New Features/Functions

• fp16 - half precision floating point support has been added -#221

• Confidence Connected ComponentsconfidenceCC -#221

• Deconvolutions -#221

  • iterativeDeconv
  • inverseDeconv

• Reduction using keys -#221

  • allTrueByKey
  • anyTrueByKey
  • countByKey
  • maxByKey
  • minByKey
  • productByKey
  • sumByKey

• Neural network based convolution and gradient functions -#221

  • convolve2GradientNN
  • convolve2NN

• Support for uniform ranges in approx1 and approx2 functions -#234

  • approx1
  • approx1_uniform
  • approx2
  • approx2_uniform

• Array class methods -#233

  • Array.logical_and
  • Array.logical_not
  • Array.logical_or

• Window.set_axes_label_format -#221

• gemm -#234

• meanvar -#221

• pad -#221

• pinverse -#221

• rsqrt -#221

• set_cublas_mode -#221

• maxRagged -#221

• New Examples

  • Logistic Regression -#203
  • FAST feature detection -#190
  • Harris feature detection -#190
  • Template Matching -#190
  • SUSAN feature detection -#190

Breaking APIs

• cov -#221
• var -#221
• stdev -#221

Fixes

• Fixed wrapper validations increate_sparse_from_host -#198
• Added a workaround forbench_cg example on less capable GPUs -#200
• Fixed missing info inArray.device_ptr function documentation -#210
• Corrected invert operation to use non-in-place bit wise inversion -#228

• Feature parity with ArrayFire v3.6. Refer to therelease notes for more information regarding upstream library improvements in v3.6.
  • anisotropic_diffusion(): Anisotropic diffusion filter.
  • topk(): Returns top-K elements given an array.
• Bug fixes:
  • Fixedsift() andgloh(), which were improperly calling the library.
• Enhancements:
  • Addedlen() method, which returnsarray.elements().
• Documentation:
  • Documented statistics API.
  • Correctedsign() documentation.
  • Modifiedhelloworld example to match C++ lib.

• Bug fixes when using v3.5 of arrayfire libs + graphics

Includes fix for arrayfire.canny.

• Feature parity with ArrayFire 3.5.

  • canny: Canny Edge detector
  • Array.scalar: Return the first element of the array
  • dot: Now support option to return scalar
  • print_mem_info: Prints memory being used / locked by arrayfire memory manager.
  • Array.allocated: Returs the amount of memory allocated for the given buffer.
  • set_fft_plan_cache_size: Sets the size of the fft plan cache.

• Bug Fixes:

  • sort_by_key had key and value flipped in documentation.

• Improvements and bugfixes from upstream include:

  • CUDA backend uses nvrtc instead of nvvm
  • Performance improvements to arrayfire.reorder
  • Faster unified backend
  • You can find more information at arrayfire'srelease notes

• Bugfix: Fixes typo inapprox1.
• Bugfix: Fixes typo inhamming_matcher andnearest_neighbour.
• Bugfix: Added necessary copy and lock mechanisms in interop.py.
• Example / Benchmark: New conjugate gradient benchmark.
• Feature: Added support to create arrayfire arrays from numba.
• Behavior change: af.print() only prints full arrays for smaller sizes.

• Fixing memory leak in array creation.
• Supporting 16 bit integer types in interop.

• Feature parity with ArrayFire 3.4 libs
  • Sparse matrix support
    • create_sparse
    • create_sparse_from_dense
    • create_sparse_from_host
    • convert_sparse_to_dense
    • convert_sparse
    • sparse_get_info
    • sparse_get_nnz
    • sparse_get_values
    • sparse_get_row_idx
    • sparse_get_col_idx
    • sparse_get_storage
  • Random Engine support
    • Three new random engines,RANDOM_ENGINE.PHILOX,RANDOM_ENGINE.THREEFRY, andRANDOM_ENGINE.MERSENNE.
    • randu andrandn now accept an additional engine parameter.
    • set_default_random_engine_type
    • get_default_random_engine
  • New functions
    • scan
    • scan_by_key
    • clamp
    • medfilt1
    • medfilt2
    • moments
    • get_size_of
    • get_manual_eval_flag
    • set_manual_eval_flag
  • Behavior changes
    • eval now supports fusing kernels.
  • Graphics updates
    • plot updated to take new parameters.
    • plot2 added.
    • scatter updated to take new parameters.
    • scatter2 added.
    • vector_field added.
    • set_axes_limits added.
• Bug fixes
  • ArrayFire now has higher priority when numpy for mixed operations.¹²
  • Numpy interoperability issues on Widnows.¹
  • Switch to a working backend by default.¹
  • Fixed incorrect behavior for Hermitian transpose and QR.¹
  • array[0:0] now returns empty arrays.¹
• Further Improvements from upstream can be read in thearrayfire release notes.

• Adding 16 bit integer support
• Adding support for sphinx documentation

v3.3.20160516

• Bugfix: Increase arrayfire's priority over numpy for mixed operations
• Added new library functions
  • get_backend returns backend name