for_stripline_coupler.U.odd.bmp.test
Added a new checksum 09f68e1466d68bf8a03e1f90eff411bb which
is produced on a Cray. 

src/non_gui/write_bitmap_out.c
MAJOR change, so this now writes data out in ONE consistant manner
that should be very portable. Note this is used only for programs 
like create_bmp_for_*, and it not used by 'atlc' itself. 

src/non_gui/write_fields_for_two_conductor_lines.c
changed the name 
extern bitmap_file_buffer to 
bmp_buff as the latter was
shorter and space was at a premium, to stop one needing huge long
lines everywhere. 

src/non_gui/read_bitmap_file_headers.c
MAJOR change, so this now writes data out in ONE consistant manner
that should be very portable. This is used only for programs 
like create_bmp_for_*, AND 'atlc' itself. 

src/non_gui/byteswap.c
Removed all functions except byteswap_doubles. 

src/non_gui/definitions.h
Added definition for try_unicos.c
Removed all entries for byteswap*, except
byteswap_doubles, which is used by readbin.c
Changed definitons of Bitmap_Head_Struct + Bitmap_File_Head_Struct
as these now use just ints for portability. 

src/non_gui/Makefile.am
Removed byteswap.c from all binaries except readbin


****** Version 4.3.8 Released August 21sth 2003 ****
NOTE THE CHANGELOG FOR VERSION 4.3.8 SHOWS 4.3.7 AT THE TOP. THIS WAS AN
ERROR. THE DETAILS ARE VALID FOR 4.3.8. 
configure.ac
Added the lines
AC_CHECK_DECL(_SC_PHYS_PAGES,,)
AC_CHECK_DECL(_SC_PAGESIZE,,)
to enable me to check if _SC_PHYS_PAGES and _SC_PAGESIZE
are defined for debugging purposes only. No use of them is 
made, but it help to determine why code might not work later
on, if its known whether these are defined or not. Problems
reported by someone indicate problems, which would be easier
to track with this option enabled. 
Added a check for errno.h, as it was needed on Suse. Prior to this
many files had #ifdef HAVE_ERRNO_H, but obviously the inclusion of
errno.h was not critical on those systems. On Suse it was. 

tests/try_portable.c
check the return value or uname, to see if it fails or 
not. If it does fail, print message and error if errno
is defined. Replaces any spaces with underscores in the 
ouputs from any part of the utsname structure.

tests/try_linux.c
Removed any code to check for systems parameters using 
uname- it should all be found in try_portable.c, which
is called before any other try_* functions. 

****** Version 4.3.7 Released August 6th 2003  ********

README.--enable-hardware-info
New file added, to give information on what data is obtained on what
hardware platform. 

BUGS
Added a problem with test failures on IBM's AIX with thread support and
removed one about teston now working on the local host, as it now does.
There is a html version of the file in docs/html-docs/BUGS, which should
really be renamed, or some automated way of generating text from HTML,
or perhaps just text and a link. No sure how best to do this really, but
it is not entirely satisfactory. 

configure.ac
removal of code assocatiated with MPI

teston
change an example, so it shows how to configure with the
--enable-hardware-info rather than the --with--mpi, as the former is a
new option and the latter is now removed. 

README.threads
More updated information on what atlc has been tested on, and clearer
infomration on using it. 

README.help
mention I'd like information on building on MP machines others than for
AIX, IRIX, Redhat Linux and Solaris, rather than just Solaris, as the
former have all been tested

README.checks
Major updates to this, as the data in the file was based on an old
version of atlc, that failed numerous tests when multi-threaded, due to
the differing algorithms. Some data on machines has been removed, as it
some of the machines I no longer have in a running state. 

NEWS

Most other changes are just to keep the compiler a little happier in
src/non_gui, but some more significant changes in tests subdirectory. 

src/non_gui/readbin
Forced length to be case to size_t, to make gcc -Wconversion happier.

src/non_gui/design_coupler.c
changed 10 to 10.0 to make a double and so keep the compiler happier
when calling pow

src/non_gui/create_bmp_for_stripline_coupler.c
extra casts and general tidying up, to make compiler give less warnings.

src/non_gui/write_bitmap_out.c
cast to (size_t) to keep compiler happier. 

src/non_gui/memory.c
included sys/types.h  and remote the #define of size_t to an int, again
to keep the compiler happier. 

/src/non_gui/atlc.c
cast numbers to long, to be consistant with the declaration of ustring()
and free_ustring().

src/non_gui/convert_create_bmp_for_coupled_microstrip_dimensions_to_integers.c
changing numbers in the pow(x,y) to force them to be doubles, rather
than ints to keep the compiler happier. 

src/non_gui/convert_create_bmp_for_circ_in_circ_dimensions_to_integers.c
changing numbers in the pow(x,y) to force them to be doubles, rather
than ints to keep the compiler happier. 

src/non_gui/calculate_colour_data.c
changed type of image_type from char to int. Not sure why this was done,
but probably to keep compiler happier. 

src/non_gui/definitions.h
changes to enable strong type-checking in C code. 


tests/try_aix.c
Added string.h to the list of include files.
Removed need for type long long. 

tests/try_hpux.c
Removed need for type long long. 

tests/try_solaris.c
Removed need for type long long. 

tests/try_linux.c
Added file to get infomation on Linux system, but so far it 
does not get that much info - only uname stuff and the ram
in Mb. 

tests/try_tru64.c
Changed to avoid a compiler warning about not returning anything 
from the file, despite the fact it is declared to return an 
integer. 
Removed need for type long long. 

tests/benchmark.c
no longer declares opperating_system, which was an integer that was

tools/src/locatediff.c
forcing variables to be of size (size_t), to reduce compiler warnings. 
assigned a value, but never used" benchmark.c

tests/benchmark.test
now writes to text and html files more consistantly.

tests/Makefile.am
Addes teh file try_linux.c and removes @MPILIBS@, as MPI now removed
from code. 

docs/html-docs/accuracy.html
iremoval of incorrect statement that *ohm-*.bmp were not packaged, but
built. They are now packaged for my convenience. Also mention of make
check

docs/html-docs/building.html
several small changes, but main one being removal of any data about MPI

docs/html-docs/make-check.html
added new file, describing the testing process. 


