Initial commit of mkl_fft

mkl_fft is a Python interface to Intel (R) Math Kernel Library functions for
performing discrete Fourier transforms, supporting working with NumPy's
N-dimensional arrays.

Author: oleksandr-pavlyk

LICENSE.txt

@@ -0,0 +1,24 @@
+Copyright (c) 2017, Intel Corporation
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+    * Redistributions of source code must retain the above copyright notice,
+      this list of conditions and the following disclaimer.
+    * Redistributions in binary form must reproduce the above copyright
+      notice, this list of conditions and the following disclaimer in the
+      documentation and/or other materials provided with the distribution.
+    * Neither the name of Intel Corporation nor the names of its contributors
+      may be used to endorse or promote products derived from this software
+      without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

README.md

@@ -0,0 +1,54 @@
+## ``mkl_fft`` --  a NumPy-based Python interface to Intel (R) MKL FFT functionality
+
+`mkl_fft` started as a part of Intel (R) Distribution for Python* optimizations to NumPy, and is now being released 
+as a stand-alone package. It can be installed into conda environment using 
+
+```
+   conda install -c intel mkl_fft
+```
+
+---
+
+Since MKL FFT supports performing discrete Fourier transforms over non-contiguously laid out arrays, MKL can be directly
+used on any well-behaved floating point array with no internal overlaps for both in-place and not in-place transforms of 
+arrays in single and double floating point precision.
+
+This eliminates the need to copy input arrayy contiguously into an intermediate buffer.
+
+`mkl_fft` directly supports N-dimensional Fourier transforms.
+
+More details can be found in SciPy 2017 conference proceedings:
+     https://github.com/scipy-conference/scipy_proceedings/tree/2017/papers/oleksandr_pavlyk
+
+---
+
+It implements the following functions:
+
+### Complex transforms, similar to those in `scipy.fftpack`:
+
+`fft(x, n=None, axis=-1, overwrite_x=False)`
+
+`ifft(x, n=None, axis=-1, overwrite_x=False)`
+
+`fft2(x, shape=None, axes=(-2,-1), overwrite_x=False)`
+
+`ifft2(x, shape=None, axes=(-2,-1), overwrite_x=False)`
+
+`fftn(x, n=None, axes=None, overwrite_x=False)`
+
+`ifftn(x, n=None, axes=None, overwrite_x=False)`
+
+### Real transforms
+
+`rfft(x, n=None, axis=-1, overwrite_x=False)` - real 1D Fourier transform, like `scipy.fftpack.rfft`
+
+`rfft_numpy(x, n=None, axis=-1)` - real 1D Fourier transform, like `numpy.fft.rfft`
+
+`rfft2_numpy(x, s=None, axes=(-2,-1))` - real 2D Fourier transform, like `numpy.fft.rfft2`
+
+`rfftn_numpy(x, s=None, axes=None)` - real 2D Fourier transform, like `numpy.fft.rfftn`
+
+... and similar `irfft*` functions.
+
+
+The package also provides `mkl_fft._numpy_fft` and `mkl_fft._scipy_fft` interfaces which provide drop-in replacements for equivalent functions in NumPy and SciPy respectively.

conda-recipe/bld.bat

@@ -0,0 +1,7 @@
+@rem Remember to activate Intel Compiler, or remoe these two lines to ise Microsoft Visual Studio compiler
+
+set CC=icl
+set LD=xilink
+
+%PYTHON% setup.py build --force --compiler=intelemw install --old-and-unmanageable
+if errorlevel 1 exit 1

conda-recipe/build.sh

@@ -0,0 +1,9 @@
+#!/bin/bash -x
+
+# make sure that compiler has been sourced, if necessary
+
+if [ `uname` == Darwin ]; then
+    export MACOSX_DEPLOYMENT_TARGET=10.10
+fi
+
+CFLAGS="-I$PREFIX/include $CFLAGS" $PYTHON setup.py build --force --compiler=intelem install --old-and-unmanageable

