a5b3eb3 | Nick Papior | 10 February 2020, 14:29:34 UTC | sisl release: 0.9.8 Signed-off-by: Nick Papior <nickpapior@gmail.com> | 10 February 2020, 14:29:34 UTC |
13a327b | Nick Papior | 10 February 2020, 14:14:31 UTC | updated changelog Signed-off-by: Nick Papior <nickpapior@gmail.com> | 10 February 2020, 14:14:31 UTC |
a60dfc1 | Nick Papior | 07 February 2020, 09:04:20 UTC | bug: fixed test for the new k-point Also ensured grid are doing tuple indices Signed-off-by: Nick Papior <nickpapior@gmail.com> | 07 February 2020, 09:04:20 UTC |
6c9881e | Nick Papior | 06 February 2020, 20:34:56 UTC | maint: fixed #160 by ensuring a reduced MP-grid Jonas helped with a snippet and I modified the script to MP class to work out the correct indices of the duplicate values. Signed-off-by: Nick Papior <nickpapior@gmail.com> | 06 February 2020, 20:34:56 UTC |
b3ab7c3 | Nick Papior | 05 February 2020, 12:28:12 UTC | doc: changed Sigma to Omega in Berry-flux/conductivity Signed-off-by: Nick Papior <nickpapior@gmail.com> | 05 February 2020, 12:28:12 UTC |
804c118 | Nick Papior | 05 February 2020, 12:17:03 UTC | doc: fixed electron documentation Signed-off-by: Nick Papior <nickpapior@gmail.com> | 05 February 2020, 12:17:03 UTC |
8ba38f4 | Nick Papior | 05 February 2020, 12:13:31 UTC | doc: bumped year to 2020 Signed-off-by: Nick Papior <nickpapior@gmail.com> | 05 February 2020, 12:13:31 UTC |
13b2803 | Nick Papior | 05 February 2020, 09:02:56 UTC | bug: fixed #161 (hopefully) The problem occurred if ncol was different for each orbital and thus resulted in erronerous index pointers. It was not discovered since all tests were based on same ncol or 1 orbital systems. Signed-off-by: Nick Papior <nickpapior@gmail.com> | 05 February 2020, 09:02:56 UTC |
43cd52a | Nick Papior | 04 February 2020, 20:09:03 UTC | maint: removed unnecessary np.where in MP Signed-off-by: Nick Papior <nickpapior@gmail.com> | 04 February 2020, 20:09:03 UTC |
f7d9ed7 | Nick Papior | 04 February 2020, 07:07:46 UTC | bug: fixed test with linspace Signed-off-by: Nick Papior <nickpapior@gmail.com> | 04 February 2020, 07:07:46 UTC |
f1dff69 | Nick Papior | 03 February 2020, 21:29:12 UTC | maint: added more einsum Signed-off-by: Nick Papior <nickpapior@gmail.com> | 03 February 2020, 21:29:12 UTC |
bfce1f9 | Nick Papior | 02 February 2020, 18:30:27 UTC | bug: fixed einsum import Signed-off-by: Nick Papior <nickpapior@gmail.com> | 02 February 2020, 18:30:27 UTC |
8e97d9b | Nick Papior | 02 February 2020, 18:22:03 UTC | maint: pep8 fixes Signed-off-by: Nick Papior <nickpapior@gmail.com> | 02 February 2020, 18:22:03 UTC |
a5ec872 | Nick Papior | 02 February 2020, 18:21:27 UTC | enh: added einsum to phonon.py code Signed-off-by: Nick Papior <nickpapior@gmail.com> | 02 February 2020, 18:21:27 UTC |
f7c009b | Nick Papior | 02 February 2020, 18:18:23 UTC | enh: performance increase for electron.py Replaces many .sum(...) with either a dot or einsum. This should relieve some burden and be preferred in future developments. Also added voigt_matrix to convert to/from Voigt representation. Signed-off-by: Nick Papior <nickpapior@gmail.com> | 02 February 2020, 18:18:23 UTC |
31e0a86 | Nick Papior | 31 January 2020, 18:13:38 UTC | maint: cleaned phase_dtype for clarity Also changed berry_curvature to berry_flux. This makes more sense. Signed-off-by: Nick Papior <nickpapior@gmail.com> | 31 January 2020, 18:13:38 UTC |
011defd | Nick Papior | 31 January 2020, 13:20:28 UTC | bug: fixed gamma-Hk for complex data-types Now the d-types are correct always. Also removed comments in *.c files generated from Cython (to reduce file size). Signed-off-by: Nick Papior <nickpapior@gmail.com> | 31 January 2020, 13:20:28 UTC |