****** Version 4.3.6 Released August 3rd 2003  ********
The main change is to configure.ac, to remove an unwanted
(and useless) --with-extra-benchmark-data option.

Many sources have been tidied up, to make gcc -Wconversion
happy. 

docs/html-docs/accuracy.html
Removed errors, which incorrectly stated that the lager bitmaps were
not distributed - they are now. 


docs/html-docs/make-check.html.html
Added a new file, describing how the automatic testing of atlc
is performed. 


****** Version 4.3.5 Released July 26th 2003  ********

src/non_gui/atlc.c
Changed number_of_workers from static to global, as 
it did not compile properly. 

tests/benchmark.test
changed the benchmark so speedup is printed properly
(T_sequential/T_parallel) and not the other way around. 

changed at t1 to a t2 on the line that is printed so
the code works okay when compiled multi-threaded. 

changed the conditions under which the efficiency was
computed so that it works okay when the number of cpus
is unknown. It no longer prints 'Inf', but instead 'unknown'


tests/benchmark.c
Now calls an AIX specific routine, which does nothing!

tests/try_tru64.c
Much improved. Gets CPU time now, along with ram and cpu speed.

tests/try_hpux.c
Minor improvements, works pretty good now. 

****** Version 4.3.4 Never released  ********

README.aix 
Removed. Works fine under AIX now. Two issues were 
resolved:
use of char when I needed signed char. Rounding error
caused one pixel to change in colour by one graylevel

src/non_gui
Changed many files in src/non_gui that referenced
char **cell_type; 
I was not aware declared in this way, there is no
standard for whether cell_type consists of signed or 
unsigned characters. Hence it was not portable. This
caused a very large number of failures on AIX. This 
has now been corrected. One failure remains (see above).

Added a file src/non_gui/memory.c to include funtions
scmatrix and free scharmatrix, which  allocate and 
deallocate matrixes with signed characters. 

src/non_Gui/Makefile.am
Removed unnecessary entries of GSL_LIBS, which causes
the SGI linker to generate a warning, that the gsl
library was not used to resolve any functions. 




****** Version 4.3.3 Released July 17th 2003  ********
README.aix 
Added a new file, highlighting problems on AIX.

configure.ac 
Added a --enable-debug option, although little use is made
of this in the code. 

Added --with-extra-benchmark-data which should be an optional
featuure, not an optional package as the configure script says
it is. This needs correcting at a later date. 

tests/benchmark.c
changed the file so that the header file sys/processor.h only
got included if the file existed and processor_info() was 
defined

#ifdef HAVE_PROCESSOR_INFO /* function processor_info() is in Solaris */
#ifdef HAVE_SYS_PROCESSOR_H /* and in OSX */
#include <sys/processor.h> /* But not on Tru64 */
#endif /* <sys/processor.h> does not exist on OSX */
#endif

and made sure that areas of the code that needed aspects of
processor_info always had this header file. The reason is that OSX seems
to have the fuction processor_info(), but does not have the header file
and don't seem to have it defined anywhere. 

Made tests/benchmark.c write benchmark results to tests.log

tests/benchmark.c now gathers information about the processors
on an HP-UX box. 

src/non_gui/Makefile.am
Added function error_check()

src/non_gui/reaadbin.c
Changed an fopen(,"r") to fopen(,"rb"); so it works on
Windoze based systems. 

docs/html-docs
BUGS - removed and replaced by docs/html-docs/BUGS.html

****** Version 4.3.2 Released June 12th 2003  ********

Top level

testson. 
Added a new file 'teston' which allows for rapid testing 
on a number of remote machines. Developers might find this
of general use - it is not specific to atlc. 

README
Correected a couple of typos. Added 'tests' to the list of
directories, as that was missing. Also tools and tools/src.

definitions.h
Remoted a few lines which were commented out
Put void as a function argument - ie foo(void) instead of foo();



tests subdirectrory.

tests/benchmark.c 
This should not be able to obtain the number of cpus on a wider range of
machines, and get the cpu speed on at least some Alpha systems. 

Should be able to determine the RAM on Solaris and Tru64 UNIX.

tests/benchmark.tests
Supplied with the new information from tests/benchmark.c
this should now display more sensible information. 

****** Version 4.3.1 Released June 7th 2003  ********
Unfortunately, the ChangeLog for 4.3.1 was never updated. 


****** Version 4.3.0 Released April 5th 2003  ********

Top level directory

Added a new directory 'm4' to hold autoconf macros.
Moved the top level autoconf macros to 'm4'

Added file 'benchmark' to run a test. It takes 
about 1:17 on Ultra 80 with no thread support. 

ChangeLog
Updated the ChangeLog of course!!

configure.in changed to configure.ac
Renamed configure.in to configure.ac, as this is now the preferable
name for the old configure.in
Significan chanes made to this. Now there is an m4 macro to check for
gsl library and the optimal way to link for threads and the optimal 
way to link for MPI. 
Now accepts gsl version of 0.8 and above. It works on 0.8 but not 0.5. I
don't quite know where the border line lines, but requesting 0.8 dones
not seem unreasonable.

Added a check for automake 2.57 or above. 

Replaced 'dnl' witrh '#' for a comment.

Added a macro gsl.m4 
http://www.ugcs.caltech.edu/info/gsl/usage_4.html
to check for a sepcific version of gsl and to add
entries on CFLAGS and LD_FLAGS

Directory atlc/src/no_gui
atlc/src/no_gui/usage_atlc.c
Removed references to WITH-MP and replaced with ENABLE-POSIX-THREADS
as this was a bug. Now changed code so that if the -t option is
disabled, the code prints the reason (Windoze or not configured
with --with-threads). Previously it gave two possible reasons.


atlc/src/no_gui/atlc.c
removed thr_setconcurrancy and replaced with pthread_setconcurrancy
as this is portable, whereas the other is Solaris specific. 

atlc/src/do_fd_calculation.c
This now calls function finite_difference_single_threaded or
finite_difference_multi_threaded, depending on how the software is
configured and the presence of -t0 option to atlc. 

atlc/src/non_gui/finite_difference_multi_threaded.c
The name of the function has been changed from finite_difference to
finite_difference_multi_threaded - no longer are there two functions
with the same name. The correct one function is called from
do_fd_calculation.c, depending on whether the software is configured
with threads or whether the user has elected to run a single threaded
algorithm on a multi-threaded system, by passing the option -t0 to atlc,
which indicates it will use the single threaded algorithm. 

