[cig-commits] [commit] devel's head updated: Merge pull request #18 from mpbl/devel (099a4d3)
cig_noreply at geodynamics.org
cig_noreply at geodynamics.org
Wed Apr 9 08:57:53 PDT 2014
Repository : ssh://geoshell/specfem3d_globe
Branch 'devel' now includes:
c36d69a Initial revision
27b5bcb *** empty log message ***
e2d7c38 *** empty log message ***
59f786a *** empty log message ***
df86c5d *** empty log message ***
e335c9e *** empty log message ***
09d5406 *** empty log message ***
66538f2 *** empty log message ***
efb9600 *** empty log message ***
5e3f813 *** empty log message ***
a1abcf1 *** empty log message ***
f3c8cc4 *** empty log message ***
9f5dbbf *** empty log message ***
92c7e8a *** empty log message ***
c26ed07 *** empty log message ***
ad47499 *** empty log message ***
a18f1ae *** empty log message ***
3c829e7 *** empty log message ***
055aeed *** empty log message ***
f794f08 *** empty log message ***
b4da4a7 *** empty log message ***
f969d11 *** empty log message ***
5da1e57 *** empty log message ***
9b848ab *** empty log message ***
c0fc395 *** empty log message ***
8c4c3f2 *** empty log message ***
b023521 *** empty log message ***
4e38d4e *** empty log message ***
4932e7f *** empty log message ***
2dba700 *** empty log message ***
8cfebd9 *** empty log message ***
343b110 *** empty log message ***
be2aa87 *** empty log message ***
fc11339 *** empty log message ***
04f2401 *** empty log message ***
d88df85 added maximum depth of oceans to suppress problem with trenches
596423c fixed problem with radial mesh on the ES - now uniform from d220 to surface
edb3314 added max of topography = 5000 meters
b0f86cb added AbSoft compiler to Makefile. Fixed deallocate(rmass_oceans) bug found by Jean-Paul Ampuero.
426eafc suppressed low-res movies for Keken, and replaced with high-res
6e0f360 allowed receivers to be buried in global code
40171df added OpenDX f90 routine to compute x y z position of hypocenter
e6f9bc3 fixed i6 bug in movies van Keken
6439710 removed old fluid-solid version with iterations
8af31c2 switched to new fluid-solid version with no iterations
9ccf196 saved backup of older fluid-solid version with iterations
b914bf3 added parallel version of fully anisotropic solver for Sebastien
7692a64 Moved all movie-related parameters to the read_parameters file and updated all related calls.
4b3316f Fixed a bug in create_movie_AVS_DX.f90 related to counting the four corners of an element on the surface.
d73eeb7 Fixed the movie code to work at full resolution. Modified the regular code to display velocity in the outer core for movies.
8e672f5 Added GMT capabilities to the create_movie code.
d7f23c9 merged regional and global codes
c747e49 Updated the README file to reflect the use of finite sources and movie capabilities.
8c3de0a fixed bug in get_attenuation (need fictitious inner core even for regional)
0046e0d Moved the absorbing boundary conditions from compute_forces_crust_mantle to specfem3D. Check that both the inlined and the regular version are stable for 25 minutes and give identical answers
f1de25f Implemented absorbing boundaries in regional code for the outer core and the crust-mantle. Tested that the code is stable. Coupling at the CMB appears to be fine.
9c1f4c7 We now have working versions of the both the regular and the inlined versions with absorbing boundary conditions. The regular and inlined versions give identical results on the cluster and are stable for 10,000 timesteps.
427c9e0 added list of Gurnis's Beowulf nodes in mymachines_gurnis
ee1345f fixed inner core bug detected on the ES in Japan
738d501 removed unused array veloc() from compute_forces_crust_mantle
a5b0ad2 added Funaro routines from Bernhard Schubert removed old anisotropy code from Sebastien
58a11bb fixed bug in ES inlining directives for Stacey absorbing conditions
1ff0e36 added clean backup of anisotropic code for Sebastien (OMP July 2003, serial, for IBM p690 at CNES)
10e2df8 added comments about CMTSOLUTION units to get_cmt.f90
cacc12b added Par_file for 507 nodes (4056 processors) on the ES in Japan
59636f7 fixed version number (now all files have v 3.4 label)
a557992 updated Makefile for Intel ifc v7.1
df93c72 added comment about vectorized loop for NSOURCES
6bb9547 Fixed a mismatch in types between combine_AVS_DX.f90 and write_AVS_DX_global_chunks_data.f90. CV: ----------------------------------------------------------------------
abb36ea vectorized versions of sources and receivers detection for the ES in Japan
470e2d6 modified Makefile for vectorized source and detection routine for ES in Japan. Modified go_* scripts for MPICH compiled with ifc v7.1
767e4d7 more modifications to vectorize the NSOURCES code for the ES in Japan
ad2ad69 changed default value of SUPPRESS_STOPS_INLINE to .true. for ES in Japan
ee39dbf added script show_all_max_amplitudes.csh to monitor max amplitude of results
39d21f0 station name up to 8 characters in write_seismograms.f90
9ce9acf changed code to allow t_cmt(i) = 0 to be for any CMT source, not only first
760fed2 removed unused test from locate_source.f90
8a26fe4 changed hdur(1) to minval(hdur) to save seismograms
e7a45d0 added convention for definition of reference frame in comments and README
5b2ce1e fixed format problem for double precision seismograms
3c9174e added file STATIONS_tsuboi_tokachi-oki from Tsuboi-san
65a8de8 added comment about bug in euler_angles. One of the formulas is not general enough, it does not work (I think) for a chunk of size <> 90 degrees
fa597cc Fixed a bug in crustal_model detected by Brian Savage. The Moho depth was not properly smoothed before placing it on the mesh.
85fac63 added rename_files_IBM_f_2_f90.csh script to DATA/util
ccaf280 added Denali, Alaska, source CMTSOLUTION_DENALI
a737a39 Fixed bug introduced by Jeroen for sign of coupling condition at the CMB. Also removed old inlined loops on NSOURCES for ES in Japan.
8ac4b63 finished removing old inlined NSOURCES loops in inlined version of solver for the ES
cc24450 modified regional code to allow for rectangular slice for Lev Vinnik Sdiff study
c2f1b16 fixed time step problem in some Par_files
defe347 added movie code for Tsuboi for ES in Japan in DATA/util/tsuboi_movie_feb2004
2e22baa modified constants.h to be consistent with Sdiff regional code for Vinnik
f75a97a modified write_seismograms.f90 to handle any length of station and network name
c6dc19e changed Par_file and constants.h not to save movies and OpenDX files by default
fbf2dfa modified movie code in the inlined version of the solver to check that half-duration of the source is ok for movies
7755587 modified go_mesher and go_solver to be more robust in terms of detecting NCHUNKS
1e75060 added Par_file for ES movies based on NEX=320
1c3d1cb modified the code to handle 2 chunks (NCHUNKS = 2). Will still need to improve the code in the future to allow for 2 chunks with angles <> 90 degrees, and to assemble the second corner between chunks (in the current version for simplicity only the first corner is assembled)
3a86f58 added test that dynamic memory allocation goes well in create_movie_AVS_DX.f90
4040afc updated CMTSOLUTION info for multiple sources in README file
85edcc9 fixed small bug related to length of station names
2366b7a converted information printed about distance to single precision in locate_receivers.f90
f22fed8 updated Makefile to take into account new Intel ifort v8.0 compiler options
1a9455a fixed one more detail in Makefile
000127a modified format() statements in read_parameter_file.f90 to suppress fixed length of fields
6b6c0d9 3-D Attenuation added
e2d0dc7 added 3D anisotropic code for copper crystal for Jose Carcione in DATA/util
297b72b updated anisotropy code for Carcione (copper crystal) in DATA/util
1e1e06f updated Makefile options for Intel ifort compiler and updated go_* scripts
8421fb0 fixed non-declared length problem in format() in read_parameter_file.f90
12984c3 updated anisotropic code for Jose Carcione in DATA/util
daefd6d distinguish between scalar and vector MPI buffers in assemble_MPI*.f90 in order to avoid fictitious additional dimension (NDIM,...) when using a scalar buffer
0b68d15 Fixed number of dimensions to factor_common
6410b12 changed NSEIS to large values in all the Par_file* in order to save seismograms only once at the end of the simulation to avoid NFS problems
9d5c2dc updated the README file for the global code to inform users that values of NEX_XI or ETA below 64 can make the mesher exit with an error message (negative Jacobian)
a8abcaf removed white spaces, cleaned listing and cleaned attenuation_model.f90 to compile without warnings with Intel ifort
77efef4 added test to check that sphere can be cut into slices without getting negative Jacobian (in practice, check that both NEX_XI and NEX_ETA are >= 48)
cfd8da7 added comment about the fact that 3D attenuation model is not included and that currently only PREM attenuation is available
17c642c cleaned attenuation listing
9b90a63 fixed typo in comment
f42e769 added smoothed etopo-4 topo/bathy model, and modified routines accordingly
ce308b5 cleaned the code for 3D attenuation and better topo/bathy model (etopo-4) in order to release version 3.5 in a few days (still need to fix the code for any chunk size)
3952fa5 Cleaned code to have chunk of any size and modified corresponding lines in Par_file. Still need to clean euler_angles.f90
60673b8 released version 3.5 of global code and version 1.2 of basin code
87e9cee fixed small bug in calculation of exact surface of mesh (not correct if NCHUNKS /= 6)
eb8d86b suppressed derive_parameters.f90, which was not used anymore
7ae5df2 added white spaces to align tabs in loops
9e5c9c9 added output of number of time steps to output_mesher.txt
e46d08e fixed problem in default values in constants.h
97dded0 cleaned comments before releasing new versions of both codes (v1.2 and v3.5)
4ac2ae4 fixed vertical edges problem for Stacey found by Qinya in get_absorb.f90
e2b5e31 suppressed obsolete regional/global if statement in get_attenuation_model.f90
05b04a9 modify line-break for absoft compiler (-W132)
647a044 updated Makefile options for Absoft compiler
1966065 added comment in anisotropic_mantle_model.f90
61ba5c8 added line to suppress Montagner's model if needed in anisotropic_mantle_model.f90
fe93ce4 fixed bug in IASPEI radii (R80 and RMIDDLE_CRUST) found by Jeroen (previously set to 1., which was dangerous) in read_parameter_file.f90
7af7b1f moved the check of SAVE_AVS_DX_MESH at the beginning of the file to after the read_parameter_file subroutine since it is no longer in the constants file.
8860234 added deallocate(Qmu_store) and deallocate(tau_e_store) to create_regions_mesh.f90 to fix small bug found by Bernhard Schubert on Hitachi compiler
00f762f ROCEAN now set to PREM value if IASPEI is used to avoid problem with fictitious ROCEAN = 1.0d0 in make_ellipticity.f90
0654c27 added support for g95 compiler in Makefile
92790ac incorporate movie 3-D output statements added by Peter van Keken
547a949 Now the code computes seismograms for a source that is equivalent to a triangle with half duration given in CMT file. Previously the half duration from the CMT file was used directly when computing the Gaussian.
813515a Following modification to the method of determining the stress and strain relaxation times in the standard linear solid system for attenuation implementation. Old method was a SVD inversion , new version is a simplex inversion. Simplex methods contained within attenuation_simplex.f90 Modifications made to attenuation.f90 to mirror attenuation_prem.c output. Added run_attenuation2.sh to make it easier to run the new attenuation code.
518e8e9 Modified the attenuation_conversion code for 3D attenuation to use a simplex method for determining the stress and strain relaxation times in a standard linear solid. The previous version used a SVD inversion which worked well for large values of Q but not for small values of Q (Q < 80). No changes to existing code should be necesary, although meshing may be slightly longer due to the increased compuatation.
5d8e3fa added .keepdir to keep some empty directories such as obj/
47a4a29 Inserted new tau_mu's into get_attenuation_model.f90 based on Brian Savage's simplex inversion for the standard linear solids that will give the nearest to constant Q values.
872e666 changed "nodes" to "processors" in go_* scripts
a8dc3d0 modified compute_arrays_source.f90 to compile much faster under Portland pgf90 (used to take 20 minutes!)
a1c97d7 initialized memory variables before time loop in specfem3D*.f90
67f5782 Changed elat and elon to lat and long in order to be consistent in all the subroutines. fixed small bug in initialization of deviatoric tensor epsilondev for memory variables in global code. Added bak directory.
15b3400 fixed OUTPUT_FILES/timestamp* OUTPUT_FILES/starttime*txt in Makefile
fdd3827 moved inlined version of the solver to DATA/util/scaling_nodes_ES. Won't be maintained anymore.
2005de9 added read_value_parameters.f90 to global code to ignore comments and white lines in Par_file
0037460 fixed small details in two comments
abbbf32 updated run3d.csh scripts and also OpenDX stuff in DATA/util/opendx_AVS_GMT
76a116d clean implementation of Lacq gas field MNT in basin code
90caa33 removed SPECFEM3D_GLOBE/DATA/util/wasiu_code/wasiu_code_26feb03.tar.bz2
f93dcce Fixed subroutine name and allocation deallocation issues Default 3D attenuation now donw by simplex inversion Somewhat slower than SVD inversion but produces better results
323deb5 Modified Files: anisotropic_mantle_model.f90: in line 555 change the reading format from (21f5.3) to (21f7.3) DATA/Montagner_model/Adrem119 Added Files: DATA/Montagner_model/Adrem119_old
3255ee3 removed old script in bak in Makefile
adc4af3 According to Seiji Tsuboi's Email, they use DT = 0.072 for V3.3 in 1944 processor case. Modify DT from 0.05 to 0.072.
33eeb63 updated Makefiles for IDRIS machines (IBM Power4 machines at www.idris.fr)
facf20a removed rename_files_IBM_f_2_f90.csh and rename_files_IBM_f90_2_f.csh and changed Makefile for IBM accordingly
eb6164f changed Makefiles for IBM Power 4
fad6dad suppressed constraint of aligning values in Par_file on the 34th column
affd0cd fixed portability problem that made reading of Par_file crash on IBM Power4
0a0f08a fixed some comments
e133fea exit_mpi.f90 -- added the if conditions for closing IMAIN
b67232f modify go_solver to run as front job
25c2afb changed details in global code: switched order of parameters in calls to subroutines to use NEX_XI,NEX_ETA instead of NEX_ETA,NEX_XI and NPROC_XI,NPROC_ETA instead of NPROC_ETA,NPROC_XI. Also changed minor details in attenuation code by Brian Savage: suppressed semicolons, changed real(8) to double precision etc.
0bbb3f6 define r_earth and r_ocean differently under gravity = .true. to avoid problems in make_gravity subroutine in the solver
55d6e35 Add full anisotropic tensor (total 21 elements) rotation in the subroutine anisotropic_mantle_model.f90.
00708d3 Vala: Added HDUR_MOVIE to parameter file, and modified all files that call read_parameter_file. If HDUR_MOVIE=0.d0 in Par_file, a HDUR_MOVIE appropriate for the resolution of the run is computed in read_parameters_file.f90. Also fixed the bug where hdur(1) is being called, by introducing the variable t0 that is the start time of the simulation. The source decay rate was changed in constants.h and the 5/3 to convert to gaussian hdur was removed from compute_source_time_function and the division by source decay rate was moved to specfem3D.f90.
e4701f2 suppressed old CDIR NODEP directives that remained from Japanese Earth Simulator implementation
efc3371 removed "format(" from 3D global code
b61a48f Add the auxiliary programs to convert the mesh files created by the specfem package into the format known to Paraview (unstructured grid files)
72aa655 renamed save_arrays.f90 to save_arrays_solver.f90
c493718 Added an Output section at the bottom to report various parameters internal to the code, but needed for computing the memory requirements. A new file OUTPUT_FILES/derived_values.h will be created, which is then read in by memory.pl to estimate the memory usage for the solver and mesher.
415fe23 Added in yet another attenuation subroutine describing the PREM attenuation structure. It is called from get_attenuation_model() when the MIN_ and MAX_ATTENUATION_PERIODS are determined automatically. This code conincides with the automatic radial meshing for very dense meshes and CHUNKS less than 90 degrees.
631890f Code to automatically determine the attenuation parameters for period ranges which do not match the current one defined which makes up most of this file. This code will be called for very dense meshes and CHUNKS less than 90 degrees.
f577da0 Brian's subroutine for automatical determination of radial spacing (NER's)
58b2a54 removed mymachines_gurnis
f57f20b moved FIX_UNDERFLOW_PROBLEM to section that should never be modified in constants.h
5ed47a4 renamed SAVE_AVS_DX_MESH_FILES to SAVE_MESH_FILES
9077998 converted solver from static memory allocation to dynamic
302eeaf changed details in statistics that are printed about the mesh
3d9f4db fixed bug introduced by Brian Savage: writing output to IMAIN while it is not open yet
75cf318 write mesh statistics to IMAIN instead of IOUT
17dfef5 suppressed extra white spaces that resulted in "line too long" error
b0bb47e removed old stuff from DATA/util
be40d44 removed old stuff from DATA/util
77a75a6 removed old stuff from DATA/util
c115d48 removed old stuff from DATA/util
0316dfc (Re)Pyrization Part I: Read parameters from a Pyre inventory.
eb0d2d7 Fixed a bug in read_value_string(), and improved the error message.
3e89ec1 (Re)Pyrization Part II: Partition inventory items among various Pyre components which mirror the concepts in the underlying Fortran code.
39861a1 Doh! Sync'd read_value_parameters with my changes to the Pyrized version.
49550ef Improved error reporting for bogus user input; unfortunately this required patching Pyre ("PyrePatches.py"). Incorporated my launchers from CitComS.
079ae28 Added a codec to allow the use of INI-format ".cfg" parameter/preference files, which users will undoubtedly prefer to XML.
bc5f941 Command line overrides .cfg files, and not vice-versa. Allow users to give .cfg/.pml files on the command line. Handle more exceptions in _richCast() (bool, dimensional). Changed RECORD_LENGTH_IN_MINUTES to a dimensional.
41a95cd Added example parameter file and preferences file for the Pyrized version; they are derived from the traditional "Par_file".
2d3ae5e Link Pyrized version with MPICC instead of MPI90: ifort preempts our 'main' with its own.
6f8fc0d Call MPI_Init() from "workers" only -- and from Python instead of main() -- in order to avoid SIGPIPE with MPICH-GM ("<MPICH-GM> Error: Need to obtain the job magic number in GMPI_MAGIC !"). Created a tiny Pyrex MPI module for this purpose.
fa4a694 Running jobs on Pangu: added "bsub-options" to LauncherLSF; added Fortran function check_parameters.
8ed3988 added routines to convert DX files to AVS in DATA/util/opendx_AVS/
712855f Allow the Pyrized version to override the hard-coded pathnames in the code ('DATA', 'OUTPUT_FILES') so that users can specify an alternate location for files, and even effortlessly change the output directory between runs.
013c9f1 Check input/output pathnames from Python.
52aa8e5 Reworked InputPile handling.
a85358e Work-around for bizzare performance hit incurred by mixed-language programming with ifort v9 on Pangu (Pyrized version was 3x+ slower than classic version).
2defef1 Use CIT_FC_MAIN instead AC_FC_MAIN, which doesn't work with ifort v9.
313f504 Validate parameters in Python.
f0b73db Eliminated default time limit for batch jobs (port of fix for CitcomS issue8).
12a39cc Checked-in Pyrex-generated code, so that users don't need to have Pyrex installed to build the Pyrized version. (The Pyrex source is quite stable.)
6557105 Refactored Python scripts so that they can run standalone on a web server (without the Fortran/C code present).
a65c697 Fix for compile-time errors (bugs?) reported by the NAGWare f95 compiler:
397e999 Reference Model inventory items by their Python name, not their Pyre name.
50c0c9c added LaTeX manual under DATA/util/manual
d2b3ca0 added mesh_scaling_processors_ES_japan.xls to DATA/util/scaling_nodes (Excel file to compute mesh size and scaling on large machines)
4f46e31 removed -Mstandard from the Portland pgf90 Makefile because gives errors that are due to a bug in version 6 of pgf90 (the Portland technical support confirmed that the problem comes from their compiler, not from SPECFEM3D)
6ff3369 create the adjoint branch off the 502 version of the global code, and merge the adjoint changes I made into the branch.
72a80f1 Fix the hdur/hdur_gaussian problem according to the newest update
1010438 fix the bug in combine_paraview_data.f90 where arg() should have array length of 6 instead 5
ea50840 update HDUR_MOVIE and IOVTK variables
a1a82bf Fix Makefile, add combine_paraview_data
0bc6d4e Fix for possible bug: initialize 'A_array_rotation' and 'B_array_rotation' to zero.
921c104 combine_paraview_data.f90: fix the strings after stop function to aviod compilation errors for ifort
fa2308b Reverted r517 by komatits ("converted solver from static memory allocation to dynamic").
add8139 fixed problem introduced by Brian Savage: writing output to IMAIN going nowhere
b236fcc fixed order of parameters from NEX_ETA,NEX_XI to NEX_XI,NEX_ETA
2dd9b5a fixed two bugs detected by Bernhard Schubert: AVS point order was incorrect, and one statement was incorrect when using less than 6 chunks
7f673aa Suppress ifort's warning about unused variables with "-warn unused".