697b12e | Nick Papior | 31 January 2020, 10:41:16 UTC | doc: improved front-page Signed-off-by: Nick Papior <nickpapior@gmail.com> | 31 January 2020, 10:41:16 UTC |
6b31caf | Nick Papior | 30 January 2020, 13:05:56 UTC | enh: added conductivity calculation (still need to check it) Signed-off-by: Nick Papior <nickpapior@gmail.com> | 30 January 2020, 13:05:56 UTC |
f0dc234 | Nick Papior | 30 January 2020, 12:10:54 UTC | doc: added references Signed-off-by: Nick Papior <nickpapior@gmail.com> | 30 January 2020, 12:10:54 UTC |
2232eca | Nick Papior | 26 January 2020, 20:55:25 UTC | bug: fixed berry curvature dtype Also added warning for non-orthogonal Berry curvature Signed-off-by: Nick Papior <nickpapior@gmail.com> | 26 January 2020, 20:55:25 UTC |
b9d848b | Nick Papior | 25 January 2020, 06:55:01 UTC | bug: fixed OVerlap matrix init This fixes #159 by doing a different paradigm. The original Overlap class did not have the extra dimension argument in __init__. This caused problems since many down-stream classes used __class__(...) with the `dim` argument. Now we have re-instantiated the dim argument such that the class is compatible with all other classes. Also fixed some erroneous setter and getters in the classes. Also added a test made by @jonaslb. Signed-off-by: Nick Papior <nickpapior@gmail.com> | 25 January 2020, 06:55:01 UTC |
a41e805 | Nick Papior | 22 January 2020, 12:23:15 UTC | bug: fixed the case where degenerate is None Signed-off-by: Nick Papior <nickpapior@gmail.com> | 22 January 2020, 12:23:15 UTC |
840595a | Nick Papior | 22 January 2020, 12:17:15 UTC | enh: added Berry curvature calculation Added calculation of Berry curvature for states. This may be used to calculate conductivity (to be done). For now it calculates the Berry curvature using: X. Wang, J. R. Yates, I. Souza, D. Vanderbilt, "Ab initio calculation of the anomalous Hall conductivity by Wannier interpolation", PRB, *74*, 195118 (2006) and is thus limited by the number of unoccupied states. The more un-occupied states the better. However, since the Berry curvature is scaled with the eigenvalue difference it truncates relatively fast. Note: this is untested and requires some benchmarks. Signed-off-by: Nick Papior <nickpapior@gmail.com> | 22 January 2020, 12:17:15 UTC |
49b9f5b | Nick Papior | 18 January 2020, 20:18:22 UTC | enh: added SparseGeometry.add, changed `align` to `offset` This enables direct addition of two sparse matrices without having to worry about the couplings. I.e. both sparse matrices will be added as is. Note that if the lattice vectors are not commensurate then supercell couplings may not be transferred as one might expect. Currently there is no warning for this, but possibly we could warn if couplings are being discarded. The `align` keyword argument in SparseGeometry.append/prepend has now changed to `offset` (also for Geometry.append/prepend) This will enable a consistent argument that makes sense for its purpose and how it may be used. For Geometry it may either be a string (none or min) or a vector. The `offset` argument has also been added to Geometry.add method. Added tests for SparseGeometry.add (with and without axis specification). Signed-off-by: Nick Papior <nickpapior@gmail.com> | 18 January 2020, 20:18:22 UTC |
3516469 | Nick Papior | 14 January 2020, 19:06:32 UTC | doc: fixed wrong default and new lines in examples Signed-off-by: Nick Papior <nickpapior@gmail.com> | 14 January 2020, 19:06:32 UTC |
