WEEKLY REPORT · ISO 2026-W45

PHARE, 2 Nov – 8 Nov 2026

Commits on master
17 ▼ -11 vs last week
PRs merged
15 ▼ -7 vs last week
Issues opened
12 ▲ +4 vs last week
Issues closed
2 ▼ -7 vs last week
Reviews
38 ▼ -23 vs last week
CI builds
1,926 ▼ -25 vs last week

The week at a glance

each dot is one event · hover for details, click to open
Mon 2MTue 3TWed 4WThu 5TFri 6FSat 7SSun 8S
Commits on master
17 PRs merged
15 PRs opened
32 Issues opened
12 Issues closed
2 Reviews
38 CI failures
232
Daily numbers as a table
DayCommitsIssues openedIssues closedPRs openedPRs mergedPRs closedReviewsCommentsCI buildsCI failed
Mon 2 Nov520331669123
Tue 3 Nov4118425420459
Wed 4 Nov460846163130764
Thu 5 Nov3103305732614
Fri 6 Nov0201001627721
Sat 7 Nov1011101016824
Sun 8 Nov0008024355327

Issues

12 opened · 2 closed

Open issues

The axis starts at 130 so the weekly changes stay visible.

Labels on new issues

bug 🔥
1amr
1triage
1MHD
1no label
11

How closed issues were closed

fixed by a PR 1fixed by a commit 1
  • #1515 One process-wide scratch buffer for field data instead of hidden static vectors completed fixed by PR #1516 corerefactoringperformance
  • #1496 Projection stencils (momentsToEx, BxToEx, ...) are hand-written per quantity pair and never checked against the centerings completed fixed by commit 75189b49e corerefactoring
All 12 new issues
  • #1577 Ideal Ohm term does not conserve energy: Ki + WB grows in a uniform undamped plasma @acrd
  • #1575 The max on the signed ion flux corrupts it when patches are G+1 cells apart, giving wrong E @acrd
  • #1571 mpi::collectArrays copies elements as bytes (unused until #1171) @PhilipDeegan
  • #1569 TestGridLayout cannot make a two level test: dl is always 1 / cells, the AMR box and origin are fixed @PhilipDeegan
  • #1563 dispersion test: unseeded runs make the amplitude check flaky (o3, high_k aic m=4) @PhilipDeegan
  • #1562 res/eg: finish the 1D ion beam notebook @PhilipDeegan
  • #1559 Emergency dumps: write computed diagnostics (momentum tensor, MHD V/P) from the crash state @PhilipDeegan
  • #1558 Charge density shows dips at coarse/fine patch boundaries with interp_order=1 (flat_finest_field) @acrd
  • #1557 Adaptive time step with AMR: reflux weights assume ratio² substeps that SAMRAI does not take, and related fixes @igirault bug 🔥amrtriageMHD
  • #1553 pharesee: merged finest coords extend one node past the domain @PhilipDeegan
  • #1546 Schedules of a removed finest level keep that level's data allocated @nicolasaunai
  • #1544 Messengers: make the B init schedule a Refiner like the other init fills @UCaromel

Pull requests

15 merged · 32 opened · 11 closed unmerged

Open PRs

The axis starts at 20 so the weekly changes stay visible.