3fc9857 added comment about routine handling unused variables
b817682 removed -fast and -Munroll for Portland pgf90, because that compiler is not very stable and these options create problems on some machines
ebd0367 Minor.
70fe818 Merged Qinya Liu's adjoint code -- r570:600 from the 'adjoint' branch -- to the trunk.
3025e0d renamed file
d0b0fe1 changed copyright information in new files added by Qinya
3c5eb00 added classical files with list of developers, copyright, known bugs etc.
f2240ff moved SIMULATION_TYPE and SAVE_FORWARD to Par_file. Also removed unused variables in several other routines.
1ad3c27 updated BibTeX file and changed citations accordingly in the manual
0fee12c Update the scaling and dimension of the kernel variables. Now output kernel is physically in s/km^3 unit.
d843c0f Fix some minor errors to eliminate the warning messages from ifort compilation
2771454 fix bug of scale_disp
a992622 Added SIMULATION_TYPE and SAVE_FORWARD to Pyre code.
de2594a Added 'simulation-type' and 'save-forward'.
59377ec change back save_arrays() to save_arrays_solver() according to the newest version in the main branch
75d3c54 clean this empty file
16ebaaa fix NPROC_ETA,NPROC_XI for read_parameter_file.f90
811a23c Suppressed NGLLSQUARE and NGLLCUBE. Cleaned extra white spaces at end of lines in listings using a script. Changed length of station and network names in LaTeX manual.
e106811 removed NGLLSQUARE. reduced size of postscript figure in Manual.
502002c changed jtromp at caltech.edu to jtromp AT caltech.edu to avoid scanning of source code by spam engines
dbbd43c Delete PvK MOVIE_VOLUME implementation.
ac313aa added comment about how to report bugs.
d0807a0 added scripts to run on the "pangu" cluster at Caltech.
7c6d5f4 added paragraph about how to define the "machines" file on "pangu"
c3ccf2d removed extra white spaces in listing
e5186a0 fixed important topography bug detected by Bernhard Schubert: longitude blocks were inverted in topography/bathymetry file
169be95 - added details about how to check the mesh in the manual.
700eae4 changed transversly to transversely (there was a typo)
dc90e93 change exit_mpi to exit_MPI in specfem3D.f90 rename the saved array files to start with 'proc'
099e598 Further optimize the loops in the 'it' main loop.
c820b30 get rid of dot_product in the kernel calculations
7ffa04b fix bugs in combine_paraview_data.f90 which fails to treat cases of low-resolution mesh of the inner_core.
44e722a - added section to additionally output SAC format in write_seismograms.f90 (- creates seismogram files named according to ASCII files + label ".sac" - leads to only 50% more disk space for seismogram output as time information is not written explicitly for every station and component)
22b691b moved Users' Manual from DATA/util/manual to USERS_MANUAL
171c86f cleaned white spaces and comments
e6ba805 moved DATA/util/scripts_pangu_Caltech to SCRIPTS_SCHEDULER
6384bf0 committed Jeroen's changes to the manual and my changes to the README and BibTeX files.
7ae0aad Restructured Pyrized version to easily allow running mesher+solver in a single batch (LSF) job.
7c0ef87 Minor bug fix: must set scheduler.nodes in addition to launcher.nodes.
0ad1aba Last-minute fixes before meeting :-)
bda857b A handful of fixes and work-arounds.
22d7d80 Use Cheetah instead of Opal for HTML generation.
9e18c40 More web interface work.
5eb6a94 Whoops -- one more template file.
75ae405 added options for IBM MareNostrum at the Barcelona SuperComputing Center (Spain)
fb4c70e added create_header_file to "make all"
d71e83f suppressed non-standard "call flush(IMAIN)"
e4401ed added program to filter the STATIONS file to include stations located in a given region only
ebeb827 Revamped build system -- standardizing it somewhat -- in anticipation of building and testing Specfem under CIG's BuildBot. Users must now run 'configure' before building for the first time -- 'Makefile', 'constants.h', and 'precision.h' are now generated by 'configure'.
4d86175 - created and added regional_c1_1.html.tmpl, regional_c2_1.html.tmpl, regional_c3_1.html.tmpl - modified global.html.tmpl - modifed regional.html.tmpl
6f70002 Factored-out Fortran-compiler-flag guessing code into 'flags.guess', so that it can be updated without having Autotools installed. Also, svn:ignore all the files which are now generated.
3cefd6b Bug fixes, tabs -> spaces for flags.guess.
fbc6556 Merged r664 by komatits from Basin code:
bcd64e2 Significant re-work of Pyrized version (still in progress).
3e2cb64 changed (i4) to (i5) in Qinya's display of mesh topology, otherwise problem when using more than 999 processors.
64f8e35 added routine to display mesh topology (how processor numbers are assigned in the cubed sphere)
c04a706 Reverted accidental change.
386f47f merged the "do l=1,NGLLY" and "do l=1,NGLLZ" loops because we know that NGLLX = NGLLY = NGLLZ and therefore the compiler can optimize further if we merge the three loops in one.
94cdfdc I'm getting there: currently, it gets to mpirun, and then bombs.
150b470 removed derived_values.h problem introduced by Brian Savage
bea6452 replaced reading of files by all the nodes from directory OUTPUT_FILES with call to MPI_BCAST to avoid bottleneck on very large number of nodes.
4c40d82 changed i3.3 and i4.4 to i6.6 to allow for up to 999,999 processors to be used instead of 999 or 9,999.
9589561 changed "slave" to "worker" in all the source files
5951652 Specfem can launch now! But the output files are still a mess. Need to add scratch dir cleanup and collection of seismos.
4d4efad fixed definition of IASP91 model in the upper mantle and crust, which was completely wrong
e85bb95 added models 1066a and ak135, and changed call structure of iasp91
f9fd72e added tools to display models 1066a and ak135
8b1b300 removed semicolons from source code because they make the code difficult to read
1b6af00 Fixed build under version 8 of the Intel compiler.
4a88401 Silly work-around for hang during compile with ifort v8.1.
a447236 Fixed infinite loop introduced by my previous check-in (it is true that I don't really know Fortran).
3c04c82 The scratch dir is now cleaned-up when the job completes, and a tarball of everything in the output-dir is staged back to the login node.
6bbbbce fixed integer*8 problem in that include file from MPICH
7045a0c Added a codec for Par_files.
bc390f6 moved three of the six loops to a separate subroutine because (although the code was correct) some compilers used to take forever trying to unroll the six levels of loops in that file.
d654ece wrote better version of routines for Earth models 1066a and ak135 (model is now defined and stored once and for all)
a75d3f4 fixed a potential problem: Q values for attenuation could be zero near the CMB and ICB because of possible roundoff errors.
7a30580 fixed potential problem for Vs in outer core in models 1066a and ak135 due to possible roundoff errors on depth
5a4a7e4 fixed radii values for 1066a and ak135, which were incorrect, now uses new values provided by Ying Zhou <yingz at gps.caltech.edu>
4e69bdb added comment about bug for ROCEANS in the case of the 1066a model
8266358 fixed some radii for ak135 in main code, and for PREM in display routine
922c54d changed a comment
d1aa7ee Added CIT_FC_MPI_HEADER, which, if necessary, will automatically generate a 'mpif.h' sanitized for Fortran-9x. (Only works with MPICH at present... not sure if other MPI implementations have a problem.)
7db1656 Reset AC_LANG back to C for Pyrized build.
a33ce1b fixed problem with ROCEAN for model 1066A, which has no oceans
fd37fe0 Added "call system()" calls for MareNostrum in Barcelona. Also added an MPI cascade to avoid overloading the GPFS file system there. Moved statements and comments about beginning of time loop to the right location. Renamed isender to sender and ireceiver to receiver.
2037bf9 Removed the last bottlenecks related to reading files on all the nodes from the remotely-mounted home file system in directory DATA.
e75ee26 fixed deadlocking problem in my MPI implementation of Bernhard Schubert's SAC seismograms routine. added file get_event_info.f90 (which contains both a serial and a parallel version) to get information needed for the SAC headers without deadlocking.
2263931 added non-dimensionalization to models 1066a and ak135 (it had been forgotten)
b245543 modified routines for models 1066a and ak135 to make sure behavior is correct around the ICB and CMB (the previous version had a bug: it could take a point above and a point below and interpolate between them, leading to a wrong value)
04d73c0 removed old trick used to fix a potential roundoff problem in the interpolation code across the CMB and ICB
6c7920e moved creation of central cube buffers from main program to a separate subroutine (create_central_cube_buffers.f90)
fe40c1e fixed bug in scaling of radius for models 1066a and ak135. added flag to avoid saving seismograms to local disks on MareNostrum. removed useless space in create_regions_mesh.f90.
f68fea2 Pyrized version can now collect seismograms. (Finally!)
17cd116 Removed 'values_from_mesher.h' from the repository. Use a Makefile rule instead.
ffcd3ee suppressed old dummy .keepdir files that remained from the conversion from CVS to SubVersion.
2ff3588 Don't archive seismograms during dry runs. Also, the mesher always writes 'values_from_mesher.h', so set its pathname correctly in all cases.
83bcb42 Use OUTPUT_FILES variable instead of hard-coded path (for Pyrized version).
bc932ed The 'local-path' option is now called 'scratch-dir'.
491d73a Final bug fixes: create solver header file if it doesn't exist; minor fix in ParFileCodec.
a5f3573 for model IASP91, following a suggestion by Ying Zhou, changed ROCEAN = 6368000.d0 RMIDDLE_CRUST = 6356000.d0 to ROCEAN = 6371000.d0 RMIDDLE_CRUST = 6351000.d0
98a8495 Modified code to use less memory for large adjoint runs (large meaning many adjoint sources and receivers and many processors). Changes involve: *reading adjoint sources in in chunks *writing seismograms out in chunks for adjoint runs (not storing the entire trace *locating sources in chunks
6acad2d Split MPI_GATHER lines in locate_sources routine that generated errors. Removed a variable (iline) from comp_subarrays_adjoint_src that wasn't used.
c43272b Fixed typo.
db340ba New Attenuation Routines for 1D Attenuation Models
f29d4a1 Use eggs for Python dependencies. Also, made some minor tweaks to the Fortran so that it compiles with g95.
598d13c Initial Version.
fc315a9 Removed *.pyc files.
81c74b5 Convert seismograms to SAC binary format using asc2sac() (the *.sac files produced by the Fortran code are "alphanumeric" SAC files, which isn't what is wanted).
413c4e0 Fixed typo.
c0aedc1 - Added UserInfo, Mesh, Model, Simulation tables in the database (models.py) - For now, there is no user authentication (e.g. login). Let's use 'test_user' for testing purposes. - Added simulation creation, viewing, deletion functionalities.
3c0d591 Initial Version
7d21197 Minor cleanup of HTML.
501724a More HTML cleanup. Pluralized some URLs and removed references to "mysite".
6e68708 Model clean-up: made 'simulation_type' an integer choice; removed parameters that don't make sense for the web.
aaae6de Split global/regional form template into separate files using template inheritance.
84452f1 Beautified HTML.
9c536d0 Added skeleton code for PML file download.
f3bf3dd - views.py - skipped hdur_movie field when writing to db. hdur_movie valued will be set to 1000 which is default value. - models.py - added ntstep_between_output_info, ntstep_between_output_seismos fields to simulation table. - set 1000 as hdur_movie field's default value. - simulation_base.html - set 100 as timestep's default displayed value. - detail.html - nothing significant. (changed the html table border size) - simulation.pml - now this page displays values read from db.
87cb2f4 - Created [mesh_info,model_info,movie_info,output_format_info].html for user information - Modified [views,urls].py to support user information displays above - Modified [simulation_base,simulation_form_regional].html to have button which will pop up a info window - Created Daemon.py which will read ready records from database, fetch xml files from web server, and finally save it to a local file.
0b51260 - Fixed a bug where cur = conn.cursor() was used twice within the for loop, and it caused only first record to be downloaded.
2f7e1c8 Added an 'events' sub-application for collecting CMTSOLUTION data ('events' is actually a component of the CIG framework).
81afe1f Put most object files into an archive library, 'libspecfem.a', and thereby let the linker decide which object files are needed to link which programs. (That's the linker's job!)
e234e72 Pyrized create_movie_AVS_DX. (Internalized its rank() and swap_all() subroutines to avoid multiply defined symbol conflict.)
4e22408 Changed the url path of http://.../simulations/sid.pml to http://.../simulations/sid/parameters.pml
ed985ff - views.py - Fixed indentation error (replaced spaces -> tabs)
6d5a1a9 - Added stations.txt, events.txt for Daemon interce. - Added userinfo_form.html for user registration interface. - Modified home.html, detail.html, base.html for simpler interface with user login feature.
a5d08ac - Added session (user login) feature with simpler interface.
3612fba Incorporated my new 'stations' mini-app for selecting seismo stations (STATIONS).
30451d8 Got inline editing working for UserInfo's OneToOneField. The "edit registration" page now works again, but the "new registration" page is still broken.
5c37733 Registration is now fully functional. I've given up on inline editing; registration now uses a custom form.
9d2a0ad Added 'simulation_list.py' and made related changes to the daemon.
91bbe2f Reworked the "Mesh" section of the "new simulation" form. There are actually several constraints that vary with the number of chunks; as the number of chunks increases, the more constraints there are. (It is more complicated than simply regional vs. global.)
ad4e93b Consolidated site structure somewhat.
c89fe42 added program to permute point order in a CUBIT mesh to generate an OpenDX file
9df9e48 Replaced templates 'events.txt' and 'stations.txt' with Python code which does precise formatting. Fixed bug where 'absorbing_conditions' wasn't being set.
1bbc692 Added a form for the daemon to change simulation status and upload output. Added a link for the user to download the output.
0e1b175 Completed the sketch of how the daemon will work.
2289257 Gave the daemon the ability to post "multipart/form-data", so that it can upload the simulation output to the web server.
955bb18 Connected the daemon to the Pyrized Specfem code; it did a dry run of a simulation. Moved 'Daemon.py' out of the 'web' subdirectory to avoid a name conflict with 'Specfem3DGlobe'.
14da573 Fixed unclosed HTML comment.
8ccc88b Avoid specific fonts & sizes so that the site is readable in all browsers.
a1e017e Changed response of simulation status form to text/plain Python repr() (either 'OK' or errors) so the Daemon can log/parse it.
9fe7b3b Zero "time shift" of CMTSOLUTION, as required by Specfem. Currently, it is not possible to specify a kinematic rupture from the web portal.
90be348 Changes leading up to the first working run of the daemon.
f81d9ca Archive all output files, not just seismograms. Also fixed glitch in Models.py.
1506301 Trap validation errors, log them, and keep going.
a4da5b5 Fixed mislabeled input.
a03a35d Log & keep going if output tarball is missing.
804ed14 updated the copyright information to release version 3.6
e8149b7 changed comments and updated the copyright information to release version 3.6
9e415b1 more updated comments for release of version 3.6
bfaf567 - models.py - removed some models (iaspei, ak135, 3d isotropic) and changed MODEL_TYPES enum values - renamed 'isotropic prem' -> 'isotropic PREM' - renamed 'transversely isotropic prem' -> 'transversely isotropic PREM' - added 's20rts' model
5bdbce4 - model_info.html - removed description for iasp91 and ak135 models
9867d4c - models.py - removed '3d anisotropic' model
35459a1 - model_info.html - fixed a typo (transverserly -> transversely).
a3abbbe Initial re-work of events & stations, in response to feedback from Tromp's group. Users can now upload events and stations, manage multiple station lists, and create multi-source events (kinematic ruptures).
870a306 - models.py - removed 'adjoint' and 'both forward and adjoint' from SIMULATION_TYPES
bc9d6d5 "Surface" -> "Create movie". Hide 'simulation_type'.
d2063f1 Updated get_type_id() to reflect new MODEL_TYPES.
509a14b Removed random garbage inserted by... Subversion?
286ffa0 Fix the inconsistent naming scheme of save_forward_array file
0fcf504 move the scripts/programs directory name to UTILS/
85af046 add utility programs into UTILS/
8647c0f - Added model types (1066a, iasp91, ak135, crust2.0+PREM) and deleted model types (3d attenuation).
c6c737b Removal of myrank variables and general attenuaiton code cleanups
edf6cf9 model options updated: ------------ isotropic_prem transversely_isotropic_prem crust_2.0_plus_transversely_isotropic_prem iasp91 1066a ak135 s20rts 3D_anisotropic 3D_attenuation -----------
b8f8b33 added a comment about the call to read_aniso_mantle_model
aeb8455 add auxilliary programs for
29e249a remove unnecessary temp.m file
01a0844 Added create_movie_GMT_global, and modified the Makefile.in accordingly. Note that this code is not yet pyrized.
3eb3ea5 Moving web portal to its own project directory......done.
3f1889d
2f3ee27 update the output of sr.vtk for multiple sources and multiple receivers.
c47317c fix usage of change_simulation_type.pl
dcf4e77 for the manual: - add kernel figure - post-processing scripts - simpler scheduler scripts
5d6c5f5 modify globe_slice_number.pl to add the calculation of the normal of the source-receiver plane
609efb3 add more comments for the scripts
86593a6 Remove HOWTO, since everything is documented in the manual now.
89c5afa Version 3.5 -> 3.6 on handful more files.
55e1817 add more utility scripts
fa6e2ac remove binary executables from the SVN sever
a4ddc93 modifications to take care of warning message.
fad9eba remove mesh2vtu under DATA/util
ddf86a1 correct scheduler bash file name
d6101fa more modifications
b91f0ac add lsf scripts for kernel simulation
5725b9b fix plot_source_time_function.txt more changes in the global manual
c4e7070 added KoVi98 to list of possible BibTeX references to cite
2c3877f added DOI information to BibTeX references
8b7c75f add bib
59692b6 fix bug in plot_source_time_function.txt
283dc0b fix kernel comments for UTILS/run_lsf
acdf376 fix bug in run_lsf
8bcc70d fix attenuation_conversion(myrank...) bug
0eb5418 filter stations for 1 chunk run
8f64c11 Improved robustness by adding 'preparing' and 'error' states, forking the launch process (which also builds the solver), and putting a try-except around poll().
ea17c3f fix bug in locate_receiver.f90 (broadcast nu!)
2babe6f Added one more state: 'finishing'. The upload of output data at the end of a simulation can take a long time... the daemon needs to fork.
0b5c6a3 Fixed bug in "move old sim dir out of the way" code.
8ab3939 Brought model names into sync with code. Reaffirms my conviction: one language, one code.
5435c55 small modification in Qinya's code to exclude stations that are outside the mesh: I now print a different message depending if NCHUNKS = 6 or not.
4adafbf use PREM to get the density profile for ellipticity and gravity (fine for other 1D reference models)
4537f01 renamed run_lsf to run_lsf.bash
2ef14e7 added code to suppress trailing carriage return (ASCII code 13) if any (e.g. if input text file coming from Windows/DOS)
5a32592 fix globe go script
011074e added support for GNU gfortran compiler
25c865e Fix for different 'flags.guess' behavior when FC is guessed vs. when it is specified explicitly.
fd854b8 Regenerated 'configure' using the latest version of Autoconf (v2.61) so that 'configure' will search for 'g95', and prefer 'ifort' to 'ifc' when checking for the Intel Fortran driver.
e026608 Updated the Pyrization to use the latest and greatest: the newly released (v0.8.1.x), egg-based "CIG Edition" of Pythia/Pyre; and "merlin". (I have not tested these changes yet.)
f4cc9ca fixed bug: gfortran was not correctly detected by "configure"
654570c fix bug 'cut'
aeea044 added to-do list
feb9bc2 updated the to-do list
5dc63c7 added program to convert AVS files containing "quads" to OpenDX format
e9e652f updated the to-do list
f227cd6 added an item about the way we should set the attenuation absorption band
e7edbed added program to shrink elements in an OpenDX file
cd25bc9 updated the todo list
719d45f updated the todo list
f64d397 task reordering
6862ad0 Major Bug fix for Attenuation Code - Now Uses idoubling flag to determine specific region of attenuation - Scaling of Velocities in the Inner Core now uses correct values
8f67e38 modified the todo list
0c1cad3 added handling of First Touch in Linux to todo list
658a9bc added doubling_brick directory to DATA/util
5f08719 updated the todo list
39f035f have changed data reading method. DM.
29ecd9a added scripts used for implementation of the new doubling brick. DM.
abc802d updated the todo list
b96279d moved DATA/util to UTILS
04dd872 moved DATA/oldstuff to UTILS and suppressed DATA/util
3e6016d suppressed a line cut in source code
81e3c46 added scripts to run at IDRIS in Paris (France) using MPI or OpenMP
a34036d updated the todo list
6aa7fa3 added simpler Makefile that does not use "configure", in case "configure" does not give a fully functional Makefile (e.g. at IDRIS in Paris, France)
7040d55 cleaned and improved the LoadLeveler script for MPI runs at IDRIS in Paris (France)
d101c26 added copy of seismograms to LoadLeveler script
49bb737 moved filter_station_list.f90 to UTILS
838cfc1 added data and stack memory limits to the script
55a88ad added automatic handling of current submission path to LoadLeveler script
41d60f3 added serial version of LoadLeveler script
36ffd34 thread-safe modification of the code : all modules have been removed.
ff92e85 improved the LoadLeveler script and added comments
3613f4f fix VTK output bug on the NSOURCES + nrec
b53de39 added some scripts in UTILS
e90e44b changed 3 to NDIM
de317f6 updated BibTeX information
57b5a12 fixed bug in scaling of non-dimensional velocity vector for surface movies
ecaccff updated the todo list
0c5c32c updated options for IBM xlf
5f4bc4c changed options for IBM xlf
468b380 AM_V%Qs is now an integer in place of a double precision
80d9837 MAJ des tâches effectuées par David Michéa au 05/03/2007.
