changelog

麻省理工的计算光子晶体的程序

	don't carry over to the next line (I guess each line is executed
	in its own subshell).

2001-02-19  Steven G. Johnson  <[e-mail]>

	* doc/analysis-tutorial.html: punctuation alteration

	* doc/analysis-tutorial.html: slight rewording

	* doc/analysis-tutorial.html: noted that the real part is
	sometimes zero

	* doc/analysis-tutorial.html: noted units for resolution in
	mpb-data

	* doc/analysis-tutorial.html: use -S 3 instead of -X 3 -Y 3

	* doc/tri-rods-eps-2.gif, doc/tri-rods-ez1.gif,
	doc/tri-rods-ez2.gif, doc/tri-rods-ez3.gif, doc/tri-rods-ez4.gif,
	doc/tri-rods-ez5.gif, doc/tri-rods-ez6.gif, doc/tri-rods-ez7.gif,
	doc/tri-rods-ez8.gif: updated mode pictures (since mpb-data now
	centers its output at the origin)

	* config.guess, config.sub: updated to latest version

	* ChangeLog: updated

	* doc/analysis-tutorial.html, mpb-ctl/examples/diamond.ctl,
	mpb-ctl/mpb.scm.in, src/maxwell/maxwell_eps.c: a couple
	compilation fixes, and noted use of define-param in the manual

	* mpb-ctl/examples/check.ctl: whoops, got rid of premature exit

	* mpb-ctl/examples/check.ctl: improved accuracy in 1d tests...but
	there still seem to be occasional discrepancies.  What is going
	on?

	* doc/user-ref.html: clarification

	* TODO: updated

2001-02-18  Steven G. Johnson  <[e-mail]>

	* doc/installation.html: fixed autoconf URL

	* src/maxwell/maxwell.c: whoops, non-constant array initializers
	are not allowed

2001-02-17  Steven G. Johnson  <[e-mail]>

	* doc/installation.html: mention parallel HDF5

	* doc/installation.html: fixed typo

	* doc/user-ref.html, doc/installation.html: noted mpbi-mpi in
	manual

	* mpb-ctl/mpb.c: fixed counting statistics for mpbi-mpi

	* src/maxwell/maxwell_op.c: fixed field output for 2d mpbi-mpi

	* mpb-ctl/Makefile.in: fixed copyright year

	* src/matrixio/matrixio.c: don't call H5Dwrite at all for empty
	dataspaces (e.g. in leftover processes under MPI)

	* src/maxwell/maxwell_op.c: bug fixed for mpbi-mpi field output

	* mpb-ctl/examples/hole-slab.ctl: There sees to be some funny
	rounding problem on the Origin (at least) that causes it to get
	confused about whether x=0 and/or y=0 is inside the eps block.

	* doc/user-ref.html: fixed broken <code> tag

	* mpb-ctl/mpb.c: undid erroneous cleanups (although they wouldn't
	have technically caused problems)

	* ChangeLog: updated

2001-02-16  Steven G. Johnson  <[e-mail]>

	* NEWS: minor rewordings

	* NEWS: updated

	* utils/mpb-data.c: Fixed a problem with MODF_POSITIVE that caused
	mpb-data to crash on an Alpha; thanks to Dominique Caron for the
	bug report.

2001-02-14  Steven G. Johnson  <[e-mail]>

	* doc/user-ref.html: note that MPB exits with an error if
	run-te/tm are used nonsensically, or if epsilon is not
	positive-definite.

	* NEWS: fixed hyphenation

	* configure.in: bumped version

	* mpb-ctl/mpb.c, mpb-ctl/mpb.scm.in, src/matrices/scalar.h,
	src/matrixio/fieldio.c, src/matrixio/matrixio.c,
	src/matrixio/matrixio.h, src/maxwell/maxwell.c,
	src/maxwell/maxwell.h, src/maxwell/maxwell_op.c, NEWS: Rewrote
	field output for inversion-symmetry case.  We now write the field
	to the file in a few chunks, rather than trying to expand the
	half-complex rfftw output to the full array.  This paves the way
	for a parallel mpbi.  As a side effect, we also circumvent a bug
	in the old makefull routines that caused artifacts to appear in
	the output for 3d cells (reported by Michelle Povinelli).

