fredRos
commented
7 years ago Turns out to be irrelevant when profiling
Closed mtazzari closed 7 years ago
fredRos
commented
7 years ago Turns out to be irrelevant when profiling
fredRos
commented
7 years ago Through git bisect I saw that commit a3e4003 breaks several unit tests related to rotix, for example test_loss. @mtazzari Could you look into fixing that? Now I'm not sure other changes I make can be tested properly.
mtazzari
commented
7 years ago Sorry, I don't get what you are trying to do. commit a3e4003 just changed the way du was computed: see line 551 in the commit page. I don't understand what rotix tests were broken and how this is an issue now. My latest commit have all the tests passing. To make things faster, we shall talk.
fredRos
commented
7 years ago on my machine with the latest master. The errors are the same as with a3e4003
python/py.test.sh -s python_package/tests/test_galario.py --gpu=0
============================= test session starts ==============================
platform linux -- Python 3.6.1, pytest-3.0.7, py-1.4.33, pluggy-0.4.0
rootdir: /home/beaujean/workspace/protoplanetary/galario/build3, inifile:
collected 33 items
python_package/tests/test_galario.py FF................FFsize:256, minuv:113.91199493408203, maxuv:19333.672265625
.size:2048, minuv:17.164953054780607, maxuv:19563.900024738796
.size:256, minuv:113.91199493408203, maxuv:19333.672265625
.size:2048, minuv:17.164953054780607, maxuv:19563.900024738796
.size:256, minuv:113.91199493408203, maxuv:19333.672265625
.size:2048, minuv:17.164953054780607, maxuv:19563.900024738796
.size:2048, minuv:17.164953231811523, maxuv:19563.899999999998
.size:2048, minuv:17.164953054780607, maxuv:19563.900024738796
.size:2048, minuv:17.164953231811523, maxuv:19563.899999999998
.size:2048, minuv:17.164953054780607, maxuv:19563.900024738796
.size:2048, minuv:17.164953231811523, maxuv:19563.899999999998
.size:2048, minuv:17.164953054780607, maxuv:19563.900024738796
.size:4096, minuv:17.164953054780607, maxuv:19563.900024738796
.
=================================== FAILURES ===================================
________________________________ test_rotix[SP] ________________________________
x_id = array([ True, True, True, True, True, True, True, True, True, True], dtype=bool)
y_id = array([False, False, False, False, False, False, False, False, False, False], dtype=bool)
hasval = '+inf'
def chk_same_position(x_id, y_id, hasval='nan'):
"""Handling nan/inf: check that x and y have the nan/inf at the same
locations."""
try:
> assert_array_equal(x_id, y_id)
../../../../.local/miniconda3/envs/galario3/lib/python3.6/site-packages/numpy/testing/utils.py:729:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
x = array([ True, True, True, True, True, True, True, True, True, True], dtype=bool)
y = array([False, False, False, False, False, False, False, False, False, False], dtype=bool)
err_msg = '', verbose = True
def assert_array_equal(x, y, err_msg='', verbose=True):
"""
Raises an AssertionError if two array_like objects are not equal.
Given two array_like objects, check that the shape is equal and all
elements of these objects are equal. An exception is raised at
shape mismatch or conflicting values. In contrast to the standard usage
in numpy, NaNs are compared like numbers, no assertion is raised if
both objects have NaNs in the same positions.
The usual caution for verifying equality with floating point numbers is
advised.
Parameters
----------
x : array_like
The actual object to check.
y : array_like
The desired, expected object.
err_msg : str, optional
The error message to be printed in case of failure.
verbose : bool, optional
If True, the conflicting values are appended to the error message.
Raises
------
AssertionError
If actual and desired objects are not equal.
See Also
--------
assert_allclose: Compare two array_like objects for equality with desired
relative and/or absolute precision.
assert_array_almost_equal_nulp, assert_array_max_ulp, assert_equal
Examples
--------
The first assert does not raise an exception:
>>> np.testing.assert_array_equal([1.0,2.33333,np.nan],
... [np.exp(0),2.33333, np.nan])
Assert fails with numerical inprecision with floats:
>>> np.testing.assert_array_equal([1.0,np.pi,np.nan],
... [1, np.sqrt(np.pi)**2, np.nan])
...
<type 'exceptions.ValueError'>:
AssertionError:
Arrays are not equal
<BLANKLINE>
(mismatch 50.0%)
x: array([ 1. , 3.14159265, NaN])
y: array([ 1. , 3.14159265, NaN])
Use `assert_allclose` or one of the nulp (number of floating point values)
functions for these cases instead:
>>> np.testing.assert_allclose([1.0,np.pi,np.nan],
... [1, np.sqrt(np.pi)**2, np.nan],
... rtol=1e-10, atol=0)
"""
assert_array_compare(operator.__eq__, x, y, err_msg=err_msg,
> verbose=verbose, header='Arrays are not equal')
../../../../.local/miniconda3/envs/galario3/lib/python3.6/site-packages/numpy/testing/utils.py:871:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
comparison = <built-in function eq>
x = array([ True, True, True, True, True, True, True, True, True, True], dtype=bool)
y = array([False, False, False, False, False, False, False, False, False, False], dtype=bool)
err_msg = '', verbose = True, header = 'Arrays are not equal', precision = 6
equal_nan = True
def assert_array_compare(comparison, x, y, err_msg='', verbose=True,
header='', precision=6, equal_nan=True):
__tracebackhide__ = True # Hide traceback for py.test
from numpy.core import array, isnan, isinf, any, all, inf
x = array(x, copy=False, subok=True)
y = array(y, copy=False, subok=True)
def safe_comparison(*args, **kwargs):
# There are a number of cases where comparing two arrays hits special
# cases in array_richcompare, specifically around strings and void
# dtypes. Basically, we just can't do comparisons involving these
# types, unless both arrays have exactly the *same* type. So
# e.g. you can apply == to two string arrays, or two arrays with
# identical structured dtypes. But if you compare a non-string array
# to a string array, or two arrays with non-identical structured
# dtypes, or anything like that, then internally stuff blows up.
# Currently, when things blow up, we just return a scalar False or
# True. But we also emit a DeprecationWarning, b/c eventually we
# should raise an error here. (Ideally we might even make this work
# properly, but since that will require rewriting a bunch of how
# ufuncs work then we are not counting on that.)
#
# The point of this little function is to let the DeprecationWarning
# pass (or maybe eventually catch the errors and return False, I
# dunno, that's a little trickier and we can figure that out when the
# time comes).
# Note: Put a lock around warning filter (comment at lock definition)
with _assert_comparison_lock, suppress_warnings() as sup:
sup.filter(DeprecationWarning, ".*==")
sup.filter(FutureWarning, ".*==")
return comparison(*args, **kwargs)
def isnumber(x):
return x.dtype.char in '?bhilqpBHILQPefdgFDG'
def chk_same_position(x_id, y_id, hasval='nan'):
"""Handling nan/inf: check that x and y have the nan/inf at the same
locations."""