520bfb4 deleted debug lines that i had forgotten. DM.
3412e76 added central_cube directory in UTILS and IC_analyse.pl in central_cube. DM.
db51c0c added two little scripts to visualize inflated central cube. DM.
9017c66 renamed file
479d8ce updated the todo list
4959e8c i have modified define_superbrick.pl in order to add boundaries detection for the superbrick. DM.
2bc1e4b updated the todo list
3a3df35 suppressed an old (and false) version of count_nglob_analytical.pl
74daa84 EXAMPLES/readme
e831524 EXAMPLES/NEX_480/Q
2d8ac45 NEX_480 del
ced9f07 MAJ todo_list
2f63877 deleted the huge EXAMPLES directory (more than 7 gigabytes, including very large binary files) committed by Ying and which made local copies crash
8c81ecf reverted wrong version committed by David following the crash resulting from the huge binary files committed by Ying. The bug was: double precision, dimension(:), pointer :: Qr ! Radius integer, dimension(:), pointer :: Qs ! Steps instead of: double precision, dimension(:), pointer :: Qr, Qs ! Radius and Steps
a35130a changed order of BibTeX information in header of main programs
6a58702 cleaned some Numerical Recipes routines (suppressed some goto's etc)
9037ed3 updated the todo list
622dcff archived scripts and programs used to create the new doubling brick; suppressed old versions
bdbc3c9 again suppressed old versions of scripts and programs used to create the new doubling brick
b9483af added a routine to study radial thickness of the elements in the new mesh; also updated the todo list
1e9e5ef I have changed : double precision :: Qs to integer :: interval_Q in type attenuation_model_variables to avoid confusion with a real attenuation factor.
40544ee updated the todo list
40da6ba updated the todo list
c641b49 committed Jeroen's modifications for gravity in the fluid outer core to fix the small problem found by Tarje Nissen-Meyer
cef0808 added a test to exit if user tries to turn rotation on with gravity off
db2eec0 fixed a bug (undeclared variable) introduced by Jeroen in compute_forces_outer_core.f90 when he fixed the gravity/rotation problem found by Tarje Nissen-Meyer.
9bd0e72 updated comment about displacement potential used in the outer core
fbaaf28 committed new changes made by Jeroen in the fluid outer core with gravity, following a discussion with Tarje about a problem he found.
7ba557e Updated the Pyrization to Pythia v0.8.1.3b2; successful dry run.
92d94d5 release of version 4.0:
f7a6376 Updated the Pyrization for v4.0 and Pythia v0.8.1.3b4; successful run.
a397395 correct values of dT updated for the following cases :
3a96f56 updated DT for 3D case with NEX=320
6aa8f5e added comments to the routine
4cbbca8 changed the first two sections of Brian's 3D attenuation code (out of four sections that need to change)
0d806f3 finished fixing Brian Savage's 3D attenuation code to get rid of the old way of detecting the d80 discontinuity for attenuation
6cda9b3 EXAMPLES directory added
dbc8e96 mv PDF to EXAMPLES
8616a19 mv readme to EXAMPLES
d1ddbf6 mv readme.stations to EXAMPLES
a46ad12 removed unused variables from modified 3D attenuation routines
2ff8bbb suppressed attenuation_lookup_index subroutine by Brian Savage, now unused.
9c98858 updated values of Dt with an additional security margin of 5%. Also updated values of Dt for NEX > 480.
b21af13 Updated for newly-released Pythia v0.8.1.3.
e7914bf fixed bug in specfem3D.f90 : NSOURCES was read by all processes.
cea4641 cleaned the arrays before the gather to fix a problem detected in Vala's NSOURCES_SUB modification
15ee9f8 updated the todo list
716196b updated the todo list
8deace0 added one more comment in the todo list
fdf5d93 changed limit of 100000 sources to 1 million for adjoint simulations
d8e65c4 added files with location of CMB and ICB for gnuplot scripts
bab4e03 added code to display old mesh sampling in v3.6, in order to be able to compare with v4.0
7091f28 values of DT updated.
2b2e882 Removed program 'create_movie_GMT_global' from library.
2f7b40f deleted simpler_Makefile since Leif has now improved the "configure" script
49f7be1 updated the todo list
9521b27 fixed bug in the solver memory size calculation : static_size & dynamic_size are now in double precision.
439e190 fixed bug : declaration of NUMBER_OF_MESH_LAYER in place of NUMBER_OF_MESH_LAYERS
f184f3c added program to test the cache size of a processor and its effect on CPU time
dbbf6ab - suppressed AB, AC and BC chunks in solver - removed old comments left by David - fixed a bug for PREM transverse anisotropy in new d80 mesh layer
c3e0a23 - fixed a small bug (variable used before being assigned a value, but then multiplied by zero therefore no big deal) - suppressed old comments left by David
1a6cf45 fixed a few other small bugs (other variables used before being assigned a value, but then multiplied by zero therefore no big deal)
765d256 removed the redundant calculation of the memory needed for the solver from save_header_file. static_size & dynamic_size are now passed as parameters to save_header_file by create_header_file. Call to save_header_file from meshfem3D use 0.d0,0.d0 as value for static_size,dynamic_size.
934d8c7 added SUPPRESS_CRUSTALMESH flag.
cdd85d3 back to the revision 1345 (r1346 & r1347 overwrite Dimitri's previous modifications)
77f1453 readded SUPPRESS_CRUSTALMESH flag
be81a10 suppressed useless files saved to the disk (and which were not used anywhere anymore)
df9f36e set the flag SUPPRESS_CRUSTAL_MESH to .false. by default.
61064fe added routine memory_eval to compute the amount of memory needed to run the solver. severals dynamic arrays in the solver set to static allocation. modified calculation of NER_80_MOHO in case of SUPPRESS_CRUSTLAMESH = .true. in read_compute_parameters.f90
382d4ad fixed a bug : pasted spaces instead of a tab in a rule
96d4fc5 read_compute_parameters is now called by the master and results are broadcasted to the other MPI processes. removed an unused variable from get_absorb.
d84a84d fixed a minor bug in create_regions_mesh
2c29a9f fixed bug detected at IDRIS in Paris on an IBM / AIX machine: name aliasing such as create_movie_AVS_DX: xcreate_movie_AVS_DX did not work and made "make all" crash, therefore I got rid of the aliases.
9178a44 fixed 2 bugs
a27f453 added parameter ATTENUATION_3D_VAL in values_from_mesher.h, replaced ATTENUATION_3D by ATTENUATION_3D_VAL in the solver.
e8f78b5 renamed parameters from save_header_file : PARAMETER_ become PARAMETER_VAL yet.
f5319c0 renamed model name : 3D_crust_2.0_plus_transversely_isotropic_prem by transversely_isotropic_prem_plus_3D_crust_2.0
01dfe0b replaced variables in if statements by precalculated parameters of values_from_mesher.h for code which belong to the critical part of the code (time loop).
97f6117 fixed bug in specfem3D.
91ec140 Fixed Pyrized build w.r.t. recent Makefile changes.
731a9d1 changed -O2 to -O3 for pgf90
09b58df Removed old html templates.
de39dc4 Bumped version number in configure script.
acbfc81 changed some default options in flags.guess
cdbf816 changed some default options in flags.guess
054ac4d Regenerated 'configure' (it was bogus).
fe4d28a better versions of LoadLeveler submission scripts for MareNostrum (IBM Power in Barcelona, Spain)
54207c2 added a system call to clean the local scratch space on MareNostrum (Barcelona, Spain)
d3f1fc4 modified prompt in bash_profile
1aba85e Turned the 'flags.guess' case statement inside-out so that it switches on compiler, then on OS/platform (instead of vice-versa). This allows e.g., xlf95 to be recognized on both AIX and Linux. (The only ambiguous compiler command name is 'f90', which is assumed to be different compilers depending upon the platform.)
8686aaf removed call to MPI_FINALIZE
fb3894f modified the todo list
dc27efe added -qflttrap=en:ov:zero:inv to the IBM options
b1f8d8f changed -qnostrict to -qstrict for IBM
9006004 added -mcmodel=medium for Intel ifort, Portland pgf90 and gfortran
20e9046 suppressed -mcmodel=medium as a default (because may be slower), but left it as a comment; added -i-dynamic for Intel ifort
e42be79 updated the todo list
7446af0 bug temporary fixed : anisotropic elements are created first again yet, should write something more general later because of the problem of cuthill McKee sorting.
457de7e fixed bug in specfem3D : source line truncated l1643
745ab47 Fixed r1433.
7d45d58 removed useless UNUSED_compute_field_gradient subroutine
fbdef6a added system call to suppress local directory at the end of the simulation on MareNostrum
99c3d9a added output of information if it == NSTEP
aab9a1c - added sorting of ibool to reduce cache misses; performance improvement is very significant (factor between 1.5 on most machines and 3.3 on some) - added cleaning of /scratch disk space on MareNostrum in Barcelona - suppressed MPI_FINALIZE from exit_MPI
af395b9 added -q64 for xlf
c0929b1 modified access to /scratch on MareNostrum in Barcelona
b742fde Addition of radial meshing for single chunk widths less than 90 degrees and for chunks greater than 1248 elements wide.
3a1a1e7 correction of the bugs announced by Qinya, adaptation of the DT for NEX=160 (simulation became unstable with the topography flag on)
f5c69ed updated the todo list
fa4456c updated the todo list
fa1ab0c updated the todo list
095c67c updated the todo list
aa9b5c3 updated the todo list
36f1bf5 fixed compiler warning in Qinya's code to display the mesh topology when NCHUNKS < 6
60f6fbb converted single precision to double precision everywhere, and %g to %lf in scanf statements
3ecd77b converted hprime derivation matrix to its transpose to conform to usual definition as in Deville et al. (2002)
9380011 - added possibility of writing SAC binary format in write_seismograms.f90 this involved adding a new function in write_c_binary.c - added flags OUTPUT_SEISMOS_ASCII_TEXT, OUTPUT_SEISMOS_SAC_ALPHANUM, and OUTPUT_SEISMOS_SAC_BINARY to Par_file, which are used to decide on the output format - added flag ROTATE_SEISMOGRAMS_RT which is used to decide whether the horizontal components of the seismograms are rotated to radial and transverse directions - added file get_backazimuth.f90 which computes the backazimuth from source and receiver coordinates needed to rotate the horizontal components - added get_backazimuth.f90 as object and target in Makefile.in - changed read_compute_parameters.f90 according to the new flags in Par_file - updated all calls to read_compute_parameters.f90 and the respective MPI_BAST call in specfem3D.f90 (not in meshfem3D.f90, the new flags are not used there) - Added information on new output flags to the manual
5513037 Updated Pyrized version w.r.t. bernhard's r1529 changes.
b2f6592 main commit of modifications done during our stay at the BSC to optimize the code.
b3d3208
e6f8a4e Jeroen added a new integer variable THREE_D_MODEL which enables us to distinguish between various 3D model. The new 1D Harvard model "REF" is now available as "1Dref", and the 3D models s362ani, s362wmani, s362ani_prem and s29EA will soon be available.
9c9dd69 I have tested the new 1D model REF, which is working fine. The 3D Harvard models S362ANI, S362WMANI, S362ANI_PREM and S2.9EA have also been incorporated. I will continue to test these, including the topographic variations on the 410 km and 650 km discontinuities. I needed to add the new REF models to the write_AVS_DX_global_chunks_data.f90 routine.
ce66fbd Minor modifications to the Par_file, read_compute_parameters.f90 and add_topography_410_650.f90.
3447737 converted Harvard model s362ani Fortran77 source code to Fortran90
9ca78fc Moved the new f90 source code for the 3D model S362ANI to s362ani.f90 and fixed the Makefile.
cb1e840 Minor modifications to ensure that the define_model_* routines get called only when the appropriate REFERENCE_1D_MODEL is on. Also made the real(kind=4) in s362ani.f90 consistent with the type inthe calls to this routine.
499beac deleted useless examples; kept only one
b4611ee moved some example files
ad95cb7 updated the todo list
9706a0d updated the todo list again
9f13b13 Added a routine moho_stretching.f90 to slightly squish the mesh underneath the oceans and stretch it underneath the continents to get better sampling of the crust. This is accomplished based upon a degree-12 Moho map. Once paraview reads the new file structure we should make some mesh crss sections with the wave speeds superimposed to make sure we want to use this stretching. Tests suggests the stretched mesh gives slightly better dispersion.
9474706 Added routines add_topography_cmb and add_topography_icb that accommodate topography on the CMB and ICB, respectively. The user will have to provide the subroutines subtopo_cmb and subtopo_icb that assign the topography. I tested the routines by putting the smooth Moho variations from read_smooth_moho on both boundaries, and saw sensible results, e.g., changes in the ScS and sScS traveltimes. These Moho variations involve larger topography than is believe to exist on the CMB or ICB, so we should be all set.
0b1b772 renamed IFLAG_IN_CENTRAL_CUBE to IFLAG_MIDDLE_CENTRAL_CUBE everywhere. Also removed useless white spaces at the end of all the lines.
9f87ec8 updated the todo list
8d91def Renamed add_topography_moho.f90 moh-Stretching.f90, since what the routine does is stretch th moho, not honor it.
4402f4c changed the name or spelling of some flags in constants.h, changed some comments as well, and updated the todo list
cae4f37 suppressed 3dmodl.h and mod.h because they contained "common" blocks; now use regular subroutine arguments instead
aecc053 moved program_test_s362ani.f90 to UTILS/s362ani
7a6a96b added MPI_BCAST to Jeroen's routine to read the new Harvard model, and inlined some model definition statements
eb7fab7 moved files to UTILS/s362ani
bcfdc29 renamed file
40a3bee changed program name and path to data file
22505b2 updated the Makefile
1b7b3b9 updated the todo list
4733775 removed unused variables
57927fa used array constructors instead of inlined array indices to define the model in order for the code to compile much faster
97c8537 reverted Jeroen's changes for points at the ICB or the CMB because my original code was correct; added a more precise comment about this
a35cf5a added #include <stdlib.h> to suppress a warning about "exit" being undeclared
a71bd1f merged calls to assemble_MPI_vector for crust_mantle and for the inner core in one call in order to divide the number of MPI messages by 2; updated the todo list
4640db6 changed some variable names that were longer than 31 characters
8a1f17b suppressed unused getcurrmodels subroutine, which produced compilation errors with IBM xlf
a7fe66e also suppressed unused putcurrmodels subroutine, which produced a compilation error with IBM xlf
3f8aa05 fixed bug in second call to assemble_MPI_vector in specfem3D.f90; updated copyright information in write_c_binary.c; added comment about options for the C compiler on IBM in flags.guess
bd08107 made a comment for IBM xlf with xlc or gcc more precise
fc65ff1 Fixed a small bug in the assignment of REF radii near the ICB and CMB in model_ref.f90.
2956d37 Converted read_compute_parameters.f90 to its previous state.
deff806 updated the todo list
9c7a4c9 updated the README file and deleted obsolete text files
1f0e86f added a test in create_regions_mesh.f90 to completely suppress Cuthill-McKee sorting if needed; updated some comments; removed useless white spaces in C codes
1ded4f9 renamed a few variables and updated several comments
ada8a31 fixed a bug introduced by David Michea about dummy variables to call read_arrays_solver that were declared with a real size (i.e. large arrays wasting memory) instead of a small fictitious dummy size.
1c662ac - fixed bug in David Michea's routine to compute static memory size: all the numbers needed to be in double precision to avoid integer overflow - removed dynamic memory size calculation (which was too complex and not crucial) - converted all memory sizes displayed to gigabytes rather than megabytes - updated some comments - updated the todo list
71018f5 cleaner implementation to assign mesh parameters in the case of PREM, PREM_one_crust or a 3D model
f943212 fixed a bug in one of the static memory size estimates
fe93d59 rho_vp_crust_mantle, rho_vs_crust_mantle and vp_outer_core are now allocated only if absorbing conditions are used; rmass_ocean_load and updated_dof_ocean_load are now allocated only if oceans are used
1bff810 reverted attenuation routines back to v3.6 because new routines in v4.0 did not work
6b7dc7a - went back to attenuation v4.0 even if it has a bug, because switching back to attenuation v3.6 did not work either - suppressed dead code and dead routine in Brian Savage's attenuation routines - removed unused variables - added the LSF scripts to the main directory
a41bae9 - fixed incorrect index flag calculation in get_attenuation_index - fixed missing parentheses around IFLAG_220_80, which induced incorrect comparison - suppressed useless external read_crustal_model,read_mantle_model, crustal_model etc. - use f90 syntax everywhere instead of f77 syntax for comparison operators
6b73f96 updated a comment
02cb8de fixed the allocation size of eps_trace_over_3_inner_core and b_eps_trace_over_3_inner_core, which was NSPEC_CRUST_MANTLE instead of NSPEC_INNER_CORE, thus wasting a lot of memory
bc6b903 removed unused commented lines
2c134e2 removed a comment
73c5fa1 updated the todo list
4718b3a updated the todo list
14b594b removed a useless print statement left by David Michea
34efa21 suppressed two problems with variables used before being assigned a value
4e4e51c avoided automatic creation of a dynamic array when calling pspline() with a sub-array
956cd21 removed FINAL argument
6537ac1 fixed a misplaced if statement that created a deadlock when intermediate seismograms were saved to disk; also used a local variable to make RECV statement clearer
85c1896 updated the todo list
48352e7 added output of more detailed information about memory consumption per node, and added output of total number of time steps
bb02aee added comments about the (chi/rho)grad(rho) term added by Jeroen in the fluid when gravity is turned off
163d4f6 added missing spaces for arguments of shmux command
1936bb5 Fixed a mismatch between types in gt3dmodl and rd3dmodl.
a62aafe replaced nspec_reg with NSPEC(iregion_code) for clarity
fa11659 added more options for IBM xlf
9d0a0b3 Fixed a bug assocaited with adjoint simulations. The arrays epsilondev_crust_mantle and epsilondev_crust_inner_core were not properly being allocated. I also set the Max and Min attenuation periods to better values in read_compute_parameters.
0231e1c added STATIONS_SUBSET_35
9f2b789 added more details about optimal memory size at Caltech and BSC
0d2bbda changed /scratch/komatits to /scratch/komatits_new because of write permissions related to new account on MareNostrum in Barcelona
5ea7354 moved these lines:
d7108ab now create the mesh in each region in two passes to save memory and use approximately the same amount of memory in the mesher as in the solver (the mesher used to use a lot more than the solver and be the main limitation in terms of memory on a given machine)
64ca2fc removed storage of the determinant of the Jacobian matrix (array jacobianstore); it is now recomputed every time it is needed in order to save memory
f77f640 added more precise timing information
3c54513 added a few white lines to the output of timing information
a1f4f57 changed "total time" to "total run time" in output
58b1d3b Properly initialized "gamma" in add_topography_410_650.f90. Moved to static memory allocation for the SAVE_STRAIN arrays.
2c685fd replaced Ying Zhou's v3.6 examples with Anne Sieminski's v4.0 examples
bd19dba fixed the bug I introduced in one of Brian Savage's attenuation routines in release 1688: one of his parameters was used as both input and output, not input only
406d962 Changed the adjoint arrays from dynamic to static. Needs to be carefully tested by Qinya & Anne.
153fa8f More modifications associated with the (static) adjoint simulations. Now the code combine_paraview_data.f90 still needs to be fixed to read the new *_solver_data_1.bin files.
00a7d86 Modified combine_paraview_data.f90 to accommodate the new *solver_data_1.bin files written by the mesher and solver. This meant getting access to ibool and xstore, ystore & zstore is a bit more tricky. Qinya should check carefully...
8f92288 - added size of static arrays for adjoint runs to memory size estimate - added more details in timing information - general definition and calculation of ratio to use to divide the central cube, to allow for easier suppression of the crustal mesh - added more debugging options for the Intel ifort compiler - added a full STATIONS file called STATIONS_FULL_758
e4eece4 updated the todo list
87d1693 Changed the convolution for a Gaussian to be consistent with the new convention.
eb2caa8 better explanations about the decay rate when convolving with a Gaussian
5d82c1d made the convolution routine run faster
4890c93 fixed a small bug: N_j was too small in the case of a Gaussian
3bfe83d updated the todo list
db2a59b commented out an option for IBM xlf that slowed down the code
64a047e added a program to generate a list of stations every half degree along a great circle
d8c0c34 updated the todo list
93f5b21 better programs to create a station list for a great circle
a1f94a4 updated the todo list
d4cce27 removed two obsolete lines that were commented out
81bfa37 - fixed truncated station names bug - updated some comments and information messages - better code to generate artificial stations along the equator - better CMT file for Sdiff runs for Lev Vinnik
6910d89 added option to save all the seismograms in one large combined file to avoid overloading of shared non-local file systems such as GPFS
6f21050 updated the todo list
908b8fd added programs to convert from strike dip rake to CMTSOLUTION or Aki/Richards convention for moment tensors
db395f1 fixed a bug in save_arrays_solver.f90 for simulations with NCHUNK=1
f63c26f added control of multilevel cuthill mckee by the flag MULTI_LEVEL_CUTHILL in get_perm_cuthill_mckee.f90
f42607d removed debug lines i had forgotten.