c895a85 | Nick Papior | 13 January 2020, 13:29:09 UTC | doc: added documentation of se2scat Signed-off-by: Nick Papior <nickpapior@gmail.com> | 13 January 2020, 13:29:09 UTC |
b048264 | Nick Papior | 13 January 2020, 13:26:28 UTC | doc: fixed -1 in real-space SE documentation Signed-off-by: Nick Papior <nickpapior@gmail.com> | 13 January 2020, 13:26:28 UTC |
d35fe4e | Nick Papior | 10 January 2020, 18:52:46 UTC | enh: added SparseCSR.diagonal This also highlighted that I needed to fix the __getitem__ method to correctly return scalars for scalar value requests. This is now fixed so that a diagonal retrieval will always return with the correct dimensionality. This fixes #158 Signed-off-by: Nick Papior <nickpapior@gmail.com> | 10 January 2020, 18:52:46 UTC |
2736ce0 | Nick Papior | 10 January 2020, 18:45:47 UTC | bug: fixed typo in 01b8f71ab7 Signed-off-by: Nick Papior <nickpapior@gmail.com> | 10 January 2020, 18:45:47 UTC |
7d359b1 | Nick Papior | 10 January 2020, 17:36:31 UTC | enh: added diagonal method to extract diagonal from sparse matrices Signed-off-by: Nick Papior <nickpapior@gmail.com> | 10 January 2020, 17:36:31 UTC |
03f334b | Nick Papior | 10 January 2020, 17:22:48 UTC | Merge branch 'jb-overlapObj' | 10 January 2020, 17:22:48 UTC |
8208e7f | Nick Papior | 10 January 2020, 17:22:29 UTC | enh: major clean-up in siesta_nc This clean-up was inspired by Jonas Bertelsen (JB). 1) All crt_grp calls in the read_* methods have been changed into explicit group lookups. This ensures that the group exists prior to extracting the group. 2) Added a _write_settings method which writes common (fake) settings to the file. These settings are forcefully read by TBtrans and hence they need to be present, although they are not used. (also fixed a bug such that BZ_displ is actually a float). 3) Offloaded the write_overlap into a separate method. This is necessary since all matrices may have an overlap component. I.e. we re-use the same method to write the overlap for all write_<matrix> methods. 4) Used the new _write_sparsity in all write_<> methods to reduce duplicate code. Now it is much cleaner what happens in each of the write methods. 5) Removed erroneous finalize calls in the write_* methods. We should never force a user to finalize their matrix, perhaps they wish to change it. 6) Updated JB's commit to allow Py2 (untested). The fromsp changed interface to not allow the P argument. This is just to ensure users don't call it erroneously, it also made the overload a bit clearer, I think. Signed-off-by: Nick Papior <nickpapior@gmail.com> | 10 January 2020, 17:22:29 UTC |
01b8f71 | Nick Papior | 08 January 2020, 10:41:06 UTC | maint: minor updates to bloch to reduce calls Signed-off-by: Nick Papior <nickpapior@gmail.com> | 08 January 2020, 10:41:06 UTC |
0c9b065 | Nick Papior | 04 January 2020, 13:44:12 UTC | maint: cleaned test for clarity Signed-off-by: Nick Papior <nickpapior@gmail.com> | 04 January 2020, 13:44:12 UTC |
d3c045a | Jonas Lundholm Bertelsen | 03 January 2020, 09:31:39 UTC | Deterministic ordering of tests for parallel runs When running the tests with pytest-xdist, failure could occur because a set intersection was used to generate them, which is not deterministically ordered, so different processes did not have the same (ordering of) tests. This changes the _my_intersect function to iterate over a list rather than a set. | 03 January 2020, 21:59:49 UTC |
8dfc652 | Jonas Lundholm Bertelsen | 03 January 2020, 10:03:29 UTC | Test writing S object | 03 January 2020, 10:03:29 UTC |
528f602 | Jonas Lundholm Bertelsen | 03 January 2020, 08:45:47 UTC | Write overlap functionality | 03 January 2020, 09:24:04 UTC |