Changed the algorithm considerably. This has several advantages over the
old one.
a) The algorithm should be suitable with minor modification for MPI use.
b) The code now produces exactly the same results whether run for single
or multi-threaded. finite_difference_single_threaded.c needed to be
changed to enable this.
d) A barrier is now used for syncronisation. 
This algorithm is apparently not very cache friendly, so it could do
with some teaking, but at least its a decent starting point, producing
the same data from single threaded or multi-threaded use. 

atlc/src/non_gui/do_fd_calculation.c
This now calls one of two functions.

a) Code is now inline, rather than a function call.
Directory atlc/tools/src:

atlc/tools/src/Makefile.am
Change bin_PROGRAM to noinst_PROGRAMS so that the programs
used only for testing are not installed. They serve no useful
function apart from when running the tests in atlc/tests.

atlc/tools/src/memory.c
Removed 'memalign' since it seems to be in some libraries, but with
no header files. Also, there seems to be POSIX specific versions of 
it and finally, the compilers will force this with the right
switches. Hence overall, it seems more hassle than its worth to 
have 'memalihn' anywhere. For this reason, the check was removed
from configure.ac too. 

src/gui/Makefile.am
Commented out most of the things, so a C++ compiler is not needed. The 
# files in this directory servered no useful purpose, but the icons at 
# least might be useful. Whether or not a GUI will be added is anyone's
# guess. 

MPI code has been almost fully removed, to start this from scratch again.

****** Version 4.2.12 Released 30/3/2003  ********

Top level:
renamed configure.in to configure.ac, as I gather this is now the 
prefereed name for the file. 

****** Version 4.2.11 Released 30/3/2003  ********
Many many changes!! Not all are documented!
Started to the use autoconf macro acx_pthread.m4 to configure POSIx
threads properly. 

Implemented an md5 checksum program for testing checksums, as this is
very accurate. The one I was using (a hacked 'sum' command) is not that
good and so could miss a change in a file quite easily. 

Spent a lot of time on the test routines. These now always pass and
will work with a VPATH build too. 'make discheck' works fine now.  

Removed several bugs from the function 'do_fd_calculation' so it now
does not cause any floating point exceptions cc is configured to not
implement the IEEE standard. 

Put new algorithms for finding the electric field into find_electric_fields.c,
but they seem less accurate than the old versions, so these are not used for 
now at least. 

Change from 'char *foo' to 'const char *foo' in many files, as this
keeps Sun's compiler a bit happier. Note the Sun compiler (Sun WorkShop
6 update 2 C 5.3 2001/05/15) still moans about a lot of things that gcc
thinks is fine. 

Added the following files that were ommited when put into CVS

**top level directory changes made by drkirkby **
gtk.m4
depcomp
src/non_gui/output_from_design_coupler-Sun-Ultra-80.txt
src/non_gui/Makefile.am

removed a double entry for the header file definitions.h
in create_bmp_for_stripline_coupler_SOURCES.
Added a new binary create_bmp_for_rect_cen_in_rect_coupler


Makefile.am 
Remmoved entry for README.alpha. It's contents were out of date but we might 
want to add a file with the same name later. Who knows?

configure.in 
Added a line to check for the mpi.h header file, immaterial of whether
or not the user request MPI support. This will aid debugging. 
Changed the message about being unable to build 'make_coupler' to
'design_coupler' since the binary had been renamed.

configure  
Autogenerated file, needed updating due to change in configure.in
with reguare to MPI support (see entry for configure.in above.)

**man directory - changed made by drkirkby**
Renamed many files to take account of the new binary files names. Hence
they all needed updating. A new man page for design_coupler has 
been written. Currently there are the following man pages in
man/man1
atlc.1
create_bmp_for_circ_in_circ.1
create_bmp_for_circ_in_rect.1
create_bmp_for_microstrip_coupler.1
create_bmp_for_rect_cen_in_rect.1
create_bmp_for_rect_cen_in_rect_coupler.1
create_bmp_for_rect_in_circ.1
create_bmp_for_rect_in_rect.1
create_bmp_for_stripline_coupler.1
create_bmp_for_symmetrical_stripline.1
design_coupler.1
find_optimal_dimensions_for_microstrip_coupler.1
readbin.1

src/non_gui
usage_create_bmp_for_symmetrical_stripline.c
corrected a simple spelling error

renamed error_and_exit() to exit_with_msg_and_error_code()
and are in the process of changing every command to exit
with that, rather than with exit().     

check_parameters_of_create_bmp_for_microstrip_coupler.c
Added better error checking, as the orginal code just did one silly 
check. The improved error checking was done to try to find a 
reason there are problems on a Dec Alpha running Linux and gcc-2.95.4

do_fd_calculation.c
changed (on a termporary basis) ITERATIONS from 100 to 10
Added code so that with -vvvv we can see how many iterations were used

calculate_Zodd_and_Zeven.c
Added code to check for both the gsl libary AND
gsl header filess before considering gsl proplerly
present. A bug was noticed when using Sun's compiler whereby
the gsl library was found in /usr/local/lib, but the gsl
header files, which are in /usr/local/include/gsl were not
found. 

print_data_for_two_conductor_lines.c
Added option to print data to high-precision


Added file convert_doubles_to_integers.c although 
at present this is not working as planned. It is not
linked in with any code at present. 

usage_atlc.c no longer exits with 0, but
returns to the calling routine

atlc.c 
Now returns OKAY or 
PROGRAM_CALLED_WITH_WRONG_NUMBER_OF_ARGUMENTS
depending on how it was called. 
Add commments/changes to why the more logical 
fread was not used
changed type of cell_type from int to char to save
ram. 

