changelog

来自mit的fdtd开放性源代码meep

    M ./src/initialize.cpp -1 +1
    M ./src/integrate.cpp -1 +1
    M ./src/loop_in_chunks.cpp -1 +1
    M ./src/meep.hpp -1 +1
    M ./src/meep/mympi.hpp -1 +1
    M ./src/meep/vec.hpp -1 +1
    M ./src/meep_internals.hpp -1 +1
    M ./src/monitor.cpp -1 +1
    M ./src/mympi.cpp -1 +1
    M ./src/output_directory.cpp -1 +1
    M ./src/polarization.cpp -1 +1
    M ./src/sources.cpp -1 +1
    M ./src/sphere-quad.cpp -1 +1
    M ./src/step.cpp -1 +1
    M ./src/step_d.cpp -1 +1
    M ./src/step_h.cpp -1 +1
    M ./src/structure.cpp -1 +1
    M ./src/time.cpp -1 +1
    M ./src/update_e_from_d.cpp -1 +1
    M ./src/update_from_e.cpp -1 +1
    M ./src/vec.cpp -1 +1
    M ./tests/bench.cpp -1 +1
    M ./tests/bragg_transmission.cpp -1 +1
    M ./tests/cylindrical.cpp -1 +1
    M ./tests/flux.cpp -1 +1
    M ./tests/harmonics.cpp -1 +1
    M ./tests/integrate.cpp -1 +1
    M ./tests/known_results.cpp -1 +1
    M ./tests/one_dimensional.cpp -1 +1
    M ./tests/physical.cpp -1 +1
    M ./tests/symmetry.cpp -1 +1
    M ./tests/three_d.cpp -1 +1
    M ./tests/two_dimensional.cpp -1 +1

Mon Nov 12 19:34:43 EST 2007  [e-mail]
  * updated NEWS

    M ./NEWS +33

Mon Nov 12 18:58:32 EST 2007  [e-mail]
  * shorten long h5test

    M ./tests/h5test.cpp -32 +52

Mon Nov 12 18:30:09 EST 2007  [e-mail]
  * fix several deadlocks when using serial HDF5 with parallel Meep

    M ./src/h5fields.cpp +1
    M ./src/h5file.cpp -2 +4
    M ./tests/h5test.cpp +3

Sat Nov 10 19:54:33 EST 2007  [e-mail]
  * fix small mem leak

    M ./src/h5file.cpp -2 +4

Sat Oct 20 13:34:19 EDT 2007  [e-mail]
  * attempt to auto-detect -package name for ghc stuff, since apparently ghc keeps changing it (grr); thanks to Liang Huo for the bug report

    M ./configure.ac -1 +20
    M ./hsrc/Makefile.am -1 +1

Mon Oct  8 19:02:15 EDT 2007  [e-mail]
  * bug fix -- dft_chunk objects were improperly disposed of in ~fields due to confusion between next_in_chunk and next_in_dft, thanks to Alex McCauley for the bug report and twinkies

    M ./src/dft.cpp -6 +20
    M ./src/fields.cpp -1 +5
    M ./src/meep.hpp -1 +1

Mon Oct  8 19:00:42 EDT 2007  [e-mail]
  * eliminate compiler warnings about unused params

    M ./src/meep.hpp -6 +6

Mon Oct  8 18:22:47 EDT 2007  [e-mail]
  * set fluxes = NULL after remove_fluxes

    M ./src/fields.cpp +1

Thu Aug 16 13:33:34 EDT 2007  [e-mail]
  * version bump to 0.10.1 (not final until tag!)

    M ./configure.ac -2 +2

Thu Aug 16 13:30:18 EDT 2007  [e-mail]
  * 2D saturated polariton known_results fails on 4 processors; according to darcs trackdown this has failed since at least 2005, so it's not clear that it ever worked

    M ./tests/known_results.cpp -2 +3

Wed Aug 15 22:41:37 EDT 2007  [e-mail]
  * update_e_from_d should only update owned E's

    M ./src/update_e_from_d.cpp -2 +2

Wed Aug 15 22:41:22 EDT 2007  [e-mail]
  * comment fix

    M ./src/boundaries.cpp -1 +1

Wed Aug 15 17:28:10 EDT 2007  [e-mail]
  * flux.cpp test case accidentally put flux plane inside PML

