Release v3.4.1

[shehzan10]

v3.4.1

Installers

• Installers for Linux, OS X and Windows
  • CUDA backend now uses CUDA 8.0.
  • Uses Intel MKL 2017.
  • CUDA Compute 2.x (Fermi) is no longer compiled into the library.
• Installer for OS X
  • The libraries shipping in the OS X Installer are now compiled with Apple
    Clang v7.3.1 (previously v6.1.0).
  • The OS X version used is 10.11.6 (previously 10.10.5).
• Installer for Jetson TX1 / Tegra X1
  • Requires JetPack for L4T 2.3
    (containing Linux for Tegra r24.2 for TX1).
  • CUDA backend now uses CUDA 8.0 64-bit.
  • Using CUDA's cusolver instead of CPU fallback.
  • Uses OpenBLAS for CPU BLAS.
  • All ArrayFire libraries are now 64-bit.

Improvements

• Add [sparse array](ref sparse_func) support to \ref af::eval().
  ¹
• Add OpenCL-CPU fallback support for sparse \ref af::matmul() when running on
  a unified memory device. Uses MKL Sparse BLAS.
• When using CUDA libdevice, pick the correct compute version based on device.
  ¹
• OpenCL FFT now also supports prime factors 7, 11 and 13.
  ^{1
  2}

Bug Fixes

• Allow CUDA libdevice to be detected from custom directory.
• Fix aarch64 detection on Jetson TX1 64-bit OS.
  ¹
• Add missing definition of af_set_fft_plan_cache_size in unified backend.
  ¹
• Fix intial values for \ref af::min() and \ref af::max() operations.
  ^{1
  2}
• Fix distance calculation in \ref af::nearestNeighbour for CUDA and OpenCL backend.
  ^{1
  2}
• Fix OpenCL bug where scalars where are passed incorrectly to compile options.
  ¹
• Fix bug in \ref af::Window::surface() with respect to dimensions and ranges.
  ¹
• Fix possible double free corruption in \ref af_assign_seq().
  ¹
• Add missing eval for key in \ref af::scanByKey in CPU backend.
  ¹
• Fixed creation of sparse values array using \ref AF_STORAGE_COO.
  ^{1
  1}

Examples

• Add a [Conjugate Gradient solver example](ref benchmarks/cg.cpp)
  to demonstrate sparse and dense matrix operations.
  ¹

CUDA Backend

• When using CUDA 8.0,
  compute 2.x are no longer in default compute list.
  • This follows CUDA 8.0
    deprecating computes 2.x.
  • Default computes for CUDA 8.0 will be 30, 50, 60.
• When using CUDA pre-8.0, the default selection remains 20, 30, 50.
• CUDA backend now uses -arch=sm_30 for PTX compilation as default.
  • Unless compute 2.0 is enabled.

Known Issues

• \ref af::lu() on CPU is known to give incorrect results when built run on
  OS X 10.11 or 10.12 and compiled with Accelerate Framework.
  ¹
  • Since the OS X Installer libraries uses MKL rather than Accelerate
    Framework, this issue does not affect those libraries.

[skip arrayfire ci]