All files with symmmetrical remaned to 
symmetrical (not the 3 m's beforehand) 

removed the following files, since they were renamed
to other file names. 
convert_circ_in_rect_dimensions_to_integers.c
convert_circ_in_rect_dimensions_to_integers.c

All files with the name programme have had this changed
to program. 

create_bmp_for_stripline_coupler.c 
now the verbose level is working properly

exit_codes.h 
Added numerous new entries

design_coupler.c 
Removed the -Q option for very quite and 
implemented it so -q is quite and -qq is quiter
still. 

check_error
Now exits with UNACCEPTABLE_ERROR_WHEN_CONVERTING_TO_INTEGERS 
rather than 2 which confused matters for me. I need to tidy up
all the exit codes. 
check_parameters_of_create_bmp_for_rect_in_rect.c 
Renamed a function to check_create_bmp_for_rect_in_rect_int
since it originally had the wrong name. 

Editied create_bmp_for_stripline_coupler.c. so that it only
prints information about theoretical values if called with the 
-v option. -vv gives more information and -vvv gives even more.

calculate_impedance_for_create_bmp_for_symmetrical_stripline.c
now calls usage_calculate_impedance_for_create_bmp_for_symmetrical_stripline  

usage_create_bmp_for_symmmetrical_stripline..c
now has a new name for the function 'usage_create_bmp_for_symmmetrical_stripline()'
and mentions the additional -v option 
Also, the name of the binaries mentioned have changed, to reflect
the new naming converntion. 

calculate_Zodd_and_Zeven.c now includes #include <stdlib.h>

tests
These have been modified and should be up to date on CVS

******Version 4.2.8 Released 12 Jan 2003 ********
Copied onto Sourceforge CVS *all* the 4.2.8 files. Some 
were missing before. 

examples/Makefile.am
Changed the CLEANFILE line to DISCLEANFILES, so that
running 'make clean' will not delete the example files
created, but running 'make distclean' will. 
Added a lot of txt files (see note below). 

man
News names for man pages 

lots of changes to files name 'cvsignore

examples/twin-wire4.bmp
Added to the CVS repositry. Had prevously ignored this
large file. 

in examples diretory
Added ....bmp.txt files, as the output of running atlc
on each of the .bmp files. 

src/non_gui/usage_atlc.c
src/non_gui/usage_circ_in_rect.c
src/non_gui/usage_create_bmp_for_circ_in_circ.c
src/non_gui/usage_create_bmp_for_circ_in_rect.c
src/non_gui/usage_create_bmp_for_microstrip_coupler.c
src/non_gui/usage_create_bmp_for_rect_cen_in_rect.c
src/non_gui/usage_create_bmp_for_rect_in_circ.c
src/non_gui/usage_create_bmp_for_rect_in_rect.c
src/non_gui/usage_create_bmp_for_stripline_coupler.c
src/non_gui/usage_create_bmp_for_symmmetrical_stripline.c
src/non_gui/usage_design_coupler.c
src/non_gui/usage_find_optimal_dimensions_for_microstrip_coupler.c
src/non_gui/usage_readbin.c
All above files have been modified to print a version number if 
run incorrectly. Perhaps adding a -V option that just prints 
the version and exits would not be a bad idea, at a later date

docs/atlc9.doc docs theory2.doc and design-ideas.sxw moveed to
docs/theory/atlc9.do docs/theory/theory2.doc and docs/theory/design-ideas.sxw

html-docs/README
added more informaton about the Word and StarOffice
files giving information about the theory of the newer versions
of atlc. 

docs/theory/README 
Changed to indicate there are now 3 theory files, one PDF and 
suggesting the use of books.

docs/html-docs/theory.html
Changed to indicate there are now 3 theory files, one PDF and 
suggesting the use of books.

docs/html-docs/banner-page.html
Changed to point to docs/html-docs/theory.html, rather than the
one pdf file I wrote for QEX back in 1996.
examples/cvsignore
Removed *.bmp and added *.bin. Put cvsignore on CVS

src/non_gui/do_fd_calculation 
added a return(0) to prevent the compiler from preventing a warning.

src/non_gui/setup_arrays.c
Uncommented the check_for_shorts line, as it does do something 
very useful. 

src/non_gui/check_for_shorts.c
Remove remarks about setting a floating conductor to +1 V
as floating conductors are not supported. 

examples/README
Updated examples/README, so it is more up to date,
listing only the files in the 4.2.8 distribution
and no more. The theretical impedance for some 
examples are included, where there is an exact
theoerically answer. Sometimes the structure tries
to approximate a stucture that has an exact theoreticas
answer, but never can exactly (we can't make the bitmaps
infinitely wide!!). There are noted where appropiate.
Not all the theoretical answers are at this point in the
README. 

configure.in 
Changed to look for optimised
math library (libmopt) and optimised c 
library (libcopt) as are used on SPARC systems.

Removed one copy of the function 
get_data_interactively() from definitions.h, as
it was defined twice. 

tools/README
corrected a simple spelling error.

configure.in
Remmoved the extra docs/Makefile in it.
 
src/non_gui/setup_arrays.c
Moved line 'data->dielectrics_in_bitmap=0;' to after declaration of
local variables.  gcc 2.96 will not accept the initialization as it
was when compiling c code, and generates an error.

src/non_gui/atlc.c
Removed extraneous dereference, replacing the line '*end++;' 
with 'end++;' in the argument processing section.

examples/create
This was removed from CVS, as 'make check' already did this and
it has been removed a while ago, but not from the CVS.

src/non_gui/find_optimal_dimensions_for_microstrip_coupler.c

******Version 4.2.7 Released 8/1/03********
Corrected a bug such that -s and -S did not work.

Corrected a bug in setup_arrays.c wherby the number
of blue conductors was not set to zero at the start
so the total number reported was wrong. 

Made changes to write_fields_for_directional_couplers.c
such that the the images for even mode are now 
written. 

Further attempts to improve the documentation. 

******Version 4.2.6 Released 1/1/03********
Corrected a bug in do_do_calculation.c, which
was bought to my attention by (thanks Dan).

Added two files into the distrtibution from
former versions 'cop1.bmp' and 'rect9.bmp'
since these were metnioned in the tutorial,
but not available. 

Slightly improved the look of the homepage by
editing index2.html 

Updateed 'tutorial.html' as it was a bit out 
of date, showing old versions of files. 

Added more tests. Some more could be usefully added,
but for now this will have to do. All the 11* tests
have been dropped as for some reason they don't run
properly. 

******Version 4.2.5 Released 1/1/03********
Arragned for the Makefile.am's to have a line
CLEANFILES = ... to remove the example files
if the user does a 'make clean'.  
Running 'make distcheck' should now work

Reinstated some more tests.

