The blocked collision had three arms (MRT wrapper, channel SRT, level SRT)
and a one-caller MRT wrapper in collision_mrt.cuh. Now block_collide computes
rho/u once and has two compile-time families:
- uniform-equivalent (MRT any level, SRT channel root): collide_dispatch
arithmetic with zero force, no LES: one clamped omega, collide_mrt_no_force
or collide_srt with its zero Guo term (unchanged);
- level SRT (BLOCK_ROOT_BC 0/1): unclamped, collide_srt_no_force, shifted
momentum feq exception kept visible.
No runtime d_params.fx/fy is read. Uniform kernels unchanged (only the
unused inline wrapper is removed from collision_mrt.cuh). Source-text test
updated to the new single call sites.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The removed _boundaries list made the two body-reject patches no-ops.
Inject them via _device_boundaries (level-2 record with foreign host
geometry; non-None level-1 record). set_jet_amplitude docstring names
the public jets view.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
jet_scenario=none host tests run against the current checkout (no
blocked_base_tables), while historical runs import the pinned source (no
blocked_coarse_tables). Select whichever single name the source defines;
fail if neither or both. No src alias; pins and manifests unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Conflicts (6 files): tests took the caller-migration side, then the
358bed1 renames (_level_boundaries -> _device_boundaries); run_refined_dfg.py
kept ae53fad (same jet_interval level= repair plus blocked_coarse_tables on the
current-checkout path). Follow-up renames in 5 auto-merged tests;
blocked_fine_tables imported from refinement_tables. Frozen drivers untouched
(run_jet_circle_cv.py, check_projected_pairs.py remap_child, historical DFG path).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- No NODE<->EDGE input conversion still leaves the root-edge -> fine-local
half-cell projection (prepare_fine_boundary, fine_hits_to_edge_frame); say so.
- run_refined_dfg.py (jet_scenario != none, current checkout only): pass
level=1 to FixedPatchStepper.set_jet_amplitude on reversal; use
blocked_coarse_tables there. jet_scenario=none keeps the pinned source's
blocked_base_tables; manifests and results unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Coarse/fine name the two members of any pair (Hierarchy docstring); only the
misleading blocked_base_tables -> blocked_coarse_tables. prepare_fine_boundary
and blocked_fine_tables keep their pair-relative names.
- Depth-1 static circles use the integer-NODE frame (unchanged behavior); the
root EDGE frame applies to refined levels only. Contract text corrected.
- _kernel_configs documents why it stays distinct from Hierarchy.configs.
- Drop the stale re-export comment; two tests import table builders from
refinement_tables. Mechanical caller renames for this delta only.
No numerical code or config derivation changed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A set_jet_amplitude failure while creating the internal stream marked the
Simulation failed, but the public stream property still retried
cuda.Stream(). Guard it like the other device-touching methods and add a
regression test (no retry, close still retryable). Lifecycle wording now
names the guarded device-touching methods; cached readers and host staging
stay callable. compatibility.md records the caller-visible fail-closed change.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Simulation.set_jet_amplitude validated before upload and committed host
state after it, but after an uncertain device write nothing stopped reuse.
Record the failure in Simulation._failed and refuse every public device
call except close() (FixedPatchStepper already uses status 'failed').
Invalid input stays side-effect free and retryable. Adds host fault tests
for both owners and states the rule in docs/lifecycle.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- test_exchange_window and test_raw_exchange_gpu fine(): dense root becomes
a (12, 8) blocked root (fine(None) stays only for the dense reference test
that D2 retires);
- jets: G1/G3 compare blocked layouts against the FINE_BLOCKS[0] reference
(blocked root); the dense run is a separate leg D2 removes; G2 legacy
actuation vs Jet on the blocked reference; CPU fixtures use a blocked root;
- test_refinement_blocked_fine.raw reads every level through its layout;
- test_fine_surface_gpu: new test_fine_surface_blocked_matches_uniform_four_hops
(fine ROI after each of four hops vs the uniform 48^2 solver with the same
tolerances, blowing/suction mass sign from raw exchange, zero actuation ==
passive bit for bit); the cross-interface donor rejection passes tiles.
