0.6.0

What's Changed

• New methods and features for 0.6.0 by @henrique-borba in #64
• New routines: full, fill, positive, reciprocal, swapaxes, argmin, argmax, rollaxis, moveaxis, hstack, vstack, dstack, column_stack, concatenate, append
• NDArray::transpose now allows the user to specify permutation axes.
• NDArray::squeeze now support axis parameter.

Full Changelog: 0.5.2...0.6.0

0.5.2

What's Changed

• NDArray::poisson() 'Argument #2 ($lam) must be of type float, float g… by @SkibidiProduction in #56
• NDArray::standard_normal() 'expects at most 3 arguments' bugfix by @SkibidiProduction in #54
• GPU matmul refactoring and optimization by @SkibidiProduction in #53
• fix: destructor causing segmentation fault by @henrique-borba in #59
• fix: segmentation fault with empty arrays initialization and __toString magic method by @henrique-borba in #60
• stubs.php file for the NDArray class by @henrique-borba in #61

New Contributors

• @SkibidiProduction made their first contribution in #56

Full Changelog: 0.5.1...0.5.2

0.5.1

What's Changed

• fix: Uses cuBLAS instead of the CUDA kernel to perform matmul by @henrique-borba in #52

Full Changelog: 0.5.0...0.5.1

0.5.0

What's Changed

• NumPower 0.5.0 by @henrique-borba in #50

Full Changelog: 0.4.0...0.5.0

0.4.0

New Features

• Implementation of new manipulation methods: atleast_1d, atleast_2d atleast_3d, expand_dims and append.
• Introduction of argmin and argmax methods for CPU.
• Arithmetic operations between GPU stored NDArray and CPU stored scalars.

Bug fixes

• Fixed a problem where NDArray::transpose on GPU was causing SEGFAULT.
• Increases the stability of memory management and allocation of various components through Valgrind's detailed memcheck analyses.
• Use of an expandable buffer in the NDArrays print function, avoiding excess null characters and memory waste.
• Fixed memory leaks detected by Valgrind in all components.
• Fixed an issue where the NDArray::allclose method returned an incorrect result.
• Fixed a problem where the NDArray::sinc method returned incorrect values.
• Fixed the arginfo of the NDArray::arange method that was causing mismatch.
• Changes some tests affected by changes from this MR.
• Fixed a problem where the NDArray::allclose method accepted NDArrays stored in VRAM when there was no implementation for that.
• Fixed the NDArray::all method that used AVX2 in a suboptimal way.
• Fixed an invalid memory read when NDArray::clip received a PHP array as an argument.

Others

• General code cleaning, removal of unnecessary includes, variables and allocations.

0.3.1

What's Changed

• N-dimensional slicing and element-wise iterator by @henrique-borba in #40

Full Changelog: 0.3.0...0.3.1

0.3.0

What's Changed

• hotfix: reshape leaking memory by @henrique-borba in #36
• Serialize support for NDArray by @henrique-borba in #37
• Broadcasting updates, binominal random initializer and bug fixes by @henrique-borba in #38
• Convolve2D SciPy porting and color pallete support for GD images by @henrique-borba in #39

Full Changelog: 0.2.0...0.3.0

0.2.0

What's Changed

• GD Image loading and save fix, GPU allocation improvement and GPU slicing fixes by @henrique-borba in #34
• Broadcasting fixes and updates by @henrique-borba in #35

Full Changelog: 0.1.2...0.2.0

0.1.2

What's Changed

• fix: Fix NDArray::prod memory leak when using a PHP array as argument by @henrique-borba in #29
• style: Run astyle format (google style) by @henrique-borba in #30
• fix: Fixed double definition of envvar by @henrique-borba in #31
• Bug fix patch by @henrique-borba in #32
• NDArray_Append, slicing bug fix, AVX2 image support and AVX2 CUDA support by @henrique-borba in #33

Full Changelog: 0.1.1...0.1.2

0.1.1

What's Changed

• fix: quiet compilation warning about implicit functions by @markkimsal in #27

New Contributors

• @markkimsal made their first contribution in #27

Full Changelog: 0.1.0...0.1.1