45cd00a todo_list updated
c16b923 added calculation of the average size of a spectral element along the surface
1397b7d changed 1.6 GB to 1.7 GB in memory estimate for MareNostrum
7c87105 added program to extract individual seismograms when they are stored in one large file
2d78230 updated copyright information for future release of v4.0 on October 9, 2007
013603d added cleanup for the nodes before and after the run
94d10c0 added SUPPRESS_CRUSTAL_MESH flag. Supported by models : prem, ref, 1066a, iasp91, ak135
adaae34 put correct copyright information again (it had been erased by David Michéa)
4da938e updated the todo list
a9fba20 fixed some minors problems in read_compute_parameters.f90 : DT & NER_CRUST for 3D models
51d08bf added references in which the formula to scale mu to the right reference frequency can be found
2bd2f53 added references in which the formula to scale mu to the right reference frequency can be found in a second routine
67776b6 Update of the 3D Attenuation routines for Version 4.0
1bff9ce fixed a small mistake in a print statement
9e9c6d4 Updated notes on using the synthetics held in the EXAMPLES directory CMTSOLUTION and half duration Post Processing Long Records and those with Small Epicentral Distances
e8e58c0 fixed a bug i had introduced in read_compute_parameters.f90.
7e2724c fixed a bug for absorbing conditions : file (prname)_stacey.bin read by the solver was not created by the mesher. bug introduced when I have gathered the small *.bin files in larger files in the r1300 (first release of v4.0)
b92d5d6 Update of the Radial Auto Meshing for Version 4.0 This should affect runs with Single Chunks less than 90 degrees and Chunks with NEX > 1248
4049b2b Forgot to remove a write statment, oops
e15e0e1 Min and Max Attenuation periods should now be computed using the routine auto_attenuation_periods() in auto_ner.f90 The value of ATTENUATION_RANGE_PREDEFINED will determine if the older pre calculated period will be used (.true.) as opposed to those computed on the fly (.false.). Values are computed on the fly as the current default.
07b71da Changed model_ref.f90 to replace the ocean layer with upper crust. This solves a problem associated with attenuation.
958973e added a program to plot seismograms as record sections with "Grace" (formerly "xmgr")
d2ad06b changed grace to xmgrace
11257d6 updated the todo list
d0ff126 - updated the manual - added a program to create the table with NEX values in the manual automatically - updated the todo list - added comments to read_compute_parameters.f90 - removed "collect_seismograms" from the LSF scripts
3dd607d added program to draw regular 2D GLL element using Gnuplot for articles or presentations
9e6cd4e fix A bug in ACTUALLY_COUPLE_FLUID_ICB, and change vx,vy,vz to displ_x, displ_y, displ_z.
dde0c13 Added small single chunks for specfem v 4.0 Attenuation periods now are computed automatically for all chunk sizes and chunk numbers
a4b8111 improved the test to check that the block size is a multiple of 32: can now be 16 when one suppresses the crust with flag SUPPRESS_CRUSTAL_MESH
ded627b the seismograms are now appended and not rewritten entirely each time.
d8f66f1 Fix a bug in the process of reading adjoint source by blocks: giving the right dimension to adj_sourcearray in compute_arrays_source.f90 adj_sourcearray(NTSTEP_BETWEEN_READ_ADJSRC,NDIM,NGLLX,NGLLY,NGLLZ) instead of adj_sourcearray(NSTEP,NDIM,NGLLX,NGLLY,NGLLZ) where NSTEP=iadjsrc_len(iadj_block)
be17cfb fixed the bug about rho_vp & rho_vs calculation (both are passed in the argument list of compute_element_properties())
ee1078a Write out DT information after NSTEP to output_mesher.txt.
9ef421f added program to generate the STATIONS file to include stations every 0.1 degree along one-fourth of a great circle
8fa4f21 changed from formatted to unformatted input
cfcd019 cleaner way of storing all the seismograms in one large file, and added option to store them either in binary or in ASCII
3ee09f2 added more details to output of text messages
bbff963 fixed a bug in read_compute_parameters.f90 when auto_ner is called and NER_CRUST<2
3a8f2f0 removed element sorting in get_MPI_1D_buffers.f90 get_MPI_cutplanes_*.f90 get_jacobian_boundaries.f90.
c878b6a updated the todo list
980e48d modifications for cutting the doublingbrick in 4 sub-blocks. Allows NEX_XI to be a multiple of 16*NPROC_XI instead of 32*NPROC_XI. NEX_XI must still be a multiple of 32. user's manual has been modified too.
40ac345 updated the todo list
f38ebff Change the license to GPL v2+
c6bc43a updated the copyright information from October 2007 to November 2007
0cea276 updated the todo list
0e3d022 fixed a bug in the path of the large seismogram file when SAVE_ALL_SEISMOS_IN_ONE_FILE is used
763e64a Adding boundary kernel computations: constants.h.in -- add logical SAVE_BOUNDARY_MESH create_regions_mesh.f90, get_jacobian_discontinuities.f90 -- compute mesh info for Moho, 400, 670 save_header_file.f90 -- save NSPEC2D_Moho/400/670/CMB/ICB to values_from_mesher.h specfem3D.f90, compute_boundary_kernel.f90 -- compute boundary kernels for Moho, 400, 670, CMB, ICB read_compute_parameters.f90, create_movie_AVS_DX/GMT_global.f90, check_buffer_faces/corners_chunks.f90, create_buffer_1D/2D.f90, combine_AVS_DX.f90 create_regions_mesh.f90, create_header_file.f90 -- add HONOR_1D_SPHERICAL_MOHO to read_compute_parameters(), therefore update all the subroutines that call it.
0d8ceb1 updated copyright information
e739c47 Clean up the utilities and organize them into directories, such as Paraview, collect_database, seis_process.
1517dbe - moved variables SAVE_ALL_SEISMOS_IN_ONE_FILE and USE_BINARY_FOR_LARGE_FILE from constants.h to DATA/Par_file - added a new variable called WRITE_SEISMOGRAMS_BY_MASTER to DATA/Par_file. This new variable allows the user to choose between two modes of seismogram output. Either only the master writes all the time series files, or if set to .false. all processes write the time series of their local receivers (like in V3.6). - changed write_seismograms.f90 to incorporate the above option - adding new variables to DATA/Par_file results in changes in all subroutines calling read_compute_parameters.f90
53cb50b broadcast of SAVE_ALL_SEISMOS_IN_ONE_FILE from the master to the nodes : needed if RUN_ON_MARENOSTRUM_BARCELONA = .true.
468edc5 update of todo_list
306f6a6 fixed an error i have introduced in the last commit
4d93995 Set SAVE_BOUNDARY_MESH to be false by default.
ae31eaa added a command to clean the local /scratch space before starting the simulation when one submits a job
02367dd recommit of bug correction previously committed in R1884.
3e6fab8 replaced erf() from Numerical Recipes with erf() from Netlib.
2dd74f3 added CNRS to copyright information
41cee36 added CNRS to copyright information
4c6f063 renamed numerical_recipes.f90 to spline_routines.f90 to avoid copyright problem
3a1408a combined the two shmux commands
b696ced removed an empty line in the script
8824a23 modified the spline routines in order not to use the copyrighted version from "Numerical Recipes"
ef3fc9e added email from Jack Dongarra as a comment at the end of the Netlib routine
19a258f fixed naming of end subroutine call away from pspline and psplint to match the previous fix by dimitri
4888ac8 updated the README file
d533075 created an explicit copy before argument passing to avoid warning in create_header_file.
2645626 - modified clean_sac_headers_after_crash.sh to also run with alphanumeric SAC files.
e220869 fixed bug found by Qinya when NEX_XI/NPROC_XI = n*16 with (n>2 and mod(n,2)!=0). reduced DT for NEX_XI<=480 : 0.12 -> 0.11 because simulation was unstable.
48cf09e Fix the bug at checking the top and bottom of discontinuities: save the radii of the nodes before entering compute_element_properties()
d411700 updated the todo list
c34733c update combine_vol_data.f90 to allow different dirs for topology files and kernel files.
f55a248 todolist updated
da7e957 Delete testing comments in combine_vol_data.f90 and combine_surf_data.f90
b7f6fd9 Added an opendx program with an example file to visualize the valency in term of MPI procs of the central cube points. It was used to debug the assembly of the new shared central cube.
56ee166 merged writing of temporary and final seismograms
27d6669 updated the todo list
3bec317 suppressed the "DONE" section in the todo list
a5652ec fixed bug in syntax of MPI_ABORT. also removed some white spaces in the Makefile.
53196da fixed a small problem for the pgf90 linker
2e99ac8 quick fix by Brian Savage to solve the ScS bug; change committed by Dimitri; Brian will implement a cleaner solution in the next few weeks
ba82953 fixing the bug of calculating the div_s and b_div_s in the outer core.
c38f51d Modified the way MOVIE_VOLUME is done. Added several new parameters to the Par_file to allow saving movie in sub volume. Moved all movie writing to subroutines in write_movie_volume.f90 (similarly to write_seismograms). Added new parameter MOVIE_VOLUME_TYPE to allow for adding new output values/formats without modifying par_file and by simply adding new sub routines.
410438c added a trick to try to avoid slow runs on "pangu" at Caltech
bf81bdf better version of the trick to try to avoid slow runs on "pangu" at Caltech
87651e1 changed "sleep 300" to "sleep 30" because 300 does not prevent slow runs from occurring on "pangu"
ffbb433 added calculation of approximate date and time at which the run will finish
8502545 improved the output format
698e259 put "sleep 200" back
e2f87a4 updated comments and added indentations
c488edc change NSPEC_STRAIN_ATT to NSPEC_STR_OR_ATT
ae55ff9 added date and time estimate of end of run at a remote location, in addition to local time
2982d78 changed 9 to +9 for time difference to illustrate the fact that it can be positive or negative
b914f8c removed empty line
98ce12c suppressed output of date and time estimate of end of run if last time step
eea76f4 added "in remote time" to print statement
94f2dd4 added correct (complete) detection of leap years
3546c0e - Suppressed number of stations at the beginning of the STATIONS file - Suppressed old VINNIK files - Removed useless white spaces at the end of some lines
97c3836 suppressed nrec_dummy
f71a95b In order to reduce the bottleneck it occurs in the MPI communication pattern, the central cube is now split and shared between chunk AB & chunk AB_ANTIPODE.
3639fc1 Modify UTILS/Paraview/global_slices_number[2].pl to reflect the recent changes Dimitri made on the format of the STATIONS file.
582eb80 Fixed a bug with how 1D_ref was being read and added models 1D_ref_iso and s362iso. The elastic constants in the models are the Voight averages of the transversely isotropic models 1D_ref and s362ani respectively.
79c34ce fixed small bug: missing parentheses
958dd93 updated the todo list
735fb88 fixed minor typos and removed old comments, useless white lines and white spaces
86e1917 Add the possibility of computing 21 anisotropic kernels instead of the isotropic (P-wave, S-wave) kernels. This done by setting the new parameter ANISOTROPIC_KL in constants.h to .true.
29767ad Suppressed the fourth doubling layer. Added a flag SUPPRESS_4TH_DOUBLING = .true. by default in constants.h to do so.
dc8b165 updated generate_table_NEX_manual.f90 in order to take into account the 3 or 4 doubling cases.
1f778f6 minor correction
82ba428 replaced SUPPRESS_4TH_DOUBLING by .not. ADD_4TH_DOUBLING
65c7d34 removed an old irrelevant test.
1ce2f8a made several changes in the manual and created a better table with more values
40c4b1e better algorithm to compute the loop on variable "c"
d8e1c3f improved output about positive or negative time difference for estimate of date and time of end of run
06b29f6 fixed a typo in a comment about the definition of the potential in the fluid when the Cowling approximation is used
905152c added test for negative maximum value to detect if the simulation became unstable, because can happen with some compilers
8a121c3 Fixed NSOURCES_SUB "gather failed" bug in Vala's code, detected by Jeannot Trampert. Also removed useless empty white spaces.
9d7510e cut lines that were too long
ae22f0a moved the line that cleans a temporary array to after allocating it (!!)
e0e5657 updated section 9 of the manual (about changing the crustal, isotropic mantle and anisotropic mantle models), and updated some comments in the code accordingly
db0b397 Added the option to output displacement and velocity in volume
c1c606b fixed the last small problems detected by Vala (including an allocate statement that was far too big for the seismograms, and a problem for station H2O in RECORDHEADERS)
197a15c added comments about MPI_BCAST of model structures in the code and in the manual
f903c34 use dynamic allocation for array one_seismogram
75693b1 changing output file name for movie volume, displacement and velocity
4ac67d0 added test about modified NTSTEP_BETWEEN_OUTPUT_SEISMOS
a94f03d updated the todo list for a future version 4.1
b1315d9 added test to allocate statements for seismograms to detect if out of memory
d26c1b7 - committed the new PKP figure - moved the scripts to directory UTILS and suppressed duplicates - updated the manual accordingly
cb8d782 updated some job submission scripts
dd053b8 added INRIA to copyright
c2dced0 updated the copyright header, which was obsolete for this file
9af76da removed a useless comment
512cdce Removed the Pyrized version of the code.
2a2180e Here lies the last vestige of Pyre in the SPECFEM world. This simple Pyre script will build the code, arrange the input files, compute the required number of processors, and then launch the pure Fortran version of the code. (The SPECFEM group has Perl scripts which accomplish similar things... but those scripts aren't flexible enough for the portal.)
d418603 Allow building in a directory distinct from the source directory.
b25a765 Extended script to build the code.
852c138 Generated .h files are in build dir, not src dir.
069c3ac Executable targets now match exectuable names (e.g., specfem3D -> xspecfem3D).
06e1768 ! scale the boundary kernels properly: *scale_kl gives s/km^3 and 1.d3 gives ! the relative boundary kernels (for every 1 km) in s/km^2 moho_kl = moho_kl * scale_kl * 1.d3 d400_kl = d400_kl * scale_kl * 1.d3 d670_kl = d670_kl * scale_kl * 1.d3 cmb_kl = cmb_kl * scale_kl * 1.d3 icb_kl = icb_kl * scale_kl * 1.d3
3e81522 A call to exit_MPI in create_central_cube_buffers.f90 had only the error message argument and no myrank argument.
c9e22d8 Implement the regional models of Japan.
496e38e - added Cuthill-McKee sorting (debugged by David Michea last week) - suppressed saving iMPIcut_xi and iMPIcut_eta to disk (which had become useless) - added directory "estimate_best_values_runs" in UTILS with a program to determine the best sets of parameters to maximize memory use on a given machine - fixed a bug in the Makefile - updated the todo list
9d7f838 changed some comments
d93a0d6 added details about Cuthill-McKee ordering
95cf48f - finished routine to compute best runs in terms of memory use for a given machine - renamed "compteur" to "counter" - renamed "compteur" to "counter" - removed useless white spaces
829f899 added a temporary check to test if NSPEC is negative because at very high resolution there is a bug somewhere (maybe in auto_ner)
bab26a8 fixed a declaration that was double precision instead of integer
b5b97cd swapped two lines
3e73ead added case above 60%
4fa9923 added an option to suppress all cases with a memory occupancy below 50%
6c91f82 - fixed the iglob bug for radius_cr = xstore(iglob) - renamed SAVE_STRAIN to COMPUTE_AND_STORE_STRAIN
f560aca turned Cuthill-McKee permutation off by default, because it can be expensive for large meshes
e61dbd2 added options to suppress trapping
a6b4d9c declared a parameter (which has been moved to DATA/Par_file)
b83ab32 Fixed a problem that caused the mantle not to be filled in properly for model_ref when the crust was thinner than the PREM crust. Also set eta_aniso to always be 1.0 within the crust, and for model s362iso.
d20884c fixed a typo
1532565 changed the syntax of integer, dimension(NX_BATHY,NY_BATHY) :: ibathy_topo
3b46054 Replicate mineos.py output redirection.
6150c9b removed all sorting when Cuthill Mc Kee is not activated, this to avoid memory wasting.
21c2132 fixed a bug I have introduced in my last commit.
6d3239f reduced memory usage for the sorting in get_jacobian_boundaries() if Cuthill Mc Kee is activated.
6f0c778 updated the history of the code in the header of the routine
648cb04 suppressed PERMUTE_INNER_CORE (now set to true all the time and therefore not needed anymore)
a4912f7 fixed a minor bug in get_jacobian_boundaries.f90 removed dead code from sort_array_coordinates.f90
3d0cb56 Modified parameter MOVIE_VOLUME_COARSE to MOVIE_COARSE. This parameter will now work for both MOVIE_VOLUME and MOVIE_SURFACE to output only values at corners of elements. Note that this options so far only works for surface movies with create_movie_GMT_global, not with create_movie_AVS_DX.
4929e94 - suppressed write statements in get_perm_cuthill_mckee - changed .ne. to /= everywhere
82bacc2 Archive output, report progress.
3965a4a Add source frechet derivatives output for adjoint simulations (SIMULATION_TYPE == 2). Output file is in OUTPUT_FILES/directory with names like src_frechet.00001
0097bcc fixed bug when crustal mesh was suppressed
020f02f suppressed file that was not used anymore
479394b fixed a bug : integer limit overflow for huge meshes.
ffb83be added comment about attenuation for P waves (Q_\alpha)
a29884b changed Q_\Kappa to Q_\kappa
47134fa updated test for negative values of NSPEC
9dbbdbd modified the calculation of NGLOB2DMAX_YMIN_YMAX and NGLOB2DMAX_XMIN_XMAX to be closer of the real value and by this way to save memory
6973a3c removed some old comments about the Japanese Earth Simulator
7c8d440 fixed line that was too long
e8cae30 Fixed bug in get_jacobian_boundaries for Cuthill Mc Kee Fixed bug in assemble_MPI_vector.f90 when superbrick is cutted Fixed warning in read_compute_parameters.f90 about divider
f157382 added initialization of dummy variables to avoid a warning
2d9f924 added -C 0 option to suppress creation of core dumped if error
f7cc625 better compiler options for IBM xlf
0b2c9c5 fixed a typo in some comments
53ea947 changed -fpe3 to -fpeO for Intel ifort
6d15c3d added new submission scripts for MareNostrum in Barcelona
112485b added message saying that support for only one slice per chunk has been discontinued; the code now stops in that case
653e6a2 updated the todo list
b62c499 Added mantle heterogeneity model 'heterogen', where the user can specify an arbitrary density anomaly field. See for more information on implementation and use, DATA/heterogen/README_heterogen.
ea0cf3e Some minor fixes: -fixed bug in subroutine write_movie_volume_vector -set eta_ani=1.0 when inside crust 2.0 (in get_model.f90). -added a smaller time step for NEX = 80 (was unstable, at least for s362ani) -added a new stations file. This one includes all stations recording LHZ components, requested from IRIS June 20th, 2008.
839381d This Python script has been moved to cs/portal/trunk/northridge/backend/specfem3D.py.
fd554f1 added checks and stop statements if the routine detects that some elements are disconnected or that the mesh is composed of disconnected subsets
0410944 fixed typo
b8f1adc Added 3D mantle model QRFSI12. This is the new default when choosing ATTENUATION_3D = .true. There are two options in the Par_file to choose this model: s362ani_3DQ and s362iso_3DQ
b564bd2 small modification to files in UTILS/Profiles
667a154 Added a cap in attenuation values for 3D Qmodel QRFSI12 and small modifications to UTILS/Profiles/write_profile.f90
7a2719c updated a comment
496f53a updated a comment
20306d9 converted a parameter to upper case
b7cd98e updated compiler options for GNU gfortran
52a8ae4 changed O2 to O3 in compiler options
0c83259 added -xP option for Intel ifort
a7bdb77 improved code to estimate best parameter values for large runs
e5db3b6 fixed a small bug: potentially an infinite loop on NEX_XI for very large values
fb0b306 improved the output
17f0fd8 slightly changed the output again
6d088dd removed -i-dynamic option for Intel ifort compiler
f29d848 added -ftz option for Intel ifort compiler; it is *critical* for performance
b613790 updated an old comment for NEC
8e71779 added flag for Lahey
04ad9fd changed -no-ftz to -ftz for Intel ifort
bc33174 changed -diag-disable vec back to -vec-report0 for Intel ifort for compatibility with version 9 of the Intel compiler
c61bf10 completely suppressed Cuthill-McKee sorting because of all the stability issues we had in the last few months. Also removed useless white spaces in the process.