f1df05a | Jonas Lundholm Bertelsen | 20 December 2019, 16:52:44 UTC | read_overlap functions use overlap object | 20 December 2019, 16:52:44 UTC |
f6bf648 | Jonas Lundholm Bertelsen | 20 December 2019, 16:45:11 UTC | Added overlap object | 20 December 2019, 16:45:11 UTC |
2fb42cb | Nick Papior | 20 December 2019, 09:50:47 UTC | bug: reverted fe7295c62 for set_nsc Signed-off-by: Nick Papior <nickpapior@gmail.com> | 20 December 2019, 09:50:47 UTC |
25c0541 | Nick Papior | 20 December 2019, 09:31:46 UTC | Merge branch 'jonas/nscelements' | 20 December 2019, 09:31:46 UTC |
94535ba | Nick Papior | 20 December 2019, 09:31:18 UTC | maint: re-added check (in correct place) for n_s checks Minor adaptation of the correct fix provided by Jonas. Now it will also check that we have all super-cell indices in the translation table. Signed-off-by: Nick Papior <nickpapior@gmail.com> | 20 December 2019, 09:31:35 UTC |
ebd1d11 | Nick Papior | 20 December 2019, 09:22:04 UTC | doc: maintained some documentation for return values Signed-off-by: Nick Papior <nickpapior@gmail.com> | 20 December 2019, 09:22:04 UTC |
fe7295c | Nick Papior | 20 December 2019, 09:20:20 UTC | bug: fixed returning nsc in Geometry Signed-off-by: Nick Papior <nickpapior@gmail.com> | 20 December 2019, 09:20:20 UTC |
694219b | Nick Papior | 20 December 2019, 09:18:30 UTC | bug: fixes #153 When passing geometry=None, the geometry_align function did not work. This is now fixed. Signed-off-by: Nick Papior <nickpapior@gmail.com> | 20 December 2019, 09:18:30 UTC |
8ee8b77 | Jonas Lundholm Bertelsen | 17 December 2019, 17:14:40 UTC | Added tests for nsc changes | 17 December 2019, 17:14:40 UTC |
028580e | Jonas Lundholm Bertelsen | 17 December 2019, 16:52:01 UTC | Remove erroneous check for number of supercells | 17 December 2019, 16:52:01 UTC |
d95ba77 | Jonas Lundholm Bertelsen | 17 December 2019, 16:41:13 UTC | Fixing spgeom.set_nsc when new cell is larger | 17 December 2019, 16:41:13 UTC |
5108ac4 | Nick Papior | 11 December 2019, 09:48:57 UTC | enh: added custom scale to sgrid Signed-off-by: Nick Papior <nickpapior@gmail.com> | 11 December 2019, 09:48:57 UTC |
d430477 | Nick Papior | 11 December 2019, 09:22:08 UTC | bug: fixed interpolation of poisson, and skipping lines with comments Signed-off-by: Nick Papior <nickpapior@gmail.com> | 11 December 2019, 09:22:08 UTC |
5ddaf1c | Nick Papior | 10 December 2019, 14:28:59 UTC | maint: added device values and cleaned code Signed-off-by: Nick Papior <nickpapior@gmail.com> | 10 December 2019, 14:28:59 UTC |
0b2c936 | Nick Papior | 09 December 2019, 09:56:10 UTC | bug: fixed pickling a sparse matrix, fixes #150 Fixed #150 by correctly updating object members from the unpickling procedure. Likely there are many other objects having the same problem. Particularly the MonkhorstPack object. Signed-off-by: Nick Papior <nickpapior@gmail.com> | 09 December 2019, 09:56:10 UTC |
7d25354 | Nick Papior | 09 December 2019, 09:37:53 UTC | enh: added len(.) to SelfEnergy objects Signed-off-by: Nick Papior <nickpapior@gmail.com> | 09 December 2019, 09:37:53 UTC |
9613475 | Nick Papior | 09 December 2019, 09:36:52 UTC | enh: enhanced the Poisson solution to take into account the FFT boundary Now the Poisson solver first solves the electrode box problem. Then fixes the boundaries to the solution values, and then solves using fixed boundaries all over. This *should* be the FFT approximate solution but may in some cases be sub-optimal. Signed-off-by: Nick Papior <nickpapior@gmail.com> | 09 December 2019, 09:36:52 UTC |
