Block-structured AMR with batched fine advance, load-balance and active-box infrastructure - #1628

Draft

sbryngelson wants to merge 1034 commits into

MFlowCode:masterfrom

sbryngelson:up/mega

This pull request is big! We're only showing the most recent 250 commits

Commits on Sep 5, 2026

Ledger (77): my own hypothesis falsified -- the allocator setting recovers 7 percent of what the fused packs recover, so the cost is the map machinery not the malloc, and ledger 75's hedge is corre…
sbryngelson
committed
a4618f6
Ledger (75): all 70 AMR goldens pass on the merged tree -- the 3 chemistry-AMR tests needed the bench fftw on LD_LIBRARY_PATH, not a code fix
sbryngelson
committed
e36a680
Ledger (78): this node's intra-node MPI wait degraded 4.4x during the session -- it explains the drift, the unusable step column and probably the np16 hang, and the wait-free gather column was immune
sbryngelson
committed
4b534eb
Ledger (78): name the canary script and record that its first design was too weak to see the degradation it exists to catch
sbryngelson
committed
cede444
AMR: delete amr_rg_gather and its 35 unreachable sites -- a flag nothing has ever set, whose dead brackets owned instrument rows that could never record (-56 LOC, all 70 AMR tests green)
sbryngelson
committed
7babe17
Ledger (80): per-block cost is ~13 ms/block/step across six phases, per-dispatch not per-byte, a quarter to a third of the gap -- two drafts withdrawn on review, the clustering knobs are inert, the…
sbryngelson
committed
97eedbb
Ledger (81): negative, pre-registered, falsifier fired -- pooling the gather consume is bit-identical and saves nothing; the census shows the pooled kernels are not slower, the per-block cost in th…
sbryngelson
committed
04fb2f0
AMR instrument: five bracket-free host-time rows (h:slot/shell/own/unpk/fill) around the per-block gather consume in both waves, same clock as [mpiwait]
sbryngelson
committed
9002f3c
AMD OpenMP lane: per-file opt-in defaultmap(present:allocatable) (MFC_OMP_PRESENT_ALLOCATABLE); m_amr.fpp opts in
sbryngelson
committed
8dc669a
Ledger (82): the per-block AMR cost is amdflang's per-launch re-map of allocatable derived-type arrays; per-file defaultmap(present:allocatable) opt-in for m_amr, -8%/-18% wall, bit-identical, gated
sbryngelson
committed
00caa29
Ledger (82) correction: unallocated module arrays abort under present in every form, null components do not; m_amr re-audited under the corrected rule (comments updated)
sbryngelson
committed
256355c
AMD OpenMP lane: m_amr_registers.fpp opts in to defaultmap(present:allocatable) (audited; the reflux/capture kernels carried 5.6k mapped args per rank-step)
sbryngelson
committed
a235b5a
m_amr_registers opt-in header: fypp comments only (the formatter had wrapped them into bare Fortran comment lines and a stray #)
sbryngelson
committed
22b4fba
Ledger (83): m_amr_registers opted in -- at most 2% at cap 32, nothing at cap 64; the mapper walk was m_amr-specific and the launch-map ranking detects it but does not price it
sbryngelson
committed
11f4a77
AMR: one device-resident slab table + one GPU_UPDATE per launch for the ghost-fill and own-shell kernels, replacing eight per-launch copyin maps (same values, same order; bit-identical by construct…
sbryngelson
committed
b5b1782
Ledger (84): one device slab table + one update per launch, pre-registered and confirmed on the ghost-fill launch (-29%/-40% gfill); wall deltas are MPI-wait movement, not claimed
sbryngelson
committed
daaa80c

Commits on Sep 6, 2026

Commits on Sep 7, 2026

Commits on Sep 8, 2026

Commits on Sep 9, 2026

Commits on Sep 10, 2026

Commits on Sep 11, 2026

Commits on Sep 12, 2026

Commits on Sep 13, 2026

Commits on Sep 14, 2026

Commits on Sep 15, 2026

Commits on Sep 16, 2026

Commits on Sep 17, 2026

Commits on Sep 18, 2026

Commits on Sep 19, 2026

Commits on Sep 20, 2026

Merge branch 'master' into up/mega
sbryngelson
committed
434f983
Merge branch 'master' into up/mega
sbryngelson
committed
2a4124a

Commits on Sep 23, 2026

Fix NVHPC GPU failures: multi-rank OpenMP fault, cylindrical viscous NaN, OpenACC AMR PRESENT errors
sbryngelson
committed
7383482

Commits on Sep 24, 2026

CI: give Phoenix GPU jobs 8 MPI slots for the ppn=8 AMR test
sbryngelson
committed
546905d

Commits on Sep 29, 2026

Merge branch 'master' into up/mega
sbryngelson
committed
12a049f

Commits on Sep 30, 2026

test: stop testing 3D_reacting_mixing_layer
sbryngelson
committed
ad46e17
Copy num_gps to a local before the corrected-gps rebuild
sbryngelson
committed
5671263
Merge master into up/mega
sbryngelson
committed
864694d
Merge master into up/mega
sbryngelson
committed
b86af13
Pass pb_in/mv_in to the fine-block IB correction
sbryngelson
committed
7799bc0

Commits on Oct 1, 2026

Thread pb_in/mv_in through the fine-stage RK path too
sbryngelson
committed
7f0a816
Merge master into up/mega
sbryngelson
committed
75b40fb

Commits on Oct 5, 2026

Merge branch 'master' into fix1628
sbryngelson
committed
cc377f3