try:
assert_array_equal(x_id, y_id)
except AssertionError:
msg = build_err_msg([x, y],
err_msg + '\nx and y %s location mismatch:'
% (hasval), verbose=verbose, header=header,
names=('x', 'y'), precision=precision)
raise AssertionError(msg)
try:
cond = (x.shape == () or y.shape == ()) or x.shape == y.shape
if not cond:
msg = build_err_msg([x, y],
err_msg
+ '\n(shapes %s, %s mismatch)' % (x.shape,
y.shape),
verbose=verbose, header=header,
names=('x', 'y'), precision=precision)
if not cond:
raise AssertionError(msg)
if isnumber(x) and isnumber(y):
if equal_nan:
x_isnan, y_isnan = isnan(x), isnan(y)
# Validate that NaNs are in the same place
if any(x_isnan) or any(y_isnan):
chk_same_position(x_isnan, y_isnan, hasval='nan')
x_isinf, y_isinf = isinf(x), isinf(y)
# Validate that infinite values are in the same place
if any(x_isinf) or any(y_isinf):
# Check +inf and -inf separately, since they are different
chk_same_position(x == +inf, y == +inf, hasval='+inf')
chk_same_position(x == -inf, y == -inf, hasval='-inf')
# Combine all the special values
x_id, y_id = x_isinf, y_isinf
if equal_nan:
x_id |= x_isnan
y_id |= y_isnan
# Only do the comparison if actual values are left
if all(x_id):
return
if any(x_id):
val = safe_comparison(x[~x_id], y[~y_id])
else:
val = safe_comparison(x, y)
else:
val = safe_comparison(x, y)
if isinstance(val, bool):
cond = val
reduced = [0]
else:
reduced = val.ravel()
cond = reduced.all()
reduced = reduced.tolist()
if not cond:
match = 100-100.0*reduced.count(1)/len(reduced)
msg = build_err_msg([x, y],
err_msg
+ '\n(mismatch %s%%)' % (match,),
verbose=verbose, header=header,
names=('x', 'y'), precision=precision)
if not cond:
> raise AssertionError(msg)
E AssertionError:
E Arrays are not equal
E
E (mismatch 100.0%)
E x: array([ True, True, True, True, True, True, True, True, True, True], dtype=bool)
E y: array([False, False, False, False, False, False, False, False, False, False], dtype=bool)
../../../../.local/miniconda3/envs/galario3/lib/python3.6/site-packages/numpy/testing/utils.py:796: AssertionError
During handling of the above exception, another exception occurred:
size = 1024, real_type = 'float32', tol = 0.0001
acc_lib = <module 'galario.single' from '/home/beaujean/workspace/protoplanetary/galario/build3/python/galario/single/__init__.py'>
@pytest.mark.parametrize("size, real_type, tol, acc_lib",
[(1024, 'float32', 1.e-4, g_single),
(1024, 'float64', 1.e-13, g_double)],
ids=["SP", "DP"])
def test_rotix(size, real_type, tol, acc_lib):
nsamples = 10
maxuv = 1000.
uv = pixel_coordinates(maxuv, size).astype(real_type)
udat, vdat = create_sampling_points(nsamples, maxuv/4.8)
assert len(udat) == nsamples
assert len(vdat) == nsamples
udat = udat.astype(real_type)
vdat = vdat.astype(real_type)
ui, vi = get_rotix_n(uv, uv, udat, vdat, len(uv))
ui = ui.astype(real_type)
vi = vi.astype(real_type)
ui1, vi1 = acc_lib.acc_rotix(uv, udat, vdat)
> np.testing.assert_allclose(ui1, ui, rtol=tol)
python_package/tests/test_galario.py:286:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
x_id = array([ True, True, True, True, True, True, True, True, True, True], dtype=bool)
y_id = array([False, False, False, False, False, False, False, False, False, False], dtype=bool)
hasval = '+inf'
def chk_same_position(x_id, y_id, hasval='nan'):
"""Handling nan/inf: check that x and y have the nan/inf at the same
locations."""
try:
assert_array_equal(x_id, y_id)
except AssertionError:
msg = build_err_msg([x, y],
err_msg + '\nx and y %s location mismatch:'
% (hasval), verbose=verbose, header=header,
names=('x', 'y'), precision=precision)
> raise AssertionError(msg)
E AssertionError:
E Not equal to tolerance rtol=0.0001, atol=0
E
E x and y +inf location mismatch:
E x: array([ inf, inf, inf, inf, inf, inf, inf, inf, inf, inf], dtype=float32)
E y: array([ 458.470459, 610.98407 , 365.234619, 323.449005, 555.142395,
E 307.449402, 653.842163, 376.2453 , 428.476685, 482.963196], dtype=float32)
../../../../.local/miniconda3/envs/galario3/lib/python3.6/site-packages/numpy/testing/utils.py:735: AssertionError
________________________________ test_rotix[DP] ________________________________
x_id = array([ True, True, True, True, True, True, True, True, True, True], dtype=bool)
y_id = array([False, False, False, False, False, False, False, False, False, False], dtype=bool)
hasval = '+inf'
def chk_same_position(x_id, y_id, hasval='nan'):
"""Handling nan/inf: check that x and y have the nan/inf at the same
locations."""
try:
> assert_array_equal(x_id, y_id)
../../../../.local/miniconda3/envs/galario3/lib/python3.6/site-packages/numpy/testing/utils.py:729:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
x = array([ True, True, True, True, True, True, True, True, True, True], dtype=bool)
y = array([False, False, False, False, False, False, False, False, False, False], dtype=bool)
err_msg = '', verbose = True
def assert_array_equal(x, y, err_msg='', verbose=True):
"""
Raises an AssertionError if two array_like objects are not equal.
Given two array_like objects, check that the shape is equal and all
elements of these objects are equal. An exception is raised at
shape mismatch or conflicting values. In contrast to the standard usage
in numpy, NaNs are compared like numbers, no assertion is raised if
both objects have NaNs in the same positions.
The usual caution for verifying equality with floating point numbers is
advised.
Parameters
----------
x : array_like
The actual object to check.
y : array_like
The desired, expected object.
err_msg : str, optional
The error message to be printed in case of failure.
verbose : bool, optional
If True, the conflicting values are appended to the error message.
Raises
------
AssertionError
If actual and desired objects are not equal.
See Also
--------
assert_allclose: Compare two array_like objects for equality with desired
relative and/or absolute precision.
assert_array_almost_equal_nulp, assert_array_max_ulp, assert_equal
Examples
--------
The first assert does not raise an exception:
>>> np.testing.assert_array_equal([1.0,2.33333,np.nan],
... [np.exp(0),2.33333, np.nan])
Assert fails with numerical inprecision with floats:
>>> np.testing.assert_array_equal([1.0,np.pi,np.nan],
... [1, np.sqrt(np.pi)**2, np.nan])
...
