Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Nov 14, 2025

• fixed incorrect memory stride when deriving a vImage buffer (affecting convolve operator, possible crash/corruption)

  rsebbe

  committed
  1dddde7

  Copy full SHA for 1dddde7

Commits on Aug 1, 2025

• Float16 update: upped platforms (SPM) for easier Float16 handling, removed Float.Element for *OS26 compatibility, exposed arrays (vector, matrix, tensor) as native Float16 where available with UInt…

  Show description for 5db7c0e
  rsebbe

  committed
  5db7c0e

  Copy full SHA for 5db7c0e

Commits on Feb 19, 2025

• added minimum/maximum for arrays, multiplyAdd (+ required 4-way linear storage access).

  rsebbe

  committed
  66ca4a6

  Copy full SHA for 66ca4a6

Commits on Oct 8, 2024

• Tensor ops: added axis flipping, permutation, concatenation. Fixed slicing with negative step (range error). Exposed n. slicing shortcuts.

  rsebbe

  committed
  977373c

  Copy full SHA for 977373c

Commits on Jul 9, 2024

• Fixes Xcode 16 compilation error

  bunuelcubosoto

  committed
  1725f64

  Copy full SHA for 1725f64

Commits on Sep 22, 2023

• fix deprecated method renamed "update(from:count:)"

  

  benoit

  committed
  42bbeb3

  Copy full SHA for 42bbeb3

Commits on Feb 14, 2023

• fixed compilation errors (watch OS / numeric casts). Added matrix casting to tensor with shared data.

  rsebbe

  committed
  2b95d73

  Copy full SHA for 2b95d73

Commits on Jun 15, 2022

• fix build with Xcode 14.0 beta 1

  benoitsan

  committed
  3d23f6d

  Copy full SHA for 3d23f6d

Commits on Nov 26, 2021

• added C-based generic in-place flipping.

  rsebbe

  committed
  c5c02c0

  Copy full SHA for c5c02c0

Commits on Sep 22, 2020

• iterator experiements + Xcode 12 warnings fixes.

  rsebbe

  committed
  9d909e9

  Copy full SHA for 9d909e9

Commits on Aug 17, 2020

• added 'indexes' api to StorageAccess types. Memory traversal variants for speed testing [in set(from:)], including C implementation.

  rsebbe

  committed
  5db674e

  Copy full SHA for 5db674e

Commits on Feb 10, 2020

• support for negative index (NIndex) for all dimensional types. Tensor axis insertion (n.newaxis) and more capable axis operation parsing (rough implementation, to improve). Introduced n.all notatio…

  Show description for 905c774
  rsebbe

  committed
  905c774

  Copy full SHA for 905c774

Commits on Feb 8, 2020

• access as vImage buffers, Float16 API (through UInt16), API tuning (dimension → rank, indexRange → range), NError definition, conversion API (interleave4/detinterleave4, Float16 <-> Float), externa…

  Show description for 5ac3325
  rsebbe

  committed
  5ac3325

  Copy full SHA for 5ac3325

Commits on Dec 13, 2019

• factored matrix/vector APIs into ndarray (single implementation). Reorganized NOps code into separate files for ndarray/matrix/vector. Factory funcs have been moved from Numerics → Type for easier …

  Show description for 2d2c498
  rsebbe

  committed
  2d2c498

  Copy full SHA for 2d2c498

Commits on Dec 2, 2019

• README style + link update.

  rsebbe

  authored
  d03cf1a

  Copy full SHA for d03cf1a

• added slice image in README

  rsebbe

  authored
  d808f0e

  Copy full SHA for d808f0e

• prepared for public release.

  rsebbe

  committed
  9a6aba0

  Copy full SHA for 9a6aba0

Commits on Sep 9, 2019

• fix for older swift version.

  rsebbe

  committed
  67e1266

  Copy full SHA for 67e1266

Commits on Sep 8, 2019

• move a few operators to NDArray (for NTensor availability). Added corresponding tests.

  rsebbe

  committed
  c588e64

  Copy full SHA for c588e64

• added initial tensor implementation. Removed NFloatingPoint file, moved to NValue for type centralization.

  rsebbe

  committed
  c5719d7

  Copy full SHA for c5719d7

Commits on Jun 28, 2019

• NSlice.all substitution

  rsebbe

  committed
  60c14d8

  Copy full SHA for 60c14d8

Commits on Jun 16, 2019

• retrofix for swift 5.0

  rsebbe

  committed
  d394b6a

  Copy full SHA for d394b6a

• introducing unbounded slice (standalone ~), subscript update

  rsebbe

  committed
  e8a88ee

  Copy full SHA for e8a88ee

Commits on Jun 14, 2019

• moved to SPM

  rsebbe

  committed
  a328f4f

  Copy full SHA for a328f4f

Commits on Jun 6, 2019

• better genericity, shape synthesis

  rsebbe

  committed
  ae6e0bf

  Copy full SHA for ae6e0bf

Commits on Jun 5, 2019

• factored mask generation (comparator) into NDArray. Added NDimensionalIndex protocol for more genericity.

  rsebbe

  committed
  2fa2175

  Copy full SHA for 2fa2175

Commits on Jun 3, 2019

• moved storage abstraction into NDArray for broader capabilities implemented directly in the protocol. Factored "setter".

  rsebbe

  committed
  2710810

  Copy full SHA for 2710810

Commits on May 31, 2019

• removed factored code.

  rsebbe

  committed
  7188fea

  Copy full SHA for 7188fea

• progressively moving to more genericity for Matrix/Vector to factor more common features directly into NDArray.

  rsebbe

  committed
  3411b71

  Copy full SHA for 3411b71

• reshaping, flattening, reinterpretation (vector - matrix). Concept of "coalesceable" arrays. API naming coherence changes. simplified floating point handling (using BinaryFloatingPoint). Added more…

  Show description for 0784572
  rsebbe

  committed
  0784572

  Copy full SHA for 0784572

Commits on May 27, 2019

• New abstractions in NDimensionalType: size, init, deriving. Testing factorization of element-wise op (subtracting for vectors & matrix). Added new setter.

  rsebbe

  committed
  372d27a

  Copy full SHA for 372d27a

Commits on Apr 29, 2019

• seedable random api (based on Mersenne Twister), generic NativeIndex / NativeIndexRange api for dimensional types.

  rsebbe

  committed
  c757e96

  Copy full SHA for c757e96

Commits on Apr 20, 2019

• Swift 5 update: new hash api, unsafe pointer. Bool.yes/.no for valid unused code.

  rsebbe

  committed
  f58f61f

  Copy full SHA for f58f61f

• fix warnings

  

  Benoit

  authored and

  rsebbe

  committed
  6cddeb6

  Copy full SHA for 6cddeb6

• CeedNumerics: api udpate (compact), added vImage-based convolution (Float only).

  rsebbe

  committed
  dc888b9

  Copy full SHA for dc888b9

Pagination

Previous

Next