2001-02-11  Steven G. Johnson  <[e-mail]>

	* mpb-ctl/mpb.scm.in: Bug fix in combine-band-functions, etcetera,
	for when a thunk is encountered.  Thanks to Michelle Povinelli for
	the bug report.

2001-02-09  Steven G. Johnson  <[e-mail]>

	* mpb-ctl/mpb.c, src/matrixio/fieldio.c, src/matrixio/matrixio.c,
	src/matrixio/matrixio.h: generalized fieldio function in
	preparation for rewritten mpbi output...

	* mpb-ctl/mpb.c, mpb-ctl/mpb.scm.in, NEWS, src/matrixio/fieldio.c,
	src/matrixio/matrixio.h: eliminated output-copies in favor of
	mpb-data

2001-02-08  Steven G. Johnson  <[e-mail]>

	* mpb-ctl/mpb.c: added geometry tree debugging code

	* doc/installation.html, doc/user-ref.html, NEWS: give (more)
	thanks to Clarendon

	* src/matrices/blasglue.c, src/matrixio/matrixio.c,
	src/maxwell/maxwell_op.c: fixed problems when an MPI process has
	no portion of the eigenmodes (which can happen for inconvenient #
	processes)...matrixio is still not quite right, though

	* src/matrices/blasglue.c: don't die if passed NULL arrays for
	zero sizes

	* mpb-ctl/mpb.c, mpb-ctl/mpb.scm.in: added get/set kpoint-index
	functions

	* NEWS: updated

	* doc/index.html, doc/installation.html, doc/user-ref.html:
	documented mpb-mpi

	* mpb-ctl/examples/check.ctl: check-tolerance should be a param

	* src/maxwell/maxwell_pre.c: Fixed bug in zero-k constraint for
	parallel systems.

2001-02-07  Steven G. Johnson  <[e-mail]>

	* doc/acknowledgments.html: thanks to Clarendon

	* utils/mpb-data.1, utils/mpb-data.c: added mpb-data -T option

	* configure.in, utils/Makefile.in: don't build/install mpb-data
	when linking to MPI

	* mpb-ctl/examples/check.ctl, mpb-ctl/mpb.c, mpb-ctl/mpb.scm.in:
	expose more MPI info to Guile, and don't do 1d tests in check.ctl
	under MPI (where they are not supported)

	* mpb-ctl/mpb.c, src/matrices/eigensolver.c,
	src/matrices/eigensolver_davidson.c,
	src/matrices/eigensolver_utils.c, src/matrices/evectmatrix.c,
	src/matrices/matrices.h, src/maxwell/maxwell_eps.c,
	src/maxwell/maxwell_pre.c, src/util/mpiglue.h,
	src/util/mpi_utils.c, src/util/mpi_utils.h, tests/eigs_test.c:
	yikes, Allreduce requires sendbuf != recvbuf (although this is not
	clear from the documentation)

	* src/maxwell/maxwell_eps.c: cleanups, fixed calculation of
	eps_inv_mean for MPI

	* src/maxwell/maxwell.c: fixed indenting

	* TODO: updates

	* mpb-ctl/Makefile.in: put check target at the end so that 'make'
	does 'make all'

	* src/matrixio/matrixio.c: fprintf -> mpi_one_fprintf

	* src/maxwell/maxwell_op.c: whoops, fixed stride/dist check to
	coincide with what fftwnd_mpi really supports

	* src/config.h.in: define HAVE_FFTW to something

	* configure.in, mpb-ctl/mpb.c, src/maxwell/maxwell.c,
	src/maxwell/maxwell.h, src/util/mpi_utils.c: fixed compilation
	whoopses under MPI