Corrected a problem with some of the tests
-create-lots-of-examples-check-last-one.test' such that
a lot of example files were created with a length of 
zero bytes. This was becaused I changed the format
of all the crate_bmp_for_... files so they no longer 
printed to stdout, but -create-lots-of-examples-check-last-one.test'
relied on them printing to stdout to work.

Images of permittivity are now okay, but not sure
about the Ex, Ey, U etc. These have not been checked
for 2-conductor lines, which needs doing. 


******Version 4.2.4 Released 31/12/02********

Changed all the files create_bmp_for_....
such that they now write to a file on the command
line, rather than stdout. All programs for the generation 
of bitmaps now have the same naming converntion 
(create_bmp_for_.....) and all write to a file and 
not stdout. 

Changed man pages to reflect the changes in all the 
create_bmp_for_...

Some half-sensible bitmaps are now produced with atlc
if the -s option is not present. The binary files
(.bin) are now printed properly for couplers. It is 
quite possible this does not work for 2 conductor lines
(it has not be tested). This needs more testing. 

The files atlc-x.y.z/tests/Makefile.am now has 
lists of the files to be added to the distribution,
rathe than just *.test, as the latter caused problems
with 'make distcheck'.

The file atlc-x.y.z/docs/html-docs/index.html now
loads index2.html rather than frontpage.html.

added file atlc/docs/html/create_for_Makefile.am to 
make it easier to generate a list of files that needed
to be included in atlc/docs/html/Makefile.am

******Version 4.2.2  Released  28/12/02 *****
Less data is now printed as it was confusing I feel to show too
much informattion. An option will be added to show more, but at 
present this is not implemented. 

Added options -s and -S in 'atlc' to enable the bitmap (.bmp) and binary
(.bmp) files to not be written. The default is to write these files. 

Remove options to write data to a text file and append data to
a text file from atlc. Changes meant it got more difficult
to implement and there seems little point when > and
>> do those functions.

Changed man page on atlc, so reflect changes in the program   
itself. 

Added the program 'create_bmp_for_coupled_microstrip' which does
exactly what the name says. Run it to see what options and 
command line arguments it expects, as there is no man page at 
the minute. 

*ALL* the files that create bitmaps have been renamed to create_bmp_for_xyz,
where xyz determines the geometry. There are now the following files
for creating bitmaps:

create_bmp_for_circ_in_circ
create_bmp_for_circ_in_rect
create_bmp_for_coupled_microstrip
create_bmp_for_coupler
create_bmp_for_rect_cen_in_rect
create_bmp_for_rect_in_circ
create_bmp_for_rect_in_circ.c
create_bmp_for_sym_strip

Currently some write to stdout, some to files. I will eventually change it 
so they all write to files, as the data is binary and writing by default
to stdout is not a great idea. 

Added program 'find_optimal_dimensions_for_microstrip_coupler' which does 
an exhaustive search of different geometires for microstrip couplers. 


********************Version 4.2.0-alpha  Released  24/12/02 ********************
Made major changes to the code to make it easier to read. Hopefully
direcitonal couplers are working for multiple-dielectrics now, 
but I've no idea. The documentation is out of date with the code
which has undergone major revisions. I will update the docs and correct
and bugs as soon as I have time. 

********************Version 4.1.6  Released  22/12/02 ********************
Made it more obvious there is a bug that prevents the use of atlc
where there is more than one dielectric on a directional coupler.

Updated the documentation so removing references to make_coupler

Changed the atlc homepage a bit, as it was out of date. 

********************Version 4.1.4  Released  15/12/02 ********************
Corrected bug found in design_coupler.c that prevented it compiling
under old versions of gcc. 

Added a few more tests that fail if gsl is not included. 

Updated the documentation. Some showed the output from 
atlc 2.x, which was a lot less accurate than 4.x.

Added a line for pth in Makefile.am. Threads on machines
othe than Solaris still needs some work. 

Added the file 'IMPORTANT.html' back into the release. For
some reason this got lost and was not present in any recent 
release, despite being on the web page and still valid.

Updated the FAQ, as some information was out of date.

********************Version 4.1.1  Released  14/12/02 ********************
**************************************************************************

Changed the default number of threads a mutli-threaded version of the
prorgram runs from 2 to 4, as I now own a quad processor machine. It is
changed in definitions.h

The program 'make_coupler' has been renamed to generate_coupler_bitmap,
as it never really did make a coupler, but only generate a bitmap for
one with the dimensions you gave. 

Added a new program 'design_coupler' which is very able. It can find the
required odd and even mode impedances given the coupling factor and
frequency range, and then design you a coupler with the right
dimensions!

Added much better documentation to support the directional couplers. New
files 'calculate_zodd_and_zeven.html',
'determine_dimensions_of_couplers.html' have been created. 

No significant changes have taken place to the code for two conductor
lines. I still need to address the problem of consistancy amoungst some
of the programs. 

A new directory 'tests' has been set up and automake configured such
that running 'make check' now performms a lot of checks and creates
all the examples. 

********************Version 4.0.1  Released  21/4/02 *********************
**************************************************************************

Added copyright notices to all files and added a -C option on all 
binaries to print out information on copyright. 

Corrected an error in READED.threads, since it refered to the -p options
for number of processors, which I later changed to -t threads to be more
accurate. 

Added my e-mail addres to the AUTHORS file. 

Update the TODO list. 

Remvoed the printing of L and C in the documentation (frontpage.html and
index2.html) for couplers, as these are mode-dependant and so one value
is not enough. 

Added a CVS entry on the banner page. 

Spent some time spell checking many of the html documents. 

Made some changes in the rect_in.. circ_in.. so that the integer values
create on a machine should be exactly the same on each machine. It was
possible before that some numbers would be generated on one machine and
some on another, due to rounding errors. This hopefully stops that. 

Changed src/non_gui/tests to point out what a dodgy gcc does. 
Added a few scrips 'tests' in other directories to src/non_gui, since
someone was looking for them and could not find them. Hence I've tried to
make it a bit clearer. 

********************Version 4.0.0 Released 11/4/02  *********************
**************************************************************************
Put a bit more info in the README.windows
Removed check for "intelligent life" in configure.in, as this was not working
with autoconf 2.13.