6fe8fa3 fixed two warnings from GNU gfortran
6b41405 removed -xP from FLAGS_CHECK for Intel ifort because compiling took too long
93048f8 added action='read' or action='write' to open() statements
a4a94b4 fixed a minor bug in a test to detect unsupported small meshes with only one slice per chunk
9f565bd updated copyright information
816bde5 updated the compiler options
30bc5bb fixed a small issue (incorrect flag) when computing the local height of oceans in the case of a 3D Earth
332c280 updated the compiler flags
c9164ed switched back from -qnostrict to -qstrict for the IBM compiler because -qnostrict can be unsafe
a431c7d updated some compiler options
8c2441f removed dvpstore from a call to write AVS_DX data; added DATA/STATIONS_all_20June2008
17f4e72 removed the duplicates from the STATIONS files
541ac7c deleted "mymachines" file, which had become useless
c540eb7 fixed some incorrect stations in Jeroen's station file
62cb112 removed some old files
fa237b3 added reference to original IASP91 paper; removed comment about model from Sebastien Chevrot
13ae38e added a comment about the density model we use for IASP91 (which does not provide one by default)
c8f2c9c changed a print statement
e72ebb1 renamed lastdoubling_layer to last_doubling_layer
547ad6b changed "ACTUALLY_COUPLE_FLUID_ICB .and. NCHUNKS_VAL == 6" to "ACTUALLY_COUPLE_FLUID_ICB"
1bbdee4 added ASSUME_PERFECT_SPHERE flag to turn off the conversion of geographic to geocentric coordinates for the seismic source and the stations; i.e. assume a perfect sphere, which can be useful for benchmarks of a spherical Earth with fictitious sources and stations.
58ed52c updated the todo list
ff1ffa4 fixed a bug in the size of array bcast_logical
eee6a70 suppressed some old routines that used to be called for Cuthill-McKee sorting in previous versions
de2a974 updated the OpenDX movie network, which was from 2002 and the OpenDX syntax had changed since then
5b2fb96 added display of the mesh to OpenDX movies
d2a49a5 renamed some OpenDX files
0a20ab5 added OpenDX files to create a movie of wave propagation at the surface of a rotating Earth
837fb4b updated the todo list
eac7ce2 fixed model AK135, which was not correctly implemented
be84b03 fixed a syntax problem in the modification I made in model AK135
a5ea311 Use -fast instead of -O3 for the Intel compiler.
1852908 removed an old comment from model ak135
4727ad9 fixed a Makefile problem detected by buildbot
e833ea1 fixed a small detail in the definition of model ak135: one point was missing in the inner core compared to the original article of Kennett et al. (1995)
e8491f4 added doubling brick in CUBIT format, and OpenDX to CUBIT converters written in Fortran
384c0e6 fixed a bug detected by Eh Tan: maxval() returns an undefined value with at least one compiler when the entire array has not been assigned a value. This is not the expected behavior, it is probably a bug in this compiler, but it was easy to fix in the code therefore I did it anyway.
7aab654 updated options for the IBM xlf compiler, in particuliar now force inlining of the Deville et al. (2002) routines.
eae81d1 updated the scripts to generate ParaVer traces on MareNostrum to use the new paths given by David Vicente from BSC
19aecaf added a text file describing how to link with the ParaVer/PAPI tracing system on MareNostrum in Barcelona.
cf96022 added a missing library file to link with ParaVer/PAPI
7733550 updated the ParaVer tracing scripts and related README files for MareNostrum in Barcelona
793da08 new XML configuration for the ParaVer traces on MareNostrum in Barcelona
721ea8d changed default value of LOCAL_PATH_IS_ALSO_GLOBAL from .false. to .true. because these days shared distributed file systems are very common on parallel machines (GPFS, LUSTRE etc.)
be6f89f updated the scripts to generate ParaVer traces on MareNostrum
ba789e5 modified output of create_movie_GMT_global.f90 for 6-chunk simulations and MOVIE_COARSE option to account for cubed sphere mapping
c747a2a Fixed bug when outputting seismograms to set the source location to the CMT location, not the PDE location and, if requested, to rotate to radial and tangential using the CMT location
2386ade Added a final horizontal line to the file to test my ability to commit changes.
5646d03 more precise BibTeX information as comments in the header
f1a295a fixed a line length problem in a comment that generated a compiler warning
253f8c2 added white lines in comments
1ed4e15 Add SeismiQuery directory to update IRIS station file: readme_STATIONS_LH Default STATIONS file now is set to STATIONS_II_IU, because for usual runs, it does not make sense to use 1800+ stations For data query and corresponding synthetics calculation, use STATIONS_LH
2a9e0b1 Brian Savage fixed a bug detected by Rosa Martin from University of Granada, Spain: an array of size 1:9 instead of 1:8 was copied to a subarray of size 1:8
089b523 added Deville et al. (2002) routines for elastic solver in specfem3D.f90, compute_forces_crust_mantle.f90 and compute_forces_inner_core.f90 (only valid for NGLLX=NGLLY=NGLLZ=5)
77ced51 added more flexible file parser for DATA/Par_file (allows for any order of the input parameters, by Dennis McRitchie)
c4365a9 This version changes how attenuation is assigned on to the GLL points and it incorporates a new element stretching to honor surface topography and Moho topography for 3D models.
f849548 files and scripts related to clusters in UTILS/ have been collected in subdirectory SPECFEM3D_GLOBE/UTILS/Cluster:
7dbf1c6 files related to visualizations have been collected in SPECFEM3D_GLOBE/UTILS/Visualization
d92f48c further changes to move towards the new version 'Tiger':
7a08994 adds in directory UTILS/cut_velocity a Makefile to use `make` for compilation rather than the provided script compile_cut
81796d4 these are the main changes for the new "Tiger" version. you will have to re-configure in order to compile executables.
d32afd6 updated the description infile ./INSTALL how to compile the package
5c30b8e added the "INRIA" logo and a few other things to the manual
b164071 sac headers are updated in write_seismograms.f90
caf5b74 updated the manuals
5584e95 updated stretching constant to have rmiddle_crust at 15km; updated auto_ner.f90 to use model specific radii (affects high-resolution and/or 1-chunks smaller than 90 degrees simulations)
8f2a81a updated the todo list
9576dd1 adding feature to use a 1D crust together with 3D mantle models: simply append _1Dcrust to the model name in the DATA/Par_file, e.g. s20rts_1Dcrust, in order to have the 1D crustal model from the corresponding 1D reference model; making the distinction between elements in crust and mantle more explicit such that there is a sharp transition on all GLL points within purely crustal elements to mantle elements when assigning the velocities in case of stretched element layers
b56de36 added Deville et al. (2002) routines for fluid outer core
b48f179 updated the todo list
0ed2dfe commented out the output of function "idaywk", which I found on the Web, because it has a bug. One day we should fix it, or try to find another version on the Web.
bc7cde4 added a comment about the bug in function "idaywk"
f2908b5 updated calendar.f90 to have a correct idaywk() function output, thus the time estimate for the simulation ending should display the correct day name
f9fe74d removed the idaywk() item from the todo list
5378558 updated the version number and header information to 5.0 in the source code files
769ad85 updated copyright and date information in the source file headers
69af53f updated the configure script to create DATA/Par_file, DATA/CMTSOLUTION and DATA/STATIONS
a1353e5 Updated the data processing scripts in UTILS/seis_process
0b77afe updated header information; added new lsf-scripts; changed default CMTSOLUTION to be identical with the GlobalCMT.org solution
b337326 added a readme.txt to UTILS/seis_process and changed process_syn.pl & process_data.pl to call script phtimes.csh which uses a default binary provided by the iaspei-tau package (www.iris.edu) for phase picking
42dfde5 removed useless white spaces at the end of lines (using a script)
003e6f2 added a paragraph about cross compilers to the manual
6ab6a5b improved comments in specfem3D.f90, restored useful flags in flags.guess, and suppressed the "oldstuff" directory in tags/v5.0.0
2c24a78 removed an old Pyre configuration file
46a8b11 added an average-feature to avoid flickering movies in create_movie_GMT_global.f90
c6ac374 removed code that was specific to GPFS for MareNostrum in Barcelona
bc8fe5f cut a line that was too long
22342e0 changed two print statements
d47fdeb small change to avoid filename problems with gfortran in setup_sources_receivers.f90
f1b4b0d enables LOCAL_PATH length up to 150 characters
5bf1305 Sedimentary wavespeeds are superimposed on the mesh if sediment thickness exceeds 2 km; adds regional moho meshing routines for europe and asia; fixes model type declarations
0d6d94a the changes here include: - improved backward/reconstruction of forward wavefield
2305ae0 bug fix for no gravity case and non-Deville routine (compute_forces_outer_core.f90)
0d50714 updating few version number in SAC headers; slightly modified create_movie_GMT_global.f90
575c0f5 converts given geocentric to geographic coordinates in create_movie_GMT_global.f90
7b14676 smoothes sediment basin edges
91dc86f adding noise tomography capability, draft version
973d4de add noise_tomography.f90 and a directory
8ca802b added mantle model s40rts
9a3eaca added EXAMPLES in NOISE_TOMOGRAPHY; code cleaned up
e108be7 minor changes, ready to be merged
922c160 Added sph2utm, written by Qinya Liu. sph2utm converts spherical coordinates to UTM coordinates for source and receivers.
dc9954e minor changes on NOISE_TOMOGRAPHY
b1d8bae missing MPI_BCAST for t_shift in get_event_info.f90 is fixed
ee38841 fixes a mesher problem with IBM compilers found by Rosa Martin in create_regions_mesh.f90; bug fix for GLL models in model_gll.f90
a96b8e6 add seismic processing utility asc2sac, which has been modified to use only in-house subroutines and sac libraries.
1585760 updated the todo list
f9c5848 updated the todo list regarding attenuation and the time scheme
dad5af5 modifications, bug fixes related to the creation of OpenDX and AVS files and fixes of a few wrong references found in the code
4a0fb40 NOISE: merging changes from trunk (rev:16569)
8eb25b4 NOISE: merging changes from trunk (rev:16910,17024,17026)
aed6f08 NOISE: merging changes from trunk (rev:17049,17055,17056,17067) (Done)
265a101 merging NOISE stuff to trunk
9bfa806 adding one more parameter NOISE_TOMOGRAPHY in default DATA/Par_files
5e69100 adds transverse isotropic kernel calculations; minor bug fixes, e.g. in SAC origin time
dbda06f added a check for Deville's products: verify that NGLL == 5
9fdfd52 fixed a typo in two comments
70d3c19 updated the todo list
4aff659 adds 3D mantle model from Masayuki Obayashi; minor modifications in models GLL and PPM; minor code bug fix in model s362ani
3a5627f updated the manual and the BibTeX refences included in the source code
dd05bd1 made the README file shorter (simply refer to the user manual)
c5755bb renamed the README file
1693576 updated the README file
d5fe923 Updates to UTILS/Profiles/ to work with new version of code
5d92636 Updates to UTILS/Profiles/README
12b29a8 new version of param_reader.c written by Laurent Delphin from LDVenture, Dunkerque, France in order to compile on Mac OS X (Snow Leopard).
3da5b3b fixed a typo
65c36f9 committed a new version of param_reader.c by Dennis McRitchie that works on Mac OS. Also updated the citation section of the manual.
d2b19da moved calendar routines from get_cmt.f90 to calendar.f90
23fd6cf updated the manual and AUTHORS file
4cde5cb Add adjoint source derivatives for hdur and tshift.
6c62e74 removed useless white spaces
9742e38 - fixed Makefile.in problem (calendar.f90 missing in get_cmt.f90) - removed unused variables from noise_tomography.f90
a6cfe8a cut a line that was too long (gave a warning with Intel ifort)
34b3071 changed the copyright information and updated the manual to prepare for the future release of version 5.1, which will contain non-blocking MPI.
99c5e15 new Adobe Illustrator cover for the manual; complete list of authors
a4789e1 channel codes of seismograms are changed. Instead of LH(E/N/Z), IRIS convention is used to name band codes (governed by sampling rate of synthetics) and instrument codes (X), i.e., MX(E/N/Z), BX(E/N/Z), etc.
e19bd96 added a dummy file to remind me to remove all the warnings that could be found with IBM xlf before the next official release
ef4c2fe fixed all the warnings detected with IBM xlf
4cd4739 final list of authors; fixed a typo in a comment
4e60c13 better compilation options for IBM xlf
3be9239 fixed the warnings detected by Mila using MPICH and gcc
bea1639 Scale time shift and half duration derivatives
34ce8a4 moved Hom Nath Gharti to the right position in the list of developers (his last name is Gharti). Added checkpointing/restarting to the todo list.
0f90558 moved USER_MANUAL to directory "doc"
cfbde1e new BibTeX file
e2f89f0 added ref to LaWaBe05
f7e5b23 fixed "New York" (removed comma)
cc3e8f1 merged different BibTeX files
65859f6 added DOI to Kom11
5e90182 updates model S40RTS; adds calculations for an approximate hessian for kernel simulations; updates mesh2vtu files
9952f59 fixed a line that was longer than 132 characters; used better flags for gfortran
6aa573c shortest and longest periods at which simulations are accurate are added to SAC headers
fd4e804 Appendix D and Appendix E are updated in the manual
8f916f3 process_data,syn.pl -- interpolate after filter/convolution/deconvolution
7b0285f added a reference to AcKo11
d69446b done modifying the Sesame paper; made similar modifications in the introduction of both manuals
5c7abf6 for SPECFEM3D_GLOBE: adds USER_T0 handling to fix start time; adds force point sources; updates writing out seismograms, adds new files write_output_ASCII.f90 and write_output_SAC.f90
e83ace0 made changes to the introduction of the manuals
1309ad7 added a reference to GeRe09 and also to DaNa98
d095aeb removed Emiljana from the current list of authors, since she will contribute to the GPU version but has not worked on the current CPU version
f03e762 generated new small versions of the covers for the Web site, and removed old comments from the list of authors.
486dae3 updates some comments in the code for mesh generation
eae8f3e updates crust handling for PPM models to take default crust2.0 instead of eucrust; fixes a bug in call to model_eucrust()
c949ee3 Fix bug on rotation matrix for source derivatives
f16b2bf fixes DT bug in compute_add_sources_adjoint(); adds feature to output a source mask, zeroing out values close to sources
992a1c6 updates reading in adjoint source files; updates manual and PPM models
e8b507f added DG references to the manual
95964f5 updated the BibTeX file
7c7c7d7 updated the headers to version 5.1, to prepare for the future release of non blocking MPI + new seismogram naming convention next week
b3e6934 updates source variable names in SPECFEM3D_GLOBE/
4f2ca77 removed Emiljana from a comment
1ec98c8 - added non blocking MPI to the outer core for the forward problem only for now - added scripts to submit jobs to CEA CCRT / GENCI in Paris, France
9e93afe added references to non blocking MPI articles
d1efb03 added non blocking MPI for crust_mantle, inner_core and central_cube for the forward problem; i.e., done adding non blocking MPI for the whole forward problem.
7356769 fixed typos in ref to Daniel Peter's GJI article
7f39db1 fixed unallocated arrays for the central cube in the case of fewer than 6 chunks (these arrays are now used in the call to the compute_forces* subroutines in order to handle non blocking MPI)
1f19ce9 Update processing scripts.
b422436 added non blocking MPI to adjoint calculations
c386770 checked that the new code (version 5.1) compiles without a single warning with IBM xlf
70afea9 moved all the source files to the new "src" directory and all the setup files to the new "setup" directory
1d9a7ef Fix configure and make
92aa8a1 updated the copyright information in the Makefile
58854f7 correct email address for Shawn Smith
58a4cd8 moves source files in SPECFEM3D_GLOBE/ into corresponding subdirectories for mesher and solver; updates Makefiles and configure
7607379 fixes bugs for adjoint simulations in specfem3D.f90; updates example scripts
8eed0c1 bug fix in array initialization; puts compute_adj_source_frechet routine into separate source file src/specfem3D/compute_adj_source_frechet.f90 (compilation of this routine with -O3 may take a while)
87cf84a adds EPcrust from Irene Molinari & Andrea Morelli (2011)
093be7a fixes bug in array allocations; puts GLL model constants to constants.h
d2894f5 updates create_movie_GMT_global routine to avoid flickering
a029261 fix EXAMPLES/noise_tomography to fit updates made in other subroutines. write SURFACE_MOVIE using the same c rountine for absorbing boundaries (presumably much faster and more effective, since there is only one file on each PROC for the whole simulation instead of one file for each time step)
c14bcbe NOISE minor changes
c16b9ab added matlab script to make a movie from a set of jpg files; also has options to add text or other annotations to the jpg files
d98103c change number of files defined in src/shared/write_c_binary.c (from 9 to 10), new index 9 corresponds to SURFACE_MOVIE used in NOISE_TOMOGRAPHY
cdb7ca1 NOISE GLOBAL bugs fixed
51bafb5 another update on NOISE_TOMOGRAPHY in SPECFEM3D_GLOBE
a027c8b added better flags to debug with Intel ifort
dfc6975 updates create_movie_AVS_DX.f90; updates saving source and receiver vtk files; updates allocation checks and fixes compiler warnings; adds movie surface outputs for displacements if MOVIE_VOLUME_TYPE is set to 5
63c595a adds constant USE_FULL_TISO_MANTLE to simulate transverse isotropy in all mantle elements (more expensive, computation takes ~45% longer)
c60bc18 fixes a check for anisotropic elements in meshfem3D.f90
40fcdef added vtk file for global plate boundaries of Peter Bird (GGG 2003); the surface positions of the boundaries are computed assuming some ellipticity, but not necessarily the one used in a particular SPECFEM3D_GLOBE simulation.
b5e961b updates noise routines to take pre-allocated surface movie array
521c751 clean up dummy variables in new version of NOISE_TOMOGRAPHY
477ff2e adds UTILS/adjoint_sources directory with subdirectories to create traveltime and amplitude adjoint sources
2ac24ce updates Makefiles for UTILS/adjoint_sources; adds a check in routine read_parameters_noise()
d26fab5 change order of old & new implementations of NOISE_TOMOGRAPHY
80833a6 adds sac2asc processing utility in UTILS/seis_process/sac2asc/ directory
c955ae9 updates EXAMPLES/ folder and adds new traveltime/amplitude kernel example
fc0afdf added amplitude kernel example with some minor modifications in the manual
f5ec572 manual is updated
1972ebf removes trailing spaces from source code files; adds transverse isotropy for STW 1D reference models to have tiso between moho and 670km depth
22e969f added a note about removing the Unix stack size limit
dc240b5 rephrased and shortened the paragraph about Discontinuous Galerkin
4e5f687 made history of the code more precise
2af76cf extended routines convtime() and invtime() to work beyond the year 2020, up to the year 3000. Also hardwired -O1 to the Makefile for routine compute_arrays_source.f90 otherwise it takes forever to compile with some versions of Intel ifort.
b6bba1e added evolution of the code to the manual
620b51f added references to Patera 1984, Seriani et al., and Faccioli et al.
0eddd1e added a comment about how to generate movies using GMT.
771b54b added a GMT script to create movies for Japan (plot_gmt_movie_file_Japan.pl)
6280047 updates headers and version numbers to version 5.1.1
c561286 combine_vol_data.f90 modified to deal with visualization issues from fictitious elements in central cube of the inner core
4fb7c2e minor change in combine_vol_data.f90, when AVERAGE_GLOBALPOINTS = .true. (not averaging fictitious elements)
31afeef deleted -qxflag=dvz because it requires handler function __xl_dzx and thus linking will fail
e49e7de updated param_reader.c to allow for multi words when reading a string
257b94c Set FTZ if available
c170ab6 updated the comments
33bf135 removed -fno-trapping-math because I now implement by own FTZ forcing routine in C
04c6c83 changed -O3 to -O2 by default for gfortran because of an internal error in gfortran (some versions crash with -O3 because of the number of arguments in some subroutine calls; they should not crash, but they do; reducing the optimization level to -O2 fixes the problem)
84a43cb added a comment to the call to force_ftz()
99d6ad9 added a comment
5e42790 committed Mila's modifications to merge by GPU mesh coloring routines with the official mesher of SPECFEM3D_GLOBE, in order to maintain a single code and prepare for the future official GPU version of the SPECFEM3D_GLOBE solver
25e990d removed -qsigtrap option for IBM xlf compiler
43df87b changed MPI_ISSEND to MPI_ISEND according to a suggestion by Jesus Labarta
ab97476 improved two comments
8bfe6e5 updated PeKoLuMaLeCaLeMaLiBlNiBaTr11
e7758aa added "if(imodulo_NGLOB_CRUST_MANTLE >= 1) then" tests to avoid compilation warnings with IBM xlf
3003c6a added "deallocate(perm)" following the suggestion of Marcin Zielinski from SARA in the Netherlands
b75fa17 added a second comment about fixing the RK4 scheme for attenuation one day.
b4b4844 changed directory 'NOISE_TOMOGRAPHY' to 'OUTPUT_FILES/NOISE_TOMOGRAPHY'
852fb9c added an explanation about the time scheme we use in the code in the elastic case
58e2c09 optimizes some element calculations and adds pre-processing; additional compiler flag -D_HANDOPT would turn on hand-optimized loops; renames specfem3D.f90 to specfem3D.F90, and so with compute_forces_crust_mantle_Dev.F90 and compute_forces_inner_core_Dev.F90; adds new file compute_element.F90; updates movie volume outputs; sets smaller NSOURCES_SUBSET_MAX in constants.h to avoid memory problems with large simulations
49a380c removed an old comment
3b81396 fixed a tiny bug on IBM BlueGene: expressions such as call MPI_IRECV(buffer_all_cube_from_slices(:,:,nb_msgs_theor_in_cube) were not recognized as pointing to a contiguous block of memory, thus a temporary copy was made by the compiler, resulting in incorrect results for non blocking MPI calls. Changed that to call MPI_IRECV(buffer_all_cube_from_slices(1,1,nb_msgs_theor_in_cube) to force the compiler to use a pointer to the beginning of the memory block instead of making a temporary copy.
