Repository navigation
Fix TDOS/PDOS energy grid misalignment, unify DOS output file naming, and improve plot-tools TDOS - #8144
Open
DanielZhao0432 wants to merge 6 commits into
Open
Fix TDOS/PDOS energy grid misalignment, unify DOS output file naming, and improve plot-tools TDOS#8144DanielZhao0432 wants to merge 6 commits into
DanielZhao0432 wants to merge 6 commits into
Conversation
…h PDOS grid cal_dos() recorded each energy bin [E, E+de) at its right edge, so the TDOS energy grid started at emin+de while the PDOS text grid starts at emin. Both files now share the same emin + n*de grid, allowing point-by-point comparison of TDOS and PDOS output. Only the energy column shifts by one step; dos/dos_int/dos_smear values are unchanged. Verified: ctest --test-dir build -V -R MODULE_DOS (3/3 passed).
…eom}.txt
- write_dos_pw: dos.txt/dos1/dos2.txt -> doss{spin}g{geom}.txt, the same
scheme as LCAO; the geometry index is now always tagged by passing the
real istep in ctrl_output_pw
- write_dos_lcao / write_pdos_text: drop the _nao/_pw basis suffix
(pdoss{spin}g{geom}.txt); the now-unused basis parameter of
write_pdos_text is removed
- docs (parameters.yaml, input-main.md, dos.md, output-specification.md),
examples/13_dos and tests/performance/catch_properties.sh updated
accordingly
Verified: ctest --test-dir build -V -R MODULE_DOS (3/3 passed).
…pairing, x-axis label - TDOS plots the Gaussian-smeared column (dos_smear, unit 1/eV) by default instead of the raw per-bin histogram; a new smearing switch (plot(..., smearing=False) or "smearing": false in config.json) restores the raw column, and utils.gaussian_smearing() lets users apply their own smearing width to it - TDOS accepts per-spin file pairs of the new text format: tdosfile may be a list of both spin files, a single spin file (paired in both directions, so doss2* works too), or a directory; energy grids are checked before merging. nspin=4 single spinor-summed files work as a single channel - the x-axis label is set centrally in DOSPlot._set_figure (bottom-most subplot only), so TDOS figures now show 'Energy(eV)' and the duplicated set_xlabel calls in PDOS branches are removed
…dos-issues # Conflicts: # docs/advanced/input_files/input-main.md # docs/parameters.yaml
Collaborator
Author
|
@ABACUS-PR-BOT please review this PR. |
DanielZhao0432
added
Bugs
ABACUS-PR-BOT
left a comment
ABACUS-PR-BOT
left a comment
Collaborator
There was a problem hiding this comment.
Static review of 5d88d5b with a fresh independent reviewer. Trusted source-governance and whitespace checks passed; builds, runtime/plotting tests and CI verification were not performed. Inline feedback separates the compatibility defect from nonblocking follow-ups.
Comment on lines
+213
to
+215
| // label the bin [e_old, curr_energy) with its left edge so that the | ||
| // TDOS grid (emin + n*de_ev) matches the PDOS grid point-by-point | ||
| ene.push_back(e_old); |
Collaborator
There was a problem hiding this comment.
Non-blocking: an exact endpoint case still has different grid lengths: emin=0, emax=1, de=0.5 produces two TDOS rows (0, 0.5) versus three PDOS points (0, 0.5, 1). The strict while loop predates this PR, so this is a residual limitation of the alignment fix. Please qualify the point-by-point claim and track a common integer-indexed grid plus a row-count/endpoint regression; the code correction can follow separately.
…pairing robustness - guard get_subplotspec()/is_last_row() behind getattr/hasattr so plotting still works on matplotlib < 3.4 (setup.py declares no minimum matplotlib version) - canonicalize a two-file tdosfile list so the spin-1 channel always comes first (a reversed list no longer swaps the up/down labels) - directory discovery now pairs spin files by identical name suffix (doss1g1.txt only pairs doss2g1.txt, never doss2g2.txt) and falls back to the legacy PW dos.txt - docs: the raw dos column is described as states-per-bin (not 1/eV); the smearing switch note now covers its effect on the shift=true gap estimate; PDOS file patterns corrected to pdoss1*/pdoss2* Verified with abacus_dos_test cases: reversed list equals forward list, mixed-geometry directory stays single-spin, legacy dos.txt directory loads, TDOS/PDOS plots unchanged.