The dense hook/ledger/audit probes in test_fine_surface_four_hops and the
dense SoA fault tests stay until D2 retires them by name.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- actual-TG-rate case: the unshifted leg uses [(16, 16), (24, 24)];
- partial-advance: blocked storage, fine owner storage read through the layout;
- new test_bounded_equivalence_blocked: the same eight-step scalar-reference
checks (rest + seeded nonequilibrium, synchronization, leaf drift/moments,
hop-2 validity mask, coalesce sources, 8 corner packets) and launch-failure
publication guard on blocked tiles, via the shared eight_steps helper.
The dense audit build (CELERIS_REFINEMENT_AUDIT counters) stays in
test_bounded_equivalence_and_ownership until D2 retires it with the dense
kernels.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every FixedPatchStepper in test_refinement_followup passes explicit tiles
(whole-level tiles on the GPU cases, (1, 1) in the pre-CUDA margin
contract). Assertions kept; dense-specific probes rewritten:
- fault fakes use a blocked BlockLayout2D instead of _layouts [None, None];
- slot/reservoir: the fine module sees its tile window incl. halo
(fx+2, fy+2) with slot fields zeroed; the root's walls are
block_root_bc 1 (fine 0) instead of dense Y_WALL_BC/TOP_WALL_BC macros;
- raw owner storage is read through each level's layout (to_dense);
- the off-centre allocation bound counts blocked buffers.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Extends DenseBlockedRouteParity (test-only, on 1573942) with the residual
dense-supported combinations named before any dense-route deletion:
- test_mrt_bodies: MRT on a periodic root with a passive circle, an
actuated circle (+/-) and a velocity Jet with an amplitude switch
(MRT slot/reservoir is rejected on every route: on_site_velocity is
SRT-only);
- test_two_bodies: two circles on a wider patch (4, 4, 28, 20), periodic
and slot/reservoir, Jet on one + passive other, and two actuated (+/-);
routes D/D, D/B [None, (8, 8)], B/B [(12, 8), (8, 8)].
Same gate: exact FP32 bits and NaN masks at 0, 1, 2, 8 and after
re-initialize; D/B vs B/B raw-exchange ledgers equal for every body.
Docstring table updated; no tolerance, no src change.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Both public examples move off the dense refinement route, inside the scope
proven by the dense/blocked parity gate (1573942: two-level unshifted SRT,
periodic or slot/reservoir root, one circle/actuation or body-free):
- fixed_patch_jet: LEVEL_BLOCKS = [(12, 8), (8, 8)] (48x48 root, 56x40
fine window), recorded in its report config;
- fixed_patch_circle: --base-block (default 12 8) and --fine-block
(default 16 20); the dense 'omit --fine-block' option is gone.
Host test test_example_tiles: each example's tiles (read from the example
source without importing it) build every layout and table up to the first
CUDA call, for the circle in both coarse boundary modes.
No src change.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
same_summary applied np.array_equal(equal_nan=True) to string metadata
('last_base_step'), a TypeError on the first D/B vs B/B ledger compare.
Numbers (bool/int/float/complex) still compare exactly with equal shapes
and NaN-equality; other values (window labels) compare by dtype and value.