6ca58ab added a section about how to compile on an IBM BlueGene
9cb6bf5 removed "in part"
02cb80f updates some source code comments and initializes crustal flag
3a91071 created a shell script to automatically create a Gnuplot script to compare two sets of seismograms (for instance to validate a small modification or bug fix in the code by checking that the seismograms computed are unchanged)
250edf8 fixes a problem at poles for horizontal outputs in create_movie_GMT_global.f90
d9f54cb adds choice for horizontal outputs in create_movie_AVS_DX.f90
c05ce0a updated the IBM xlf flags
8ddd09e added Manh Ha's modifications to save header files for the C / StarSs versions of the code. changed max number of colors from 10000 to 1000. removed useless white spaces.
5b126ff fixed a bug in the calculation of NGLOB2DMAX for the header files for the C and StarSs versions of the code
9086bc7 added a comment about -D_HANDOPT for the IBM xlf compiler
bafac31 improved the paragraph about single precision in the manual
448ba8c added doc/LICENSE
e78bff5 added flag DO_BENCHMARK_RUN_ONLY
dabfde4 committed the modifications made by Laurent Leger at IDRIS (France) in order to compile on an IBM BlueGene
5b931c2 added flag SET_INITIAL_FIELD_TO_1_IN_BENCH for scaling benchmarks
78b1465 added unrolling options for Intel ifort
c16fbd5 added a "bash" script to launch the solver using "nohup"
71fb008 improved the Gnuplot script to compare seismograms
86f6892 added EXAMPLES/small_benchmark_run_to_test_a_new_machine
517cba0 added a comment about -D_HANDOPT for the IBM xlf compiler
b584262 added default values for AR and RANLIB
18d6106 added -debug to the list of flags
50cc8f9 added flags -Wl,-relax and -qunroll=yes for IBM xlf
fc23f81 updated SAC headers. KUSER slots are rearranged and KHOLE is added based on IRIS convention.
c55bf7f removed flag -fpe0 for Intel ifort because it can prevent vectorization in some cases
d33c9c4 updates version to 5.1.2
76e77ba suppressed DATA/Par_file.in and DATA/SOURCE.in handled by "configure", since it was both useless and dangerous (modifications made by the user would be erased by another "configure" run in the future)
bba8f13 updated the description of LOCAL_PATH_IS_ALSO_GLOBAL
1a434d7 updated the BibTeX file
f6ba3e0 updated a comment about the 2010 BSSA paper by Brian Savage
d18a31c cleaned the scanned chunk notes
63ebed6 cleaned the scanned chunk notes even more
2cd58c4 removed -qsave from the default flags.guess
042a4b3 updated the manual to describe the options one may need to change in flags.guess
c74a665 added comments / warnings about using _HANDOPT
6aa5886 added more details about _HANDOPT
c2b1453 frees dynamically allocated memory at end of simulation
e822d22 updated the BibTeX file
8aea6c9 minor modifications to the LaTeX manual
c97fe60 adds Makefile to UTILS/lib; updates UTILS/seis_process/ascii2sac.csh script
2979a76 changed the name of "rmass" to make it clear that it is also used as dummy temporary storage later in that routine
3aed5f4 adds modules and cuda routines
5072c9d removes obsolete file assemble_MPI_vector_block.f90
709f91e fixed a problem in the explanation about how time is non-dimensionalized. Also removed "Cray".
e1eee49 updated the todo list
7d53c2a updates stacey and sources routines
dd817b4 updates noise sources; updates reference solutions for regional_Greece_small example; fixes rotation arrays
5921d1f updates time loop and writing seismograms
8dab90c updates kernel routines
3497555 updates Makefile.in
0b86ea0 adds cuda routines for coupling inner_core/outer_core, outer_core/crust_mantle and crust_mantle/ocean
a8378b7 fixes the previous commit in which some git messages remained
555f9ce updates coupling routines
87b7e66 includes updates ensuring forward & adjoint simulations run ok for regional_Greece_small example with all options switched on
3fef583 updates iterate_time() routine
9bcbafa added (partial) OpenMP support from Marcin Zielinski at SARA (The Netherlands)
d2030f2 added more details about "relocation truncated to fit"
7404841 added a picture of the historic cubed sphere from 2000 at Harvard
cb5f194 renamed "UTILS" to "utils" to be consistent with SPECFEM3D
9fc48a6 re-added UTILS, which had been lost when a SVN commit went wrong earlier today
caec056 renamed "UTILS" to "utils" to be consistent with SPECFEM3D
3f80f4e moved "utils/chunk_notes_scanned" to "doc"
2719b3a added a scan to doc/chunk_notes_scanned/all_chunk_notes_scanned.pdf
6c40f88 updates routines for central cube debugging
e66cc87 fixed the write(IMAIN,'(i5)') problem detected by Craig P Steffen in the case of very large runs
9a60992 suppressed the print statements to display the topology of the mesh because the records printed become too long for very large meshes and make the code crash
655cd92 adds another local path parameter to separate mesh files from temporary storage file; updates locate receivers routine; updates read topo file; updates check norm routines on cuda
93b195a $OMP CRITICAL statements by Marcin Zielinski from SARA (Netherlands) for OpenMP
3d01a06 updates model_gapp2.f90 routines; updates small comments about adjoint simulations in the manual
bf79d45 Removed some large arrays that were useless. Adjusted the size of other large arrays when some options are off (MOVIE_VOLUME, TOPOGRAPHY etc). Also fixed the memory size estimate in memory_eval.f90.
d0d66ff fixed the estimate of additional memory size needed when attenuation is turned on.
32e76d8 added a message about how to get a more precise estimate of static memory size used by the solver
4c427b9 fixed a bug I had introduced in attenuation yesterday: I had forgotten to convert some arrays from double to single precision. also put display of exact surface back when chunk has a size of 90 degrees x 90 degrees.
5bfef35 removed a few white spaces in a print statement
595c5d9 put calculation of the exact area back for the top surface
4df86d7 fixed one last bug introduced by my changes yesterday
860c38c updates mpi interfaces; adds support for nproc_xi==1
f5ae1af updated the todo list
4bc24c0 added more details regarding 3D attenuation to the todo list
edf498d fixed the display of the exact volume of the inner core
5be1b80 fixes attenuation bug in cuda code by using 1st order Taylor expansion for epsilondev_loc variable
a84cc0a puts mpi interface detection to mesher; removes unused routines
e820918 updates routines for attenuation debugging; 1st order Taylor expansion used for epsilon_dev_loc variable
4bdbac7 updates memory evaluation and source location routine; updates tabs and whitespaces
70f1f5b updates routines for attenuation handling: adds "ATTENUATION_NEW" flag in Par_file for switching between old/new version
3a93d2e adds synchronizations when reading mesh files
86651fa merged the to-do list into that of SPECFEM3D in order to keep a single to-do list for all the flavors of SPECFEM
70966b3 fixed a few minor details
9ec2160 added a note about the merged to-do list
dc9c340 fixed the USE_COMPONENT bug of create_movie_AVS_DX()
450940c more precise definition of the size of some MPI buffers; also made some of these array sizes static. Also added option -DUSE_SERIAL_CASCADE_FOR_IOs for slow or not very powerful shared file systems.
a89bd7e restored a line that I had commented out temporarily for tests
0ea09e7 better IBM compiler options. also updated the manual
f288a05 more precise sentence about the memory size estimate
dd65de5 more precise OpenMP description
6f46640 fixed a typo
345c207 added new developers
952c42e new developers
5b6abe3 added comments about the non symmetric sigma
b8292d9 updated some references in the manual, and fixed a statement that was not fully correct in the PDF file of the time scheme we use
3d5a416 Add AR, ARFLAGS, and RANLIB checks to configure.ac.
9735ee7 Correct configure.ac version number.
49a90b4 Add support for out-of-source builds.
37a4ae3 Fix output of blank lines.
d3716bd renamed directory to VTK_ParaView
9d4a54b moved some files
e2a9a85 added directory utils/Visualization/VTK_ParaView
842a8f1 switched to -xHost for Intel ifort
dd415fe fixed an undeclared variable in OpenMP
a18b0c3 clarified the way parameters are non-dimensionalized internally in the code
563319c fixes Stacey boundary conditions bug by adding a C*deltat/2 contribution to the mass matrices on Stacey edges; updates routines for attenuation debugging
75cd26a fixes Stacey boundary conditions bug by adding a C*deltat/2 contribution to the mass matrices on Stacey edges
2437aea updates attenuation routines without Deville optimizations for efficiency purposes
6e7d31a added -assume byterecl
32922af fixes test for negative mass matrix
a834c86 updates mesher routines; updates Par_files in examples
47070d6 adds setup_inner_outer.f90 and setup_MPI_interfaces.f90
061a365 added support for the Cray compiler; fixed a warning detected on Cray
c713fa2 change for the Cray compiler
d322bb8 improved a modification I had made to suppress a warning with the Cray compiler
26e1cf9 added some optimization flags for Cray
3a8b08b fixed a bug in auto_ner(), and also a warning on Cray
b71b835 fixed a warning given by the Cray compiler
ee4f284 better compiler flags for Cray
5b315b5 updated the manual
8bb750b updated the header files
2b9c5bc updated some references in the comments
126c2da fixed an undeclared variable
2ba5a9b updated the manual
e0f173f updates meshing routines; updates configuration files; adds EXAMPLE/global_s362ani_shakemovie
b6221ca updates mesher routines and create_regions_mesh() modules; adds initialize_cuda.cu file
e07d81d sets automatically the size of attenuation arrays according to the model defined in Par_file
33b8c24 updates cuda routines to account for 1D and 3D attenuation arrays
e9478bc updates texture usage for cuda routines
6e1fe12 updates mesh ordering for setting up MPI interfaces
b590278 fixed BibTeX errors and updated a few references
8efb5bc cleans up header file and mesher variables; separates OCEANS and TOPOGRAPHY
64caa9c fixed some problems in the BibTeX file
d7b99f0 updates module usage for models; avoids 3D moho stretching when no topography is used
b2e5388 changes meshing order to setup mesh coloring; uses internal arrays for mpi buffers to avoid additional mesh file outputs
65d7e87 adds mesh coloring; uses modules for reference models
d827603 updates module usage for attenuation models; updates coupling terms; adds flush statement for main file output
8a63ba6 updates ocean coupling; updates mesh coloring for attenuation arrays; adds flag in constants.h.in to use 3D attenuation by default
e79154e fixed a pointer arithmetic warning in src/shared/write_c_binary.c
7f6150e fixes inner core coupling; fixes acceleration updates with rotation
00683e8 updates reference solutions for regional_Greece_small/ and global_s362ani_small/ in EXAMPLES/; removes unneeded hprimewgll arrays for CUDA version
a0d7bab improved the sentence about mcmodel=medium
4067a79 added directory "mesh_numbering_convention"
1900571 added the path to xcreate_header_file
6540512 bug fix in get_event_info() routine; updates compute_arrays_source() routine arguments
b78d547 added doc/how_to_properly_detect_all_MPI_or_PML_edges.pdf
77f3f7b bug fix for iboolfaces indexing; fixes broadcast of eventname in get_event_info() routine; adds moho honoring on global simulations only in case topography is set as well
facdc6e updates current version number to 5.1.5
8119c90 merges solver_data_1.bin, solver_data_2.bin and array_dims.txt into solver_data.bin; re-calculates NSTEP and adds t0 initial time to reach full record length; bug fix in rotation for outer core Deville routine
a313098 bug fix in file unit id in combine_vol_data_vtk.f90; bug fix for default attenuation model in get_model_parameters.f90; renames USE_3D_ATTENUATION to USE_3D_ATTENUATION_ARRAYS and only uses ATTENUATION_3D flag to depict 3D attenuation models in mesher setup; adds stations to regional_Greece_small example and changes default model for faster meshing; slight change in duration and attenuation flags in global_s362ani_small example; increase NMAX trace length in UTILS/adjoint_sources/ examples
be5a99d switched to std03 for Intel ifort
395c118 added wrapfig.sty; better cover page
5ee631e added better IBM BlueGene configuration options
6eaaf1f simplified the IBM BlueGene section
23537ac updated the manual
2ed3485 updated the small icon picture
67b9664 Remove an old TODO item that's finished a while ago.
4ec407d Correct calculation of kernels on a regular grid.
11cd68f removed two unused variables
a428497 removed useless semicolons
fe2157e Move SAVE_REGULAR_KL flag into the Par_file.
7b76e2e fixed model ak135f_no_mud, which was not correct
cb2cb1c Remove SAVE_REGULAR_KL from constants.h
52042e9 Remove old Makefile rules.
7a1c954 Fix some Makefile rules.
57d9809 Print ocean percentage only when required.
a02ce8a Use non-recursive make and fix dependencies.
eb281bb better options for Portland pgf90
0dc23c8 removed an extra -ftz
f485b93 updates cuda routines to be compatible with version 5 addressing; updates configuration files to work with --with-cuda=cuda5 option; adds visual_vtk routines; shortens some variable names
3f4e09a added -shared-intel
e46af2a added pgf95
39e0871 Prevent a crash when the Par_file is invalid.
5ff90fe Fix module output directory.
9f669b7 Fix some Makefile bits.
82ca933 Fix up Par_file reading.
9aeb6ca added a comment about the factor of 2 in the scaling used to shift "mu" to the right frequency in the Earth model used
de29f40 Print some information at the end of ./configure.
3f04127 removed FLAGS_NO_CHECK
1803772 Fix missing argument to initialize_simulation.
c16a6d2 Fix time estimates for multi-run simulations.
c720d5c removed three calls to "system" (they do not conform to the Fortran2003 standard)
a7511c7 switched to Fortran2003 in the manual and in flags.guess
69dba96 fixed a typo
850b108 Allow setting an unlimited NUMBER_OF_RUNS.
2e0ed8e changed USE_ATTENUATION_MIMIC to USE_PHYSICAL_DISPERSION_ONLY
daca55a added "_noDev" to name of compute_forces() routines for clarity, and also to be consistent with names in SPECFEM3D_Cartesian
c49f96a bug fix for anisotropic/transverse isotropic kernels in cuda routines; update in combine_vol_data_vtk for file output creation
c4aba82 renamed USE_PHYSICAL_DISPERSION_ONLY to PARTIAL_PHYS_DISPERSION_ONLY to avoid confusion (since there is only average physical dispersion shift, once and for all before the run, but no physical dispersion then during the run; i.e. it is a rough approximation)
c59b907 removed underflow trapping in gfortran debugging option because underflows happen in any normal SPECFEM run
a73e659 made a comment about PARTIAL_PHYS_DISPERSION_ONLY more precise
8a7bfa3 added Elliott Sales de Andrade to the list of authors; also removed the names of the authors from the cover page because it was becoming too long to fit.
1a10840 Fix example Par_files.
f1023ad Force Makefile error when xcreate_header_file fails.
28126f3 Use relative paths in VTK perl files.
02096f1 Fix incorrect parentheses.
302ad80 Fix mesh2vtu Makefile.
efae171 Simplify mesh2vtu code.
60ba473 Add support for byteswapping the input mesh file.
d8dba58 Remove mention of mesh2vtu Perl scripts from docs.
d7a7197 Remove references to AR, ARFLAGS, and RANLIB.
840bb04 Fix float value conversion.
5139c86 Fix some Fortran2003 incompatibilities.
cc4477d better flags.guess
d0ae1c5 Set the list of file to be ignored by git (for users using git-svn). new file: .gitignore
d490db2 Allow the configuration to search for -ladiosf and -lmxml (required by adios) modified: configure.ac
0fe83aa Read and Broadcast the "ADIOS_ENABLED" flags from the Par_file modified: src/shared/broadcast_compute_parameters.f90 modified: src/shared/read_parameter_file.f90
b04f025 Declare and store the 'ADIOS_ENABLED' flags. modified: src/specfem3D/specfem3D_par.F90
379d4f5 Routines to initialize and finalize ADIOS. Initialize_simulation also calls the routines to read and broadcast 'ADIOS_ENABLED' new file: src/specfem3D/adios_manager.F90 modified: src/specfem3D/finalize_simulation.f90 modified: src/specfem3D/initialize_simulation.f90
0552e83 Variables to be written in the header of the ADIOS output file are store in par_header.xml. This xml file is used in write_par_header_ADIOS.F90. new file: setup/adios_xml_files/par_header.xml new file: src/specfem3D/write_par_header_ADIOS.F90
23efd88 Makefile.in for specfem3D take ADIOS-dependant files into account modified: src/specfem3D/Makefile.in
c0e4563 Auto-generate the ADIOS stubs from the configuration file into src/specfem3D Calls ADIOS gpp.py script to do so. modified: Makefile.in
a0a06da update in .gitignore in order to avoid 'git status' to complain modified: .gitignore
974cfae Not sure why the configure file is in the svn repo. modified: configure
9b5b8d6 Change adios_init to call the non xml API. modified: ../../src/specfem3D/adios_manager.F90
2c141c0 Do not use xml file anymore to define variable. Do not use the autogenerated adios code as well. modified: ../../src/specfem3D/write_par_header_ADIOS.F90
49a965c Remove the autogeneration of adios includes from the xml configuration file modified: Makefile.in
f5fd7f5 Helpers for easier definitions of ADIOS variables moved on a separate file. These helpers are called by write_par_header_ADIOS (need to be cleaned). new file: src/specfem3D/adios_helpers.f90 modified: src/specfem3D/Makefile.in modified: src/specfem3D/write_par_header_ADIOS.F90
93c5662 added more details about SIMULATION_TYPE == 2
b6c5c27 updated a C-PML reference
aa579d4 Bug corrected. "\" was missing before adios_helpers.f90 in the file list. modified: src/specfem3D/Makefile.in
3ed3616 Formatting and comments added to write_par_header_ADIOS modified: src/specfem3D/write_par_header_ADIOS.F90
0289a5c added -fpe0 for Intel ifort
0a2dfff added -ffpe-trap=invalid,zero,overflow for gfortran
b23a0be switched from -O2 to -O1 for the C compiler to avoid potential problems with Intel icc (our C routines are so small that we do not need optimization anyway)
070e501 Do not track vi garbagge files modified: .gitignore
ebeb856 Routine that write the adios header renamed and a little bit factorized modified: flags.guess modified: src/specfem3D/Makefile.in modified: src/specfem3D/initialize_simulation.f90 renamed: src/specfem3D/write_par_header_ADIOS.F90 -> src/specfem3D/write_specfem_adios_header.F90
66686eb Subroutine to write ADIOS header splitted and file doxygen-ated modified: src/specfem3D/write_specfem_adios_header.F90
dd9f43e Configuration file for doxygen or doxywizard. Help to produce nice auto-documentation new file: doc/Doxyfile
78fc877 Generate a macro to config.h to store configuration flags. Convert config.h to config.fh to insert defines in F90 files modified: configure.ac
62e9fc1 routine to write the adios header factorized, use config.fh to save configuration files and program name+version modified: configure modified: setup/config.h.in modified: src/specfem3D/write_specfem_adios_header.F90
0090b69 clearer flags.guess
459130c added white lines for clarity; also removed white spaces
3839a9d suppressed the -fpe0 option for Intel ifort; it was making the code unstable on some Intel machines
cab9778 better compiler options for IBM BlueGene
5e44b26 more detailed options for the IBM xlf compiler
5102b6a better flags for IBM xlf
db93c8c acknowledge Yvon Maday in the manual
019de71 better flags for IBM xlf
2085c2f switched to unified syntax for endif, enddo, else if; also removed extra white spaces
2314e07 rename qsort() to bubble_sort()
56a2fde removed useless comments
1cc8235 removed _HANDOPT from the main program; we know how to implement better ways of vectorizing the code, we will do that later, once the code is restructured to undo attenuation
356ddc1 done suppressing _HANDOPT, which was unused; we know how to vectorize the code better, we will do that once it is restructured to undo attenuation
5e4c482 got rid of USE_NONBLOCKING_COMMS, since it has been the default and has been fully validated for several years now, and thus we should always use it
f0ba911 added a second way of outputting PDF files from Gnuplot
74fc4e2 switched back to PostScript by default in the Gnuplot script
a3609fe added undo_att/this_branch_was_created_with_revision_number_21957.txt
ec051d1 improved version of trunk/doc/time_scheme_we_use_in_the_code_from_Hughes_1987.pdf
103869a Forward arrays (intermediate and finals) are written with ADIOS. intermediate forward arrays are read with adios. ADIOS_FOR_FORWARD_ARRAYS flags addded to Par_file. new file: ../../src/specfem3D/read_forward_arrays_adios.F90 new file: ../../src/specfem3D/save_forward_arrays_adios.F90 modified: ../../DATA/Par_file modified: DATA/Par_file modified: ../../src/shared/broadcast_compute_parameters.f90 modified: ../../src/shared/read_parameter_file.f90 modified: ../../src/specfem3D/Makefile.in modified: ../../src/specfem3D/adios_helpers.f90 modified: ../../src/specfem3D/initialize_simulation.f90 modified: ../../src/specfem3D/read_forward_arrays.f90 modified: ../../src/specfem3D/save_forward_arrays.f90 modified: ../../src/specfem3D/specfem3D_par.F90 modified: ../../src/specfem3D/write_specfem_adios_header.F90
3afef0f Read the ADIOS_FOR_FORWARD_ARRAYS flag and branch in forward function to call adios routines.