1a285eb | Nick Papior | 28 November 2019, 08:31:32 UTC | enh: allowed other file prefixes for Siesta_grid Signed-off-by: Nick Papior <nickpapior@gmail.com> | 28 November 2019, 08:31:32 UTC |
e26de52 | Nick Papior | 28 November 2019, 08:23:11 UTC | bug: fixed TSV.nc write-out and docs in poisson_explicit Signed-off-by: Nick Papior <nickpapior@gmail.com> | 28 November 2019, 08:23:11 UTC |
87c4200 | Nick Papior | 27 November 2019, 20:32:18 UTC | doc: fixed Gamma in documentation Signed-off-by: Nick Papior <nickpapior@gmail.com> | 27 November 2019, 20:32:18 UTC |
80f9c57 | Nick Papior | 16 November 2019, 08:55:23 UTC | bug: fixed corner cases of finding fermi-level Signed-off-by: Nick Papior <nickpapior@gmail.com> | 16 November 2019, 08:55:23 UTC |
f4a6bda | Nick R. Papior | 11 November 2019, 20:27:17 UTC | Merge pull request #149 from sofiasanz/master Take into account double occupancy in fermi_level | 11 November 2019, 20:27:17 UTC |
f62ae4e | Sofia Sanz | 11 November 2019, 16:09:50 UTC | fix: take into account double occupancy when the Hamiltonian is non-spin-polarized | 11 November 2019, 16:09:50 UTC |
df1f478 | Nick Papior | 28 October 2019, 19:42:29 UTC | bug: fixed fermi_level calculation The bug happens in some cases where the cubic-spline is not precise enough. In that case the precision for the spline will return the same result time and time again and the method will fail. This is now fixed. Signed-off-by: Nick Papior <nickpapior@gmail.com> | 28 October 2019, 19:42:29 UTC |
c5c8cb8 | Nick Papior | 06 October 2019, 20:15:33 UTC | test: added axyz tests Signed-off-by: Nick Papior <nickpapior@gmail.com> | 06 October 2019, 20:15:33 UTC |
b7e0683 | Nick Papior | 06 October 2019, 19:55:41 UTC | test: added more tests for Geometry Signed-off-by: Nick Papior <nickpapior@gmail.com> | 06 October 2019, 19:55:41 UTC |
b9e8cf2 | Nick Papior | 06 October 2019, 18:49:30 UTC | maint: fixed MANIFEST.in, now it should be a cleaner distro Signed-off-by: Nick Papior <nickpapior@gmail.com> | 06 October 2019, 18:49:30 UTC |
8f8c0be | Nick Papior | 01 October 2019, 19:57:00 UTC | doc: added publication Signed-off-by: Nick Papior <nickpapior@gmail.com> | 01 October 2019, 19:57:00 UTC |
2b47677 | Nick Papior | 01 October 2019, 19:49:32 UTC | maint: increased performance for read_dynamical_matrix I have removed some branch-points and a for-loop Signed-off-by: Nick Papior <nickpapior@gmail.com> | 01 October 2019, 19:49:32 UTC |
4b6c8c0 | Nick R. Papior | 01 October 2019, 19:30:14 UTC | Merge pull request #145 from jonaslb/gout_dynmat bug: dynamical matrix from gulp works for norbs%12!=0 | 01 October 2019, 19:30:14 UTC |
bef5b68 | Jonas Lundholm Bertelsen | 27 September 2019, 13:03:33 UTC | Dynamical matrix from gulp works for norbs%12!=0 | 27 September 2019, 13:03:33 UTC |
d759ae1 | Nick Papior | 26 September 2019, 13:32:05 UTC | enh: speeded up reading transiesta-grid Signed-off-by: Nick Papior <nickpapior@gmail.com> | 26 September 2019, 13:32:05 UTC |
a0847e0 | Nick Papior | 26 September 2019, 11:22:54 UTC | travis: removed bdist_wheel Currently bdist_wheel still creates linux_x86_64 but it is not supported. We need to create multilinux stuff. Signed-off-by: Nick Papior <nickpapior@gmail.com> | 26 September 2019, 11:22:54 UTC |
0b6f6b7 | Nick Papior | 26 September 2019, 11:09:02 UTC | travis: removed coveralls Signed-off-by: Nick Papior <nickpapior@gmail.com> | 26 September 2019, 11:09:02 UTC |
dfcd5ac | Nick Papior | 26 September 2019, 10:59:58 UTC | maint: reverted tag release Signed-off-by: Nick Papior <nickpapior@gmail.com> | 26 September 2019, 10:59:58 UTC |
8c4d2c6 | Nick Papior | 26 September 2019, 10:55:28 UTC | sisl release: 0.9.7 Signed-off-by: Nick Papior <nickpapior@gmail.com> | 26 September 2019, 10:55:28 UTC |