Accumulate the smeared DOS in row chunks (default 1024) instead of allocating the dense npoints x npoints Gaussian matrix, which would take ~3.2 GB at 20000 energy points. Numerical results are unchanged (identical to machine precision), verified against the on-disk smeared column and with 20000-point stress input (~0.7 GB peak RSS).
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Reminder
AGENTS.mdanddocs/developers_guide/agent_governance.md.source/changes.Linked Issue
No existing issue: the four problems were found while testing the current
tools/02_postprocessing/plot-toolsDOS/PDOS workflow against the latest ABACUS output (all of them are stably reproducible, see "What's changed").Unit Tests and/or Case Tests for my changes
cmake --build build -j $(nproc)(incremental, after each C++ commit)OMP_NUM_THREADS=1 ctest --test-dir build -V -R MODULE_DOSpython3 tools/03_code_analysis/agent_governance_check.py --base origin/develop --head HEAD --format textOMP_NUM_THREADS=1 mpirun -np 2 ./build/abacus_basic_para): PW-Al (examples/13_dos/01_pw_Al), LCAO-Si2 (examples/13_dos/02_lcao_Si2), LCAO-Fe-AFM nspin=2 (examples/03_spin_polarized/03_Fe_AFM), and the nspin=4 casetests/03_NAO_multik/scf_out_dos_spin4, followed byabacus-plot -d/-p/-ochecks on the produced files.MODULE_DOSunit tests: 3/3 passed after each C++ commit and after mergingupstream/develop.doss1g1.txt/doss2g1.txt/pdoss1g1.txt/pdoss2g1.txt); TDOS and PDOS energy grids are now point-by-point identical;totaldosref=38from the spin4 case still matches its reference (the integrate collector globsdos*.txt, so it is compatible with the new names). plot-tools: default smeared-column plot gives a smooth TDOS and ~0 gap for Al;smearing=falsereproduces the raw histogram; all threetdosfileentry styles (list / either spin file / directory) produce identical spin-up/down plots;gaussian_smearingmatches the on-disk smeared column to ~1e-5; TDOS figures now show the "Energy(eV)" x-axis label and PDOS multi-subplot layout is unchanged.dos*.txtglobs is unaffected by the rename.What's changed?
cal_dos()labeled each histogram bin[E, E+de)with its right edge, so the TDOS grid started atemin+dewhile the PDOS text grid starts atemin, and the two files could not be compared point-by-point. Bins are now labeled with their left edge, so both files share the sameemin + n*degrid. Only the energy column shifts; all data columns are unchanged.doss{spin}g{geom}.txt(TDOS) andpdoss{spin}g{geom}.txt(PDOS, LCAO only). Previously PW wrotedos.txt/dos1.txt/dos2.txtwhile LCAO wrotedoss{spin}g{geom}_nao.txt/pdoss{spin}g{geom}_nao.txt. Downstream scripts that parse the old names need to be updated. Thebasisparameter ofModuleIO::write_pdos_textbecame unused and was removed (both call sites updated).TDOSplotted the raw per-bin histogram column (unit: states per bin), which produced spiky curves and false band gaps for metals (e.g. 0.09 eV for Al). It now plotsdos_smear(unit 1/eV) by default; asmearingswitch (plot(..., smearing=False)or"smearing": falsein config.json) restores the raw column, andabacus_plot.utils.gaussian_smearing()lets users apply their own smearing width to it.TDOSonly understood the legacy 9-column single-file format, so spin-polarized TDOS could not be plotted.tdosfilenow accepts a list of both spin files, a single spin file (its counterpart is found automatically, in both directions sodoss2*works too), or a directory; energy grids are validated before merging. nspin=1/2/4 are all supported.DOSPlot._set_figure(bottom-most subplot only when axes share x), and the duplicatedset_xlabelcalls in the PDOS branches were removed.Governance Notes
out_dosdescription was updated insource/source_io/module_parameter/read_inp_out.cpp(the annotation from which the input reference is generated) and indocs/advanced/elec_properties/dos.md,docs/advanced/output_files/output-specification.md,examples/13_dos/READMEand theexamples/13_dos/*/run.shfile checks;tests/performance/catch_properties.shwas updated to the new file name.source/source_io/module_dosand the DOS output call path insource/source_io/module_ctrl/ctrl_output_pw.cpp(passing the real ionic step so the geometry index is always tagged). No ESolver/HSolver/ElecState/Hamilt logic is touched.