    M ./tests/flux.cpp -1 +1

Wed Aug 15 17:27:32 EDT 2007  [e-mail]
  * fixed bug in flux_in_box: accidentally returned flux * number of processes, instead of flux!  yikes, caused 'make check' to fail with MPI, too

    M ./src/energy_and_flux.cpp -1 +1

Wed Aug 15 16:54:25 EDT 2007  [e-mail]
  * restored legacy simplistic averaging code, modified slightly to remove potential instability, with comment explaining that this code somehow yields quadratic convergence in one of the test cases (although not in general), indicating that there may be a method behind the madness

    M ./src/anisotropic_averaging.cpp -11 +37
    M ./tests/convergence_cyl_waveguide.cpp -2 +2

Wed Aug 15 16:27:23 EDT 2007  [e-mail]
  * disable epsilon averaging for pixels that include metals (eps < 0), since this can lead to an 0 < average epsilon < 1, which causes instabilities by changing the Courant condition (apparently)

    M ./libctl/structure.cpp -4 +18
    M ./src/anisotropic_averaging.cpp -1 +21

Wed Aug 15 15:44:07 EDT 2007  [e-mail]
  * run h5topng only once on multiple CPUs, avoid race condition

    M ./libctl/meep.scm.in -2 +4

Wed Aug 15 15:43:22 EDT 2007  [e-mail]
  * fix deadlock for output-hfield etc. reported by Lingling Tang (deadlocks when using non-MPI HDF5)

    M ./libctl/meep.scm.in -1 +5
    M ./src/h5file.cpp -8 +11

Fri Aug 10 10:53:06 EDT 2007  [e-mail]
  * update acx_blas

    M ./m4/acx_blas.m4 -1 +16

Mon Jun 25 17:29:24 EDT 2007  [e-mail]
  * use delete[] not delete for deallocating new char[] array

    M ./src/h5file.cpp -3 +3

Wed May  9 19:41:29 EDT 2007  [e-mail]
  * update hsrc/Makefile for ghc-6.6

    M ./hsrc/Makefile.am -1 +1

Wed Apr 18 18:23:01 EDT 2007  [e-mail]
  * bug fix: subpixel averaging was disabled for dimensions=1 (thanks to Mischa Megens for the bug report)

    M ./libctl/structure.cpp -1 +1

Mon Jan 15 16:45:27 EST 2007  [e-mail]
  * delete cruft

    M ./examples/pw-source.ctl -2

Thu Jan  4 20:47:38 EST 2007  [e-mail]
  * MPI_LONG_DOUBLE is optional in the MPI standard, so check before using it in MPI_Allreduce

    M ./src/mympi.cpp -2 +11

Tue Jan  2 21:30:53 EST 2007  [e-mail]
  * fixed bug that prevented output-tot-pwr from working; thanks to Vyacheslav Sokolov for the bug report

    M ./src/h5fields.cpp -1 +1
    M ./src/integrate.cpp -1 +1
    M ./src/meep.hpp -1 +1
    M ./src/vec.cpp -2 +2

Thu Nov  2 15:17:51 EST 2006  [e-mail]
  * make at-every less susceptible to rounding; thanks to L. Le Guyader for the bug report

    M ./libctl/meep.scm.in -1 +2

Fri Oct 20 18:09:36 EDT 2006  [e-mail]
  * include -march in pkgconfig flags (g++ requires it, apparently)

    M ./configure.ac +12
    M ./meep.pc.in -1 +1

Sat Sep 30 13:49:40 EDT 2006  [e-mail]
  * fix bug reported by J. L. Silva, where polarization memory was not parallelized

    M ./src/polarization.cpp -1 +1

Sat Sep 30 13:49:26 EDT 2006  [e-mail]
  * consistency

    M ./src/anisotropic_averaging.cpp +2

Mon Sep 18 21:50:06 EDT 2006  [e-mail]
  * custom_src_time should respect start_time

    M ./src/meep.hpp -1 +2

Mon Aug 28 13:54:42 EDT 2006  [e-mail]
  * bug fix in output-png+h5, thanks to Chad Husko