Added a function do_fd_calculation() and therefore added the file
do_fd_calculation.c This helps clean up atlc.c a lot, since a lot of
repetition has been removed. It needed quite a bit in it to handle the
case of coupled lines and all the different things that have to be
printed in this case. 

Changed setup_arrays.c so that it now sets a variable 'coupler' to TRUE
if there is a blue ( - 1 V) conductor. 

Got resuls out from the coupler. No idea if they are right. Pretty sure
they are wrong in fact. 

Corrected a typo in FAQ.html which meant the local copy of the QEX
paper could not be downloaded from the FAQ.

Answered that loss can't be measured in the FAQ.

Wrote a programme 'make_coupler' which writes a bitmap with two striplines 
between two parallel groundplanes as below.

------------^------------------------------------------------------------------
            |                                               
            |              <---w---><-----s----><---w-->
            H              ---------            --------
            |                                      
            |                               
------------v------------------------------------------------------------------
This has an exact analystical solution, so the file calculate_Zodd_and_Zeven.c
was created to find the exact theoretical values. This makes it possible
to compare the theory with atlc.

I got the equations from an e-mail sent to me by AA1LL / KB1CZP 
<aa1ll@email.com>. This required elliptic integrals, so its necessary to
link into the GNU scientific library, gsl to use it fully, although the
bitmap will be generated without the library.

Added information to accuracy.html on accuaracy with coupled lines. 

Changed finite_difference_multi_threaded.c & finite_difference_multi_threaded.c
to handle correctly the case of coupled lines - at least for vacuum
dielectric. 

Added a couple of tests in the file 'tests' to check the coupler 
related items. 


********************Version 3.0.7 Dated 6/4/2002**************************
**************************************************************************
Corrected the file man/makehtml, as it was not making an html page for
rect_cen_in_rect from the rect_cen_in_rect.1 man page. 

Put an error number each time a programme exits, so it is clear what
error caused it. 

Made a set of diectories called 'binaries' and subdirectories from that.
Please the latest Windows executables in binaries/Windows. 

Changed the line 
image_data=ustring(0L,width*height*3); in atlc.c
to
image_data=ustring(0L,size);
 
Removed all references to stdio.h and math.h in all files, since they
are included in definitions.h

Changed the line 
if( (image_data_fp=fopen(output_filename, "r")) == NULL)
to 
if( (image_data_fp=fopen(output_filename, "rb")) == NULL)
in atlc.c, so the bitmap is opened as a binary in Windoze.

put the line 
if((fp=fopen(temp,"w+b"))==NULL) instead of 
if((fp=fopen(temp,"w+b"))==NULL)
in get_file_pointer_with_right_filename.c
as the bitmaps were screwed up when written to disk in text mode, which
was happening under Windoze. 

Used a more complex method suggested by AD5GB to get open the bitmap
file for reading. I don't see why it should be necessary, but it keeps
Windows happy. 

defined M_PI if not already done so in definitions.h

Changed malloc call to a calloc in string.h

Corrected an error in the docs where it said in accuracy.html that the
error was 1.8% on the bottom of a table. That may have been so, but it
is a lot better now. 


********************Version 3.0.6 Dated 3/4/2002*************************
**************************************************************************
(note there was no release 3.0.5, as it got screwed up going up to
Sourceforge, after which is is impossible to overwrite the file). 


Removed all reference to the voltage v and instead replaced these by
Vij. The reason being that  Randall DuCharme <ad5gb@yahoo.com> has
advised me that the symbol v is defined in IBM's AIX operating sysytem.
Hence it would not compile without some hacking. 

Since the voltage is now called Vij, it makes sense for its indices to
be i and j and not w and h. Hence I've replaced at least some of the
v[j][h] to v[i][j]. 

Changed reference to longs and replaced by ints in write_bitmap_out.c,
since it was clear the longs would break it on 64-bit machines. 

Checked atlc one my Sun Ultra 60 with a 64-bit compiler using Sun's CC
compiler. All was okay and the generated bitmaps looked fine. 

Found a problem on the Sourforge Compile Farm with the programme
sym_strip. It crashed at the last but one line, where these is
flose(fp). Put in code to check if the file pointer was valid. This
can't be a bad thing, but it did not solve the problem. 

Another bug spotted on Sourceforge's compile farm. rect_in_rect,
circ_in_circ, circ_in_rect, and beadbin (but not atlc) all declare 
a character q, then use
q=get_options
whereas get_options returns an int, not a character. I don't know why the 
compiler does not spot that one. Anyway, it caused get_options to loop forever. 

All programme now seem to run okay on drkirkby@usf-cf-ppc-linux-1:,
which is a linux powered PowerPC RS6000. Not attempted compileing
multi-threaded. 


Attempted compiling on the Sourceforge Sun Ultra 60. This is the same as
my machine, so one might expect it to work okay. In fact, atlc 
compiles and runs okay on every machine tests (Dec Alpha, Suns
running Solaris, Suns running linux, linux PC, ... 

Correct typo, where it said 'is is' in usage_sym_strip.c

Changed the order and name or parameters in rect_cen_in_rect, so
it now

% rect_cen_in_rect W H w h Er1  > filename.bmp

Changed man page to suite. It was silly before. 

Removed note in the bitmap generators (rect_in_rect, circ_in_circ etc)
about piping into atlc, since atlc does not support reading from 
standard input. 

Changed circ_in_circ so the information it can provide about theoretical
impedances is only produced if the -v flag is given. 

corrected usage_circ_in_rect.c, as information at the bottom on usage
was all wrong. 

Added a test suite in the script src/tests. I might well add a version of
sum to the distribution, as it appears to be implementation dependant. 

Removed #include <malloc.h> as it is depricated according to gcc and
files compile fine without it. Also removed stddef.h from nrutil.c, as
that too is not needed. 

Changed the names of the variables opterr, optind, optopt and optarg to be
my_opterr, my_optind, my_optopt and my_optarg in get_options.c This is to
avoid a warning from one of the compilers (I think on a Mac) about the fact
that it is defined elsewhere. 


********************Version 3.0.4 Dated 1/4/2002*************************
**************************************************************************
NOTE 3.0.3 never released, since the upload to sourceforge failed. 