New host test: a summary with string metadata equals itself, and changing
the window label, an integer, a nested float, a scalar or a NaN-bearing
array is detected in both directions. No assertion removed, no tolerance.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Test-only, on c6d5dc1. DenseBlockedRouteParity runs the same configuration,
initial state, body/Jet schedule and tilings on D/D [None, None], D/B
[None, FINE_BLOCK] and B/B [(12, 8), FINE_BLOCK], periodic and
slot/reservoir, for body-free, passive circle, actuated circle (+/-) and a
velocity Jet with an amplitude switch: every public and owner array must
match in FP32 bits and NaN masks at steps 0, 1, 2, 8 and after
re-initialize, and the D/B and B/B raw-exchange ledgers (last base step and
last substep) must be equal. No tolerance.
test_route_capability_matrix checks before CUDA what the dense route
rejects (raw exchange on a dense body level, shifted, projected, channel,
several L1 rectangles, depth >= 3 dense). The module docstring holds the
supported/unsupported table and the parked cases (MRT covered elsewhere,
shifted and depth >= 3 have no dense route, several bodies later).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
8e1feff set root.last_raw on the single-level reference, which also makes
FixedPatchStepper._hop accumulate a ledger the reference never allocated.
The adapter now substitutes the raw output pointer in the BlockCurvedBoundary
launch (argument 10, as the deleted adapter did) and leaves last_raw unset;
Jet column and force arms still come through the root level boundary.
Comparison unchanged; no production change.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Restores the PeriodicBlockStepper safeguard dropped in 8e1feff, for all
FixedPatchStepper levels: both G buffers, halo pairs and boundary columns
of every level are compared with cuda.mem_get_info() before the stream,
any module or any buffer exists, raising MemoryError('level buffers exceed
free device memory'). Transfer tables are not counted (lower bound).
Tests: the single-level MemoryError case is back beside the param-layout
check (no stream, no allocation); host test_single_level covers depth 1
and depth 2 (no stream, build or allocation) and the wait failure after a
good step (count stays 1). Driver-mocking fixtures report ample free memory.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The second owner of one blocked level goes. Its capability is the
zero-refinement FixedPatchStepper added in f764d68/985db14 (periodic SRT
level, stationary passive circles, shifted storage, NaN spare, single wait),
which matched it array for array on GPU at 1318fc2 (3 splits x +-circle x
shifted/unshifted) and passes the host lifecycle and fault contracts.
API change for callers:
PeriodicBlockStepper(cfg, BlockLayout2D.build(ny, nx, by, bx), context=c, objects=o)
-> FixedPatchStepper(cfg, context=c, refinements=[], level_blocks=[(by, bx)],
level_objects=[tuple(o)])
download() returns a 1-tuple; static-circle tables are
_level_boundaries[0].tables (was .boundary); no free-memory preflight.
- lbm/block_stepper.py deleted; MAX_GRID_DIM lives in refinement_stepper.
- tests/validation/multigrid/test_block_stepper.py: same frozen GPU gates
(split identity, FP64 oracle, circle across seams) and pre-CUDA rejection
cases on the single level; the old-vs-new bitwise test is retired with the
old class (its result is the 1318fc2 record).
- tests/unit/test_block_step_sync.py folded into test_single_level.py
(step(n) vs n x step(1) roles and one wait, step(0), mid-batch launch
failure), which also gains bad-state rejection before upload, download
failure and activation-failure retry on close.
- test_shifted_storage and test_raw_exchange_gpu use the single level; the
raw-exchange reference feeds the AMR Jet column and raw ledger through the
root level boundary instead of patching a private kernel field.
- Pinned historical runners (taylor_green/run_case.py, run_jet_circle_cv.py)
are unchanged: they only run on trees that still ship PeriodicBlockStepper.