    M ./libctl/meep.scm.in -1 +1

Mon Aug 21 19:44:04 EDT 2006  [e-mail]
  tagged 0.10


Mon Aug 21 19:09:43 EDT 2006  [e-mail]
  * really include meep-enums.scm

    M ./libctl/Makefile.am -1 +1

Mon Aug 21 18:22:18 EDT 2006  [e-mail]
  * version bump

    M ./NEWS -4 +4
    M ./configure.ac -2 +2

Fri Aug 18 14:01:51 EDT 2006  [e-mail]
  * avoid printing redundant messages about padding

    M ./src/structure.cpp -2 +2

Tue Aug  8 20:21:33 EDT 2006  [e-mail]
  * fix test case to handle nlo != 1

    M ./tests/bragg_transmission.cpp -3 +3

Mon Aug  7 13:48:19 EDT 2006  [e-mail]
  * updated NEWS for upcoming 0.10

    M ./NEWS +49

Sun Aug  6 00:27:57 EDT 2006  [e-mail]
  * added testcase for 2nd/3rd harmonic generation

    M ./tests/Makefile.am -1 +1
    A ./tests/harmonics.cpp

Sat Aug  5 23:05:58 EDT 2006  [e-mail]
  * turned on polariton symmetry tests, with slightly increased tolerances

    M ./tests/symmetry.cpp -12 +8

Sat Aug  5 22:39:00 EDT 2006  [e-mail]
  * add nonlinear symmetry tests
  
  The nonlinear symmetry tests had been disabled because they caused
  the symmmetry.dac test to fail.  I had assumed this was a bug,
  but now I'm not so sure.
  
  First, the nonlinear calculations rely on an averaging of field values
  from adjacent points, and this averaging process will differ by rounding
  errors when using symmetries (e.g. left + right + top + bottom may 
  change to right + left + top + bottom, which differs by rounding).  So,
  I had to decrease the error threshold for the nonlinear tests.
  
  Second, a more serious concern: symmetry is not necessarily preserved
  when rounding is included.  A simple example is that exp(i m phi) modes
  in a cylindrical waveguide may be converted to exp(3 i m phi) modes
  by a Kerr nonlinearity, analogous to 3rd harmonic generation.  I *think*
  that this conversion does not occur for the trivial representation
  (analogous to m=0), but I'm not 100% sure.  So, the nonlinear tests
  (which use the trivial representation, i.e. no phase factor) should
  be okay...
  
  However, e.g. the rot4z symmetry test still fails if I use *both* TM
  and TE sources (which interact due to the nonlinearity).  Outputting
  the fields, it looks like a nonsymmetric pattern really is generated
  after a while in this case.   Perhaps this is a numerical effect,
  related to the rounding mentioned above?

    M ./tests/symmetry.cpp -30 +35

Sat Aug  5 00:59:36 EDT 2006  [e-mail]
  * turn on epsilon averaging by default in libctl interface -- the speed penalty seems to be only about 10%, and the huge accuracy gains (as well as continuous sensitivity to geometry) are more than worth it

    M ./libctl/meep.scm.in -1 +1

Sat Aug  5 00:56:16 EDT 2006  [e-mail]
  * slight optimization in handling off-diagonal epsilon
  
  Checking whether the off-diagonal inveps components are zero leads to a
  slight (< 10%) speed increase on our Pentium IVs, probably thanks to
  the avoided loads of the off-diagonal fields.  On the other hand, keeping
  a separate array of flags is definitely not a win--it is better to check
  inveps directly.

    M ./src/update_e_from_d.cpp -8 +15

Sat Aug  5 00:54:56 EDT 2006  [e-mail]
  * bug fix - set_epsilon should delete off-diagonal inveps components when averaging is not being used (otherwise, doing set_epsilon with averaging followed by set_epsilon with no averaging leads to wrong results)

    M ./src/anisotropic_averaging.cpp +2

Fri Aug  4 20:39:37 EDT 2006  [e-mail]
  * some explanatory comments for Pade approximant

    M ./src/update_e_from_d.cpp -3 +10

Fri Aug  4 03:15:24 EDT 2006  [e-mail]
  * merge chi2 (pockels) stuff from Ardavan

    M ./libctl/meep.scm.in +1
    M ./libctl/structure.cpp +52
    M ./src/meep.hpp -2 +7
    M ./src/structure.cpp -10 +54
    M ./src/update_e_from_d.cpp -12 +19