Corrected bug in write_fields.c which caused a crash under Linux
(as reported by Randall DuCharme <ad5gb@yahoo.com>) and With 
Borland C++ builder on my SunPCi card. 

Corrected a bug in atlc.c, which was causing a crash on the SunPCi
card and possibly a crash (not yet confirmed) under Solaris on 
a Sun Ultra 30 belonging to Randall DuCharme <ad5gb@yahoo.com>

Added an option to atlc.c to avoid writing the bitmap files. 

Corrected a bug in readbin.c, which caused it to report any 
number >.99 as a one, when it might well be METAL_ER. Also checked more
carefully for -1 V, by only looking from -0.999999999 to - 1.000000000001

Added the following lines to readbin and atlc.c, which are necessary
when compiling with C++ Builder version 3. I need to do likewise to the
rect_in_rect etc, but for now I have not. 

#ifdef WINDOWS
#pragma hrdstop
#include <condefs.h>
#endif

(I need to define WINDOWS when compiling on Windoze).

Corrected usage_atlc so it now does not take the filename, but knows it
is atlc. 

Added (void) to the defintion of any function that takes no arguments. I
found that while gcc is happy to accept a fuction with an argument that is
declared like this:
int somefunction()
Borland C++ Builder was not happy with that. Hence I now declare functions
as taking void when they don't take argments. This keeps Borland C++
builder happy and allows for stricter checking with gcc. 

Added a piece of code 'get_options.c' which was necessary since 
Borland C++ Builder does not have such a function.  

Stopped print_data taking an argument 'version' and then reading it from
VERSION. It now just reads from VERSION

Remved 'ER' from check_for_boundaries.c and all references to it, which
were all in comments. The Borland C++ Builder is better at spotting such
problems than gcc, so is Sun's C compiler. 

Added code to check if the upper number (eg nh) is less than or equal to
the lower number (eg nl) in functions free_* in files nrutuil.c and
string.c This stops the Borland C++ Builder complaining about the
arguments not being used and also does some sanity checking. 

********************Version 3.0.2 Dated 30/3/2002*************************
**************************************************************************
Fixed a problem with the Makefile.am's which meant not all the files were 
included and the sym_strip.1 was spelt wrong. 

********************Version 3.0.1 Dated 29/3/2002*************************
**************************************************************************
SIGNIFICANT IMPROVEMENTS TO ACCURACY !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
SIGNIFICANT IMPROVEMENTS TO ACCURACY !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
SIGNIFICANT IMPROVEMENTS TO ACCURACY !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

In a comparisson of 3 analytic models (coax, eccentric coax and
stripline), and 31 tests (roughly a third of each), the maximum error
was only 0.945 % and the RMS error is only 0.259 %. The programme now
undereads more than overreads, so I still think there may be a
systematic error, but any problems are now very small. Despite the fact
atlc is supporting multiple dielectrics, I've not managed to find any
models with exact analytic solutions to compare against atlc. When
calculationg the E-field across a dielectric, the method used does not
properly consider a boundary between two dielectrics. Hence I can expect
the error with multiple dielectrics to be a bit higher than with one.
This will be sorted out later. 

Added a man page for sym_strip and updated all other man pages. Keeping
the documentation up to date is a problem. 

Since many of the bitmaps used in the examples are large, I decided to
not bother distributing them all, but instead write a script 'create' to
make them. 'create' may be found in both  the examples and src
directories. 

Added a small script 'makehtml' in man, to create easily html version of
the man pages. 

It was necessary to take into account when a dielectric is next to
metal, as otherwise the E-fields calculated in the dielectric are in
erro when next to metal. This is done. check_for_boundaries.c has been changed
to indicate whether a dielectric has metal to the left, right, above or
below it.  

function calculate_colour_data() added to calculate the colour a pixel
should be set to - this is called by write_fields(). 

Added a file find_electric_fields.c, which has 3 functions in it -
find_Ex(), find_Ey() and find_E(). This avoids different parts of the
programme using different definitions for Ex, Ey and E. 

Changed write_fields almost completely. It now prints more files
and uses less ram, as not all the data is stored before writing to 
disk. Now only one set is stored in ram before writing to disk.
This is a bit slower, but it will save a lot of ram. 

After thinking about it a lot, I feel it is necessary to change the -1 V
conductor (which has never been used) from black to blue. Since I've
stated before that black is -1 V and blue is a dielectric with a
permittivity of 2.1, this is a bit annoying. However, it seems illogical
to have a negative electric field shown as blue, and negative voltages
as blue, yet have -1 V as black. Hence the definition of -1 V is changed
and another colour (a light blue) has been selected to reprsent the
permittivity of PTFE. 

Made a significant improvement to accuracy, but using the following
defintion for Ex
Ex=0.5*(v[i+1][j]-v[i-1][j]);
but changing it a bit when at a boundary with metal. i.e.
if((cell_type[i][j]&METAL_LEFT)==METAL_LEFT)
  Ex=(v[i+1][j]-v[i][j]);
Same is done in an obvious manner for Ey and the right hand side. This
has made a significant improvement to accuracy, so html docs have been
re-written.

Changed the definition of Ex and Ey in write_fields.c, since they were
defined at dV/dx, and dV/dy, but the electric field is defined as the
reverse of that. i.e. ex=-dV/dx and Ey is -dV/dy. 

Added some code in write_fields.c to print the energy (CV^2) too. 

********************Version 3.0.0 Dated 24/3/2002*************************
**************************************************************************
Added support for multiple dielectrics. More work needs to be done to
confirm the accuracy of this, but it is now working at least. 

Spent a lot of time thinking carefully about exactly where the voltage
are defined (top left of a pixel) and exactly where the E-field should
be calculated (centre of a pixel) and exactly what the E-field is. The
results seem no more accurate than before (still a small over-estimate
of Zo), but at least it should give anyone else chance to find a fault.
Before it was too obvious that a few problems remained on exactly what
happens on a boundardy. 

Programme should be able to cope with 4 different dielectrics in a
square - i.e. it can be really arbitrary. 

Changed the value of r in atlc.c from 1.5 to 1.95. This has made a HUGE
INCREASE IN SPEED!!! The programme is now about 6x quicker than before.