No alias.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Fault injection with the existing host fakes for FixedPatchStepper(
refinements=[]): an initialize upload failure, a BlockHaloCopy,
BlockCurvedBoundary or BlockStep launch failure, and a failing final
stream wait each raise, leave the instance failed (step/download refused)
and keep the completed step count; a failed launch never reaches the wait.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Review follow-ups for S1 (f764d68):
- a single level starts its spare buffer NaN, as PeriodicBlockStepper does,
so stale halos cannot hide (multilevel initialization unchanged);
- projected explosion is rejected for a single level;
- new tests/unit/test_single_level.py: NaN spare after initialize, per-step
launch order halo -> BFL (circles only) -> BlockStep with a null kind and
one wait per step(n), circle solids excluded from the state mask;
- the GPU bitwise comparison also covers shifted storage.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A zero-refinement FixedPatchStepper is the PeriodicBlockStepper capability:
one doubly periodic blocked SRT level (no LES, Jets or exchange
observation) whose level_objects[0] are stationary passive circles. The
circle tables come from the same checks, now blocks.static_circle_tables
(moved from block_stepper, pycuda-free); BFL, halo and BlockStep launches
are the shared helpers, with an all-ACTIVE (null) kind. Base-level bodies
stay rejected for depth >= 2. PeriodicBlockStepper and its callers are
unchanged (deleted only after S1's GPU result).
Tests: host contract rejections before CUDA for the single level; GPU
bitwise comparison against PeriodicBlockStepper for every split, with and
without the circle (the S2 gate).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
36ad523 made a manifest file missing from the checkout a 'Source mismatch',
which also broke the tests' manifest-selection helper: it reused
verify_source with hashing neutralised but still read the (renamed) files.
run_refined_dfg.source_manifest(collision, observation_mode) now owns the
frozen-manifest selection; verify_source uses it, then verifies the
checkout. The test helper calls source_manifest directly; source identity
stays owned by test_source_pin. Same errors, manifests and pins.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Fixes the regressions from the entry/compiler renames (2b3c2a7, d318911)
that Astra reproduced at d318911:
- run_refined_dfg.verify_source: a manifest file missing from the checkout
is a 'Source mismatch' (was FileNotFoundError); the current-checkout
(Jet scenario) capture hashes the manifest's harness files plus the
checkout's own 'git ls-files src', not historical src paths.
- time_window_dfg: candidate file list names kernels/blocked.cu.
- test_projected_pairs: the a135252-only harness's runtime_source is the
historical receipt value (historical_runtime_source) in the receipt
fixture and the two mocked run_case.main tests; runtime_source itself
stays tested against the a135252 package (historical_package).
Harness files, pins, manifests and thresholds are unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The package already exposed it as CelerisLab.cuda.compiler; the module now
has that name and every tip caller, doc and current-tree test imports it
directly (no alias).
Historical-revision tooling is not edited: the Taylor-Green harness
(run_case.py, hash-pinned check_projected_pairs.py, assess receipts) only
runs against a135252, whose package has cuda/compiler_v2.py, and the J
campaign runners and dated docs describe past trees. Its two current-tree
tests now import that package from git (historical_package) instead of the
tip, and the kernel-root test first checks the accepted case.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
PeriodicBlockStepper and FixedPatchStepper each built the blocked level's
kernel config (block window NX/NY, neutralized wall/slot/inlet fields,
omega/tau check) and each spelled out the raw BlockHaloCopy and
BlockCurvedBoundary argument lists. These now live once in lbm/blocks.py
(level_config, launch_halo_copy, launch_curved_boundary), so a kernel ABI
change has one host site. Both owners keep their own lifecycle; no new
class, no behaviour or default change. PeriodicBlockStepper now checks its
level config before any device work (same errors). Dense refinement levels
keep their own config check.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
kernel_v2.cu -> uniform.cu (whole-domain NX/NY, Simulation) and
blocks.cu -> blocked.cu (one block window, every blocked level), with
InitTubeFlow_v2 -> InitTubeFlow. All live callers, include comments,
CONFIG.md, lifecycle doc and current-tree tests move together; no alias.
refinement.cu keeps its name until the dense-route decision.
Historical-revision tooling that names the old files (LES source pins,
timing/J campaign runners, TG run_case, supported_settings citations at
888c8038, development notes) is unchanged on purpose.
