TST: Save long double reference data in .mat file

scipy/fft/_pocketfft/tests/test_real_transforms.py

@@ -7,6 +7,7 @@
     assert_array_almost_equal, assert_equal, assert_allclose)
 import pytest
 from pytest import raises as assert_raises
+from scipy.io import loadmat
 
 from scipy.fft._pocketfft.realtransforms import (
     dct, idct, dst, idst, dctn, idctn, dstn, idstn)
@@ -31,13 +32,18 @@ def get_reference_data():
     #    * for every type (1, 2, 3, 4) and every size, the array dct_type_size
     #    contains the output of the DCT applied to the input np.linspace(0, size-1,
     #    size)
-    FFTWDATA_LONGDOUBLE = np.load(join(fftpack_test_dir, 'fftw_longdouble_ref.npz'))
     FFTWDATA_DOUBLE = np.load(join(fftpack_test_dir, 'fftw_double_ref.npz'))
     FFTWDATA_SINGLE = np.load(join(fftpack_test_dir, 'fftw_single_ref.npz'))
     FFTWDATA_SIZES = FFTWDATA_DOUBLE['sizes']
-
     assert len(FFTWDATA_SIZES) == FFTWDATA_COUNT
 
+    # not binary compatible on 32-bit systems so cannot use np.load
+    FFTWDATA_LONGDOUBLE = loadmat(
+        join(fftpack_test_dir, 'fftw_longdouble_ref.mat'))
+    FFTWDATA_LONGDOUBLE = {k: v.reshape((-1,))
+                           for k,v in FFTWDATA_LONGDOUBLE.items()
+                           if isinstance(v, np.ndarray)}
+
     ref = {
         'FFTWDATA_LONGDOUBLE': FFTWDATA_LONGDOUBLE,
         'FFTWDATA_DOUBLE': FFTWDATA_DOUBLE,
@@ -46,7 +52,6 @@ def get_reference_data():
         'X': X,
         'Y': Y
     }
-
     globals()['__reference_data'] = ref
     return ref
 

scipy/fftpack/tests/gen_fftw_ref.py

@@ -3,6 +3,7 @@
 from subprocess import Popen, PIPE, STDOUT
 
 import numpy as np
+from scipy.io import savemat
 
 SZ = [2, 3, 4, 8, 12, 15, 16, 17, 32, 64, 128, 256, 512, 1024]
 
@@ -62,7 +63,7 @@ def gen_data(dt):
 
 # generate long double precision data
 data = gen_data(np.float128)
-filename = 'fftw_longdouble_ref'
+filename = 'fftw_longdouble_ref.mat'
 # Save ref data into npz format
 d = {'sizes': SZ}
 for type in [1, 2, 3, 4]:
@@ -73,4 +74,4 @@ def gen_data(dt):
 for type in [5, 6, 7, 8]:
     for sz in SZ:
         d['dst_%d_%d' % (type-4, sz)] = data[type][sz]
-np.savez(filename, **d)
+savemat(filename, d)