Added a function check_for_boundaries(), which looks for changes of
index and metals around a pixel. Its not really used now, but it has the
potential for speeding the code up, since it can avoid a lot of checking
every time the FD loop executes. It makes sence to do the checking once
and use those results. 

Removed all C++ commments, as this would stop it compiling on a C
compiler. One of the Sun compilers I have is only C. 

Added a file 'fd.c' which is not in the makefile. It is included in the
C functions finite_difference_*.c. It is better to use it like this
rather than put the code inline, for speed reasons. It is called several
times. Putting it in one file insures all the times its called are
calling exactly the same code. It is too easy to have errors otherwise.
\7f


Added a programme 'thin_strip; which produces a stipline like this:

          ------^-------------------------------------
                |
                H        -----w------
                |
-infinity ------v------------------------------------ +infinity

This has a well defined impedance, which is given in "Field Theory of Guided 
Waves", Colin R.E., 2nd edition, pp 259-273, IEEE Press, (1990).

I realised the programme was taking too much memory. Reading in a 24 Mb
bitmap, resulted in the programme using 179 Mb of RAM. It was discoverd
that the variable 'cell_type' defined in atlc, was an integer array,
when a character one would have done, saving 3 bytes per pixel of RAM. I
Added functions charmatrix and freecharmatrix in nrutil.c, for allocation
and dealocation of the matrix.

Removed the '-m' option from usage_atlc.c, since the option no longer
worked - it was removed some time ago. 

Removed the '-F' option from usage_atlc.c, since the option  did not 
work. I decided not to bother fixing it, but instead of writing data
to a user-specified file name, write it to example.txt, when the 
bitmap file is example.bmp. 

Removed the handling of option 'h' in atlc.c, since the option no longer
existed. 

Printed a messge with usage_atlc, so that it indicated that if
multi-processor support was not compiled into atlc. 

Added a few commments are correct a few typos in other commments, in the
file finite_difference_multi_threaded.c

Checked the system reasonably well with multi-threading enabled on my
dual preocessor Sun Ultra 60. The programme speeds up quite a bit,
depending on the size of the job. Certainly runing with two threads
(either by defining #define MAX_CPUS 2 in definintions.h, or the -p2
command line option, seem to work best for my dual processor machine. Running
a few more threads (up to 5 or so) does not slow the system much, but as
more and more are added, there is a definate slowdown. 

renamed usage to usage_atlc, to be consistant with all the other
programmes (usage_rect_in_rect etc) 

Added quit a few comments to atlc.c, and generally made the code tidier,
with more consistant indenting. 

Added a file/function swap_conductor voltages, which converts +1V to -1V
and visa-versa. This is needed to compute the impedance of coupled
lines. 

**************************Version 2.32 Dated 9/3/2002***************************
********************************************************************************
Wrote manpage for readbin
Moved man page from man to man/man1.
Removed function free_ucmatrix, free_ucmatrix and byteswap, none of which
were used. 
Added a new ability to calculate Zo assuming a round conductor - which
it might not be. Its useful to check the software against coax, which
has well defined values. Needs -v option to see the result. 
Added a new programme (only one file) 'readbin' which analyses the .bin
files and reports on the number of each colour pixel. 
Corrected write_fields.c, so that a conductor is printed with zero
E-field. Sometimes this was not so, as there was no check before writing
the E-fields out that the pixels was not a conductor. 
Added new file/function check_for_shorts(), which causes the programme to
exit if a conductor of one voltage is in contact with a conductor of
another voltage. 
Compiled with Sun's CC compiler, which threw up a few prolems. It's more
fussy that gcc!!  Hence made these changes:
a) Replaced strncmp(bitmap_file_buffer, with strncmp((char *)
   bitmap_file_buffer, on line 41 of read_bitmap_file_headers.c. 
b) Changed atlc.c so a fread((void *)... image_data from fread(image_data...
c) Removed multiple declaration of some variables in rect_in_rect.c,
circ_in_circ.c
d) Changed fill_image_vector_with_data in rect_in_rect.c,
rect_in_circ.c, circ_in_circ.c circ_in_rect.c and definitions.h, such
that the first argument is an unsigned char *, not char*  as before. 
e) align_bitmap_image in align_bitmap_image.c changed so its arguments
are unsigned *char, not *char as before. Updated header file to reflect
this. 
f) 

In several C files, changed variables so they were initialised to -1 - it stops
compiler complaining about possible use of unitialisd variables. 
Removed multi-line fprintf statement, and made into two statements in rect_in_circ.c
Changed else if(j=D/2) to else if(j==D/2) in fill_rect_in_circ.c
Added exit(0) to end of circ_in_circ.c
Added exit(0) to end of circ_in_rect.c
Added exit(0) to end of rect_in_circ.c
Added definition of usage_circ_in_rect() to definitions.h 

Removed the note about version 2.4 from this ChangeLog, as it wasn't released. 
Corrected a serious bug in setup_arrays.c, where v was set to 0.3 instead of 0 V. 
Removed 'byteswap' from byteswap.c and definitions.h
Added code in configure.in to enable -Wall with gcc.


************Version 2.2 Dated 2/12/2001**************
*****************************************************
Changed some of the html documentation, to be more in line with the
source and man pages. 

Verion 2.1 Dated 1/12/2001
Many updates to documentation:
Added man pages for rect_cen_in_rect
Added man pages for circ_in_rect
Added man pages for rect_in_rect
Added man pages for circ_in_circ

Added binaries for:
rect_cen_in_rect
circ_in_rect
circ_in_circ
rect_in_circ - this dones NOT work, so it is configured to always exit.

Changed atlc so that it no longer uses the faster convergence algorithm.
I now know it can fail, so will remove it. 

Changed usage.c and atlc.c so that the -p option is not displayed if not 
MP enabled.
usage.c Changed the usage message about '-i fiddle' as it was wrong before, in assuming default to be
1.
atlc.c Added a '-c' option, to change the cutoff. 
atlc.c Changed the reading of the '-i fiddle' option, so its read a double and not a long

25/4/2001, verison 2.01 put on SorgeForce unix only, supports multiple
CPUs. 

Version 1.09 - Windows verison, put on my web site.

Version 1 - published in QEX, see qex-december-1996/atlc.pdf.