No arithmetic change; the PTX symbol InitTubeFlow_v2 is renamed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
generate_config/compile_kernel/load_module also recorded the last build
directory and PTX in ContextVars so that bare compile_kernel() and
load_module() calls worked: a second, implicit interface beside the
explicit build_dir/ptx_path one used by every owner. compile_kernel now
requires build_dir= and load_module requires the PTX path.
Build contract tests use explicit paths; they still check per-build
header/output isolation, interleaved generation, path escapes, refusal of a
removed snapshot, and that a failed compile keeps the previous PTX and
leaves no partial output. No kernel, numerical or launch change.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Simulation, PeriodicBlockStepper and FixedPatchStepper levels each repeated
generate_config -> compile_kernel -> load_module -> cleanup and the
d_params pack/upload. They now call compiler.build_module(cfg, entry,
arch=..., n_objects=..., block_root_bc=...) and compiler.set_params.
Simulation no longer keeps its build directory alive until close (nothing
read it after loading). Removes the unused compile_kernel_v2 alias.
Host fakes and lifecycle tests follow the new seam; the close-retry case
for build-directory cleanup goes with the removed cleanup step.
No kernel, numerical or launch change.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
refinement.cu keeps its header, includes, the curved_double.cu entry
block, the #error guard and the optional CELERIS_REFINEMENT_AUDIT
counters, then includes step/patch_step.cu, which holds patch_collide,
PatchExplode, PatchFineStep and PatchCoarseStep moved verbatim (checked
against the previous refinement.cu) in the same order.
test_refinement_mrt reads refinement.cu + step/patch_step.cu; its
assertions are unchanged. The audit-build tests include refinement.cu
from the build root, so step/patch_step.cu still resolves.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
blocks.cu keeps its header, includes and #error guards and then includes
the kernels, as kernel_v2.cu does:
- step/block_transfer.cu: BlockHaloCopy, BlockExplode,
BlockExplodeProjected, BlockCoalesce
- step/block_boundary.cu: BlockCurvedBoundary, BlockAccumulateExchange
- step/block_step.cu: BlockStep (rejects BLOCK_ROOT_BC == 2)
- step/block_channel.cu: unchanged (mode 2 only)
Kernel text moved verbatim (checked against the previous blocks.cu) and in
the same definition order, so the translation unit is the same sequence of
kernels. The two source-reading tests follow the move:
test_projected_explosion reads step/block_transfer.cu, test_refinement_mrt
reads blocks.cu + step/block_step.cu; their assertions are unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Horsea's gates showed block_ddf/block_slot change real PTX in
BlockHaloCopy, BlockExplode and BlockExplodeProjected (6/10 modules;
BlockExplodeProjected 38 -> 40 registers) although the results stayed
exact. The stack's bar is byte-identical PTX for naming-only steps, so
those three kernels return to their K3 expressions, the helpers leave
core/layout.cuh, and test_projected_explosion goes back to its K3 form.
fill_invalid stays (its kernels were byte-identical). K4 vs K3 is now
only the NaN-fill helper.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
K4 made BlockExplodeProjected address storage through block_ddf and
block_slot (core/layout.cuh). This host test compiles the kernel text
extracted from blocks.cu, so it now also extracts those two helper
definitions from core/layout.cuh (device qualifiers -> inline) instead of
failing to compile. Assertions and tolerances unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- core/cell_kind.cuh holds KIND_ACTIVE/INNER/OUTER (was a constexpr in
blocks.cu that step/block_channel.cu and the channel_root.cpp shim
relied on or redefined); blocks.cu, block_channel.cu and the shim now
include it.
- blocks.cu checks the lattice/model #error guard before the
BLOCK_ROOT_BC guard and mode includes, so an unsupported build reports
the real reason first.
- operators/block_collision.cuh includes what it uses (params, macro,
SRT/MRT collision) instead of relying on blocks.cu's include order.
Preprocessor-only: same kernels, same statements, same arithmetic.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Grok