	* doc/user-ref.html: fix

	* ChangeLog, NEWS: updated

2001-02-06  Steven G. Johnson  <[e-mail]>

	* mpb-ctl/mpb.scm.in: workaround for older versions of Guile where
	we can't get the number of arguments to a function (apparently)

	* configure.in: fixed gen-ctl-io check, removed AC_PROG_CC_STDC
	ifdef...there's something not working about ifdef that I don't
	understand

	* configure.in: call AC_PROG_CC_STDC if defined, in part so that
	autoconf 2.5 doesn't complain

	* configure.in: added MPI checks

	* src/util/mpi_utils.c: clarification

	* src/matrixio/matrixio.c: fixed misnamed function call

	* configure.in, src/config.h.in, src/matrixio/matrixio.c: support
	HDF5 file i/o without H5Pset_mpi

2001-02-05  Steven G. Johnson  <[e-mail]>

	* src/util/mpiglue.h, src/util/mpi_utils.c, src/util/mpi_utils.h:
	added critical section utility, intended for mediating output file
	access

2001-02-04  Steven G. Johnson  <[e-mail]>

	* mpb-ctl/mpb.c, mpb-ctl/mpb.scm.in: turn off printing unless we
	are the master process

	* doc/user-tutorial.html, mpb-ctl/examples/check.ctl,
	mpb-ctl/examples/tutorial.ctl, mpb-ctl/mpb.scm.in: display-many ->
	print

	* mpb-ctl/Makefile.in: make main() in new libctl call our hooks

	* mpb-ctl/epsilon_file.c, mpb-ctl/mpb.c, mpb-ctl/mpb.scm.in,
	src/matrices/eigensolver.c, src/matrices/eigensolver_davidson.c,
	src/matrices/linmin.c, src/matrices/matrices.c,
	src/matrixio/fieldio.c, src/matrixio/matrixio.h,
	src/maxwell/maxwell.c, src/maxwell/maxwell_eps.c,
	src/maxwell/maxwell_op.c, src/maxwell/maxwell_pre.c,
	src/util/mpiglue.h, src/util/mpi_utils.c, src/util/mpi_utils.h:
	many cleanups in preparation for MPI

	* mpb-ctl/mpb.c: give error if k-point is incompatible with the
	polarization

	* configure.in, tests/blastest.complex.out, tests/blastest.out,
	tests/blastest.real.out, tests/Makefile.in: fixed blastest 'make
	check' for complex scalars

2001-02-03  Steven G. Johnson  <[e-mail]>

	* mpb-ctl/mpb.c, src/maxwell/maxwell_eps.c, src/maxwell/maxwell.h:
	checks to see if dielectric tensor is okay; return error if TE/TM
	is specified but impossible

	* doc/index.html: fixed typo

2001-01-30  Steven G. Johnson  <[e-mail]>

	* configure.in: use vendor cc, too, so that Fortran and C are in
	sync (there may be linking problems, otherwise).

	* doc/developer.html, doc/license.html: noted MPB paper

2001-01-27  Steven G. Johnson  <[e-mail]>

	* mpb-ctl/mpb.c: fixed compute_energy_in_object_list for
	non-orthogonal lattices

2001-01-26  Steven G. Johnson  <[e-mail]>

	* mpb-ctl/mpb.scm.in: filename-prefix is once again a define-param
	variable

2001-01-24  Steven G. Johnson  <[e-mail]>

	* NEWS: clarified the purpose of "nothing"

	* doc/user-tutorial.html, mpb-ctl/examples/tutorial.ctl,
	mpb-ctl/mpb.c: in compute-energy-in-objects, let 'nothing'
	material punch a hole

2001-01-23  Steven G. Johnson  <[e-mail]>

	* doc/user-ref.html, doc/user-tutorial.html,
	mpb-ctl/examples/tutorial.ctl, mpb-ctl/mpb.c, mpb-ctl/mpb.scm.in,
	NEWS: added 'nothing' material

	* doc/installation.html: clarified simultaneous installation of
	mpb and mpbi

2001-01-21  Steven G. Johnson  <[e-mail]>

	* mpb-ctl/examples/check.ctl: collect error statistics

	* mpb-ctl/examples/check.ctl: Fix a bug (Bragg-sine not resetting
	default-material).  Also raise tolerance slightly.

