Generated by Cython 3.0.12
Yellow lines hint at Python interaction.
Click on a line that starts with a "+" to see the C code that Cython generated for it.
Raw output: inverter.c
+01: # To compile: #warning "Using deprecated NumPy API, disable it by #defining NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION"
__pyx_t_7 = __Pyx_PyDict_NewPresized(0); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1, __pyx_L1_error) __Pyx_GOTREF(__pyx_t_7); if (PyDict_SetItem(__pyx_d, __pyx_n_s_test, __pyx_t_7) < 0) __PYX_ERR(0, 1, __pyx_L1_error) __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
02: # python setup.py build_ext --inplace --force
03:
04: # "cimport" is used to import special compile-time information
05: # about the numpy module (this is stored in a file numpy.pxd which is
06: # currently part of the Cython distribution).
07:
08: cimport cython
09: cimport numpy as np
10:
+11: import numpy as np
__pyx_t_7 = __Pyx_ImportDottedModule(__pyx_n_s_numpy, NULL); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 11, __pyx_L1_error) __Pyx_GOTREF(__pyx_t_7); if (PyDict_SetItem(__pyx_d, __pyx_n_s_np, __pyx_t_7) < 0) __PYX_ERR(0, 11, __pyx_L1_error) __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
12: from cython.parallel import prange
13:
14: ctypedef np.float64_t DTYPE_t
15:
+16: @cython.cdivision(True)
/* Python wrapper */ static PyObject *__pyx_pw_8inverter_1diag_inverter(PyObject *__pyx_self, #if CYTHON_METH_FASTCALL PyObject *const *__pyx_args, Py_ssize_t __pyx_nargs, PyObject *__pyx_kwds #else PyObject *__pyx_args, PyObject *__pyx_kwds #endif ); /*proto*/ PyDoc_STRVAR(__pyx_doc_8inverter_diag_inverter, "\n Inverts a symmetric tridiagonal matrix\n\n Parameters\n ============\n diag_c : np.ndarray\n The central diagonal of the matrix \n diag_offset : np.ndarray\n The non-central diagonal\n output : np.ndarray\n Array to which the inverted matrix will be written to \n n_threads : int\n Number of threads to use \n "); static PyMethodDef __pyx_mdef_8inverter_1diag_inverter = {"diag_inverter", (PyCFunction)(void*)(__Pyx_PyCFunction_FastCallWithKeywords)__pyx_pw_8inverter_1diag_inverter, __Pyx_METH_FASTCALL|METH_KEYWORDS, __pyx_doc_8inverter_diag_inverter}; static PyObject *__pyx_pw_8inverter_1diag_inverter(PyObject *__pyx_self, #if CYTHON_METH_FASTCALL PyObject *const *__pyx_args, Py_ssize_t __pyx_nargs, PyObject *__pyx_kwds #else PyObject *__pyx_args, PyObject *__pyx_kwds #endif ) { __Pyx_memviewslice __pyx_v_diag_c = { 0, 0, { 0 }, { 0 }, { 0 } }; __Pyx_memviewslice __pyx_v_diag_offset = { 0, 0, { 0 }, { 0 }, { 0 } }; __Pyx_memviewslice __pyx_v_output = { 0, 0, { 0 }, { 0 }, { 0 } }; CYTHON_UNUSED int __pyx_v_n_threads; #if !CYTHON_METH_FASTCALL CYTHON_UNUSED Py_ssize_t __pyx_nargs; #endif CYTHON_UNUSED PyObject *const *__pyx_kwvalues; PyObject *__pyx_r = 0; __Pyx_RefNannyDeclarations __Pyx_RefNannySetupContext("diag_inverter (wrapper)", 0); #if !CYTHON_METH_FASTCALL #if CYTHON_ASSUME_SAFE_MACROS __pyx_nargs = PyTuple_GET_SIZE(__pyx_args); #else __pyx_nargs = PyTuple_Size(__pyx_args); if (unlikely(__pyx_nargs < 0)) return NULL; #endif #endif __pyx_kwvalues = __Pyx_KwValues_FASTCALL(__pyx_args, __pyx_nargs); { PyObject **__pyx_pyargnames[] = {&__pyx_n_s_diag_c,&__pyx_n_s_diag_offset,&__pyx_n_s_output,&__pyx_n_s_n_threads,0}; PyObject* values[4] = {0,0,0,0}; if (__pyx_kwds) { Py_ssize_t kw_args; switch (__pyx_nargs) { case 4: values[3] = __Pyx_Arg_FASTCALL(__pyx_args, 3); CYTHON_FALLTHROUGH; case 3: values[2] = __Pyx_Arg_FASTCALL(__pyx_args, 2); CYTHON_FALLTHROUGH; case 2: values[1] = __Pyx_Arg_FASTCALL(__pyx_args, 1); CYTHON_FALLTHROUGH; case 1: values[0] = __Pyx_Arg_FASTCALL(__pyx_args, 0); CYTHON_FALLTHROUGH; case 0: break; default: goto __pyx_L5_argtuple_error; } kw_args = __Pyx_NumKwargs_FASTCALL(__pyx_kwds); switch (__pyx_nargs) { case 0: if (likely((values[0] = __Pyx_GetKwValue_FASTCALL(__pyx_kwds, __pyx_kwvalues, __pyx_n_s_diag_c)) != 0)) { (void)__Pyx_Arg_NewRef_FASTCALL(values[0]); kw_args--; } else if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 16, __pyx_L3_error) else goto __pyx_L5_argtuple_error; CYTHON_FALLTHROUGH; case 1: if (likely((values[1] = __Pyx_GetKwValue_FASTCALL(__pyx_kwds, __pyx_kwvalues, __pyx_n_s_diag_offset)) != 0)) { (void)__Pyx_Arg_NewRef_FASTCALL(values[1]); kw_args--; } else if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 16, __pyx_L3_error) else { __Pyx_RaiseArgtupleInvalid("diag_inverter", 1, 4, 4, 1); __PYX_ERR(0, 16, __pyx_L3_error) } CYTHON_FALLTHROUGH; case 2: if (likely((values[2] = __Pyx_GetKwValue_FASTCALL(__pyx_kwds, __pyx_kwvalues, __pyx_n_s_output)) != 0)) { (void)__Pyx_Arg_NewRef_FASTCALL(values[2]); kw_args--; } else if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 16, __pyx_L3_error) else { __Pyx_RaiseArgtupleInvalid("diag_inverter", 1, 4, 4, 2); __PYX_ERR(0, 16, __pyx_L3_error) } CYTHON_FALLTHROUGH; case 3: if (likely((values[3] = __Pyx_GetKwValue_FASTCALL(__pyx_kwds, __pyx_kwvalues, __pyx_n_s_n_threads)) != 0)) { (void)__Pyx_Arg_NewRef_FASTCALL(values[3]); kw_args--; } else if (unlikely(PyErr_Occurred())) __PYX_ERR(0, 16, __pyx_L3_error) else { __Pyx_RaiseArgtupleInvalid("diag_inverter", 1, 4, 4, 3); __PYX_ERR(0, 16, __pyx_L3_error) } } if (unlikely(kw_args > 0)) { const Py_ssize_t kwd_pos_args = __pyx_nargs; if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_kwvalues, __pyx_pyargnames, 0, values + 0, kwd_pos_args, "diag_inverter") < 0)) __PYX_ERR(0, 16, __pyx_L3_error) } } else if (unlikely(__pyx_nargs != 4)) { goto __pyx_L5_argtuple_error; } else { values[0] = __Pyx_Arg_FASTCALL(__pyx_args, 0); values[1] = __Pyx_Arg_FASTCALL(__pyx_args, 1); values[2] = __Pyx_Arg_FASTCALL(__pyx_args, 2); values[3] = __Pyx_Arg_FASTCALL(__pyx_args, 3); } __pyx_v_diag_c = __Pyx_PyObject_to_MemoryviewSlice_ds_nn___pyx_t_8inverter_DTYPE_t__const__(values[0], 0); if (unlikely(!__pyx_v_diag_c.memview)) __PYX_ERR(0, 19, __pyx_L3_error) __pyx_v_diag_offset = __Pyx_PyObject_to_MemoryviewSlice_ds_nn___pyx_t_8inverter_DTYPE_t__const__(values[1], 0); if (unlikely(!__pyx_v_diag_offset.memview)) __PYX_ERR(0, 19, __pyx_L3_error) __pyx_v_output = __Pyx_PyObject_to_MemoryviewSlice_dsds_nn___pyx_t_8inverter_DTYPE_t(values[2], PyBUF_WRITABLE); if (unlikely(!