<type 'exceptions.ValueError'>:
AssertionError:
Arrays are not equal
<BLANKLINE>
(mismatch 50.0%)
x: array([ 1. , 3.14159265, NaN])
y: array([ 1. , 3.14159265, NaN])
Use `assert_allclose` or one of the nulp (number of floating point values)
functions for these cases instead:
>>> np.testing.assert_allclose([1.0,np.pi,np.nan],
... [1, np.sqrt(np.pi)**2, np.nan],
... rtol=1e-10, atol=0)
"""
assert_array_compare(operator.__eq__, x, y, err_msg=err_msg,
> verbose=verbose, header='Arrays are not equal')
../../../../.local/miniconda3/envs/galario3/lib/python3.6/site-packages/numpy/testing/utils.py:871:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
comparison = <built-in function eq>
x = array([ True, True, True, True, True, True, True, True, True, True], dtype=bool)
y = array([False, False, False, False, False, False, False, False, False, False], dtype=bool)
err_msg = '', verbose = True, header = 'Arrays are not equal', precision = 6
equal_nan = True
def assert_array_compare(comparison, x, y, err_msg='', verbose=True,
header='', precision=6, equal_nan=True):
__tracebackhide__ = True # Hide traceback for py.test
from numpy.core import array, isnan, isinf, any, all, inf
x = array(x, copy=False, subok=True)
y = array(y, copy=False, subok=True)
def safe_comparison(*args, **kwargs):
# There are a number of cases where comparing two arrays hits special
# cases in array_richcompare, specifically around strings and void
# dtypes. Basically, we just can't do comparisons involving these
# types, unless both arrays have exactly the *same* type. So
# e.g. you can apply == to two string arrays, or two arrays with
# identical structured dtypes. But if you compare a non-string array
# to a string array, or two arrays with non-identical structured
# dtypes, or anything like that, then internally stuff blows up.
# Currently, when things blow up, we just return a scalar False or
# True. But we also emit a DeprecationWarning, b/c eventually we
# should raise an error here. (Ideally we might even make this work
# properly, but since that will require rewriting a bunch of how
# ufuncs work then we are not counting on that.)
#
# The point of this little function is to let the DeprecationWarning
# pass (or maybe eventually catch the errors and return False, I
# dunno, that's a little trickier and we can figure that out when the
# time comes).
# Note: Put a lock around warning filter (comment at lock definition)
with _assert_comparison_lock, suppress_warnings() as sup:
sup.filter(DeprecationWarning, ".*==")
sup.filter(FutureWarning, ".*==")
return comparison(*args, **kwargs)
def isnumber(x):
return x.dtype.char in '?bhilqpBHILQPefdgFDG'
def chk_same_position(x_id, y_id, hasval='nan'):
"""Handling nan/inf: check that x and y have the nan/inf at the same
locations."""
try:
assert_array_equal(x_id, y_id)
except AssertionError:
msg = build_err_msg([x, y],
err_msg + '\nx and y %s location mismatch:'
% (hasval), verbose=verbose, header=header,
names=('x', 'y'), precision=precision)
raise AssertionError(msg)
try:
cond = (x.shape == () or y.shape == ()) or x.shape == y.shape
if not cond:
msg = build_err_msg([x, y],
err_msg
+ '\n(shapes %s, %s mismatch)' % (x.shape,
y.shape),
verbose=verbose, header=header,
names=('x', 'y'), precision=precision)
if not cond:
raise AssertionError(msg)
if isnumber(x) and isnumber(y):
if equal_nan:
x_isnan, y_isnan = isnan(x), isnan(y)
# Validate that NaNs are in the same place
if any(x_isnan) or any(y_isnan):
chk_same_position(x_isnan, y_isnan, hasval='nan')
x_isinf, y_isinf = isinf(x), isinf(y)
# Validate that infinite values are in the same place
if any(x_isinf) or any(y_isinf):
# Check +inf and -inf separately, since they are different
chk_same_position(x == +inf, y == +inf, hasval='+inf')
chk_same_position(x == -inf, y == -inf, hasval='-inf')
# Combine all the special values
x_id, y_id = x_isinf, y_isinf
if equal_nan:
x_id |= x_isnan
y_id |= y_isnan
# Only do the comparison if actual values are left
if all(x_id):
return
if any(x_id):
val = safe_comparison(x[~x_id], y[~y_id])
else:
val = safe_comparison(x, y)
else:
val = safe_comparison(x, y)
if isinstance(val, bool):
cond = val
reduced = [0]
else:
reduced = val.ravel()
cond = reduced.all()
reduced = reduced.tolist()
if not cond:
match = 100-100.0*reduced.count(1)/len(reduced)
msg = build_err_msg([x, y],
err_msg
+ '\n(mismatch %s%%)' % (match,),
verbose=verbose, header=header,
names=('x', 'y'), precision=precision)
if not cond:
> raise AssertionError(msg)
E AssertionError:
E Arrays are not equal
E
E (mismatch 100.0%)
E x: array([ True, True, True, True, True, True, True, True, True, True], dtype=bool)
E y: array([False, False, False, False, False, False, False, False, False, False], dtype=bool)
../../../../.local/miniconda3/envs/galario3/lib/python3.6/site-packages/numpy/testing/utils.py:796: AssertionError
During handling of the above exception, another exception occurred:
size = 1024, real_type = 'float64', tol = 1e-13
acc_lib = <module 'galario.double' from '/home/beaujean/workspace/protoplanetary/galario/build3/python/galario/double/__init__.py'>
@pytest.mark.parametrize("size, real_type, tol, acc_lib",
[(1024, 'float32', 1.e-4, g_single),
(1024, 'float64', 1.e-13, g_double)],
ids=["SP", "DP"])
def test_rotix(size, real_type, tol, acc_lib):
nsamples = 10
maxuv = 1000.
uv = pixel_coordinates(maxuv, size).astype(real_type)
udat, vdat = create_sampling_points(nsamples, maxuv/4.8)
assert len(udat) == nsamples
assert len(vdat) == nsamples
udat = udat.astype(real_type)
vdat = vdat.astype(real_type)
ui, vi = get_rotix_n(uv, uv, udat, vdat, len(uv))
ui = ui.astype(real_type)
vi = vi.astype(real_type)
ui1, vi1 = acc_lib.acc_rotix(uv, udat, vdat)
> np.testing.assert_allclose(ui1, ui, rtol=tol)
python_package/tests/test_galario.py:286:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
x_id = array([ True, True, True, True, True, True, True, True, True, True], dtype=bool)
y_id = array([False, False, False, False, False, False, False, False, False, False], dtype=bool)
hasval = '+inf'
def chk_same_position(x_id, y_id, hasval='nan'):
"""Handling nan/inf: check that x and y have the nan/inf at the same
locations."""