	* Makefile.in, tests/Makefile.in: print friendly message when
	(if?) make check passes

	* mpb-ctl/examples/check.ctl: added sinusoidal Bragg to make check

	* doc/user-ref.html: warn user about run-te and run-tm with
	anisotropic dielectrics

	* NEWS: give credit

	* doc/user-ref.html, mpb-ctl/examples/bragg-sine.ctl,
	mpb-ctl/mpb.c, mpb-ctl/mpb.scm.in, NEWS: added material-function
	type

	* configure.in: don't check for ctl_get_list (not there anymore)

	* doc/installation.html, doc/user-ref.html, NEWS: let's not
	capitalize 'Hermitian'

	* tests/Makefile.in: whoops, fixed $ substitution

	* Makefile.in: changed subdir order in make check

	* Makefile.in, mpb-ctl/Makefile.in, tests/Makefile.in: added 'make
	check'

	* tests/maxwell_test.c: fix -E

	* tests/maxwell_test.c: use epsilon with inversion symmetry, so it
	works with mpbi

	* tests/blastest.out: added sample output

	* tests/maxwell_test.c: allow option to fail if error is too large

	* src/maxwell/maxwell.c: Earlier, we flipped the sign of the G
	vector in order to have a consistent sign of k for mpbi and
	mpbi--it is especially important to be careful about this when you
	have a complex-Hermitian dielectric tensor that breaks (k,-k)
	time-reversal symmetry.  However, this made our planewave basis
	slightly different from before.  It used to be that we had 1 more
	positive-G planewave than negative-G (for even grid sizes), but
	the above change reverses this situation.  It doesn't matter in an
	absolute sense, since the planewave cutoff is still the same, but
	the difference was an annoyance--it caused check.ctl to fail
	because of microscopic changes in the results, and also made us
	inconsistent with the MPB paper.  This change fixes that,
	restoring the +G/-G planewave balance to what it was before
	(although the basis elements are stored in a different order, with
	the -G waves first).

	* src/maxwell/maxwell.c, src/maxwell/maxwell_op.c: fixed G sign
	convention to be consistent for mpbi and mpb

2001-01-20  Steven G. Johnson  <[e-mail]>

	* doc/developer.html, doc/user-ref.html, doc/user-tutorial.html:
	spelling correction

	* doc/developer.html, doc/introduction.html: link to
	barnesandnoble.com instead of Amazon to be GNU-ly correct

	* doc/user-tutorial.html: added link

	* doc/installation.html, doc/user-ref.html, NEWS: documented
	complex-Hermitian epsilon

	* src/maxwell/maxwell_op.c: slight change

	* configure.in, mpb-ctl/epsilon_file.c, mpb-ctl/mpb.c,
	mpb-ctl/mpb.scm.in, src/config.h.in, src/matrices/scalar.h,
	src/maxwell/maxwell_eps.c, src/maxwell/maxwell.h,
	src/maxwell/maxwell_op.c, tests/maxwell_test.c: added
	--with-hermitian-epsilon

2001-01-11  Steven G. Johnson  <[e-mail]>

	* mpb-ctl/mpb.scm.in: Modified band-range data functions so that
	they are tail recursive, lest stack overflow crashes occur when
	large numbers of bands (e.g. 600) are computed.  Thanks to Mihai
	Ibanescu for the bug report.

2001-01-10  Steven G. Johnson  <[e-mail]>

	* src/matrices/minpack2-linmin.c: use MIN, MAX, and ABS in case
	some header file defines these macros too (like AIX...grrr).
	Thanks to Angelo R. Rossi for the bug report.

2001-01-07  Steven G. Johnson  <[e-mail]>

	* doc/analysis-tutorial.html: clarification

	* doc/analysis-tutorial.html: slight clarification

	* doc/user-ref.html: slight formatting fix