__pyx_v_output.memview)) __PYX_ERR(0, 19, __pyx_L3_error) __pyx_v_n_threads = __Pyx_PyInt_As_int(values[3]); if (unlikely((__pyx_v_n_threads == (int)-1) && PyErr_Occurred())) __PYX_ERR(0, 19, __pyx_L3_error) } goto __pyx_L6_skip; __pyx_L5_argtuple_error:; __Pyx_RaiseArgtupleInvalid("diag_inverter", 1, 4, 4, __pyx_nargs); __PYX_ERR(0, 16, __pyx_L3_error) __pyx_L6_skip:; goto __pyx_L4_argument_unpacking_done; __pyx_L3_error:; { Py_ssize_t __pyx_temp; for (__pyx_temp=0; __pyx_temp < (Py_ssize_t)(sizeof(values)/sizeof(values[0])); ++__pyx_temp) { __Pyx_Arg_XDECREF_FASTCALL(values[__pyx_temp]); } } __PYX_XCLEAR_MEMVIEW(&__pyx_v_diag_c, 1); __PYX_XCLEAR_MEMVIEW(&__pyx_v_diag_offset, 1); __PYX_XCLEAR_MEMVIEW(&__pyx_v_output, 1); __Pyx_AddTraceback("inverter.diag_inverter", __pyx_clineno, __pyx_lineno, __pyx_filename); __Pyx_RefNannyFinishContext(); return NULL; __pyx_L4_argument_unpacking_done:; __pyx_r = __pyx_pf_8inverter_diag_inverter(__pyx_self, __pyx_v_diag_c, __pyx_v_diag_offset, __pyx_v_output, __pyx_v_n_threads); int __pyx_lineno = 0; const char *__pyx_filename = NULL; int __pyx_clineno = 0; /* function exit code */ __PYX_XCLEAR_MEMVIEW(&__pyx_v_diag_c, 1); __PYX_XCLEAR_MEMVIEW(&__pyx_v_diag_offset, 1); __PYX_XCLEAR_MEMVIEW(&__pyx_v_output, 1); { Py_ssize_t __pyx_temp; for (__pyx_temp=0; __pyx_temp < (Py_ssize_t)(sizeof(values)/sizeof(values[0])); ++__pyx_temp) { __Pyx_Arg_XDECREF_FASTCALL(values[__pyx_temp]); } } __Pyx_RefNannyFinishContext(); return __pyx_r; } static PyObject *__pyx_pf_8inverter_diag_inverter(CYTHON_UNUSED PyObject *__pyx_self, __Pyx_memviewslice __pyx_v_diag_c, __Pyx_memviewslice __pyx_v_diag_offset, __Pyx_memviewslice __pyx_v_output, CYTHON_UNUSED int __pyx_v_n_threads) { Py_ssize_t __pyx_v_x_max; PyArrayObject *__pyx_v_C_VALUES = 0; PyArrayObject *__pyx_v_inv_C_VALUES = 0; Py_ssize_t __pyx_v_row; Py_ssize_t __pyx_v_column; CYTHON_UNUSED __pyx_t_8inverter_DTYPE_t __pyx_v_temp; Py_ssize_t __pyx_v_i; __pyx_t_8inverter_DTYPE_t __pyx_v_val; __pyx_t_8inverter_DTYPE_t __pyx_v_previous; __Pyx_LocalBuf_ND __pyx_pybuffernd_C_VALUES; __Pyx_Buffer __pyx_pybuffer_C_VALUES; __Pyx_LocalBuf_ND __pyx_pybuffernd_inv_C_VALUES; __Pyx_Buffer __pyx_pybuffer_inv_C_VALUES; PyObject *__pyx_r = NULL; __pyx_pybuffer_C_VALUES.pybuffer.buf = NULL; __pyx_pybuffer_C_VALUES.refcount = 0; __pyx_pybuffernd_C_VALUES.data = NULL; __pyx_pybuffernd_C_VALUES.rcbuffer = &__pyx_pybuffer_C_VALUES; __pyx_pybuffer_inv_C_VALUES.pybuffer.buf = NULL; __pyx_pybuffer_inv_C_VALUES.refcount = 0; __pyx_pybuffernd_inv_C_VALUES.data = NULL; __pyx_pybuffernd_inv_C_VALUES.rcbuffer = &__pyx_pybuffer_inv_C_VALUES; /* … */ /* function exit code */ __pyx_r = Py_None; __Pyx_INCREF(Py_None); goto __pyx_L0; __pyx_L1_error:; __Pyx_XDECREF(__pyx_t_1); __Pyx_XDECREF(__pyx_t_2); __Pyx_XDECREF(__pyx_t_3); __Pyx_XDECREF(__pyx_t_4); { PyObject *__pyx_type, *__pyx_value, *__pyx_tb; __Pyx_PyThreadState_declare __Pyx_PyThreadState_assign __Pyx_ErrFetch(&__pyx_type, &__pyx_value, &__pyx_tb); __Pyx_SafeReleaseBuffer(&__pyx_pybuffernd_C_VALUES.rcbuffer->pybuffer); __Pyx_SafeReleaseBuffer(&__pyx_pybuffernd_inv_C_VALUES.rcbuffer->pybuffer); __Pyx_ErrRestore(__pyx_type, __pyx_value, __pyx_tb);} __Pyx_AddTraceback("inverter.diag_inverter", __pyx_clineno, __pyx_lineno, __pyx_filename); __pyx_r = NULL; goto __pyx_L2; __pyx_L0:; __Pyx_SafeReleaseBuffer(&__pyx_pybuffernd_C_VALUES.rcbuffer->pybuffer); __Pyx_SafeReleaseBuffer(&__pyx_pybuffernd_inv_C_VALUES.rcbuffer->pybuffer); __pyx_L2:; __Pyx_XDECREF((PyObject *)__pyx_v_C_VALUES); __Pyx_XDECREF((PyObject *)__pyx_v_inv_C_VALUES); __Pyx_XGIVEREF(__pyx_r); __Pyx_RefNannyFinishContext(); return __pyx_r; } /* … */ __pyx_tuple__22 = PyTuple_Pack(13, __pyx_n_s_diag_c, __pyx_n_s_diag_offset, __pyx_n_s_output, __pyx_n_s_n_threads, __pyx_n_s_x_max, __pyx_n_s_C_VALUES, __pyx_n_s_inv_C_VALUES, __pyx_n_s_row, __pyx_n_s_column, __pyx_n_s_temp, __pyx_n_s_i, __pyx_n_s_val, __pyx_n_s_previous); if (unlikely(!