try:
assert_array_equal(x_id, y_id)
except AssertionError:
msg = build_err_msg([x, y],
err_msg + '\nx and y %s location mismatch:'
% (hasval), verbose=verbose, header=header,
names=('x', 'y'), precision=precision)
> raise AssertionError(msg)
E AssertionError:
E Not equal to tolerance rtol=1e-13, atol=0
E
E x and y +inf location mismatch:
E x: array([ inf, inf, inf, inf, inf, inf, inf, inf, inf, inf])
E y: array([ 458.470451, 610.984082, 365.23462 , 323.449008, 555.142405,
E 307.449384, 653.842193, 376.245319, 428.47669 , 482.963208])
../../../../.local/miniconda3/envs/galario3/lib/python3.6/site-packages/numpy/testing/utils.py:735: AssertionError
______________________________ test_loss[SP_par1] ______________________________
nsamples = 100, real_type = 'float32', complex_type = 'complex64', rtol = 0.0001
atol = 0.001
acc_lib = <module 'galario.single' from '/home/beaujean/workspace/protoplanetary/galario/build3/python/galario/single/__init__.py'>
pars = {'wle_m': 0.003, 'x0_arcsec': 0.4, 'y0_arcsec': 4.0}
@pytest.mark.parametrize("nsamples, real_type, complex_type, rtol, atol, acc_lib, pars",
[(100, 'float32', 'complex64', 1e-4, 1e-3, g_single, par1),
(1000, 'float64', 'complex128', 1e-14, 1e-10, g_double, par1)],
ids=["SP_par1", "DP_par1"])
def test_loss(nsamples, real_type, complex_type, rtol, atol, acc_lib, pars):
# try to find out where precision is lost
wle_m = pars.get('wle_m', 0.003)
x0_arcsec = pars.get('x0_arcsec', 0.4)
y0_arcsec = pars.get('y0_arcsec', 10.)
# generate the samples
maxuv_generator = 3.e3
udat, vdat = create_sampling_points(nsamples, maxuv_generator, dtype=real_type)
# compute the matrix size and maxuv
size, minuv, maxuv = matrix_size(udat, vdat)
uv = pixel_coordinates(maxuv, size)
# create model complex image (it happens to have 0 imaginary part)
reference_image = create_reference_image(size=size, kernel='gaussian', dtype=complex_type)
ref_complex = reference_image.copy()
###
# shift - FFT - shift
###
cpu_shift_fft_shift = np.fft.fftshift(np.fft.fft2(np.fft.fftshift(reference_image)))
shift_acc = reference_image.copy()
acc_lib.fftshift_fft2d_fftshift(shift_acc)
np.testing.assert_allclose(cpu_shift_fft_shift.real, shift_acc.real, rtol, atol)
np.testing.assert_allclose(cpu_shift_fft_shift.imag, shift_acc.imag, rtol, atol)
###
# phase
###
factor = sec2rad/wle_m*maxuv
fourier_shifted = Fourier_shift_static(cpu_shift_fft_shift, x0_arcsec, y0_arcsec, wle_m, maxuv)
acc_lib.apply_phase_2d(shift_acc, x0_arcsec*factor, y0_arcsec*factor)
# lose some absolute precision here --> not anymore
# atol *= 2
np.testing.assert_allclose(fourier_shifted.real, shift_acc.real, rtol, atol)
np.testing.assert_allclose(fourier_shifted.imag, shift_acc.imag, rtol, atol)
# but continue with previous tolerance
# atol /= 2
###
# rotation indices
###
uroti, vroti = get_rotix_n(uv, uv, udat, vdat, size)
ui1, vi1 = acc_lib.acc_rotix(uv.astype(real_type), udat.astype(real_type), vdat.astype(real_type))
> np.testing.assert_allclose(ui1, uroti, rtol, atol)
E AssertionError:
E Not equal to tolerance rtol=0.0001, atol=0.001
E
E (mismatch 100.0%)
E x: array([ 118.854355, 147.450668, 101.372635, 93.537834, 136.980347,
E 90.53791 , 155.486557, 103.437141, 113.230522, 123.446747,
E 137.839371, 112.262718, 125.376747, 104.865051, 136.284317,...
E y: array([ 118.0326 , 146.431198, 100.671753, 92.891121, 136.033279,
E 89.911934, 154.41153 , 102.721985, 112.447655, 122.593246,
E 136.886368, 111.486542, 124.509895, 104.140022, 135.342056,...
python_package/tests/test_galario.py:499: AssertionError
______________________________ test_loss[DP_par1] ______________________________
nsamples = 1000, real_type = 'float64', complex_type = 'complex128'
rtol = 1e-14, atol = 1e-10
acc_lib = <module 'galario.double' from '/home/beaujean/workspace/protoplanetary/galario/build3/python/galario/double/__init__.py'>
pars = {'wle_m': 0.003, 'x0_arcsec': 0.4, 'y0_arcsec': 4.0}
@pytest.mark.parametrize("nsamples, real_type, complex_type, rtol, atol, acc_lib, pars",
[(100, 'float32', 'complex64', 1e-4, 1e-3, g_single, par1),
(1000, 'float64', 'complex128', 1e-14, 1e-10, g_double, par1)],
ids=["SP_par1", "DP_par1"])
def test_loss(nsamples, real_type, complex_type, rtol, atol, acc_lib, pars):
# try to find out where precision is lost
wle_m = pars.get('wle_m', 0.003)
x0_arcsec = pars.get('x0_arcsec', 0.4)
y0_arcsec = pars.get('y0_arcsec', 10.)
# generate the samples
maxuv_generator = 3.e3
udat, vdat = create_sampling_points(nsamples, maxuv_generator, dtype=real_type)
# compute the matrix size and maxuv
size, minuv, maxuv = matrix_size(udat, vdat)
uv = pixel_coordinates(maxuv, size)
# create model complex image (it happens to have 0 imaginary part)
reference_image = create_reference_image(size=size, kernel='gaussian', dtype=complex_type)
ref_complex = reference_image.copy()
###
# shift - FFT - shift
###
cpu_shift_fft_shift = np.fft.fftshift(np.fft.fft2(np.fft.fftshift(reference_image)))
shift_acc = reference_image.copy()
acc_lib.fftshift_fft2d_fftshift(shift_acc)
np.testing.assert_allclose(cpu_shift_fft_shift.real, shift_acc.real, rtol, atol)
np.testing.assert_allclose(cpu_shift_fft_shift.imag, shift_acc.imag, rtol, atol)
###
# phase
###
factor = sec2rad/wle_m*maxuv
fourier_shifted = Fourier_shift_static(cpu_shift_fft_shift, x0_arcsec, y0_arcsec, wle_m, maxuv)
acc_lib.apply_phase_2d(shift_acc, x0_arcsec*factor, y0_arcsec*factor)
# lose some absolute precision here --> not anymore
# atol *= 2
np.testing.assert_allclose(fourier_shifted.real, shift_acc.real, rtol, atol)
np.testing.assert_allclose(fourier_shifted.imag, shift_acc.imag, rtol, atol)
# but continue with previous tolerance
# atol /= 2
###
# rotation indices
###
uroti, vroti = get_rotix_n(uv, uv, udat, vdat, size)
ui1, vi1 = acc_lib.acc_rotix(uv.astype(real_type), udat.astype(real_type), vdat.astype(real_type))
> np.testing.assert_allclose(ui1, uroti, rtol, atol)
E AssertionError:
E Not equal to tolerance rtol=1e-14, atol=1e-10
E
E (mismatch 100.0%)
E x: array([ 1003.243414, 1241.545963, 857.562428, 792.272409,
E 1154.293343, 767.272998, 1308.511762, 874.766646,
E 956.378163, 1041.513347, 1161.451931, 948.3131 ,...