01a5bf1 | Nick Papior | 26 September 2019, 10:22:35 UTC | doc: cleaned documentation for the eigenvalues Signed-off-by: Nick Papior <nickpapior@gmail.com> | 26 September 2019, 10:22:35 UTC |
4e501b0 | Nick Papior | 25 September 2019, 10:49:11 UTC | doc: changed python/pip -> python3/pip3 Signed-off-by: Nick Papior <nickpapior@gmail.com> | 25 September 2019, 10:49:11 UTC |
e985d9d | Nick Papior | 24 September 2019, 08:44:54 UTC | doc: updated CHANGELOG Signed-off-by: Nick Papior <nickpapior@gmail.com> | 24 September 2019, 08:44:54 UTC |
5d096e6 | Nick Papior | 22 September 2019, 19:22:57 UTC | bug: fixed siesta out reading geometries Signed-off-by: Nick Papior <nickpapior@gmail.com> | 22 September 2019, 19:22:57 UTC |
6326a9b | Nick Papior | 20 September 2019, 09:00:49 UTC | bug: fixed missing line-ending Signed-off-by: Nick Papior <nickpapior@gmail.com> | 20 September 2019, 09:00:49 UTC |
f0d43aa | Nick Papior | 20 September 2019, 08:53:24 UTC | maint: cleaned out siesta for read_data to be more versatile Signed-off-by: Nick Papior <nickpapior@gmail.com> | 20 September 2019, 08:53:24 UTC |
68cb0ce | Nick Papior | 20 September 2019, 06:42:17 UTC | enh: enabled reading fermi-level from EIG and TSHS files Signed-off-by: Nick Papior <nickpapior@gmail.com> | 20 September 2019, 06:42:17 UTC |
980df49 | Nick Papior | 19 September 2019, 13:06:28 UTC | doc: fixed math alignment Signed-off-by: Nick Papior <nickpapior@gmail.com> | 19 September 2019, 13:06:28 UTC |
dbfcb57 | Nick Papior | 19 September 2019, 10:30:41 UTC | enh: enabled reading the fermi-level from the siesta output I haven't fully decided what name this should have. Currently it is named read_fermi_level which is probably justified because it is used extensively. However, it could be argued that this should be read through read_energy(fermi_level=True) or something like this. For now I'll keep this. It currently allows reading the fermi-level from the TSDE, nc files. This fixes #126. Signed-off-by: Nick Papior <nickpapior@gmail.com> | 19 September 2019, 10:30:41 UTC |
572882e | Nick Papior | 19 September 2019, 07:27:28 UTC | maint: ensured looping sparse structures is consistent Now the physical sparse objects loop in the same way as SparseCSR (which makes sense) and scipy.sparse. This means that there is broken backwards compatibility. Also enabled fdfSile to accept strings for order= instead of only lists. Makes it more flexible. Signed-off-by: Nick Papior <nickpapior@gmail.com> | 19 September 2019, 07:27:28 UTC |
3232355 | Nick Papior | 19 September 2019, 06:21:09 UTC | doc: updated publication Signed-off-by: Nick Papior <nickpapior@gmail.com> | 19 September 2019, 06:21:09 UTC |
e48260d | Nick Papior | 18 September 2019, 13:23:22 UTC | enh: enabled STS and STM.LDOS files from siesta Signed-off-by: Nick Papior <nickpapior@gmail.com> | 18 September 2019, 13:23:22 UTC |
b532f99 | Nick Papior | 18 September 2019, 12:38:09 UTC | maint: flake8 for unit/base Signed-off-by: Nick Papior <nickpapior@gmail.com> | 18 September 2019, 12:38:09 UTC |
e0f8735 | Nick Papior | 18 September 2019, 12:21:47 UTC | maint: removed last fragments of OS dependent directory seps Signed-off-by: Nick Papior <nickpapior@gmail.com> | 18 September 2019, 12:21:47 UTC |
d8a8dbd | Nick Papior | 18 September 2019, 09:27:17 UTC | travis: added travis deployment Signed-off-by: Nick Papior <nickpapior@gmail.com> | 18 September 2019, 09:27:17 UTC |
dcf72c3 | Nick Papior | 18 September 2019, 07:41:57 UTC | maint: fixed paths in tests for omni-OS Signed-off-by: Nick Papior <nickpapior@gmail.com> | 18 September 2019, 07:41:57 UTC |