__pyx_tuple__22)) __PYX_ERR(0, 16, __pyx_L1_error) __Pyx_GOTREF(__pyx_tuple__22); __Pyx_GIVEREF(__pyx_tuple__22); /* … */ __pyx_t_7 = __Pyx_CyFunction_New(&__pyx_mdef_8inverter_1diag_inverter, 0, __pyx_n_s_diag_inverter, NULL, __pyx_n_s_inverter, __pyx_d, ((PyObject *)__pyx_codeobj__23)); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 16, __pyx_L1_error) __Pyx_GOTREF(__pyx_t_7); if (PyDict_SetItem(__pyx_d, __pyx_n_s_diag_inverter, __pyx_t_7) < 0) __PYX_ERR(0, 16, __pyx_L1_error) __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
17: @cython.boundscheck(False)
18: @cython.wraparound(False)
19: def diag_inverter(const DTYPE_t[:] diag_c, const DTYPE_t[:] diag_offset, DTYPE_t[:,:] output, const int n_threads):
20: """
21: Inverts a symmetric tridiagonal matrix
22:
23: Parameters
24: ============
25: diag_c : np.ndarray
26: The central diagonal of the matrix
27: diag_offset : np.ndarray
28: The non-central diagonal
29: output : np.ndarray
30: Array to which the inverted matrix will be written to
31: n_threads : int
32: Number of threads to use
33: """
+34: cdef Py_ssize_t x_max = diag_c.shape[0]
__pyx_v_x_max = (__pyx_v_diag_c.shape[0]);
+35: cdef np.ndarray[DTYPE_t, ndim=1] C_VALUES = np.empty(x_max)
__Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_np); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 35, __pyx_L1_error) __Pyx_GOTREF(__pyx_t_2); __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_empty); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 35, __pyx_L1_error) __Pyx_GOTREF(__pyx_t_3); __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0; __pyx_t_2 = PyInt_FromSsize_t(__pyx_v_x_max); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 35, __pyx_L1_error) __Pyx_GOTREF(__pyx_t_2); __pyx_t_4 = NULL; __pyx_t_5 = 0; #if CYTHON_UNPACK_METHODS if (unlikely(PyMethod_Check(__pyx_t_3))) { __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_3); if (likely(__pyx_t_4)) { PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3); __Pyx_INCREF(__pyx_t_4); __Pyx_INCREF(function); __Pyx_DECREF_SET(__pyx_t_3, function); __pyx_t_5 = 1; } } #endif { PyObject *__pyx_callargs[2] = {__pyx_t_4, __pyx_t_2}; __pyx_t_1 = __Pyx_PyObject_FastCall(__pyx_t_3, __pyx_callargs+1-__pyx_t_5, 1+__pyx_t_5); __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0; __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 35, __pyx_L1_error) __Pyx_GOTREF(__pyx_t_1); __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0; } if (!(likely(((__pyx_t_1) == Py_None) || likely(__Pyx_TypeTest(__pyx_t_1, __pyx_ptype_5numpy_ndarray))))) __PYX_ERR(0, 35, __pyx_L1_error) __pyx_t_6 = ((PyArrayObject *)__pyx_t_1); { __Pyx_BufFmt_StackElem __pyx_stack[1]; if (unlikely(__Pyx_GetBufferAndValidate(&__pyx_pybuffernd_C_VALUES.rcbuffer->pybuffer, (PyObject*)__pyx_t_6, &__Pyx_TypeInfo_nn___pyx_t_8inverter_DTYPE_t, PyBUF_FORMAT| PyBUF_STRIDES| PyBUF_WRITABLE, 1, 0, __pyx_stack) == -1)) { __pyx_v_C_VALUES = ((PyArrayObject *)Py_None); __Pyx_INCREF(Py_None); __pyx_pybuffernd_C_VALUES.rcbuffer->pybuffer.buf = NULL; __PYX_ERR(0, 35, __pyx_L1_error) } else {__pyx_pybuffernd_C_VALUES.diminfo[0].strides = __pyx_pybuffernd_C_VALUES.rcbuffer->pybuffer.strides[0]; __pyx_pybuffernd_C_VALUES.diminfo[0].shape = __pyx_pybuffernd_C_VALUES.rcbuffer->pybuffer.shape[0]; } } __pyx_t_6 = 0; __pyx_v_C_VALUES = ((PyArrayObject *)__pyx_t_1); __pyx_t_1 = 0;
+36: cdef np.ndarray[DTYPE_t, ndim=1] inv_C_VALUES = np.empty((x_max))
__Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_np); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 36, __pyx_L1_error) __Pyx_GOTREF(__pyx_t_3); __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_empty); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 36, __pyx_L1_error) __Pyx_GOTREF(__pyx_t_2); __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0; __pyx_t_3 = PyInt_FromSsize_t(__pyx_v_x_max); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 36, __pyx_L1_error) __Pyx_GOTREF(__pyx_t_3); __pyx_t_4 = NULL; __pyx_t_5 = 0; #if CYTHON_UNPACK_METHODS if (unlikely(PyMethod_Check(__pyx_t_2))) { __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_2); if (likely(__pyx_t_4)) { PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2); __Pyx_INCREF(__pyx_t_4); __Pyx_INCREF(function); __Pyx_DECREF_SET(__pyx_t_2, function); __pyx_t_5 = 1; } } #endif { PyObject *__pyx_callargs[2] = {__pyx_t_4, __pyx_t_3}; __pyx_t_1 = __Pyx_PyObject_FastCall(__pyx_t_2, __pyx_callargs+1-__pyx_t_5, 1+__pyx_t_5); __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0; __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 36, __pyx_L1_error) __Pyx_GOTREF(__pyx_t_1); __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0; } if (!(likely(((__pyx_t_1) == Py_None) || likely(__Pyx_TypeTest(__pyx_t_1, __pyx_ptype_5numpy_ndarray))))) __PYX_ERR(0, 36, __pyx_L1_error) __pyx_t_7 = ((PyArrayObject *)__pyx_t_1); { __Pyx_BufFmt_StackElem __pyx_stack[1]; if (unlikely(__Pyx_GetBufferAndValidate(&__pyx_pybuffernd_inv_C_VALUES.rcbuffer->pybuffer, (PyObject*)__pyx_t_7, &__Pyx_TypeInfo_nn___pyx_t_8inverter_DTYPE_t, PyBUF_FORMAT| PyBUF_STRIDES| PyBUF_WRITABLE, 1, 0, __pyx_stack) == -1)) { __pyx_v_inv_C_VALUES = ((PyArrayObject *)Py_None); __Pyx_INCREF(Py_None); __pyx_pybuffernd_inv_C_VALUES.rcbuffer->pybuffer.buf = NULL; __PYX_ERR(0, 36, __pyx_L1_error) } else {__pyx_pybuffernd_inv_C_VALUES.diminfo[0].strides = __pyx_pybuffernd_inv_C_VALUES.rcbuffer->pybuffer.strides[0]; __pyx_pybuffernd_inv_C_VALUES.diminfo[0].shape = __pyx_pybuffernd_inv_C_VALUES.rcbuffer->pybuffer.shape[0]; } } __pyx_t_7 = 0; __pyx_v_inv_C_VALUES = ((PyArrayObject *)__pyx_t_1); __pyx_t_1 = 0;
37:
+38: cdef Py_ssize_t row = 0
__pyx_v_row = 0;
+39: cdef Py_ssize_t column = 0
__pyx_v_column = 0;
+40: cdef DTYPE_t temp = 0
__pyx_v_temp = 0.0;
41:
+42: cdef Py_ssize_t i = 0
__pyx_v_i = 0;
+43: cdef DTYPE_t val = 1
__pyx_v_val = 1.0;
+44: cdef DTYPE_t previous = 1
__pyx_v_previous = 1.0;
45:
+46: C_VALUES[0] = diag_c[0]
__pyx_t_8 = 0; __pyx_t_9 = 0; *__Pyx_BufPtrStrided1d(__pyx_t_8inverter_DTYPE_t *, __pyx_pybuffernd_C_VALUES.rcbuffer->pybuffer.buf, __pyx_t_9, __pyx_pybuffernd_C_VALUES.diminfo[0].strides) = (*((__pyx_t_8inverter_DTYPE_t const *) ( /* dim=0 */ (__pyx_v_diag_c.data + __pyx_t_8 * __pyx_v_diag_c.strides[0]) )));
+47: inv_C_VALUES[0] = 1/C_VALUES[0]