conda-recipe/meta.yaml

@@ -0,0 +1,61 @@
+{% set version = "1.0.0" %}
+{% set buildnumber = 14 %}
+
+### If you change the iccver here, you must also set the path correctly in build.sh / bld.bat!!!
+{% set iccver = "16.0.3" %}    [unix or py3k]
+{% set iccver = "13.1.5" %}    [win and py27]
+
+package:
+    name: mkl_fft
+    version: {{ version }}
+
+source:
+    git_url: http://github.com/IntelPython/mkl_fft
+    git_tag: v{{version}}
+
+build:
+    number: {{buildnumber}}
+    features:
+      - intel
+    always_include_files:
+      - {{ SP_DIR.replace('\\', '/') if win else SP_DIR }}/mkl_fft/__init__.py
+      - {{ SP_DIR.replace('\\', '/') if win else SP_DIR }}/mkl_fft/_float16_utils.py
+      - {{ SP_DIR.replace('\\', '/') if win else SP_DIR }}/mkl_fft/_pydfti.*
+      - {{ SP_DIR.replace('\\', '/') if win else SP_DIR }}/mkl_fft/_numpy_fft.py
+      - {{ SP_DIR.replace('\\', '/') if win else SP_DIR }}/mkl_fft/_scipy_fft.py
+      - {{ SP_DIR.replace('\\', '/') if win else SP_DIR }}/mkl_fft/setup.py
+      - {{ SP_DIR.replace('\\', '/') if win else SP_DIR }}/mkl_fft/tests/test_fft1d.py
+      - {{ SP_DIR.replace('\\', '/') if win else SP_DIR }}/mkl_fft/__pycache__/*
+      - {{ SP_DIR.replace('\\', '/') if win else SP_DIR }}/mkl_fft/tests/__pycache__/*
+
+requirements:
+    build:
+      - python
+      - setuptools
+      - intelpython
+      - mkl-devel               [not nomkl]
+      - icc_rt
+      - cython
+      - numpy x.x
+    run:
+      - python
+      - mkl                     [not nomkl]
+      - icc_rt >={{iccver}}
+      - intelpython
+      - numpy x.x
+
+test:
+    commands:
+      - nosetests -v mkl_fft
+    requires:
+      - nose
+    imports:
+      - mkl_fft
+      - mkl_fft._numpy_fft
+      - mkl_fft._scipy_fft
+
+about:
+    home: http://github.com/IntelPython/mkl_fft
+    license: BSD
+    license_file: LICENSE.txt
+    summary: NumPy-based implementation of Fast Fourier Transform using Intel (R) Math Kernel Library

conda-recipe/run_test.py

@@ -0,0 +1,31 @@
+#!/usr/bin/env python
+# Copyright (c) 2017, Intel Corporation
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions are met:
+#
+#     * Redistributions of source code must retain the above copyright notice,
+#       this list of conditions and the following disclaimer.
+#     * Redistributions in binary form must reproduce the above copyright
+#       notice, this list of conditions and the following disclaimer in the
+#       documentation and/or other materials provided with the distribution.
+#     * Neither the name of Intel Corporation nor the names of its contributors
+#       may be used to endorse or promote products derived from this software
+#       without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+# DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+# FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+# DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+# SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+# CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+# OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+from ._pydfti import (fft, ifft, fft2, ifft2, fftn, ifftn, rfft, irfft,
+                      rfft_numpy, irfft_numpy, rfftn_numpy, irfftn_numpy)
+
+__all__ = ['fft', 'ifft', 'fft2', 'ifft2', 'fftn', 'ifftn', 'rfft', 'irfft',
+           'rfft_numpy', 'irfft_numpy', 'rfftn_numpy', 'irfftn_numpy']

mkl_fft/__init__.py

@@ -0,0 +1,48 @@
+#!/usr/bin/env python
+# Copyright (c) 2017, Intel Corporation
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions are met:
+#
+#     * Redistributions of source code must retain the above copyright notice,
+#       this list of conditions and the following disclaimer.
+#     * Redistributions in binary form must reproduce the above copyright
+#       notice, this list of conditions and the following disclaimer in the
+#       documentation and/or other materials provided with the distribution.
+#     * Neither the name of Intel Corporation nor the names of its contributors
+#       may be used to endorse or promote products derived from this software
+#       without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+# DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
+# FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+# DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+# SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+# CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+# OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+from numpy import half, float32, asarray, ndarray, longdouble, float64
+
+__all__ = ['__upcast_float16_array']
+
+def __upcast_float16_array(x):
+    if hasattr(x, "dtype"):
+        xdt = x.dtype
+        if xdt == half:
+            # no half-precision routines, so convert to single precision
+            return asarray(x, dtype=float32)
+        if xdt == longdouble and not xdt == float64 :
+            raise ValueError("type %s is not supported" % xdt)
+    if not isinstance(x, ndarray):
+        __x = asarray(x)
+        xdt = __x.dtype
+        if xdt == half:
+            # no half-precision routines, so convert to single precision
+            return asarray(__x, dtype=float32)
+        if xdt == longdouble and not xdt == float64:
+            raise ValueError("type %s is not supported" % xdt)
+        return __x
+    return x