add README.md to interfaces

CHANGELOG.md

@@ -9,9 +9,11 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ### Added
 * Added Hermitian FFT functions to SciPy interface `mkl_fft.interfaces.scipy_fft`: `hfft`, `ihfft`, `hfftn`, `ihfftn`, `hfft2`, and `ihfft2` [gh-161](https://github.com/IntelPython/mkl_fft/pull/161)
 * Added support for `out` kwarg to all FFT functions in `mkl_fft` and `mkl_fft.interfaces.numpy_fft` [gh-157](https://github.com/IntelPython/mkl_fft/pull/157)
+* Added `fftfreq`, `fftshift`, `ifftshift`, and `rfftfreq` to both NumPy and SciPy interfaces [gh-179](https://github.com/IntelPython/mkl_fft/pull/179)
 
 ### Changed
-* NumPy interface `mkl_fft.interfaces.numpy_fft` is aligned with numpy-2.* [gh-139](https://github.com/IntelPython/mkl_fft/pull/139), [gh-157](https://github.com/IntelPython/mkl_fft/pull/157)
+* NumPy interface `mkl_fft.interfaces.numpy_fft` is aligned with numpy-2.x.x [gh-139](https://github.com/IntelPython/mkl_fft/pull/139), [gh-157](https://github.com/IntelPython/mkl_fft/pull/157)
+* To set `mkl_fft` as the backend for SciPy is only possible through `mkl_fft.interfaces.scipy_fft` [gh-179](https://github.com/IntelPython/mkl_fft/pull/179)
 
 ## [1.3.14] (04/10/2025)
 
@@ -86,24 +88,6 @@ transform improves multi-core utilization which may offset the performance loss
 
 Added `scipy.fft` backend, see #42. Fixed #46.
 
-
-```python
->>> import numpy as np, mkl_fft, mkl_fft._scipy_fft as mkl_be, scipy, scipy.fft, mkl
-
->>> mkl.verbose(1)
-# True
-
->>> x = np.random.randn(8*7).reshape((7, 8))
->>> with scipy.fft.set_backend(mkl_be, only=True):
->>>     ff = scipy.fft.fft2(x, workers=4)
->>> ff2 = scipy.fft.fft2(x)
-# MKL_VERBOSE Intel(R) MKL 2020.0 Product build 20191102 for Intel(R) 64 architecture Intel(R) Advanced Vector Extensions 2 (Intel(R) AVX2) enabled processors, Lnx 2.40GHz intel_thread
-# MKL_VERBOSE FFT(drfo7:8:8x8:1:1,bScale:0.0178571,tLim:1,desc:0x5629ad31b800) 24.85ms CNR:OFF Dyn:1 FastMM:1 TID:0  NThr:16,FFT:4
-
->>> np.allclose(ff, ff2)
-# True
-```
-
 ## [1.0.15]
 
 Changed tests to not compare against numpy fft, as this broke due to renaming of `np.fft.pocketfft` to

README.md

@@ -43,34 +43,45 @@ This eliminates the need to copy input array contiguously into an intermediate b
 
 `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
+More details can be found in [SciPy 2017 conference proceedings](https://github.com/scipy-conference/scipy_proceedings/tree/2017/papers/oleksandr_pavlyk).
 
 ---
 
-`mkl_fft` implements the following functions:
+The `mkl_fft` package offers interfaces that act as drop-in replacements for equivalent functions in NumPy and SciPy. Learn more about these interfaces [here](https://github.com/IntelPython/mkl_fft/blob/master/mkl_fft/interfaces/README.md).
+
+While using these interfaces is the easiest way to leverage `mk_fft`, one can also use `mkl_fft` directly with the following FFT functions:
 
 ### complex-to-complex (c2c) transforms:
 
-`fft(x, n=None, axis=-1, overwrite_x=False, fwd_scale=1.0, out=out)` - 1D FFT, similar to `scipy.fft.fft`
+`fft(x, n=None, axis=-1, overwrite_x=False, fwd_scale=1.0, out=None)` - 1D FFT, similar to `scipy.fft.fft`
 
-`fft2(x, s=None, axes=(-2, -1), overwrite_x=False, fwd_scale=1.0, out=out)` - 2D FFT, similar to `scipy.fft.fft2`
+`fft2(x, s=None, axes=(-2, -1), overwrite_x=False, fwd_scale=1.0, out=None)` - 2D FFT, similar to `scipy.fft.fft2`
 
-`fftn(x, s=None, axes=None, overwrite_x=False, fwd_scale=1.0, out=out)` - ND FFT, similar to `scipy.fft.fftn`
+`fftn(x, s=None, axes=None, overwrite_x=False, fwd_scale=1.0, out=None)` - ND FFT, similar to `scipy.fft.fftn`
 
 and similar inverse FFT (`ifft*`) functions.
 