__pyx_t_8 = 0; __pyx_t_9 = 0; *__Pyx_BufPtrStrided1d(__pyx_t_8inverter_DTYPE_t *, __pyx_pybuffernd_inv_C_VALUES.rcbuffer->pybuffer.buf, __pyx_t_9, __pyx_pybuffernd_inv_C_VALUES.diminfo[0].strides) = (1.0 / (*__Pyx_BufPtrStrided1d(__pyx_t_8inverter_DTYPE_t *, __pyx_pybuffernd_C_VALUES.rcbuffer->pybuffer.buf, __pyx_t_8, __pyx_pybuffernd_C_VALUES.diminfo[0].strides)));
48:
+49: for i in range(1, x_max):
__pyx_t_10 = __pyx_v_x_max;
__pyx_t_11 = __pyx_t_10;
for (__pyx_t_12 = 1; __pyx_t_12 < __pyx_t_11; __pyx_t_12+=1) {
__pyx_v_i = __pyx_t_12;
+50: C_VALUES[i] = diag_c[i] - diag_offset[i-1]**2/C_VALUES[i-1]
__pyx_t_8 = __pyx_v_i;
__pyx_t_9 = (__pyx_v_i - 1);
__pyx_t_13 = (__pyx_v_i - 1);
__pyx_t_14 = __pyx_v_i;
*__Pyx_BufPtrStrided1d(__pyx_t_8inverter_DTYPE_t *, __pyx_pybuffernd_C_VALUES.rcbuffer->pybuffer.buf, __pyx_t_14, __pyx_pybuffernd_C_VALUES.diminfo[0].strides) = ((*((__pyx_t_8inverter_DTYPE_t const *) ( /* dim=0 */ (__pyx_v_diag_c.data + __pyx_t_8 * __pyx_v_diag_c.strides[0]) ))) - (pow(((__pyx_t_8inverter_DTYPE_t)(*((__pyx_t_8inverter_DTYPE_t const *) ( /* dim=0 */ (__pyx_v_diag_offset.data + __pyx_t_9 * __pyx_v_diag_offset.strides[0]) )))), 2.0) / (*__Pyx_BufPtrStrided1d(__pyx_t_8inverter_DTYPE_t *, __pyx_pybuffernd_C_VALUES.rcbuffer->pybuffer.buf, __pyx_t_13, __pyx_pybuffernd_C_VALUES.diminfo[0].strides))));
+51: inv_C_VALUES[i] = 1/C_VALUES[i]
__pyx_t_13 = __pyx_v_i;
__pyx_t_9 = __pyx_v_i;
*__Pyx_BufPtrStrided1d(__pyx_t_8inverter_DTYPE_t *, __pyx_pybuffernd_inv_C_VALUES.rcbuffer->pybuffer.buf, __pyx_t_9, __pyx_pybuffernd_inv_C_VALUES.diminfo[0].strides) = (1.0 / (*__Pyx_BufPtrStrided1d(__pyx_t_8inverter_DTYPE_t *, __pyx_pybuffernd_C_VALUES.rcbuffer->pybuffer.buf, __pyx_t_13, __pyx_pybuffernd_C_VALUES.diminfo[0].strides)));
}
52:
+53: for row in prange(x_max, nogil=True, num_threads=n_threads):
{
#ifdef WITH_THREAD
PyThreadState *_save;
_save = NULL;
Py_UNBLOCK_THREADS
__Pyx_FastGIL_Remember();
#endif
/*try:*/ {
__pyx_t_10 = __pyx_v_x_max;
{
#if ((defined(__APPLE__) || defined(__OSX__)) && (defined(__GNUC__) && (__GNUC__ > 2 || (__GNUC__ == 2 && (__GNUC_MINOR__ > 95)))))
#undef likely
#undef unlikely
#define likely(x) (x)
#define unlikely(x) (x)
#endif
__pyx_t_12 = (__pyx_t_10 - 0 + 1 - 1/abs(1)) / 1;
if (__pyx_t_12 > 0)
{
#ifdef _OPENMP
#pragma omp parallel
#endif /* _OPENMP */
{
#ifdef _OPENMP
#pragma omp for lastprivate(__pyx_v_column) lastprivate(__pyx_v_i) lastprivate(__pyx_v_previous) firstprivate(__pyx_v_row) lastprivate(__pyx_v_row) lastprivate(__pyx_v_val) __pyx_t_10 = __pyx_v_x_max;
{
#if ((defined(__APPLE__) || defined(__OSX__)) && (defined(__GNUC__) && (__GNUC__ > 2 || (__GNUC__ == 2 && (__GNUC_MINOR__ > 95)))))
#undef likely
#undef unlikely
#define likely(x) (x)
#define unlikely(x) (x)
#endif
__pyx_t_12 = (__pyx_t_10 - 0 + 1 - 1/abs(1)) / 1;
if (__pyx_t_12 > 0)
{
#ifdef _OPENMP
#pragma omp parallel
#endif /* _OPENMP */
{
#ifdef _OPENMP
#pragma omp for lastprivate(__pyx_v_column) lastprivate(__pyx_v_i) lastprivate(__pyx_v_previous) firstprivate(__pyx_v_row) lastprivate(__pyx_v_row) lastprivate(__pyx_v_val) num_threads(__pyx_v_n_threads)
#endif /* _OPENMP */
for (__pyx_t_11 = 0; __pyx_t_11 < __pyx_t_12; __pyx_t_11++){
{
__pyx_v_row = (Py_ssize_t)(0 + 1 * __pyx_t_11);
/* Initialize private variables to invalid values */