2bc5c29 Ability to read forward arrays for adjoint simulations. No conditional groupsize definition for saving forward arrays otherwise it messes with the buffersize. modified: ../../src/specfem3D/read_forward_arrays.f90 modified: ../../src/specfem3D/read_forward_arrays_adios.F90 modified: ../../src/specfem3D/save_forward_arrays_adios.F90
dcb526d MPI arrays are written as adios files in the mesher and read as such in the solver. New flags to allow the user to choose wether he wants regular or adios files. Common routines moved in the src/shared/ directory.
2fd27b2 procxxx_regy_solver_data.bin files saved and read with ADIOS as regy_solver_data.bp. New files added to Makefile.in. Modifications for meshfem3d/ and specfem3d/.
243f2f0 added description of utils/strike_dip_rake_to_CMTSOLUTION.c
13db4eb attenuation and boundary databases written with ADIOS.
d92c54d ADIOS_FOR_SOLVER_MESHFILES flag added. solver meshfiles are saved with adios with the prerequisite that arrays solvers are written with ADIOS as well.
907f587 stacey mesh files saved (meshfem3d) and read (specfem3d) with ADIOS modified: ../../src/meshfem3D/Makefile.in modified: ../../src/meshfem3D/create_regions_mesh.f90 modified: ../../src/meshfem3D/get_absorb.f90 new file: ../../src/meshfem3D/get_absorb_adios.f90 modified: ../../src/meshfem3D/save_arrays_solver_adios.F90 modified: ../../src/specfem3D/read_mesh_databases.f90 modified: ../../src/specfem3D/read_mesh_databases_adios.f90 modified: ../../src/specfem3D/specfem3D.F90
501d289 Added some Doxygen comments in the files using ADIOS.
4f3fde1 updated the manual
c49a6c8 split the coupling routines into forward and adjoint, calling them twice with only one set of arrays in each call
b3bd0ce done with part1 (forward problem)
96565b0 fixed a section number
c4a8ae6 added the case SIMULATION_TYPE == 2 to Part 1 (only SIMULATION_TYPE == 3 will be handled by Part 2, since it is the only case in which we need to use the b_* backward arrays)
2578647 done with Part 2 dealing with SIMULATION_TYPE == 3
8c109a6 fixed a bug about how the adjoint source was put back; also started to split the main time loop into two sub-loops; made sure that the code split in two parts now gives the same results as the original adjoint code
7f3b528 introduced logical flag UNDO_ATT, added two versions of the main time loop, and added part1_undo_att.f90 and part2_undo_att.f90
ff0b118 added an if statement on UNDO_ATT to the modification of NSTEP
9294511 ADIOS for points and elements AVS/DX Data.
e77d7c7 Flag added for using ADIOS for AVs/DX data.
7c2e18f ADIOS used to write global faces data for AVS/DX.
7761966 write_AVS_DX_* are outputing a single ADIOS file per region.
16037b7 added original_code_now_unused_is_there_only_as_a_reference.F90 (not compiled, not used, but useful as a reference for the code restructuring that we need to do)
c57b12c rename EXAMPLES/small_benchmark_run_to_test_a_new_machine to EXAMPLES/small_benchmark_run_to_test_more_complex_Earth
cbb2515 - changed src/specfem3D/locate_receivers.f90 to work in the same way as src/specfem3D/locate_sources.f90; that is, stations are now read in in subsets. The size of the subsets is given by NREC_SUBSET_MAX specified in constants.h.in. This should reduce memory requirements when reading in a large number of stations (tested for ~50000 stations).
e64e64a added touch src/specfem3D/specfem3D.F90
199eb87 added empty part3_classical.f90 and made part1_undo_att.f90 and part2_undo_att.f90 empty as well for now
513dd34 added -DUNDO_ATT and converted some file names from *.f90 to *.F90 accordingly
9a906a9 re-added src/specfem3D/check_simulation_stability.f90 (I had erroneously deleted it)
c448e82 change read_compute_parameters.f90 into read_compute_parameters.F90 in src/shared/rules.mk
884906c split the declaration_part_for_backward_wavefield_simulation in specfem3D
6681f95 creat array for storing displ
e151dec split the code for backward wave field
8424bbf take out the part of kernel computation from part2_classical.F90
3656da0 modify the time loop
13e38a3 the first edition of undoatt framework
22bedb9 restructuring the code
7f6626b restructuring the code and copy part1_classical.F90 to part1_undo_att.F90, part2_classical.F90 to part2_undo_att.F90
3545021 the second edition of restructuring.
0052a04 Multiple selection bounding box. One perform read for: modified: ../specfem3D/read_arrays_solver_adios.f90
f8b6ff5 arrays_solver written in a single adios file modified: ../meshfem3D/save_arrays_solver_adios.F90
1985c9e remove write seismogram for backward wave field simulation
54f8e25 set PARTIAL_PHYS_DISPERSION_ONLY = .false. when using UNDO_ATT
162843c fix error of it_begin and it_end
ef1fa00 add one comment "besure that we set PARTIAL_PHYS_DISPERSION_ONLY = .false. in UNDO_ATT"
6c25d2d fix one error in computation of it in compute_add_sources_backward
4454760 fix one error in wrong use of &
dc73b9d fix one error in computation of time
a903242 fix the problem that ifdef can not used in part1_undo_att.F90 and part2_undo_att.F90
0131458 fix one error by moving the save kernel out of time loop
29648a6 fix the error of time value
3e91700 add part2_undo_att_changed_from_forward.F90
91a1f89 try to fix the time error
6d466e4 fix one error of not storing b_accel_outer_core
60a89b9 fix one type error
c42d237 remove one unused variable in /save_forward_arrays.f90
c2bb787 added undo_att/src/compute_optimized_dumping_undo_att
2953fc8 added #ifdef UNDO_ATT
bc1fc18 committed compute_optimized_dumping_undo_att.F90; also renamed NT_500 to NT_DUMP
4b78c27 renamed "_as_bwf" to "_buffer"
5c45044 converted UNDO_ATT back from a compilation flag to a parameter in setup/constants.h
c133b00 removed -DUNDO_ATT
397a9fb cleaned UNDO_ATT flag in the last remaining files
c2e2cb6 remove one comment.
9b86434 put back one useful variable
a7660e3 remove useless subroutine
12513f6 remove useless include program
90f51d1 remove epsilondev_* first in save_forward_arrays.f90 read_forward_arrays.f90
deca6b7 fix an error about including removed program
e094ea5 fix unused variables
1ee9e72 changed 32 GB to 100 GB in a stop statement
7dfe965 remove the epsilonedev and eps_trace
bf6aa83 updated memory_eval() to suppress the epsilon arrays that have been removed by Zhinan to save as much memory as possible to undo attenuation
ce7e9a9 updated compute_optimized_dumping_undo_att() because we do not save the epsilon arrays any more
63da1dc now force SIMULATION_TYPE = 3 in this serial code; also display the output in TB in addition to GB
f38f0af added display of percentage of memory initially used
14a5044 fix one error in computation of strain inside one element
70399e1 remove useless comments
a1d0c91 set PARTIAL_PHYS_DISPERSION_ONLY to be .true. when undo_att is .true., fix the problem in write out seismogram when NT_DUMP is odd integer
fa28f1a fix one error arise in misdelete one temp variable
c429ecc fix the problem concerning with the value of PARTIAL_PHYS_DISPERSION_ONLY
b64303e replaced UNDO_ATT with UNDO_ATTENUATION
1745f15 added checks for UNDO_ATTENUATION and PARTIAL_PHYS_DISPERSION_ONLY
69d3918 added: # to undo attenuation for sensitivity kernel calculations or forward runs with SAVE_FORWARD # use one (and only one) of the two flags below. UNDO_ATTENUATION is much better (it is exact) # but requires a significant amount of disk space for temporary storage. PARTIAL_PHYS_DISPERSION_ONLY = .true. UNDO_ATTENUATION = .false. NT_DUMP = 100 # how often we dump restart files to undo attenuation, only needed when using UNDO_ATTENUATION to the Par_file
0d111e8 added trunk/yyyy_will_need_to_add_UNDO_ATT_NT_DUMP_to_all_EXAMPLE_Par_files
eadaea7 renamed displ_crust_mantle_store_buffer to b_displ_crust_mantle_store_buffer for clarity
2c09113 changed NT_DUMP to NT_DUMP_ATTENUATION
acb420a removed UNDO_ATTENUATION parameters from constants.h.in
b383abe cut three lines that were too long
0d48067 fixed an error in the syntax of one of the commands to visualize sensitivity kernels
67d8ef5 Refactor crust map loading.
4afe1a7 Remove gotos and fix indenting in lecmod function.
d57ca16 Open model files in a consistent manner.
b7163ac Simplify flags.guess logic a little bit.
7e5de4a Use AC_CANONICAL_HOST to test systems.
b78129b Allow setting default CFLAGS in flags.guess.
36ba503 Remove deprecated AC_HELP_STRING call.
4a5acad Remove extra argument in save_header_file call.
a92192e Fix arguments to exit_MPI.
7cfa997 Add missing argument to save_regular_kernels_crust_mantle.
87fbc33 Remove two unused variables.
542db58 Remove mention of SCOTCH from user manual.
91fe190 better print statements for disk space
584795d fix one error by put UNDO_ATTENUATION in broadcast_compute_parameters program
96a293e added a comment about parameters that need to be broadcast with MPI_Bcast()
89d3b91 add another two new parameter in broadcast_compute_parameters
b2c986b added a reference about the dispersion analysis of the SEM
a7112fd added three Discontinuous Galerkin references
e8ab58c new BibTeX file
f9c194f removed white spaces
98fcc45 new undo_att/flags.guess from the trunk
4195742 merged changes from the trunk; no conflict was detected
a34533f removed yyyy_will_need_to_add_UNDO_ATT_NT_DUMP_to_all_EXAMPLE_Par_files
70235d9 done merging undoing of attenuation into the trunk; three new parameters have been added to the Par_file for that
94ed4fd merged the high-order time scheme LDDRK4-6 into the trunk
d5cc6a6 moved the LDDRK4-6 logical flag before the constants
7be16d5 added INCREASE_CFL_FOR_LDDRK and RATIO_BY_WHICH_TO_INCREASE_IT to constants.h
b19c7b8 added branches/SUNFLOWER_ADIOS/this_branch_was_created_from_revision_r21488_of_the_SUNFLOWER_branch.txt
5edb779 merged branches/SUNFLOWER_ADIOS into branches/SPECFEM3D_GLOBE_SUNFLOWER
9c2ff5c removed SPECFEM3D_GLOBE_SUNFLOWER/this_branch_was_created_from_revision_r21488_of_the_SUNFLOWER_branch.txt
62566b0 added SPECFEM3D_GLOBE_SUNFLOWER/this_branch_was_created_from_revision_r19620_of_the_trunk.txt
1f12e36 merged "doc" from the trunk
569adb2 renamed UTILS to utils
258cf74 updated utils/update_headers_change_word_f90.pl
c520fd5 merged "utils" from the trunk
9bbbfe5 updated "EXAMPLES" from the trunk; will need to re-add some GPU parameters to all the Par_files later
4135de7 added trunk/YYYY_will_need_to_put_GPU_parameters_back_into_all_the_Parfiles_including_for_the_EXAMPLES
cc88b93 merged or copied all the files from the root of the trunk
63e25dc started to merge using "meld"
a9193cf started to merge using "meld"
3071c08 more merging
050030a more merging
6fb6d70 svn added some files
acf1e79 svn added some files
465882f done merging new files in "shared"
1a30b14 done merging new files in "shared"
86d703b done merging new files in "create_header_file"
981348a done adding new files in "meshfem3D"
a3fc7e5 done adding new files in "meshfem3D"
30e7b78 done merging new files in "specfem3D"
ef164d6 done merging new files in "specfem3D"
cb90807 started to do a few obvious merges
60923b5 started to do a few obvious merges
6bb1af9 removed white spaces, changed end if to endif, changed end do to enddo, change elseif to else if
c85a8e2 removed white spaces, changed end if to endif, changed end do to enddo, change elseif to else if
2579f4e merged more obvious things
0bfe775 merged more obvious things
7ecf39b more obvious merges
a22df10 more obvious merges
ff29152 more obvious merges
a09f432 more obvious merges
68220d9 more obvious merges
ba237cf more obvious merges
13d7d64 done merging all the easy changes in src/auxiliaries
f962846 done merging all the easy changes in src/auxiliaries
bb243b5 done doing all the easy merges in src/create_header_file
95c8270 done merging all easy modifications in "src/meshfem3D"
258b2c8 done merging all easy modifications in "src/meshfem3D"
ccc32ca done with all the easy merges in "src/shared"
f5f546d done with all the easy merges in "src/shared"
750d59e made a few easy merges in src/specfem3D
b9136be made a few easy merges in src/specfem3D
01833a1 added doc/an_excellent_code_branch_comparison_is_meld.txt
ebc52cb added doc/an_excellent_code_branch_comparison_is_meld.txt
a191aea more easy merges in src/specfem3D
dbc6968 more easy merges in src/specfem3D
a1e8a51 done making all the easy merges in src/specfem3D; also got rid of all the ADIOS files that I had put in the trunk
8c7adb1 done making all the easy merges in src/specfem3D; also got rid of all the ADIOS files that I had put in the trunk
a8cf9a5 removed all new files specific to the branch that I had erroneously added to the trunk (all the files that appeared in green in "meld")
126c880 fixed all compilation warnings that resulted from the changes I made in the last two days (merging some straightforward modifications and small bug fixes from the GPU version back into the trunk)
b7aad31 fixed all compilation warnings that resulted from the changes I made in the last two days (merging some straightforward modifications and small bug fixes from the GPU version back into the trunk)
50a312d re-added branches/SPECFEM3D_GLOBE_SUNFLOWER/this_branch_was_created_from_revision_r19620_of_the_trunk.txt
8be2fe6 fixed a small bug related to the "DEBUG" flag that I had introduced when merging some simple changes from the GPU branch back into the trunk
0bbdfae fixed a small bug related to the "DEBUG" flag that I had introduced when merging some simple changes from the GPU branch back into the trunk
7074b24 removed old 'idoubling' comments, now that idoubling is not used in crust_mantle any more
9fdcbff added typical values of installed memory per core at Princeton and at ORNL
5da809e more precise comment left for Daniel Peter about attenuation arrays
c601240 renamed locate_reg_points() to locate_regular_points()
6e784e8 added amount of memory installed per core on the machine used by Christina Morency
6ca4924 switched from "include mpif.h" to "use mpi"
4b6d7f5 fixed a typo
b7bc593 switched to double precision for model AK135
6472855 fixed a missing white space at the end of the file
f1c0c5f moved all the part*.f90 files to src/specfem3D
6157f63 moved all the part*.f90 files to src/specfem3D
1b173fe renamed src/shared/broadcast_compute_parameters.f90 to src/shared/broadcast_computed_parameters.f90
342ab5e renamed src/shared/broadcast_compute_parameters.f90 to src/shared/broadcast_computed_parameters.f90
6a4017d added all the new and future parameters to the Par_file; modified all the routines that read it accordingly; also renamed broadcast_compute_parameters() to broadcast_computed_parameters()
2b953b2 defined NM_KL_REG_PTS_VAL in order to reduce total memory size when regular grid kernel output is off
7c1cb7e added NSPEC_CRUST_MANTLE_ADJOINT_NOISE to save memory
ae331af added equivalence(R_memory_crust_mantle,ibathy_topo) to save memory
f1dc02a first step of saving memory when 3D attenuation is off
fe1dadd to save a lot of memory, now save the 1D attenuation arrays in (1,1,1) format instead of (NGLLX,NGLLY,NGLLZ) and set ATT1 = ATT2 = ATT3 = 1
b7520a6 suppressed b_div_displ_outer_core to save memory, since it was unused
50ba1bb to save memory, changed vector_accel_outer_core, vector_displ_outer_core and b_vector_displ_outer_core from full arrays to a local version only. Also got rid of deviatoric kernel check option "deviatoric_outercore", which was unused but had a major bug in its implementation (matrix product elements were computed before the previous loop was finished i.e. before the initial array was fully filled. Thus also got rid of array "beta_kl_outer_core".
7828b0e added equivalence(mu_kl_crust_mantle, xix_crust_mantle) equivalence(kappa_kl_crust_mantle, xiy_crust_mantle) equivalence(rhonotprime_kl_crust_mantle, xiz_crust_mantle) to save a significant amount of memory
4f9a28b made hess_kl_crust_mantle() a static array instead of allocatable, to conform with the rest of the code
40518b6 added hess_kl_crust_mantle and Sigma_kl_crust_mantle to memory_eval()
3fe4ff7 removed two old comments
b00adcc added intent(in) and intent(out) for sensitivity kernel rotation routine
d013b10 logos and improved cover for the manual
c9880eb added two more logos
6595c51 improved splcon() in model_s362ani.f90 and added a comment about memory copies that can potentially slow down the code; also removed some constants that were already defined in setup/constants.h
7f167fe split the save_kernel_crust_mantle() routine into three routines for clarity: iso, tiso and aniso
639dba5 renamed SAVE_TRANSVERSE_KL to SAVE_TRANSVERSE_KL_ONLY for clarity
c433d75 switched to static memory allocation with equivalence() statements for final temporary arrays used for isotropic kernel calculations
5e98a39 switched to static memory allocation in save_kernels_crust_mantle_tiso(); equivalence statements not added yet
9576df6 added all the equivalence() statements for large temporary arrays used in save_kernels()
ccf1b9b fixed the size of cijkl_kl_crust_mantle when ANISOTROPIC_KL is .false.; also added cijkl_kl_crust_mantle to the calculation of total memory size in memory_eval()
27613ab added check that we do not have both PARTIAL_PHYS_DISPERSION_ONLY and UNDO_ATTENUATION, since they are mutually exclusive
7f70ef1 deleted figures/Logo_INRIA.jpg
2b4447c deleted figures/Logo_INRIA.jpg
a3f99c3 removed useless white lines
420d39d added two new flags RECOMPUTE_STRAIN_DO_NOT_STORE and EXACT_MASS_MATRIX_FOR_ROTATION to the Par_file
67970d0 programing the code for flag RECOMPUTE_STRAIN_DO_NOT_STORE = .false.
03d95a3 add RECOMPUTE_STRAIN_DO_NOT_STORE,EXACT_MASS_MATRIX_FOR_ROTATION in broadcast_computed_parameters
a196f15 finish changes concerning flag EXACT_MASS_MATRIX_FOR_ROTATION
4f9744d fixed a typo
55fe787 fixed a typo
1ec9787 added -DFORCE_VECTORIZATION to flags.guess
ca436fc renamed compute_kernels.f90 to compute_kernels.F90 in order to be able to use compiler directives
1c46ec6 started to vectorize the code using the FORCE_VECTORIZATION flag; also moved rotate_kernels_dble() from compute_kernels.F90 to save_kernels.f90
9a5f8d7 manually unrolled loops in compute_strain_product() to speed up the code
a55365e switched to a Deville implementation for compute_kernels_outer_core(); also vectorized all its loops
293d3dc better maximum percentages
402b137 moved the "if (ANISOTROPIC_KL) then" test outside of the nested loops to improve performance and be able to vectorize later
5bf157d completely removed flag RECOMPUTE_STRAIN_DO_NOT_STORE, which turned out not to be a good idea
5b8d21e got rid of "if(COMPUTE_AND_STORE_STRAIN) then" test in compute_kernels() because COMPUTE_AND_STORE_STRAIN is always set to true for kernel calculations i.e. when SIMULATION_TYPE /= 1
5259ee8 got rid of expensive temporary array copies
dcfaf51 got rid of b_epsilondev_loc(), and fixed a small bug I had just introduced in a subroutine call
72744c3 inlined the call to compute_strain_product()
c79e8dc converted array prod(21) to 21 variables
3444c21 done vectorizing compute_kernels.F90; added utils/test_shape_for_vectorization_if_not_first_index.f90
1607b4a removed "if(COMPUTE_AND_STORE_STRAIN) then" test from inside the "if(ATTENUATION) then" statements because in the case of attenuation, COMPUTE_AND_STORE_STRAIN is always true. Also removed useless white spaces.
