Dev - #11
Merged
Merged
Dev#11
Conversation
…oncurrency manual unlock (photonics)
…ousarray/require with rvalue overloads and flag parsing
…TA handling per CLAUDE spec
… bit_cast, fix __FLT16 macro, consteval, bfloat16 quant, drop misleading SIMD comment
…p-allocate enqueue task, cache dlopen probes, dedup device count, fix managed free, overflow-safe matmul, MADV guard, missing includes
…ft not claim O(N^2) DFT success
…, document hybrid deque and global() singleton
…numbers with named constants
…reductions - Negative indexing (ptrdiff_t, _norm_idx) for operator()/at/get/operator[] and chained proxy subscripts with bounds checks. - Correct vector<bool> storage: const_reference iterators, contiguity offsets, copy compaction, tobytes/byteswap fixes. - NumPy arithmetic semantics: true_divide, floored div/mod with div0 errors, checked shifts, complex abs/var/mean/min/max and NaN propagation. - View-preserving heterogeneous in-place ops via _inplace_op/_inplace_scalar with broadcasting. - Reduction/sorting fixes: empty-slice guards, None-flatten, stable radix sort, templated searchsorted. - Add test_ndarray coverage for the above.
- Reject non-positive dims in unravel_index/ravel_multi_index (fixes wrap and div-by-zero). - Fix ravel_multi_index Fortran branch to iterate low-to-high order.
- Require values.size() == 1 or == a.size() with scalar broadcast (fixes heap-OOB on short values). - Honor view offsets in pointer paths; bypass raw-pointer path for bool.
- Skip zero-size inputs in concatenate; early-return empty result in stack.
- Hoist is_contiguous() with raw-pointer fast path plus strided fallback for all ops. - Constrain real overloads against complex; add pow_elem and analytic sqrt of negatives. - Single-pass log2/log10/logn with validated bases; register and cover with test_emath.
- Add ddof to var/std/nanvar/nanstd; ddof>=n returns NaN instead of throwing. - cov with ddof>=k returns NaN matrix; corrcoef maps zero-variance rows to NaN.
- Add require_* domain guards with NumPy-parity degenerate returns for ~20 distributions. - Delete PCG64/Philox/SFC64 aliases over mt19937_64; honest MT19937 with seed_seq spread. - Split default_rng(seed) into default_rng() plus seed_default_rng().
- Replace front-gate-only apply() with full stride-k state-vector update for all gates. - Add Y/Z/S/T/RY/RZ/CZ/SWAP/Toffoli; fix CNOT to LSB-first; validate qubits, shapes and state size. - Remove dead guard-bytes check; document in-house evolution.
- Explicit read/write_le16/32 for npy header lengths; truncated header length check. - Any payload short-read now throws; add truncated-payload regression test.
- keepdims fills the value instead of returning zeros; validate and dedup axes. - Dispatch matrix orders per slice; enforce 2-axis Nuc, reject >2 axes and 1-axis Fro.
- Exact bigint valuation/inverse/expansion/teichmuller without long long truncation. - Guard prec<0 and p^prec int64 overflow; split euclidean_volume from p^-v_p(det) volume. - Fix p_adic_norm to non-Archimedean max; drop bogus noexcept.
- Guard TwiddleCache with recursive_mutex for ThreadPool use. - Reject s<=0 (other than -1) instead of wrapping to a giant allocation.
- Split count(axis) scalar-total bug into count_axis(); count(axis) now throws directing to it. - Masked dot uses zero-filled accumulation with contamination mask; put hard-mask checks dst. - Add test_masked_array coverage.
- Replace exponential gcd-of-minors enumeration with Bezout diagonalization; drop minors caps. - smith_normal_form delegates to the new diagonalizer; negative Betti now throws. - Add torsion/divisibility/zero-matrix and 14x14 perf-smoke SNF tests.
- Replace pattern-table cup with cochain-level Alexander-Whitney over rationals. - Add cup_pairing_rank; Poincare/intersection read the real cup block with torsion guards. - Homotopy: cup-rank mismatch is conclusively false; Whitehead branch and out-of-range groups provisional. - Cover T2/wedge cup distinction and T2 Poincare skew-unimodularity.
- Add Euclidean, genus-g, lens-space and connected-sum manifolds with factories. - Product Kunneth torsion, de Rham torsion kill, S0 as two points, general sphere volume/curvature. - Faithful-or-placeholder to_simplicial contract; cover volumes, Kunneth cases and cross-checks.