__pyx_v_column = ((Py_ssize_t)0xbad0bad0);
__pyx_v_i = ((Py_ssize_t)0xbad0bad0);
__pyx_v_previous = ((__pyx_t_8inverter_DTYPE_t)__PYX_NAN());
__pyx_v_val = ((__pyx_t_8inverter_DTYPE_t)__PYX_NAN());
/* … */
/*finally:*/ {
/*normal exit:*/{
#ifdef WITH_THREAD
__Pyx_FastGIL_Forget();
Py_BLOCK_THREADS
#endif
goto __pyx_L7;
}
__pyx_L7:;
}
}
54: #for row in range(x_max):
+55: previous = 1
__pyx_v_previous = 1.0;
+56: for column in range(row+1, x_max):
__pyx_t_15 = __pyx_v_x_max;
__pyx_t_16 = __pyx_t_15;
for (__pyx_t_17 = (__pyx_v_row + 1); __pyx_t_17 < __pyx_t_16; __pyx_t_17+=1) {
__pyx_v_column = __pyx_t_17;
57: # non diagonal terms, without accounting for w_ii
+58: val = previous * diag_offset[column-1]*inv_C_VALUES[column-1] * (-1)
__pyx_t_13 = (__pyx_v_column - 1);
__pyx_t_9 = (__pyx_v_column - 1);
__pyx_v_val = (((__pyx_v_previous * (*((__pyx_t_8inverter_DTYPE_t const *) ( /* dim=0 */ (__pyx_v_diag_offset.data + __pyx_t_13 * __pyx_v_diag_offset.strides[0]) )))) * (*__Pyx_BufPtrStrided1d(__pyx_t_8inverter_DTYPE_t *, __pyx_pybuffernd_inv_C_VALUES.rcbuffer->pybuffer.buf, __pyx_t_9, __pyx_pybuffernd_inv_C_VALUES.diminfo[0].strides))) * -1.0);
+59: previous = val
__pyx_v_previous = __pyx_v_val;
+60: output[row, column] = val
__pyx_t_9 = __pyx_v_row;
__pyx_t_13 = __pyx_v_column;
*((__pyx_t_8inverter_DTYPE_t *) ( /* dim=1 */ (( /* dim=0 */ (__pyx_v_output.data + __pyx_t_9 * __pyx_v_output.strides[0]) ) + __pyx_t_13 * __pyx_v_output.strides[1]) )) = __pyx_v_val;
+61: output[column, row] = val
__pyx_t_13 = __pyx_v_column;
__pyx_t_9 = __pyx_v_row;
*((__pyx_t_8inverter_DTYPE_t *) ( /* dim=1 */ (( /* dim=0 */ (__pyx_v_output.data + __pyx_t_13 * __pyx_v_output.strides[0]) ) + __pyx_t_9 * __pyx_v_output.strides[1]) )) = __pyx_v_val;
}
62:
+63: output[row, row] = inv_C_VALUES[row]
__pyx_t_9 = __pyx_v_row;
__pyx_t_13 = __pyx_v_row;
__pyx_t_8 = __pyx_v_row;
*((__pyx_t_8inverter_DTYPE_t *) ( /* dim=1 */ (( /* dim=0 */ (__pyx_v_output.data + __pyx_t_13 * __pyx_v_output.strides[0]) ) + __pyx_t_8 * __pyx_v_output.strides[1]) )) = (*__Pyx_BufPtrStrided1d(__pyx_t_8inverter_DTYPE_t *, __pyx_pybuffernd_inv_C_VALUES.rcbuffer->pybuffer.buf, __pyx_t_9, __pyx_pybuffernd_inv_C_VALUES.diminfo[0].strides));
+64: if row != x_max - 1: # compute main diagonal values!
__pyx_t_18 = (__pyx_v_row != (__pyx_v_x_max - 1));
if (__pyx_t_18) {
/* … */
}
}
}
}
}
}
#if ((defined(__APPLE__) || defined(__OSX__)) && (defined(__GNUC__) && (__GNUC__ > 2 || (__GNUC__ == 2 && (__GNUC_MINOR__ > 95)))))
#undef likely
#undef unlikely
#define likely(x) __builtin_expect(!!(x), 1)
#define unlikely(x) __builtin_expect(!!(x), 0)
#endif
}
+65: val = 0
__pyx_v_val = 0.0;
+66: for i in range(row+1, x_max):
__pyx_t_15 = __pyx_v_x_max;
__pyx_t_16 = __pyx_t_15;
for (__pyx_t_17 = (__pyx_v_row + 1); __pyx_t_17 < __pyx_t_16; __pyx_t_17+=1) {
__pyx_v_i = __pyx_t_17;
+67: val = val + (inv_C_VALUES[i]) * output[row, i]**2
__pyx_t_9 = __pyx_v_i;
__pyx_t_8 = __pyx_v_row;
__pyx_t_13 = __pyx_v_i;
__pyx_v_val = (__pyx_v_val + ((*__Pyx_BufPtrStrided1d(__pyx_t_8inverter_DTYPE_t *, __pyx_pybuffernd_inv_C_VALUES.rcbuffer->pybuffer.buf, __pyx_t_9, __pyx_pybuffernd_inv_C_VALUES.diminfo[0].strides)) * pow((*((__pyx_t_8inverter_DTYPE_t *) ( /* dim=1 */ (( /* dim=0 */ (__pyx_v_output.data + __pyx_t_8 * __pyx_v_output.strides[0]) ) + __pyx_t_13 * __pyx_v_output.strides[1]) ))), 2.0)));
}