E y: array([ 945.199198, 1169.71438 , 807.946813, 746.43425 ,
E 1087.509896, 722.881219, 1232.805768, 824.155654,
E 901.045409, 981.254964, 1094.254314, 893.446963,...
python_package/tests/test_galario.py:499: AssertionError
===================== 4 failed, 29 passed in 19.32 seconds =====================
I have tested on Python 2 and 3 on my linux and all 33 tests pass.
The code was on the latest master commit bb8f3a2d36e41877c77786b32a43fe113e18b1f9 Author: Frederik Beaujean Frederik.Beaujean@lmu.de Date: Fri Jul 7 13:43:02 2017 +0200 [docs] Extract numpy docstring into html
Perhaps you need to cleanup your build directory or so..?
On 7 Jul 2017, at 12:57, Frederik Beaujean notifications@github.com wrote:
============================= test session starts ============================== platform linux -- Python 3.6.1, pytest-3.0.7, py-1.4.33, pluggy-0.4.0 rootdir: /home/beaujean/workspace/protoplanetary/galario/build3, inifile: collected 33 items
python_package/tests/test_galario.py FF................FFsize:256, minuv:113.91199493408203, maxuv:19333.672265625 .size:2048, minuv:17.164953054780607, maxuv:19563.900024738796 .size:256, minuv:113.91199493408203, maxuv:19333.672265625 .size:2048, minuv:17.164953054780607, maxuv:19563.900024738796 .size:256, minuv:113.91199493408203, maxuv:19333.672265625 .size:2048, minuv:17.164953054780607, maxuv:19563.900024738796 .size:2048, minuv:17.164953231811523, maxuv:19563.899999999998 .size:2048, minuv:17.164953054780607, maxuv:19563.900024738796 .size:2048, minuv:17.164953231811523, maxuv:19563.899999999998 .size:2048, minuv:17.164953054780607, maxuv:19563.900024738796 .size:2048, minuv:17.164953231811523, maxuv:19563.899999999998 .size:2048, minuv:17.164953054780607, maxuv:19563.900024738796 .size:4096, minuv:17.164953054780607, maxuv:19563.900024738796 .
=================================== FAILURES =================================== ____ test_rotix[SP] ____
x_id = array([ True, True, True, True, True, True, True, True, True, True], dtype=bool) y_id = array([False, False, False, False, False, False, False, False, False, False], dtype=bool) hasval = '+inf'
def chk_same_position(x_id, y_id, hasval='nan'): """Handling nan/inf: check that x and y have the nan/inf at the same locations.""" try:assert_array_equal(x_id, y_id)../../../../.local/miniconda3/envs/galario3/lib/python3.6/site-packages/numpy/testing/utils.py:729:
x = array([ True, True, True, True, True, True, True, True, True, True], dtype=bool) y = array([False, False, False, False, False, False, False, False, False, False], dtype=bool) err_msg = '', verbose = True
def assert_array_equal(x, y, err_msg='', verbose=True): """ Raises an AssertionError if two array_like objects are not equal. Given two array_like objects, check that the shape is equal and all elements of these objects are equal. An exception is raised at shape mismatch or conflicting values. In contrast to the standard usage in numpy, NaNs are compared like numbers, no assertion is raised if both objects have NaNs in the same positions. The usual caution for verifying equality with floating point numbers is advised. Parameters ---------- x : array_like The actual object to check. y : array_like The desired, expected object. err_msg : str, optional The error message to be printed in case of failure. verbose : bool, optional If True, the conflicting values are appended to the error message. Raises ------ AssertionError If actual and desired objects are not equal. See Also -------- assert_allclose: Compare two array_like objects for equality with desired relative and/or absolute precision. assert_array_almost_equal_nulp, assert_array_max_ulp, assert_equal Examples -------- The first assert does not raise an exception: >>> np.testing.assert_array_equal([1.0,2.33333,np.nan], ... [np.exp(0),2.33333, np.nan]) Assert fails with numerical inprecision with floats: >>> np.testing.assert_array_equal([1.0,np.pi,np.nan], ... [1, np.sqrt(np.pi)**2, np.nan]) ... <type 'exceptions.ValueError'>: AssertionError: Arrays are not equal <BLANKLINE> (mismatch 50.0%) x: array([ 1. , 3.14159265, NaN]) y: array([ 1. , 3.14159265, NaN]) Use `assert_allclose` or one of the nulp (number of floating point values) functions for these cases instead: >>> np.testing.assert_allclose([1.0,np.pi,np.nan], ... [1, np.sqrt(np.pi)**2, np.nan], ... rtol=1e-10, atol=0) """ assert_array_compare(operator.__eq__, x, y, err_msg=err_msg,verbose=verbose, header='Arrays are not equal')../../../../.local/miniconda3/envs/galario3/lib/python3.6/site-packages/numpy/testing/utils.py:871:
comparison =
x = array([ True, True, True, True, True, True, True, True, True, True], dtype=bool) y = array([False, False, False, False, False, False, False, False, False, False], dtype=bool) err_msg = '', verbose = True, header = 'Arrays are not equal', precision = 6 equal_nan = True def assert_array_compare(comparison, x, y, err_msg='', verbose=True, header='', precision=6, equal_nan=True): __tracebackhide__ = True # Hide traceback for py.test from numpy.core import array, isnan, isinf, any, all, inf x = array(x, copy=False, subok=True) y = array(y, copy=False, subok=True) def safe_comparison(*args, **kwargs): # There are a number of cases where comparing two arrays hits special # cases in array_richcompare, specifically around strings and void # dtypes. Basically, we just can't do comparisons involving these # types, unless both arrays have exactly the *same* type. So # e.g. you can apply == to two string arrays, or two arrays with # identical structured dtypes. But if you compare a non-string array # to a string array, or two arrays with non-identical structured # dtypes, or anything like that, then internally stuff blows up. # Currently, when things blow up, we just return a scalar False or # True. But we also emit a DeprecationWarning, b/c eventually we # should raise an error here. (Ideally we might even make this work # properly, but since that will require rewriting a bunch of how # ufuncs work then we are not counting on that.) # # The point of this little function is to let the DeprecationWarning # pass (or maybe eventually catch the errors and return False, I # dunno, that's a little trickier and we can figure that out when the # time comes). # Note: Put a lock around warning filter (comment at lock definition) with _assert_comparison_lock, suppress_warnings() as sup: sup.filter(DeprecationWarning, ".*==") sup.filter(FutureWarning, ".