 ### real-to-complex (r2c) and complex-to-real (c2r) transforms:
 
-`rfft(x, n=None, axis=-1, fwd_scale=1.0, out=out)` - r2c 1D FFT, similar to `numpy.fft.rfft`
+`rfft(x, n=None, axis=-1, fwd_scale=1.0, out=None)` - r2c 1D FFT, similar to `numpy.fft.rfft`
 
-`rfft2(x, s=None, axes=(-2, -1), fwd_scale=1.0, out=out)` - r2c 2D FFT, similar to `numpy.fft.rfft2`
+`rfft2(x, s=None, axes=(-2, -1), fwd_scale=1.0, out=None)` - r2c 2D FFT, similar to `numpy.fft.rfft2`
 
-`rfftn(x, s=None, axes=None, fwd_scale=1.0, out=out)` - r2c ND FFT, similar to `numpy.fft.rfftn`
+`rfftn(x, s=None, axes=None, fwd_scale=1.0, out=None)` - r2c ND FFT, similar to `numpy.fft.rfftn`
 
 and similar inverse c2r FFT (`irfft*`) functions.
 
-The package also provides `mkl_fft.interfaces.numpy_fft` and `mkl_fft.interfaces.scipy_fft` interfaces which provide drop-in replacements for equivalent functions in NumPy and SciPy, respectively.
+The following example shows how to use `mkl_fft` for calculating a 1D FFT.
+
+```python
+import numpy, mkl_fft
+a = numpy.random.randn(10) + 1j*numpy.random.randn(10)
+
+mkl_res = mkl_fft.fft(a)
+np_res = numpy.fft.fft(a)
+numpy.allclose(mkl_res, np_res)
+# True
+```
 
