mctpy 0.0.1__py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
mctpy/__init__.py ADDED
@@ -0,0 +1,16 @@
1
+ from .solver import correlator, mean_squared_displacement, non_gaussian_parameter
2
+ from .asymptotics import beta_scaling_function
3
+ from .fsolver import nonergodicity_parameter, eigenvalue
4
+ from .schematic import f12model,f12gammadot_model,sjoegren_model,bosse_krieger_model,f12gammadot_tensorial_model
5
+ from .standard import simple_liquid_model, tagged_particle_model, tagged_particle_q0, tagged_particle_ngp
6
+ from .standard_2d import simple_liquid_model_2d, tagged_particle_model_2d
7
+ from .mixture import mixture_model
8
+ from .shear import isotropically_sheared_model
9
+ from .abp import abp_model_2d
10
+ from .util import CorrelatorStack, exponents, evscan, filon_integrate, filon_cos_transform, filon_sin_transform
11
+
12
+ from .__util__ import model_base
13
+
14
+ import mctpy.granular
15
+
16
+ import mctpy.structurefactors
mctpy/__util__.py ADDED
@@ -0,0 +1,154 @@
1
+ import numpy as np
2
+ import numba as nb
3
+
4
+ # https://stackoverflow.com/questions/61509903/how-to-pass-array-pointer-to-numba-function
5
+ # we need this to keep references in jit-compiled models that
6
+ # reference underlying data of a base model
7
+ @nb.extending.intrinsic
8
+ def address_as_void_pointer(typingctx, src):
9
+ """ returns a void pointer from a given memory address """
10
+ from numba.core import types, cgutils
11
+ sig = types.voidptr(src)
12
+
13
+ def codegen(cgctx, builder, sig, args):
14
+ return builder.inttoptr(args[0], cgutils.voidptr_t)
15
+ return sig, codegen
16
+
17
+ # take 3-tuple of addres, shape, dtype
18
+ @nb.njit
19
+ def nparray(ast,i=None):
20
+ return nb.carray(address_as_void_pointer(ast[0]),ast[1],ast[2])
21
+
22
+ def void(nparray):
23
+ return nparray.ctypes.data, nparray.shape, nparray.dtype
24
+
25
+
26
+ class model_base (object):
27
+ def __init__ (self):
28
+ self.dtype = np.dtype(float)
29
+ def __len__ (self):
30
+ return 1
31
+ def matrix_dimension (self):
32
+ return 1
33
+ def scalar (self):
34
+ return True
35
+ def hopping (self):
36
+ return None
37
+ def phi0 (self):
38
+ return np.ones(len(self))
39
+ def phi0d (self):
40
+ return np.zeros(len(self))
41
+ def Wq (self):
42
+ return np.ones(len(self))
43
+ def Aq (self):
44
+ return np.ones(len(self))
45
+ def Bq (self):
46
+ return np.ones(len(self))
47
+
48
+ def set_base (self, array):
49
+ #self.phi = void(array)
50
+ self.phi = array
51
+ def cache (self):
52
+ if 'base' in dir(self): return False
53
+ return True
54
+
55
+ def kernel_extra_args (self):
56
+ return []
57
+
58
+ def get_kernel (self):
59
+ if not self.cache() or not '__m__' in dir(self):
60
+ self.__m__ = self.make_kernel()
61
+ return self.__m__
62
+ def get_dm (self):
63
+ if not self.cache() or not '__dm__' in dir(self):
64
+ self.__dm__ = self.make_dm()
65
+ return self.__dm__
66
+ def get_dmhat (self):
67
+ if not self.cache() or not '__dmhat__' in dir(self):
68
+ self.__dmhat__ = self.make_dmhat()
69
+ return self.__dmhat__
70
+ def get_dm2 (self):
71
+ if not self.cache() or not '__dm2__' in dir(self):
72
+ self.__dm2__ = self.make_dm2()
73
+ return self.__dm2__
74
+
75
+ def make_kernel (self):
76
+ @nb.njit
77
+ def dummy(m, phi, i, t):
78
+ return
79
+ return dummy
80
+ def make_dm (self):
81
+ @nb.njit
82
+ def dummy(m, phi, dphi):
83
+ return
84
+ return dummy
85
+ def set_C (self, f):
86
+ # can use this to pre-calculate stability matrix before make_dm
87
+ return
88
+ def make_dmhat(self):
89
+ @nb.njit
90
+ def dummy(m, f, ehat):
91
+ return
92
+ return dummy
93
+ def make_dm2 (self):
94
+ @nb.njit
95
+ def dummy(m, phi, dphi):
96
+ return
97
+ return dummy
98
+
99
+ def dq (self):
100
+ return 1.0
101
+
102
+ def h5save (self, fh):
103
+ return
104
+
105
+ class loaded_model(model_base):
106
+ def __init__ (self, h5data):
107
+ model_definition = h5data['model']
108
+ for attr,val in model_definition.attrs.items():
109
+ self.__dict__[attr] = val
110
+ for field,val in model_definition.items():
111
+ self.__dict__[field] = np.array(val)
112
+ def __len__ (self):
113
+ return self.M
114
+
115
+ @nb.njit
116
+ def np_gradient(f,k):
117
+ df_dk = np.zeros_like(f)
118
+ df_dk[1:-1] = (f[2:]-f[0:-2])/(k[2:]-k[0:-2])
119
+ df_dk[0] = (f[1]-f[0])/(k[1]-k[0])
120
+ df_dk[-1] = (f[-1]-f[-2])/(k[-1]-k[-2])
121
+ return df_dk
122
+
123
+ # numba does not support np.isclose with complex arrays
124
+ # we need to patch around that
125
+ # note that we also include a test like np.isclose().all()
126
+ # since we need that, and we can overload to make sure it
127
+ # also works with scalars (needed in the regula_falsi when njit'ed)
128
+
129
+ def np_isclose_all_impl(a, b, rtol, atol):
130
+ return np.isclose(a, b, rtol=rtol, atol=atol).all()
131
+
132
+ @nb.extending.overload(np_isclose_all_impl)
133
+ def np_isclose_all_impl_overload(a, b, rtol, atol):
134
+ if not isinstance(a, nb.types.Array):
135
+ if not a == nb.complex128:
136
+ def np_isclose_scalar (a, b, rtol, atol):
137
+ return np.isclose (a, b, rtol, atol)
138
+ return np_isclose_scalar
139
+ else:
140
+ def np_isclose_scalar (a, b, rtol, atol):
141
+ return np.isclose (a.real, b.real, rtol, atol) \
142
+ & np.isclose (a.imag, b.imag, rtol, atol)
143
+ return np_isclose_scalar
144
+ if a.dtype == nb.complex128:
145
+ def np_isclose_complex(a, b, rtol, atol):
146
+ return (np.isclose(a.real, b.real, rtol=rtol, atol=atol) \
147
+ & np.isclose(a.imag, b.imag, rtol=rtol, atol=atol)).all()
148
+ return np_isclose_complex
149
+ else:
150
+ return np_isclose_all_impl
151
+
152
+ @nb.njit
153
+ def np_isclose_all(a, b, rtol=1e-5, atol=1e-8):
154
+ return np_isclose_all_impl(a, b, rtol, atol)