*==") return comparison(*args, **kwargs) def isnumber(x): return x.dtype.char in '?bhilqpBHILQPefdgFDG' def chk_same_position(x_id, y_id, hasval='nan'): """Handling nan/inf: check that x and y have the nan/inf at the same locations.""" try: assert_array_equal(x_id, y_id) except AssertionError: msg = build_err_msg([x, y], err_msg + '\nx and y %s location mismatch:' % (hasval), verbose=verbose, header=header, names=('x', 'y'), precision=precision) raise AssertionError(msg) try: cond = (x.shape == () or y.shape == ()) or x.shape == y.shape if not cond: msg = build_err_msg([x, y], err_msg + '\n(shapes %s, %s mismatch)' % (x.shape, y.shape), verbose=verbose, header=header, names=('x', 'y'), precision=precision) if not cond: raise AssertionError(msg) if isnumber(x) and isnumber(y): if equal_nan: x_isnan, y_isnan = isnan(x), isnan(y) # Validate that NaNs are in the same place if any(x_isnan) or any(y_isnan): chk_same_position(x_isnan, y_isnan, hasval='nan') x_isinf, y_isinf = isinf(x), isinf(y) # Validate that infinite values are in the same place if any(x_isinf) or any(y_isinf): # Check +inf and -inf separately, since they are different chk_same_position(x == +inf, y == +inf, hasval='+inf') chk_same_position(x == -inf, y == -inf, hasval='-inf') # Combine all the special values x_id, y_id = x_isinf, y_isinf if equal_nan: x_id |= x_isnan y_id |= y_isnan # Only do the comparison if actual values are left if all(x_id): return if any(x_id): val = safe_comparison(x[~x_id], y[~y_id]) else: val = safe_comparison(x, y) else: val = safe_comparison(x, y) if isinstance(val, bool): cond = val reduced = [0] else: reduced = val.ravel() cond = reduced.all() reduced = reduced.tolist() if not cond: match = 100-100.0*reduced.count(1)/len(reduced) msg = build_err_msg([x, y], err_msg + '\n(mismatch %s%%)' % (match,), verbose=verbose, header=header, names=('x', 'y'), precision=precision) if not cond:raise AssertionError(msg)E AssertionError: E Arrays are not equal E
E (mismatch 100.0%) E x: array([ True, True, True, True, True, True, True, True, True, True], dtype=bool) E y: array([False, False, False, False, False, False, False, False, False, False], dtype=bool)../../../../.local/miniconda3/envs/galario3/lib/python3.6/site-packages/numpy/testing/utils.py:796: AssertionError
During handling of the above exception, another exception occurred:
size = 1024, real_type = 'float32', tol = 0.0001 acc_lib = <module 'galario.single' from '/home/beaujean/workspace/protoplanetary/galario/build3/python/galario/single/init.py'>
@pytest.mark.parametrize("size, real_type, tol, acc_lib", [(1024, 'float32', 1.e-4, g_single), (1024, 'float64', 1.e-13, g_double)], ids=["SP", "DP"]) def test_rotix(size, real_type, tol, acc_lib): nsamples = 10 maxuv = 1000. uv = pixel_coordinates(maxuv, size).astype(real_type) udat, vdat = create_sampling_points(nsamples, maxuv/4.8) assert len(udat) == nsamples assert len(vdat) == nsamples udat = udat.astype(real_type) vdat = vdat.astype(real_type) ui, vi = get_rotix_n(uv, uv, udat, vdat, len(uv)) ui = ui.astype(real_type) vi = vi.astype(real_type) ui1, vi1 = acc_lib.acc_rotix(uv, udat, vdat)np.testing.assert_allclose(ui1, ui, rtol=tol)python_package/tests/test_galario.py:286:
x_id = array([ True, True, True, True, True, True, True, True, True, True], dtype=bool) y_id = array([False, False, False, False, False, False, False, False, False, False], dtype=bool) hasval = '+inf'
def chk_same_position(x_id, y_id, hasval='nan'): """Handling nan/inf: check that x and y have the nan/inf at the same locations.""" try: assert_array_equal(x_id, y_id) except AssertionError: msg = build_err_msg([x, y], err_msg + '\nx and y %s location mismatch:' % (hasval), verbose=verbose, header=header, names=('x', 'y'), precision=precision)raise AssertionError(msg)E AssertionError: E Not equal to tolerance rtol=0.0001, atol=0 E
E x and y +inf location mismatch: E x: array([ inf, inf, inf, inf, inf, inf, inf, inf, inf, inf], dtype=float32) E y: array([ 458.470459, 610.98407 , 365.234619, 323.449005, 555.142395, E 307.449402, 653.842163, 376.2453 , 428.476685, 482.963196], dtype=float32)../../../../.local/miniconda3/envs/galario3/lib/python3.6/site-packages/numpy/testing/utils.py:735: AssertionError ____ test_rotix[DP] ____
x_id = array([ True, True, True, True, True, True, True, True, True, True], dtype=bool) y_id = array([False, False, False, False, False, False, False, False, False, False], dtype=bool) hasval = '+inf'
def chk_same_position(x_id, y_id, hasval='nan'): """Handling nan/inf: check that x and y have the nan/inf at the same locations.""" try:assert_array_equal(x_id, y_id)../../../../.local/miniconda3/envs/galario3/lib/python3.6/site-packages/numpy/testing/utils.py:729:
x = array([ True, True, True, True, True, True, True, True, True, True], dtype=bool) y = array([False, False, False, False, False, False, False, False, False, False], dtype=bool) err_msg = '', verbose = True
def assert_array_equal(x, y, err_msg='', verbose=True): """ Raises an AssertionError if two array_like objects are not equal. Given two array_like objects, check that the shape is equal and all elements of these objects are equal. An exception is raised at shape mismatch or conflicting values. In contrast to the standard usage in numpy, NaNs are compared like numbers, no assertion is raised if both objects have NaNs in the same positions. The usual caution for verifying equality with floating point numbers is advised. Parameters ---------- x : array_like The actual object to check. y : array_like The desired, expected object. err_msg : str, optional The error message to be printed in case of failure. verbose : bool, optional If True, the conflicting values are appended to the error message. Raises ------ AssertionError If actual and desired objects are not equal. See Also -------- assert_allclose: Compare two array_like objects for equality with desired relative and/or absolute precision. assert_array_almost_equal_nulp, assert_array_max_ulp, assert_equal Examples -------- The first assert does not raise an exception: >>> np.testing.assert_array_equal([1.0,2.33333,np.nan], ... [np.exp(0),2.33333, np.nan]) Assert fails with numerical inprecision with floats: >>> np.testing.assert_array_equal([1.0,np.pi,np.nan], ... [1, np.sqrt(np.pi)**2, np.nan]) ... <type 'exceptions.ValueError'>: AssertionError: Arrays are not equal <BLANKLINE> (mismatch 50.0%) x: array([ 1. , 3.14159265, NaN]) y: array([ 1. , 3.14159265, NaN]) Use `assert_allclose` or one of the nulp (number of floating point values) functions for these cases instead: >>> np.testing.assert_allclose([1.0,np.pi,np.nan], ... [1, np.sqrt(np.pi)**2, np.nan], ... rtol=1e-10, atol=0) """ assert_array_compare(operator.__eq__, x, y, err_msg=err_msg,verbose=verbose, header='Arrays are not equal')../../../../.local/miniconda3/envs/galario3/lib/python3.6/site-packages/numpy/testing/utils.py:871:
comparison =
x = array([ True, True, True, True, True, True, True, True, True, True], dtype=bool) y = array([False, False, False, False, False, False, False, False, False, False], dtype=bool) err_msg = '', verbose = True, header = 'Arrays are not equal', precision = 6 equal_nan = True def assert_array_compare(comparison, x, y, err_msg='', verbose=True, header='', precision=6, equal_nan=True): __tracebackhide__ = True # Hide traceback for py.test from numpy.core import array, isnan, isinf, any, all, inf x = array(x, copy=False, subok=True) y = array(y, copy=False, subok=True) def safe_comparison(*args, **kwargs): # There are a number of cases where comparing two arrays hits special # cases in array_richcompare, specifically around strings and void # dtypes. Basically, we just can't do comparisons involving these # types, unless both arrays have exactly the *same* type. So # e.g. you can apply == to two string arrays, or two arrays with # identical structured dtypes. But if you compare a non-string array # to a string array, or two arrays with non-identical structured # dtypes, or anything like that, then internally stuff blows up. # Currently, when things blow up, we just return a scalar False or # True. But we also emit a DeprecationWarning, b/c eventually we # should raise an error here. (Ideally we might even make this work # properly, but since that will require rewriting a bunch of how # ufuncs work then we are not counting on that.) # # The point of this little function is to let the DeprecationWarning # pass (or maybe eventually catch the errors and return False, I # dunno, that's a little trickier and we can figure that out when the # time comes). # Note: Put a lock around warning filter (comment at lock definition) with _assert_comparison_lock, suppress_warnings() as sup: sup.filter(DeprecationWarning, ".*==") sup.filter(FutureWarning, ".*==") return comparison(*args, **kwargs) def isnumber(x): return x.dtype.char in '?bhilqpBHILQPefdgFDG' def chk_same_position(x_id, y_id, hasval='nan'): """Handling nan/inf: check that x and y have the nan/inf at the same locations.""" try: assert_array_equal(x_id, y_id) except AssertionError: msg = build_err_msg([x, y], err_msg + '\nx and y %s location mismatch:' % (hasval), verbose=verbose, header=header, names=('x', 'y'), precision=precision) raise AssertionError(msg) try: cond = (x.shape == () or y.shape == ()) or x.shape == y.shape if not cond: msg = build_err_msg([x, y], err_msg + '\n(shapes %s, %s mismatch)' % (x.shape, y.shape), verbose=verbose, header=header, names=('x', 'y'), precision=precision) if not cond: raise AssertionError(msg) if isnumber(x) and isnumber(y): if equal_nan: x_isnan, y_isnan = isnan(x), isnan(y) # Validate that NaNs are in the same place if any(x_isnan) or any(y_isnan): chk_same_position(x_isnan, y_isnan, hasval='nan') x_isinf, y_isinf = isinf(x), isinf(y) # Validate that infinite values are in the same place if any(x_isinf) or any(y_isinf): # Check +inf and -inf separately, since they are different chk_same_position(x == +inf, y == +inf, hasval='+inf') chk_same_position(x == -inf, y == -inf, hasval='-inf') # Combine all the special values x_id, y_id = x_isinf, y_isinf if equal_nan: x_id |= x_isnan y_id |= y_isnan # Only do the comparison if actual values are left if all(x_id): return if any(x_id): val = safe_comparison(x[~x_id], y[~y_id]) else: val = safe_comparison(x, y) else: val = safe_comparison(x, y) if isinstance(val, bool): cond = val reduced = [0] else: reduced = val.ravel() cond = reduced.all() reduced = reduced.tolist() if not cond: match = 100-100.0*reduced.count(1)/len(reduced) msg = build_err_msg([x, y], err_msg + '\n(mismatch %s%%)' % (match,), verbose=verbose, header=header, names=('x', 'y'), precision=precision) if not cond:raise AssertionError(msg)E AssertionError: E Arrays are not equal E
E (mismatch 100.0%) E x: array([ True, True, True, True, True, True, True, True, True, True], dtype=bool) E y: array([False, False, False, False, False, False, False, False, False, False], dtype=bool)../../../../.local/miniconda3/envs/galario3/lib/python3.6/site-packages/numpy/testing/utils.py:796: AssertionError
During handling of the above exception, another exception occurred:
size = 1024, real_type = 'float64', tol = 1e-13 acc_lib = <module 'galario.double' from '/home/beaujean/workspace/protoplanetary/galario/build3/python/galario/double/init.py'>
@pytest.mark.parametrize("size, real_type, tol, acc_lib", [(1024, 'float32', 1.e-4, g_single), (1024, 'float64', 1.e-13, g_double)], ids=["SP", "DP"]) def test_rotix(size, real_type, tol, acc_lib): nsamples = 10 maxuv = 1000. uv = pixel_coordinates(maxuv, size).astype(real_type) udat, vdat = create_sampling_points(nsamples, maxuv/4.8) assert len(udat) == nsamples assert len(vdat) == nsamples udat = udat.astype(real_type) vdat = vdat.astype(real_type) ui, vi = get_rotix_n(uv, uv, udat, vdat, len(uv)) ui = ui.astype(real_type) vi = vi.astype(real_type) ui1, vi1 = acc_lib.acc_rotix(uv, udat, vdat)np.testing.assert_allclose(ui1, ui, rtol=tol)python_package/tests/test_galario.py:286:
x_id = array([ True, True, True, True, True, True, True, True, True, True], dtype=bool) y_id = array([False, False, False, False, False, False, False, False, False, False], dtype=bool) hasval = '+inf'
def chk_same_position(x_id, y_id, hasval='nan'): """Handling nan/inf: check that x and y have the nan/inf at the same locations.""" try: assert_array_equal(x_id, y_id) except AssertionError: msg = build_err_msg([x, y], err_msg + '\nx and y %s location mismatch:' % (hasval), verbose=verbose, header=header, names=('x', 'y'), precision=precision)raise AssertionError(msg)E AssertionError: E Not equal to tolerance rtol=1e-13, atol=0 E
E x and y +inf location mismatch: E x: array([ inf, inf, inf, inf, inf, inf, inf, inf, inf, inf]) E y: array([ 458.470451, 610.984082, 365.23462 , 323.449008, 555.142405, E 307.449384, 653.842193, 376.245319, 428.47669 , 482.963208])../../../../.local/miniconda3/envs/galario3/lib/python3.6/site-packages/numpy/testing/utils.py:735: AssertionError __ test_loss[SP_par1] __