Every PR that moved this week

mergedopen draft or closed first human review
Time open (log scale)
1 h1 d1 wk1 mo1 yr
Lines changed, by area (log scale)
101001k10k
1560 test-diagnostics removes its diagnostics when all tests pass, ctest ran out of disk on CI@PhilipDeegan
1556 recentering: quantity type from GridLayout, drop weightPointDim@PhilipDeegan
1554 Make the second-order AMR refiners readable: named fine sites, one generic divB touch-up@UCaromel
1547 Fix red master: load-balancing test still passes the LoadBalancer 'active' argument removed in #1308@nicolasaunai
1545 Stop py3_init-densities failing by chance: thresholds measured over many runs@nicolasaunai
1541 Hybrid dispersion test now checks ω(k) of both branches in 80 s, instead of checking nothing for 2+ hours@nicolasaunai
1540 Lighter 1D functional tests: drop td1d and translat1d, td1dtagged checks and plots more in less time@nicolasaunai
1532 Give MHD test scripts unique names so the nightly runs and logs them correctly@nicolasaunai
1530 Seed particles per cell instead of per patch@nicolasaunai
1516 single source for static buffers @PhilipDeegan
1505 vtkhdf: stop writing each dump twice, one write per rank (17-31x faster dumps at 1500-3000 ranks)@igirault
1501 Drop box iterator@PhilipDeegan
1499 Derive projection stencils from centerings instead of 58 hand-written tables@igirault
1492 Perf/mhd merge schedules@UCaromel
1419 Build inter-level schedules once per regrid; fix crash on level-0 rebalance with 3+ hybrid levels@UCaromel
1586 Optional LLNL mpiP for MPI profiling@PhilipDeegan
1585 Per scope phlop timers, memory and hardware counters, and a build comparison@PhilipDeegan
1584 New updater: copy mode pushes structure of arrays batches, vectorised@PhilipDeegan
1583 [4/5] Remove the ion border max: hybrid fields no longer depend on how a level is cut into patches@UCaromel
1582 [5/5] Hybrid: patches sharing a node start from the same E and B (fill level-0 B ghosts at init)@UCaromel
1581 [3/5] Border sums add the contributions to a node in the same order on every patch, so all copies have the same bits@UCaromel
1580 [2/5] Deposit each level ghost particle once, before the border sums, so refined-level moments do not depend on the patch layout@UCaromel
1579 [1/5] Fix a level ghost particle duplicated at patch corners on the coarse–fine edge@UCaromel
1578 Zero copy particle restarts@PhilipDeegan
1576 Border sums: pair patches G+1 apart, fields no longer depend on the patch size (#1575)@PhilipDeegan
1574 Delete the catch-all core/utilities/types.hpp, each piece moves to a header named after what it does@PhilipDeegan
1573 mpi_utils: tests, fixed broken gather/broadcast functions @PhilipDeegan
1572 Type traits become C++20 concepts, each in a header named after what it describes@PhilipDeegan
1570 Adaptive dt with AMR: each level at its own stable dt, flux sums weighted by substep duration@igirault
1568 New Updater does not copy vector on first move@PhilipDeegan
1567 Updater refactor: Boris as free functions, the current updater as IonUpdater0@PhilipDeegan
1566 Test groundwork: TestGridLayout constructors instead of make, field comparison fixtures@PhilipDeegan
1565 New updater: domain particles are never outside the ghost box on exception@PhilipDeegan
1564 Drop typename where C++20 makes it redundant@PhilipDeegan
1561 Pin phlop to v0.0.45 so phlop changes can't break PHARE builds@PhilipDeegan
1555 vtkhdf: write the hybrid momentum_tensor diagnostic@UCaromel
1552 electrons: polytropic closure transporting s = Pe/n^gamma@rochSmets
1551 electrons: pressure closure hierarchy and factory@rochSmets
1550 resources manager: runtime resources may be std::variant@rochSmets
1549 pharein: input checks for the electron entries, optional min_initial_density@rochSmets
1548 diagnostics: electron pressure output, set on every level at initialisation@rochSmets
1543 Conservation tests that check energy and momentum, fast on every PR@nicolasaunai
1490 Polytropic electron closure@rochSmets
1485 CI: run builds, CodeQL and coverage as one workflow so each PR push is one row on the Actions page@PhilipDeegan
1417 amr: skip every other regrid of the finest level@PhilipDeegan
1416 (WIP) Tiling gpu@PhilipDeegan
1396 do no duplicate levelghosts across overlapping patch ghost boxes@PhilipDeegan
1332 WIP / experimental multi precision particles@PhilipDeegan
1214 expose TileClustering tile_size and allow_remote_tile_extent params@UCaromel
1166 Updater does not copy vector@PhilipDeegan
coreamrpythontestssimulator/iobuild/cidocother

People

human activity inside the week · darker = more

Who did what

CommitsPRs openedPRs mergedReviews givenIssues openedIssues closedCommentsLines on master
@PhilipDeegan 416476·33 +662 −953
@nicolasaunai 936111214 +3,529 −2,828
@UCaromel 37361·3 +1,712 −1,009
@igirault 112141·3 +216 −100
@acrd ····3·3 ·
@rochSmets ·5····1 ·

Bots: coderabbitai (46 reviews, 32 comments), github-advanced-security (14 reviews, 0 comments), github-actions (0 reviews, 12 comments).