4f2152f more complete version of test_shape_for_vectorization_if_not_first_index.f90
b8f4c97 got rid of compute_element_strain_att_noDev(), which was unused
ecd8ce6 got rid of subroutine compute_element_att_stress() in order to be able to vectorize the code later; also changed "fortran" to "Fortran" in all comments
9bd1db7 renamed COMPUTE_AND_STORE_STRAIN to COMPUTE_AND_STORE_STRAIN_VAL for clarity
febbae7 added new flag PARTIAL_PHYS_DISPERSION_ONLY_VAL to values_from_mesher.h, and switched to it in the solver to improve performance
f336232 got rid of single precision versus double precision calculation for the gravity terms, which was not really useful and made the code more complex (and more difficult to vectorize)
47e2695 vectorized compute_element_strain_att_Dev() and compute_element_strain_undo_att_Dev()
2e6ad6e done vectorizing compute_element.F90
6daec1a remove useless array copies in specfem3D.F90
2033ff8 added a stop statement for a movie_volume option that is not implemented when UNDO_ATTENUATION is on in order to save memory
43e71a9 renamed four files from *.f90 to *.F90 so that they are processed with the preprocessor
ad57cce done vectorizing part 1; I now get some warnings from Intel ifort because include files are apparently not processed by the FPP preprocessor even if they are named *.F90
32189c2 changed NSPEC_CRUST_MANTLE_ATTENUAT to NSPEC_CRUST_MANTLE_ATTENUATION for clarity and to conform with the convention used in the inner core
74dca5d switched back to *.f90 instead of *.F90 for all the part*.F90 include files because they cannot be processed with the Intel ifort preprocessor apparently
c0623cc renamed save_header_file.f90 to save_header_file.F90 to be able to preprocess it
e4fc2b1 found a trick to overcome the problem of some compilers refusing to preprocess include files; I now define a new flag in save_header_file.F90; thus now really done vectorizing the two part1_*.f90 files
4c1f04e done vectorizing part2_*.f90; thus now all included files are fully vectorized
ea3e409 switched to static memory allocation for mask_source() and used an "equivalence" statement in order to reuse memory for the mask; thus now saving the source mask has no additional cost in terms of memory in the case of large runs
1b4dafa switched to static memory allocation for epsilondev() and eps_trace_over_3_crust_mantle(), and uncommented their memory size calculation in memory_eval()
8b8ec3b converted NGLOB_XY_CM,NGLOB_XY_CM_BACKWARD,NGLOB_XY_IC,NGLOB_XY_IC_BACKWARD from variables to constants in values_from_mesher.h
27d7c81 converted the new mass matrices for EXACT_MASS_MATRIX_FOR_ROTATION from dynamic to static memory allocation
68839b6 added the new mass matrices for EXACT_MASS_MATRIX_FOR_ROTATION to memory_eval()
6b6c15e cleaned compute_forces() routines in order to be able to vectorize them later
0c4f7b3 got rid of a useless memory copy for epsilondev()
19358f9 renamed compute_forces_outer_core_Dev.f90 to compute_forces_outer_core_Dev.F90 to be able to preprocess it
b0652fb vectorized many loops in compute_forces(); but not done with all the loops yet
8df59f9 removed statement that forced EXACT_MASS_MATRIX_FOR_ROTATION to be set to .true. when ABSORBING_CONDITIONS is on
e6c22a0 vectorized more loops in compute_forces()
fe973d2 added fac1,fac2,fac3 to compute_forces_outer_core_Dev() to conform to standard used for crust_mantle() and inner_core(); added _CUSTOM_REAL to constants in compute_forces_inner_core_Dev()
a8a7c38 done vectorizing all compute_forces()
a7fb975 added _CUSTOM_REAL to some constants
4545dfd suppressed a useless array memory copy in compute_element_att_memory_ic()
b56780d done vectorizing compute_element_att_memory_ic()
9de8996 done vectorizing compute_element_att_memory_cm()
01d96a4 renamed compute_kernels.f90 to compute_kernels.F90
f7e50f9 added ATTENUATION_1D_WITH_3D_STORAGE_VAL to setup/constants.h.in
fec2345 avoid potential useless array copies in model_s362ani
8e465af added ATTENUATION_1D_WITH_3D_STORAGE to the Par_file
ab9451a removed ATTENUATION_1D_WITH_3D_STORAGE from setup/constants.h.in
4d6760c changed "requires" to "will require" in print statements
1ee3574 renamed some files to merge with "meld"
b8a6a1d performed obvious merges from the trunk to the GPU branch
80aba9e started to merge with "meld"
5a7be0f started to merge with "meld"
8d90cbd done with all the easy merges in src/meshfem3D
66febfe done with all the easy merges in src/meshfem3D
e4e733c done with all the easy merges in src/shared
67841b4 done with all the easy merges in src/shared
078961c done with all the easy merges in src/specfem3D
5f18e06 done with all the easy merges in src/specfem3D
9e328d0 replaced all include 'mpif.h' with use mpi
6ae536e done with easy merges in src/meshfem3D
1c521e5 done with easy merges in src/meshfem3D
437d187 done with easy merges in src/shared
13e8f72 done with easy merges in src/shared
40f4897 done with easy merges in src/specfem3D
01fd519 done with easy merges in src/specfem3D
2c06340 fixed two undeclared variables (following the merge of lines coming from the GPU branch)
5d95bd7 added PGI vectorization directives
ca198d3 added PGI vectorization directives
47d3043 added all the executables to the default Makefile (two were missing)
cf98b11 added all the executables to the default Makefile (two were missing)
478cdb0 better comment about the memory size available on Titan at ORNL
d4fd049 slightly modified a comment
70bb786 made the position of "use mpi" suitable for automatic replacement with "include mpif.h" based on a script
a3e804d made the position of "use mpi" suitable for automatic replacement with "include mpif.h" based on a script
e462c77 added replace_use_mpi_with_include_mpif_dot_h.pl
9680add added replace_use_mpi_with_include_mpif_dot_h.pl
0072994 added a comment for Daniel about the fact that buffers to undo attenuation should remain on the host in the case of GPU_MODE, they would not fit on the device, and the host has more free memory anyway
22ba48e added GPU_MODE support to compute_optimized_dumping_undo_att.f90
af19f2a copied compute_optimized_dumping_undo_att.f90 from trunk to branch
f16c488 fixed the bug detected by Min Chen about NGLOB_XY_CM_BACKWARD and NGLOB_XY_IC_BACKWARD that were not correct for EXACT_MASS_MATRIX_FOR_ROTATION if the code was not recompiled and the mesher rerun when switching from SIMULATION_TYPE == 1 to SIMULATION_TYPE == 3
dfbb99f converted COMPUTE_AND_STORE_STRAIN_VAL (a constant in values_from_mesher.h) back to COMPUTE_AND_STORE_STRAIN (a variable computed in the solver) because it depends on SIMULATION_TYPE and thus cannot safely be stored in values_from_mesher.h if users do not recompile the code when switching from SIMULATION_TYPE == 1 to SIMULATION_TYPE == 3 (most users do not, and they are not supposed to)
d4cdc9c fix one error in computation of b_rmass
b0c111a improved some comments
a6c898c fixed the "negative rmassx matrix term" bug detected by Min Chen, as well as a few similar other issues in create_mass_matrices()
b7aab60 merged changes made in the trunk to fix small bugs mentioned by Min Chen into the branch
36ddf74 merged changes made in the trunk to fix small bugs mentioned by Min Chen into the branch
55d2662 switched from 90% or 92% to 85% because the system apparently needs more than 8% of the memory of a node for itself (at least on some machines)
7e4f69d switched from 90% or 92% to 85% because the system apparently needs more than 8% of the memory of a node for itself (at least on some machines)
ac2788e fixed a small bug (missing subroutine argument) in sac2asc.c, added missing header files and simplified the Makefile
fa9d010 fixed a small bug (missing subroutine argument) in sac2asc.c, added missing header files and simplified the Makefile
ec64edf made "-mcmodel=medium -shared-intel" optional for Intel ifort, since some releases of Intel ifort v12 create codes that are not correct when this option is used (that problem comes from the compiler, not from our code; it has been fixed in Intel ifort v13, and some releases of v12 are also OK)
a6bc990 made "-mcmodel=medium -shared-intel" optional for Intel ifort, since some releases of Intel ifort v12 create codes that are not correct when this option is used (that problem comes from the compiler, not from our code; it has been fixed in Intel ifort v13, and some releases of v12 are also OK)
c5d6114 changed if(ATTENUATION_3D_VAL .or. ATTENUATION_1D_WITH_3D_STORAGE_VAL) then to if(ATTENUATION_VAL .and. (ATTENUATION_3D_VAL .or. ATTENUATION_1D_WITH_3D_STORAGE_VAL)) then to avoid out-of-bound array access when ATTENUATION is off
c92c6eb adds missing flags to all example's Par_file; bug fixes
edb3429 merges commit 22724 to SUNFLOWER branch; updates model_crustmaps routines
b91447a updates Makefiles and rules for proper compilation; adds modules for shared parameters; adds adios stubs
2564224 removes obsolete files in src/specfem3D/ assemble_MPI_central_cube_block.f90,read_arrays_buffers_solver.f90,assemble_MPI_scalar_block.f90,assemble_MPI_central_cube.f90,assemble_MPI_vector_block.f90,fix_non_blocking_flags.f90; adds file parallel.f90 for mpi wrappers; renames acoustic and elastic calling routine files; renames check_simulation_stability to check_stability for compliance with SPECFEM3D_Cartesian; adds mass matrix arrays for rotation correction
8dbcb1c adds new files compute_element_strain.F90 and update_displacement_Newmark.f90; prepares some more undoatt routines; updates restart file handling for restarting/checkpointing capabilities
58e24e1 added a sentence about asking for exclusive use of the compute nodes
55d8bff added a sentence about asking for exclusive use of the compute nodes
c081203 separates forward and backward routines; renames endings to read_mesh_databases.F90 and compute_forces_outer_core_Dev.F90 to use preprocessor directives; adds regular kernel routines
464ea9c updates locate receivers routine; updates model ak135 modifications; removes some unused variables from save_header_file output
bf33248 updates movie surface routines in solver, and auxiliary programs create_movie_AVS_DX and create_movie_GMT_global to read in new moviedata format
c99b713 adds LDDRK time scheme routines; adds files compute_element_att_memory.F90, update_displacement_LDDRK.f90, iterate_time_undoatt.F90 and deletes create_central_cube_buffers.f90 in solver
d1de5ba updates undoatt routines, seismogram outputs and force vectorization routines
00f2ff6 updates cuda routines; renames update_displacement_cuda.cu
9b9df06 updates cuda routines for forward&backward wavefields
3367774 added Ainsworth
fb9abe9 added Ainsworth
9c0f78a removed all trailing white spaces
a74ec48 fixed all Intel ifort warnings (mostly unused variables)
36d1347 new option for Cray compiler to ignore a warning about temporary array copies
854e46e new option for Cray compiler to ignore a warning about temporary array copies
7b0d120 fixed all the warnings on Cray with the "ftn" compiler (except warnings about temporary array copies). In particular, renamed sync_all() to synchronize_all() because sync_all() seems to be a reserved function on Cray.
052247f made a message from the mesher more precise
1935800 added SPECFEM3D_GLOBE_SUNFLOWER/yyyyyyyyy_we_can_get_rid_of_all_the_check_buffers_programs_when_new_version_is_released
643411b added trunk/yyyyyyyyy_we_can_get_rid_of_all_the_check_buffers_programs_when_new_version_is_released
6438b61 switched to "use constants" everywhere instead of include "constants.h"
5bf2c31 fixed a few (but not all) problems in src/meshfem3D/add_topography_410_650.f90, and commented the call to it in src/meshfem3D/compute_element_properties.f90
a39b612 added a stop statement to prevent the add_topography_410_650() routine from being used, since it has a bug that makes the mesher crash
73690a7 updates cuda debugging for backward wavefields
66bc390 renamed branches/SPECFEM3D_GLOBE_SUNFLOWER to trunk, i.e. made the GPU branch become the new trunk
2862c45 suppressed this_branch_was_created_from_revision_r19620_of_the_trunk.txt
ea3cf63 removed all check_buffers_* programs, which are now unused in the new merged GPU and MPI version
10c2384 added comments about -mcmodel=medium -shared-intel
0b31158 added --shell-escape to pdflatex
c8f7791 added more details about the mcmodel=medium problem
7e3e7ef renamed YYYY_will_need_to_put_GPU_parameters_back_into_all_the_Parfiles_including_for_the_EXAMPLES to YYYY_will_need_to_put_new_GPU_and_ADIOS_parameters_into_all_the_Parfiles_including_for_the_EXAMPLES
6f4773a renamed YYYY_will_need_to_put_new_GPU_and_ADIOS_parameters_into_all_the_Parfiles_including_for_the_EXAMPLES to YYYY_will_need_to_put_new_GPU_and_ADIOS_parameters_into_all_the_Parfiles_for_the_EXAMPLES
4459d7c done adding the new GPU and ADIOS parameters into all the Parfiles for the EXAMPLES; also fixed all the STATIONS files that were still using the old format (with the number of stations on the first line of the file)
b7de540 new version of utils/seis_process/asc2sac that does not require any external library
3e7e645 link to the new version of asc2sac.c
7072620 removed _new
94855da removed the new asc2sac because it does not work
42d24e1 switched to a working "asc2sac" that calls SAC to perform the conversion
8343818 new version of utils/seis_process/ascii2sac.csh
3e9cb4e fixed the NEX = 96 instability problem found by Zhinan Xie
1ee6fc1 fix the value of NSTEP when using UNDO_ATTENUATION in setup_sources_receivers.f90
f9827e5 updates compilation rules for running parallel make
2e8473b mentioned the fact that the three benchmarks of Anne Sieminski are fine for mcmodel=medium
d5e4c74 fix the name of output fime tim_stamp* in check_stability.f90
4e4456b updates adios routines and configuration
c999fd4 build: Make flags.guess smarter on Cray systems.
2ff7f4e build: Fixup many minor merge issues.
80fcee0 autoconf: Update Fortran module macro names.
161d621 autoconf: Test for compiler flag to define macros.
fef3f1f build: Update for macro changes in autoconf repo.
beac33e autoconf: Test for MPI module instead of header.
fbb276f autoconf: Change vec from --with to --enable.
a8000e4 build: Fix `make clean' command.
2b3f713 build: Rename clashing variable names in Makefiles.
7a65817 Use constants module instead of header in auxiliaries.
67ab98b build: Update 'constants' dependencies.
adb793e build: Fix CUDA5 rules.
1d2b911 conditional allocation fixed in create_regions_mesh
f342402 remove conditional deallocation for nimin, ... in create_regions_mesh
0991bc5 ipass allocation for global simulation fixed in create_regions_mesh.
795f7e0 new bibliography.bib
8b61719 Remove extra variables from module statement.
669db1a fixed the test about the total number of receivers found: it did not take into account the fact that some stations located outside of the mesh can be excluded when the mesh consists of a single chunk
18a4f16 changed the default to no_vectorization (want_vec=no) to be able to debug with range checking options (and to be able to turn vectorization off)
5874cae bug fix for receivers outside of chunk; add vectorization to element routines; rewrites some routines to compile with ifort and -warn all without warnings/errors
5018b6a updates vectorization routines to reduce code redundancy
2e0eea9 switches indexing of arrays epsilondev_loc, rho_s_H, sum_terms, R_xx,.. and factor_common for better vectorization; removes equivalence statements, uses Deville subroutine calls instead; updates gradient calculations for acoustic kernels in outer core
3506a43 updates eccentricity constant for conversions between geocentric & geographic latitudes; adds TI to mantle elements just below actual moho for s362ani models
8768150 re-adds 410/660 topography routine for s362ani models
e9e70a8 fix hessian dimension when writting it with adios.
4a43aff updates memory copies to avoid compiler warning and cleans up code; adds flag CUDA_FLAGS to configure script for passing specific cuda compilation flags
a1acafa common routines for combine_vol_data w/ and w/o vtk in a new file.
cea8db6 combine_vol_data.f90 merged in combine_vol_data_vtk.f90
285f266 combine_vol_data_* merge tested.
21231f0 combine vol data w/ conditional compilation for vtk and/or adios.
ea853c1 renamed todo_list_please_dont_remove.txt to todo_list_please_dont_remove_but_convert_to_GIT_issues_when_we_switch_to_GIT.txt
8a76d60 added a comment to todo_list_please_dont_remove_but_convert_to_GIT_issues_when_we_switch_to_GIT.txt
89f1797 added all the issues found since this summer (included) to the to-do list
ac7da25 adds asynchronuous memory copies between CPU and GPU; bug fix in adjoint sources on GPU; additional tweaks in getting values for crust2.0 arrays; adds explicit output format for VTK cell types
4c4980e adds textures as an option for GPU simulation; adds kernel launch optimization for kepler architectures
132bd9d adds (preliminary) crustal model CRUST1.0 (see: http://igppweb.ucsd.edu/~gabi/crust1.html) which can be selected in constants.h
a1424b8 uses asynchronuous reading of adjoint sources by use of a separate posix thread, trying to overlap file i/o with kernel computations; adds flag IO_ASYNC_COPY to constants.h.in; adds new files file_io_threads.c and read_adjoint_sources.f90; renames IOVTK to IOUT_VTK
a00a808 moves middle abscissa check to routine zwgljd()
95c5efe uses a limited search range for receiver detection
64b594a Vicious bug fix. Multi GPU per nodes implied multiple synchronizations of the same device when checking error after the call to deviceCount.
4452f01 adds hybrid initialization for GPU mode (turned off by default); re-adds error check on GPU initialization
5dee00e James' ASDF format for seismograms
9b8cd5a Cleaned ASDF routines from James.
0abac09 exit_MPI called on error.
28af897 minor formatting
1598656 bug fix: variable local dimension <= constant offset in adios mpi arrays.
11a43e9 better flags.guess for Intel ifort
2669412 removed two unused variables in src/specfem3D/compute_forces_viscoelastic_calling_routine.F90 and src/specfem3D/compute_forces_acoustic_calling_routine.F90
a5f0ddb temporary fix for an 'optional' argument forgotten by Matthieu; I have contacted him about this
8496c52 changed -gen-interface to -gen-interfaces
4f9e622 local dimensions revert to constant offset. otherwise the solver fails to read the correct values from the adios files.
e693162 Update from James: integer kind change to long for adios.
af7c445 readd cover, manual
e8c4360 m4 submodule
6990887 examples submodule
a4c38e8 data submodule
cbe2ff7 Copy out ignored file rules from SVN.
bdcb9d9 Remove old todo list.
5e338b5 Merge pull request #1 from QuLogic/master
5b53f58 explicit interface in write_seismograms.f90 to be able to use optional arguments as requested by DK
9ad95ce Merge pull request #2 from mpbl/devel
553861c Merge pull request #2 from mpbl/devel
14a5045 Merge pull request #3 from mpbl/devel
8e6616f Updated ASDF headers for scale factor, azimuth, and incident angle
a2ed28e Merge remote-tracking branch 'upstream/devel' into devel
1a6a6eb Merge pull request #4 from jas11/devel
f2c60d7 updated some comments in README and in configure
ae74771 Merge pull request #5 from komatits/devel
9221b3e Fixed call of write_seismograms in iterate_time and an allocate issue with ASDF that was causing the output to hang when more than 6 processors were used to run the solver.
3a17ce8 Fixes bug with ASDF where the sovler hangs when too many processors were used.
32290d1 Removed print statements used for debugging the ASDF output
0b6766c Merge pull request #6 from jas11/devel
486137b reverted changes to flags.guess unintentionally made by James
38bf62f Merge pull request #7 from komatits/devel
81b8b05 switched to the right version of EXAMPLES
59cf4ee Merge pull request #8 from komatits/devel
74fa36e save_arrays_solver_adios bug corrected: local dimensions vary along interfaces.
91ee696 Merge remote-tracking branch 'upstream/devel' into devel
e01ed88 Merge pull request #9 from mpbl/devel
f3ab45b added the logo of the Del Duca Foundation
64c52e0 Merge pull request #10 from komatits/devel
7082777 double precision literals in correct format, use myrank from specfem3d_par module added.
21e5885 Merge pull request #11 from mpbl/devel
a38b4a5 tried to patch
26b149a changed a file
ed8f2b4 Merge pull request #13 from komatits/devel
398ae1b fixed all the warnings given by the Portland and Cray compilers on a Cray at ORNL. In particular, renamed all the variables or functions that had the same name as Fortran intrinsics (although that is correct in Fortran, it can be confusing and it also confuses syntax highlighting, and the Portland compiler prints a warning for each of them)
98bda06 Merge pull request #15 from komatits/devel
f32b71b Update flags.guess
026de77 fixed three warnings / errors from the gfortran compiler about undefined behavior when aggressive loop vectorization is used by the compiler
82f4981 Merge pull request #16 from komatits/devel
3e83caf added a return statement after phtread_exit to please the copmiler.
64e1b38 Merge pull request #17 from mpbl/devel
c8b107d fixes solver_data_mpi.bp read in adios.
099a4d3 Merge pull request #18 from mpbl/devel
More information about the CIG-COMMITS
mailing list