 ---
 

mkl_fft/interfaces/README.md

@@ -0,0 +1,104 @@
+# Interfaces
+The `mkl_fft` package provides interfaces that serve as drop-in replacements for equivalent functions in NumPy and SciPy.
+
+---
+
+## NumPy interface - `mkl_fft.interfaces.numpy_fft`
+
+This interface is a drop-in replacement for the [`numpy.fft`](https://numpy.org/devdocs/reference/routines.fft.html) module and includes **all** the functions available there:
+
+* complex-to-complex FFTs: `fft`, `ifft`, `fft2`, `ifft2`, `fftn`, `ifftn`.
+
+* real-to-complex and complex-to-real FFTs: `rfft`, `irfft`, `rfft2`, `irfft2`, `rfftn`, `irfftn`.
+
+* Hermitian FFTs: `hfft`, `ihfft`.
+
+* Helper routines: `fftfreq`, `rfftfreq`, `fftshift`, `ifftshift`. These routines serve as a fallback to the NumPy implementation and are included for completeness.
+
+The following example shows how to use this interface for calculating a 1D FFT.
+
+```python
+import numpy
+import mkl_fft.interfaces.numpy_fft as numpy_fft
+
+a = numpy.random.randn(10) + 1j*numpy.random.randn(10)
+
+mkl_res = numpy_fft.fft(a)
+np_res = numpy.fft.fft(a)
+numpy.allclose(mkl_res, np_res)
+# True
+```
+
+---
+
+## SciPy interface - `mkl_fft.interfaces.scipy_fft`
+This interface is a drop-in replacement for the [`scipy.fft`](https://scipy.github.io/devdocs/reference/fft.html) module and includes **subset** of the functions available there:
+
+* complex-to-complex FFTs: `fft`, `ifft`, `fft2`, `ifft2`, `fftn`, `ifftn`.
+
+* real-to-complex and complex-to-real FFTs: `rfft`, `irfft`, `rfft2`, `irfft2`, `rfftn`, `irfftn`.
+
+* Hermitian FFTs: `hfft`, `ihfft`, `hfft2`, `ihfft2`, `hfftn`, `ihfftn`.
+
+* Helper functions: `fftshift`, `ifftshift`, `fftfreq`, `rfftfreq`, `set_workers`, `get_workers`. All of these functions, except for `set_workers` and `get_workers`, serve as a fallback to the SciPy implementation and are included for completeness.
+
+The following example shows how to use this interface for calculating a 1D FFT.
+
+```python
+import numpy, scipy
+import mkl_fft.interfaces.scipy_fft as scipy_fft
+
+a = numpy.random.randn(10) + 1j * numpy.random.randn(10)
+
+mkl_res = scipy_fft.fft(a)
+sp_res = scipy.fft.fft(a)
+numpy.allclose(mkl_res, sp_res)
+# True
+```
+
+---
+
+### Registering `mkl_fft` as the FFT backend for SciPy
+
+`mkl_fft.interfaces.scipy_fft` can be registered as a backend for SciPy. To learn more about how to control the backend [see the SciPy documentation](https://docs.scipy.org/doc/scipy/reference/fft.html#backend-control). The following example shows how to set `mkl_fft` as the FFT backend for SciPy using a context manager.
+
+```python
+import numpy, scipy, mkl
+import mkl_fft.interfaces.scipy_fft as mkl_backend
+x = numpy.random.randn(56).reshape(7, 8)
+
+# turning on verbosity to show `mkl_fft` is used as the SciPy backend
+mkl.verbose(1)
+# True
+
+with scipy.fft.set_backend(mkl_backend, only=True):
+    mkl_res = scipy.fft.fft2(x, workers=4)  # Calls `mkl_fft` backend
+# MKL_VERBOSE oneMKL 2024.0 Update 2 Patch 2 Product build 20240823 for Intel(R) 64 architecture Intel(R) Advanced Vector Extensions 512 (Intel(R) AVX-512) with support for INT8, BF16, FP16 (limited) instructions, and Intel(R) Advanced Matrix Extensions (Intel(R) AMX) with INT8 and BF16, Lnx 2.00GHz intel_thread
+# MKL_VERBOSE FFT(drfo7:8:8x8:1:1,input_strides:{0,8,1},output_strides:{0,8,1},bScale:0.0178571,tLim:1,unaligned_output,desc:0x557affb60d40) 36.11us CNR:OFF Dyn:1 FastMM:1 TID:0  NThr:4
+
+sp_res = scipy.fft.fft2(x, workers=4)  # Calls default SciPy backend
+numpy.allclose(mkl_res, sp_res)
+# True
+```
+
+The previous example was only for illustration purposes. In practice, there is no added benefit to defining a backend and calculating FFT, since this functionality is already accessible through the scipy interface, as shown earlier.
+To demonstrate the advantages of using `mkl_fft` as a backend, the following example compares the timing of `scipy.signal.fftconvolve` using the default SciPy backend versus the `mkl_fft` backend on an Intel® Xeon® CPU.
+
+```python
+import numpy, scipy
+import mkl_fft.interfaces.scipy_fft as mkl_backend
+import timeit
+shape = (4096, 2048)
+a = numpy.random.randn(*shape) + 1j*numpy.random.randn(*shape)
+b = numpy.random.randn(*shape) + 1j*numpy.random.randn(*shape)
+
+t1 = timeit.timeit(lambda: scipy.signal.fftconvolve(a, b), number=10)
+print(f"Time with scipy.fft default backend: {t1:.1f} seconds")
+# Time with scipy.fft default backend: 51.9 seconds
+
+with scipy.fft.set_backend(mkl_backend, only=True):
+    t2 = timeit.timeit(lambda: scipy.signal.fftconvolve(a, b), number=10)
+
+print(f"Time with OneMKL FFT backend installed: {t2:.1f} seconds")
+# Time with MKL FFT backend installed: 9.1 seconds
+```

mkl_fft/interfaces/_scipy_fft.py

@@ -63,47 +63,9 @@
     "ihfftn",
     "get_workers",
     "set_workers",
-    "DftiBackend",
 ]
 