nsamples = 100, real_type = 'float32', complex_type = 'complex64', rtol = 0.0001 atol = 0.001 acc_lib = <module 'galario.single' from '/home/beaujean/workspace/protoplanetary/galario/build3/python/galario/single/init.py'> pars = {'wle_m': 0.003, 'x0_arcsec': 0.4, 'y0_arcsec': 4.0}
@pytest.mark.parametrize("nsamples, real_type, complex_type, rtol, atol, acc_lib, pars", [(100, 'float32', 'complex64', 1e-4, 1e-3, g_single, par1), (1000, 'float64', 'complex128', 1e-14, 1e-10, g_double, par1)], ids=["SP_par1", "DP_par1"]) def test_loss(nsamples, real_type, complex_type, rtol, atol, acc_lib, pars): # try to find out where precision is lost wle_m = pars.get('wle_m', 0.003) x0_arcsec = pars.get('x0_arcsec', 0.4) y0_arcsec = pars.get('y0_arcsec', 10.) # generate the samples maxuv_generator = 3.e3 udat, vdat = create_sampling_points(nsamples, maxuv_generator, dtype=real_type) # compute the matrix size and maxuv size, minuv, maxuv = matrix_size(udat, vdat) uv = pixel_coordinates(maxuv, size) # create model complex image (it happens to have 0 imaginary part) reference_image = create_reference_image(size=size, kernel='gaussian', dtype=complex_type) ref_complex = reference_image.copy() ### # shift - FFT - shift ### cpu_shift_fft_shift = np.fft.fftshift(np.fft.fft2(np.fft.fftshift(reference_image))) shift_acc = reference_image.copy() acc_lib.fftshift_fft2d_fftshift(shift_acc) np.testing.assert_allclose(cpu_shift_fft_shift.real, shift_acc.real, rtol, atol) np.testing.assert_allclose(cpu_shift_fft_shift.imag, shift_acc.imag, rtol, atol) ### # phase ### factor = sec2rad/wle_m*maxuv fourier_shifted = Fourier_shift_static(cpu_shift_fft_shift, x0_arcsec, y0_arcsec, wle_m, maxuv) acc_lib.apply_phase_2d(shift_acc, x0_arcsec*factor, y0_arcsec*factor) # lose some absolute precision here --> not anymore # atol *= 2 np.testing.assert_allclose(fourier_shifted.real, shift_acc.real, rtol, atol) np.testing.assert_allclose(fourier_shifted.imag, shift_acc.imag, rtol, atol) # but continue with previous tolerance # atol /= 2 ### # rotation indices ### uroti, vroti = get_rotix_n(uv, uv, udat, vdat, size) ui1, vi1 = acc_lib.acc_rotix(uv.astype(real_type), udat.astype(real_type), vdat.astype(real_type))np.testing.assert_allclose(ui1, uroti, rtol, atol)E AssertionError: E Not equal to tolerance rtol=0.0001, atol=0.001 E
E (mismatch 100.0%) E x: array([ 118.854355, 147.450668, 101.372635, 93.537834, 136.980347, E 90.53791 , 155.486557, 103.437141, 113.230522, 123.446747, E 137.839371, 112.262718, 125.376747, 104.865051, 136.284317,... E y: array([ 118.0326 , 146.431198, 100.671753, 92.891121, 136.033279, E 89.911934, 154.41153 , 102.721985, 112.447655, 122.593246, E 136.886368, 111.486542, 124.509895, 104.140022, 135.342056,...python_package/tests/test_galario.py:499: AssertionError __ test_loss[DP_par1] __
nsamples = 1000, real_type = 'float64', complex_type = 'complex128' rtol = 1e-14, atol = 1e-10 acc_lib = <module 'galario.double' from '/home/beaujean/workspace/protoplanetary/galario/build3/python/galario/double/init.py'> pars = {'wle_m': 0.003, 'x0_arcsec': 0.4, 'y0_arcsec': 4.0}
@pytest.mark.parametrize("nsamples, real_type, complex_type, rtol, atol, acc_lib, pars", [(100, 'float32', 'complex64', 1e-4, 1e-3, g_single, par1), (1000, 'float64', 'complex128', 1e-14, 1e-10, g_double, par1)], ids=["SP_par1", "DP_par1"]) def test_loss(nsamples, real_type, complex_type, rtol, atol, acc_lib, pars): # try to find out where precision is lost wle_m = pars.get('wle_m', 0.003) x0_arcsec = pars.get('x0_arcsec', 0.4) y0_arcsec = pars.get('y0_arcsec', 10.) # generate the samples maxuv_generator = 3.e3 udat, vdat = create_sampling_points(nsamples, maxuv_generator, dtype=real_type) # compute the matrix size and maxuv size, minuv, maxuv = matrix_size(udat, vdat) uv = pixel_coordinates(maxuv, size) # create model complex image (it happens to have 0 imaginary part) reference_image = create_reference_image(size=size, kernel='gaussian', dtype=complex_type) ref_complex = reference_image.copy() ### # shift - FFT - shift ### cpu_shift_fft_shift = np.fft.fftshift(np.fft.fft2(np.fft.fftshift(reference_image))) shift_acc = reference_image.copy() acc_lib.fftshift_fft2d_fftshift(shift_acc) np.testing.assert_allclose(cpu_shift_fft_shift.real, shift_acc.real, rtol, atol) np.testing.assert_allclose(cpu_shift_fft_shift.imag, shift_acc.imag, rtol, atol) ### # phase ### factor = sec2rad/wle_m*maxuv fourier_shifted = Fourier_shift_static(cpu_shift_fft_shift, x0_arcsec, y0_arcsec, wle_m, maxuv) acc_lib.apply_phase_2d(shift_acc, x0_arcsec*factor, y0_arcsec*factor) # lose some absolute precision here --> not anymore # atol *= 2 np.testing.assert_allclose(fourier_shifted.real, shift_acc.real, rtol, atol) np.testing.assert_allclose(fourier_shifted.imag, shift_acc.imag, rtol, atol) # but continue with previous tolerance # atol /= 2 ### # rotation indices ### uroti, vroti = get_rotix_n(uv, uv, udat, vdat, size) ui1, vi1 = acc_lib.acc_rotix(uv.astype(real_type), udat.astype(real_type), vdat.astype(real_type))np.testing.assert_allclose(ui1, uroti, rtol, atol)E AssertionError: E Not equal to tolerance rtol=1e-14, atol=1e-10 E
E (mismatch 100.0%) E x: array([ 1003.243414, 1241.545963, 857.562428, 792.272409, E 1154.293343, 767.272998, 1308.511762, 874.766646, E 956.378163, 1041.513347, 1161.451931, 948.3131 ,... E y: array([ 945.199198, 1169.71438 , 807.946813, 746.43425 , E 1087.509896, 722.881219, 1232.805768, 824.155654, E 901.045409, 981.254964, 1094.254314, 893.446963,...python_package/tests/test_galario.py:499: AssertionError ===================== 4 failed, 29 passed in 19.32 seconds =====================
— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/mtazzari/galario/issues/7#issuecomment-313662103, or mute the thread https://github.com/notifications/unsubscribe-auth/AFvcgkpKo3twRjsWEx15PO1P_M94xw2nks5sLh0wgaJpZM4NjwHC.
mtazzari
commented
7 years ago It does not show up in the profiling for the moment, so we lower the priority for future releases.
fredRos
commented
7 years ago no more indices needed as of 1485b07
The rotix_core function can be highly simplified by noting that:
And the same applies for indv. I think we can even delete the rotix kernels and put it directly in
interpolate_core. Check if the indu and indv arrays are needed only in the interpolate. If yes, then it's a go for deleting rotix.