a455536 | Nick Papior | 18 September 2019, 06:53:08 UTC | enh: enabled float64 in eig reads Signed-off-by: Nick Papior <nickpapior@gmail.com> | 18 September 2019, 06:53:08 UTC |
ca0e9c8 | Nick Papior | 17 September 2019, 09:28:57 UTC | bug: fixed correct inheritance of axsfSile Signed-off-by: Nick Papior <nickpapior@gmail.com> | 17 September 2019, 09:28:57 UTC |
8bb7917 | Nick Papior | 16 September 2019, 13:01:54 UTC | test: enabled test for doing stuff with a non-finalized sparse matrix Signed-off-by: Nick Papior <nickpapior@gmail.com> | 16 September 2019, 13:01:54 UTC |
bf6a420 | Nick Papior | 13 September 2019, 20:47:13 UTC | maint: performance increase for .append Now everything is done in one go. ~3 times faster for the thing I tested (not too big system). So should be very much faster for very big things. No loops over rows! Signed-off-by: Nick Papior <nickpapior@gmail.com> | 13 September 2019, 20:47:13 UTC |
8ee6723 | Nick Papior | 13 September 2019, 12:37:00 UTC | bug: ensured atomic indices requests made for geometries fails for ndim>1 When requesting atomic indices for ndim>1 it has so far passed through, but it may have unwanted side-effects. I have added a check against this in _sanitize_* for ndim > 1 and fails if so. Thanks to Jonas for finding this, fixes #144. Signed-off-by: Nick Papior <nickpapior@gmail.com> | 13 September 2019, 12:37:00 UTC |
d49f070 | Nick Papior | 13 September 2019, 11:53:07 UTC | maint: updated append Now append is much smaller in code size and much easier to follow. This has two consequences: 1. It makes me more optimistic about merging different SparseGeometries 2. It makes append much slower (the larger the system, the larger penalty). However, since append shouldn't be done to often I would prefer a more resilient and easier to debug version than anything else. Besides, it can be speeded up my doing a manual merge of the two sparse matrices that are created since they only have non-zero elements in their own subspace. Signed-off-by: Nick Papior <nickpapior@gmail.com> | 13 September 2019, 11:53:07 UTC |
5d5583c | Nick Papior | 13 September 2019, 11:31:44 UTC | enh: added scale_columns to SparseCSR and performance increase for ops 1. Speeded up CSR initialization by usinf full (instead of array creation casted to numpy.array) 2. Added scale_columns which enables a scaling for only certain column values. This *should* in principle be superseed with M[:, [0, ..., ]] *= scale but for now this is easier since I don't want to go into indexing problems with matrices. 3. Added another keyword to _extend to disable the return of indices Since this requires an additional call to indices function we can do without when not needed. 4. Speeded up align using opt 3. 5. Operations for sparse matrices are now faster due to removal of duplicate code *align*. For very large matrices the largest amount of work is done looping the rows which is now only done once. Signed-off-by: Nick Papior <nickpapior@gmail.com> | 13 September 2019, 11:31:44 UTC |
fefed8f | Nick Papior | 11 September 2019, 19:56:51 UTC | maint: updated CAR files a bit with docs and clarity in code Signed-off-by: Nick Papior <nickpapior@gmail.com> | 11 September 2019, 19:56:51 UTC |
0072a9d | Nick R. Papior | 11 September 2019, 12:33:46 UTC | Merge pull request #143 from tfrederiksen/master bug: ignore geometry constraints in POSCAR/CONTCAR | 11 September 2019, 12:33:46 UTC |
b6bced3 | Thomas Frederiksen | 11 September 2019, 11:43:30 UTC | bug: ignore geometry constraints in POSCAR/CONTCAR | 11 September 2019, 11:43:30 UTC |