Who reviewed whom

reviewer ↓ · author →PhilipDeeganUCaromelnicolasaunaiigirault
igirault122··
nicolasaunai45·2
PhilipDeegan·52·
UCaromel··6·
Median first review3 h
Median time to merge3.1 d
39 open PRs waiting for a first human review · 0 with no activity for 30 days (as of 9 Nov)

Waiting for review

  • #1171Live, zero-copy access to simulation data from Python@PhilipDeegan open 245.1 d
  • #1239merge hybrid schedules to reduce MPI communications@PhilipDeegan open 127.1 d
  • #1267validate restarts against hash produced from initial patch datas@PhilipDeegan open 110.1 d
  • #1379support dump_level and fine_dump_lvl_max properly for vtk@PhilipDeegan open 35.2 d
  • #1406model temporaries as needed via model::requestTemporaries@PhilipDeegan open 28.2 d
  • #1410Exclude data rebuilt every step from restart files@PhilipDeegan open 28.1 d
  • #1431amr: level accessors, hide per patch loops behind onLevel@PhilipDeegan open 24.1 d
  • #1432Install PHARE with pip install . (scikit-build-core build backend)@igirault open 23.4 d
  • #1445Validate collisions on refined levels and document them@nicolasaunai open 21.4 d
  • #1444Ion-ion collisions in hybrid runs: Simulation(collisions=...)@nicolasaunai open 21.4 d
  • #1443Exact momentum and energy conservation for collisions with unequal weights@nicolasaunai open 21.4 d
  • #1442Binary Monte-Carlo ion-ion collision operator, in core and tested against NRL rates@nicolasaunai open 21.4 d
  • #1441Deposit ion moments after the particle exchange, so each cell's particles are on one patch first@nicolasaunai open 21.4 d
  • #1511pharesee: drop ghosts from moments, combine fields with differing ghosts@PhilipDeegan open 13.2 d
  • #1510pharesee: move flat_finest_field to run/utils, split make_interpolator@PhilipDeegan open 13.2 d
  • #1520Performance monitoring introspection@PhilipDeegan open 12.1 d
  • #1537pharesee: read vtkhdf diagnostics into hierarchies@PhilipDeegan open 9.2 d
  • #1550resources manager: runtime resources may be std::variant@rochSmets open 6.2 d
  • #1549pharein: input checks for the electron entries, optional min_initial_density@rochSmets open 6.2 d
  • #1548diagnostics: electron pressure output, set on every level at initialisation@rochSmets open 6.2 d
  • #1561Pin phlop to v0.0.45 so phlop changes can't break PHARE builds@PhilipDeegan open 5.0 d
  • #1564Drop typename where C++20 makes it redundant@PhilipDeegan open 4.9 d
  • #1568New Updater does not copy vector on first move@PhilipDeegan open 4.8 d
  • #1565New updater: domain particles are never outside the ghost box on exception@PhilipDeegan open 4.8 d
  • #1574Delete the catch-all core/utilities/types.hpp, each piece moves to a header named after what it does@PhilipDeegan open 4.2 d
  • #1573mpi_utils: tests, fixed broken gather/broadcast functions @PhilipDeegan open 4.2 d
  • #1572Type traits become C++20 concepts, each in a header named after what it describes@PhilipDeegan open 4.2 d
  • #1578Zero copy particle restarts@PhilipDeegan open 2.2 d
  • #1584New updater: copy mode pushes structure of arrays batches, vectorised@PhilipDeegan open 22 h
  • #1583[4/5] Remove the ion border max: hybrid fields no longer depend on how a level is cut into patches@UCaromel open 22 h
  • #1582[5/5] Hybrid: patches sharing a node start from the same E and B (fill level-0 B ghosts at init)@UCaromel open 22 h
  • #1586Optional LLNL mpiP for MPI profiling@PhilipDeegan open 19 h
  • #1585Per scope phlop timers, memory and hardware counters, and a build comparison@PhilipDeegan open 19 h
  • #1588Simulation data stays introspectable after Simulator.run()@PhilipDeegan open 5 h
  • #1587Code scanning cleanup: fix real alerts, filter noisy rules@PhilipDeegan open 5 h
  • #1592Selectors miss particles pushed further than the cell map around the ghost box@PhilipDeegan open 2 h
  • #1591Make PHARE and SAMRAI's own internal times consistent@igirault open 2 h
  • #1590Coverage job no longer times out on the dispersion tests@PhilipDeegan open 2 h
  • #1589dist_plot bulk lines raised for every axis choice@PhilipDeegan open 2 h