Fri Aug  4 02:21:21 EDT 2006  [e-mail]
  * rename kerr to chi3 in preparation for Ardavan chi2 merge

    M ./doc/nonlinear.tex -2 +2
    M ./libctl/structure.cpp -9 +9 r1
    M ./src/meep.hpp -1 +1 r3
    M ./src/structure.cpp r3
    M ./src/update_e_from_d.cpp r3
    M ./tests/bench.cpp r1
    M ./tests/symmetry.cpp r1

Fri Aug  4 02:10:52 EDT 2006  [e-mail]
  * kerr does not require d_minus_p

    M ./src/update_e_from_d.cpp -2 +1

Fri Aug  4 02:02:12 EDT 2006  [e-mail]
  * eliminate update_e_from_d_update.hpp in favor of much smaller, somewhat faster, and more maintainable hand-written version

    M ./src/Makefile.am -5 +2
    M ./src/meep.hpp -3
    M ./src/update_e_from_d.cpp -3 +225
    R ./src/update_e_from_d_prepare.cpp
    R ./src/update_e_from_d_sources.cpp
    R ./src/update_e_from_d_update.cpp

Thu Aug  3 17:06:49 EDT 2006  [e-mail]
  * eliminated generated update_e_from_d_prepare.hpp (a lot of pointlessly duplicated code), and sped up somewhat, and eliminated old/unused ran.* files

    M ./src/Makefile.am -8 +5
    R ./src/ran.cpp
    R ./src/ran.hpp
    M ./src/step_d.cpp -1
    M ./src/step_h.cpp -1
    M ./src/update_e_from_d_prepare.cpp -3 +34
    M ./src/update_from_e.cpp -1

Thu Aug  3 15:31:33 EDT 2006  [e-mail]
  * do correct cylindrical integration for fallback_meaneps

    M ./libctl/structure.cpp -7 +12

Thu Aug  3 14:58:31 EDT 2006  [e-mail]
  * fixed(?) failure of convergence_cyl_waveguide test

    M ./src/anisotropic_averaging.cpp -7 +7
    M ./tests/convergence_cyl_waveguide.cpp +2

Wed Aug  2 22:24:51 EDT 2006  [e-mail]
  * fixed libctl fallback to use proper cubature

    M ./libctl/structure.cpp -3 +77

Wed Aug  2 21:48:01 EDT 2006  [e-mail]
  * minor bugfix to default meaneps routine (shouldn't affect results much)

    M ./src/anisotropic_averaging.cpp -8 +12

Wed Aug  2 21:15:06 EDT 2006  [e-mail]
  * allowed user control of subpixel tolerance via subpixel-tol and subpixel-maxeval parameters, and set these to more modest values, and output progress indicator during subpixel averaging (since it may be slow in 3d)

    M ./libctl/meep-ctl-swig.hpp +1
    M ./libctl/meep.scm.in -1 +4
    M ./libctl/structure.cpp -4 +10
    M ./src/anisotropic_averaging.cpp -12 +31
    M ./src/meep.hpp -6 +17
    M ./src/meep_internals.hpp +2
    M ./src/step.cpp -2
    M ./src/structure.cpp -4 +6

Wed Aug  2 20:03:23 EDT 2006  [e-mail]
  * merged Ardavan's much-improved subpixel-averaging code, after deleting a lot of obsolete stuff.

    M ./libctl/structure.cpp -13 +183
    M ./src/Makefile.am -2 +2
    M ./src/anisotropic_averaging.cpp -101 +199
    M ./src/meep.hpp -5 +18
    M ./src/meep/vec.hpp -1 +3
    M ./src/structure.cpp -71 +9
    R ./src/threevec.hpp
    M ./src/vec.cpp -2 +10

Wed Aug  2 19:26:43 EDT 2006  [e-mail]
  * remove debugging printf

    M ./src/sources.cpp -2

Wed Aug  2 14:08:03 EDT 2006  [e-mail]
  * remove annoying restriction on symmetries requiring even number of grid points, since this is no longer required (an extra pixel is added as needed)

    M ./src/structure.cpp -2 +2

Fri Jul 28 11:36:03 EDT 2006  [e-mail]
  * use $(EXEEXT) in dependency, thanks to Christopher Kang for the bug report