 
-__doc__ = """
-This module implements interfaces mimicking `scipy.fft` module.
-
-It also provides DftiBackend class which can be used to set mkl_fft to be used
-via `scipy.fft` namespace.
-
-:Example:
-    import scipy.fft
-    import mkl_fft.interfaces._scipy_fft as mkl_be
-    # Set mkl_fft to be used as backend of SciPy's FFT functions.
-    scipy.fft.set_global_backend(mkl_be)
-"""
-
-
-__ua_domain__ = "numpy.scipy.fft"
-
-
-def __ua_function__(method, args, kwargs):
-    """Fetch registered UA function."""
-    fn = globals().get(method.__name__, None)
-    if fn is None:
-        return NotImplemented
-    return fn(*args, **kwargs)
-
-
-class DftiBackend:
-    __ua_domain__ = "numpy.scipy.fft"
-
-    @staticmethod
-    def __ua_function__(method, args, kwargs):
-        """Fetch registered UA function."""
-        fn = globals().get(method.__name__, None)
-        if fn is None:
-            return NotImplemented
-        return fn(*args, **kwargs)
-
-
 class _cpu_max_threads_count:
     def __init__(self):
         self.cpu_count = None

mkl_fft/interfaces/numpy_fft.py

@@ -24,4 +24,44 @@
 # 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_fft import *
+# Added for completing the namespaces
+from numpy.fft import fftfreq, fftshift, ifftshift, rfftfreq
+
+# pylint: disable=no-name-in-module
+from ._numpy_fft import (
+    fft,
+    fft2,
+    fftn,
+    hfft,
+    ifft,
+    ifft2,
+    ifftn,
+    ihfft,
+    irfft,
+    irfft2,
+    irfftn,
+    rfft,
+    rfft2,
+    rfftn,
+)
+
+__all__ = [
+    "fft",
+    "ifft",
+    "fft2",
+    "ifft2",
+    "fftn",
+    "ifftn",
+    "rfft",
+    "irfft",
+    "rfft2",
+    "irfft2",
+    "rfftn",
+    "irfftn",
+    "hfft",
+    "ihfft",
+    "fftshift",
+    "ifftshift",
+    "fftfreq",
+    "rfftfreq",
+]

mkl_fft/interfaces/scipy_fft.py

@@ -24,4 +24,68 @@
 # 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 ._scipy_fft import *
+
+# Added for completing the namespaces
+from scipy.fft import fftfreq, fftshift, ifftshift, rfftfreq
+
+# pylint: disable=no-name-in-module
+from ._scipy_fft import (
+    fft,
+    fft2,
+    fftn,
+    get_workers,
+    hfft,
+    hfft2,
+    hfftn,
+    ifft,
+    ifft2,
+    ifftn,
+    ihfft,
+    ihfft2,
+    ihfftn,
+    irfft,
+    irfft2,
+    irfftn,
+    rfft,
+    rfft2,
+    rfftn,
+    set_workers,
+)
+
+__all__ = [
+    "fft",
+    "ifft",
+    "fft2",
+    "ifft2",
+    "fftn",
+    "ifftn",
+    "rfft",
+    "irfft",
+    "rfft2",
+    "irfft2",
+    "rfftn",
+    "irfftn",
+    "hfft",
+    "ihfft",
+    "hfft2",
+    "ihfft2",
+    "hfftn",
+    "ihfftn",
+    "fftshift",
+    "ifftshift",
+    "fftfreq",
+    "rfftfreq",
+    "get_workers",
+    "set_workers",
+]
+
+
+__ua_domain__ = "numpy.scipy.fft"
+
+
+def __ua_function__(method, args, kwargs):
+    """Fetch registered UA function."""
+    fn = globals().get(method.__name__, None)
+    if fn is None:
+        return NotImplemented
+    return fn(*args, **kwargs)

mkl_fft/tests/test_interfaces.py

@@ -156,13 +156,21 @@ def test_scipy_fft_arg_validate():
 
 
 @pytest.mark.parametrize(
-    "func",
-    [mfi.scipy_fft.rfft2, mfi.numpy_fft.rfft2],
-    ids=["scipy", "numpy"],
+    "func", [mfi.scipy_fft.rfft2, mfi.numpy_fft.rfft2], ids=["scipy", "numpy"]
 )
 def test_axes(func):
     x = np.arange(24.0).reshape(2, 3, 4)
     res = func(x, axes=(1, 2))
     exp = np.fft.rfft2(x, axes=(1, 2))
     tol = 64 * np.finfo(np.float64).eps
     assert np.allclose(res, exp, atol=tol, rtol=tol)
+
+
+@pytest.mark.parametrize(
+    "interface", [mfi.scipy_fft, mfi.numpy_fft], ids=["scipy", "numpy"]
+)
+@pytest.mark.parametrize(
+    "func", ["fftshift", "ifftshift", "fftfreq", "rfftfreq"]
+)
+def test_interface_helper_functions(interface, func):
+    assert hasattr(interface, func)