Stale

    CI and tests

    416,760 test executions on TeamCity · 165 failed · 54 ignored
    passedfailedcanceledone tick per build, placed at its finish time
    Mon 2MTue 3TWed 4WThu 5TFri 6FSat 7SSun 8S
    GitHub Actions ci-explainer
    123 ✓ · 1 ✗ · 412 canceled review-threads
    59 ✓ · 8 ✗ · 134 canceled gate
    25 ✓ · 20 ✗ · 24 canceled CodeQL
    28 ✓ · 1 ✗ · 21 canceled Coverage
    23 ✓ · 1 ✗ · 25 canceled labels
    48 ✓ · 0 ✗ · 0 canceled CMake Ubuntu
    25 ✓ · 1 ✗ · 21 canceled TeamCity gh_pr_gcc
    79 ✓ · 24 ✗ · 15 canceled · 184,218 tests run gh_pr_clang_samraisub
    88 ✓ · 15 ✗ · 13 canceled gh_pr_clang
    76 ✓ · 25 ✗ · 15 canceled · 191,702 tests run gh_pr_gcc_samraisub
    87 ✓ · 13 ✗ · 14 canceled gh_master_gcc_nightly_samraisub_phlop
    5 ✓ · 1 ✗ · 0 canceled gh_master_clang_nightly_samraisub_phlop_asan
    5 ✓ · 1 ✗ · 0 canceled gh_master_gcc_samraisub_weekly_gcov
    0 ✓ · 0 ✗ · 4 canceled gh_master_gcc_weekly_samraisub
    1 ✓ · 0 ✗ · 0 canceled gh_master_clang_weekly_asan
    1 ✓ · 0 ✗ · 0 canceled gh_master_gcc_samraisub_weekly_caliper
    1 ✓ · 0 ✗ · 0 canceled TeamCity, next CI (shadow) Pipeline Head
    72 ✓ · 36 ✗ · 18 canceled · 14,016 tests run gcc (build + fast tests)
    73 ✓ · 28 ✗ · 19 canceled · 13,198 tests run clang ASan (build + fast tests)
    73 ✓ · 29 ✗ · 17 canceled · 13,626 tests run Nightly page
    7 ✓ · 0 ✗ · 1 canceled build (gcc Release)
    5 ✓ · 2 ✗ · 0 canceled 3D particle init, serial
    5 ✓ · 2 ✗ · 0 canceled 3D particle init, MPI
    5 ✓ · 2 ✗ · 0 canceled fast tests (all of level 10 and below but 3D init), serial + MPI
    4 ✓ · 3 ✗ · 0 canceled MHD functional
    5 ✓ · 2 ✗ · 0 canceled MHD harris 2D (periodic + with boundaries), MPI
    5 ✓ · 2 ✗ · 0 canceled hybrid functional, 1D decks
    1 ✓ · 6 ✗ · 0 canceled conservation (conserv.py), MPI
    3 ✓ · 4 ✗ · 0 canceled hybrid harris 2D, MPI
    5 ✓ · 2 ✗ · 0 canceled dispersion (dispersion.py), MPI
    4 ✓ · 3 ✗ · 0 canceled

    Tests that failed

    (empty): py3_load_balancing_mpi_n_3
    35(empty): py3_vtk_diagnostics
    8(empty): py3_test-pharesee-hierarchy
    8(empty): py3_patch_size_fields
    3(empty): py3_restarts
    3(empty): test-interpolator
    3(empty): test-diagnostics_1d_interp1_refined2
    3(empty): test-diagnostics_1d_interp1_refined3
    3(empty): test-diagnostics_1d_interp2_refined2
    3(empty): test-diagnostics_1d_interp2_refined3
    3

    Red, then green on the same commit

    9 reruns turned a red build green with no failing test: a build or infrastructure problem, not a test.

    Code on master

    17 commits · lines removed ← → lines added

    Where the code changed

    tests
    1,859
    3,393
    core
    1,873
    946
    amr
    1,052
    1,245
    build/ci
    4
    380
    simulator/io
    97
    129
    doc
    4
    25
    python
    1
    1

    Commits

    17 commits on master (also dots on the timeline)