- Mayer-Vietoris exactness from Euler chi(U)=chi(A)+chi(B)-chi(A cap B), not hardcoded true. - Leray-Serre collapses only on the Hopf path or caller-certified products; generic case inconclusive.
- WindowedLLLStrategy rename; is_modular checks the modular law; rounded integral volume. - Tree-level laplacian, scalar Lie derivative type, bounded cached evaluator, pullback convention docs.
- Device models (IonDrift/Simmons/TEAM/VTEAM/Yakopcic/Stanford), windows, mapping, DAC/ADC. - Noise/drift/faults/IR-drop/sneak/thermal; thread-safe Crossbar with program/outer-product. - Tiling/differential crossbars, Strategy backends, factory/builder presets; full test coverage.
- Remove Loihi/SpiNNaker no-op clones; add tested simulate_lif path and LifSimBackend. - Document CPU pass-through harness scope and unwired STDP; update example and bindings docstring.
- Tensor: GPU-FP32/CPU-blocked backends, 49-mult matmul_4x4_49, simulated-FP8 docs. - Accelerator: drop Loihi, ReRAM-sim crossbar model, per-size-class Auto benchmark. - Memory: *HintArray/tag_*_hint/zeros_hinted host-only tags; update benches, examples, docs, bindings.
- in_flight counter around worker and caller-help execution; wait needs empty queues and zero in-flight. - Drop trailing 1ms sleep; suppress foreign-task exceptions on caller-help path. - Add early-return, exception-propagation and post-throw reuse regression tests.
- NP_TUNE_* macros become detail::k* constexpr; add pragma once. - Cache sysconf, hardware_concurrency, simd_info and numa in function-local statics.
- cuBLAS SGEMM/DGEMM and cuFFT paths with per-thread handle/plan caches and RAII buffers. - Per-device arch predicates, cached compute-capability query, CudaStatus/last_error plumbing. - Deterministic plus live-gated backend tests; LSan suppressions for NVIDIA driver leaks.
- Add Heat3D/Boussinesq/projectile/waves/oscillator/pendulum/N-body/Boris solvers. - Add CODATA constants, relativity/EM/stat-mech/QM-TISE/nuclear/dimensionless analytic blocks. - Replace lattice/FFT-Poisson stubs with vorticity refine and spectral solve; fix p-adic unit check. - Add extended example tour and comprehensive test_physics coverage.
- Extend exact Bernoulli table to even k<=30; odd k>1 returns 0; drop bogus noexcept. - Fix Eisenstein exactness check per coefficient; add floating-point variant for k=12-type cases. - Replace j_invariant_series stub with real E4/E6 series division; add analytic j_invariant(tau). - Add modular_discriminant(tau), convergence guards, binary-exponentiation helpers. - Fix is_hecke_eigenform eigenvalue inference and truncation handling; harden ModularForm.
- Document CPU/LIF-sim neuromorphic, CPU/GPU-FP32/blocked tensor and expanded analog scope. - Sync tensor/memristor example names with the honest backend API.
- libstdc++ debug builds assert stddev > 0 at construction, so default zero-noise configs aborted. - Use a dummy stddev of 1.0 when noise is disabled; it is never sampled (all uses guarded).
- Narrow datetime catches to const std::exception& with diagnosable rethrow or documented default. - Split cuda capability-cache catch so noexcept degradation paths are explicit, not silent.
- Sync stale test counts (22/29/31 -> 49/49) and backend claims across README and docs. - Record intentional other.hpp/PQC stubs in README, API table and changelog. - CI: blocking tidy build, new clang-format check job, TSan keeps warnings-as-errors.
- Reformat the 10 files flagged by clang-format --dry-run --Werror over the CI file set. - Whitespace and line-break churn only; build clean, 49/49 green.
- Remove discarded lambda whose else-branch built a zeros array; the ad2/bd2 block below already converts correctly. - Kills the stale placeholder comment and the suppressed unused-variable warning.
- Whitney-sum euler product is exact as classes (Milnor-Stasheff 9); document the convention-0 masking caveat. - SIMD remark is deliberate design, not actionable work; point at NP_ENABLE_SLEEF.
- Running ctest from the source root drops Testing/Temporary logs; keep status clean.
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
1 participant
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.
No description provided.