68: #output[row, row] = output[row, row] + np.dot(1/C_VALUES[row+1:], np.square(output[row, row+1:]))
+69: output[row, row] = output[row, row] + val
__pyx_t_13 = __pyx_v_row;
__pyx_t_8 = __pyx_v_row;
__pyx_t_9 = __pyx_v_row;
__pyx_t_14 = __pyx_v_row;
*((__pyx_t_8inverter_DTYPE_t *) ( /* dim=1 */ (( /* dim=0 */ (__pyx_v_output.data + __pyx_t_9 * __pyx_v_output.strides[0]) ) + __pyx_t_14 * __pyx_v_output.strides[1]) )) = ((*((__pyx_t_8inverter_DTYPE_t *) ( /* dim=1 */ (( /* dim=0 */ (__pyx_v_output.data + __pyx_t_13 * __pyx_v_output.strides[0]) ) + __pyx_t_8 * __pyx_v_output.strides[1]) ))) + __pyx_v_val);
70:
+71: for row in prange(x_max, nogil=True, num_threads=n_threads):
{
#ifdef WITH_THREAD
PyThreadState *_save;
_save = NULL;
Py_UNBLOCK_THREADS
__Pyx_FastGIL_Remember();
#endif
/*try:*/ {
__pyx_t_12 = __pyx_v_x_max;
{
#if ((defined(__APPLE__) || defined(__OSX__)) && (defined(__GNUC__) && (__GNUC__ > 2 || (__GNUC__ == 2 && (__GNUC_MINOR__ > 95)))))
#undef likely
#undef unlikely
#define likely(x) (x)
#define unlikely(x) (x)
#endif
__pyx_t_10 = (__pyx_t_12 - 0 + 1 - 1/abs(1)) / 1;
if (__pyx_t_10 > 0)
{
#ifdef _OPENMP
#pragma omp parallel
#endif /* _OPENMP */
{
#ifdef _OPENMP
#pragma omp for lastprivate(__pyx_v_column) firstprivate(__pyx_v_row) lastprivate(__pyx_v_row) __pyx_t_12 = __pyx_v_x_max;
{
#if ((defined(__APPLE__) || defined(__OSX__)) && (defined(__GNUC__) && (__GNUC__ > 2 || (__GNUC__ == 2 && (__GNUC_MINOR__ > 95)))))
#undef likely
#undef unlikely
#define likely(x) (x)
#define unlikely(x) (x)
#endif
__pyx_t_10 = (__pyx_t_12 - 0 + 1 - 1/abs(1)) / 1;
if (__pyx_t_10 > 0)
{
#ifdef _OPENMP
#pragma omp parallel
#endif /* _OPENMP */
{
#ifdef _OPENMP
#pragma omp for lastprivate(__pyx_v_column) firstprivate(__pyx_v_row) lastprivate(__pyx_v_row) num_threads(__pyx_v_n_threads)
#endif /* _OPENMP */
for (__pyx_t_11 = 0; __pyx_t_11 < __pyx_t_10; __pyx_t_11++){
{
__pyx_v_row = (Py_ssize_t)(0 + 1 * __pyx_t_11);
/* Initialize private variables to invalid values */
__pyx_v_column = ((Py_ssize_t)0xbad0bad0);
/* … */
/*finally:*/ {
/*normal exit:*/{
#ifdef WITH_THREAD
__Pyx_FastGIL_Forget();
Py_BLOCK_THREADS
#endif
goto __pyx_L21;
}
__pyx_L21:;
}
}
72: # add the diagonal value to the non-diagonal values
73: # this is done this way to avoid problems when re-using output[i,j] to store
74: # data until it is needed further ahead
+75: for column in range(row+1, x_max):
__pyx_t_15 = __pyx_v_x_max;
__pyx_t_16 = __pyx_t_15;
for (__pyx_t_17 = (__pyx_v_row + 1); __pyx_t_17 < __pyx_t_16; __pyx_t_17+=1) {
__pyx_v_column = __pyx_t_17;
+76: output[row, column] = output[row, column] * output[column,column]
__pyx_t_8 = __pyx_v_row;
__pyx_t_13 = __pyx_v_column;
__pyx_t_14 = __pyx_v_column;
__pyx_t_9 = __pyx_v_column;
__pyx_t_19 = __pyx_v_row;
__pyx_t_20 = __pyx_v_column;
*((__pyx_t_8inverter_DTYPE_t *) ( /* dim=1 */ (( /* dim=0 */ (__pyx_v_output.data + __pyx_t_19 * __pyx_v_output.strides[0]) ) + __pyx_t_20 * __pyx_v_output.strides[1]) )) = ((*((__pyx_t_8inverter_DTYPE_t *) ( /* dim=1 */ (( /* dim=0 */ (__pyx_v_output.data + __pyx_t_8 * __pyx_v_output.strides[0]) ) + __pyx_t_13 * __pyx_v_output.strides[1]) ))) * (*((__pyx_t_8inverter_DTYPE_t *) ( /* dim=1 */ (( /* dim=0 */ (__pyx_v_output.data + __pyx_t_14 * __pyx_v_output.strides[0]) ) + __pyx_t_9 * __pyx_v_output.strides[1]) ))));
+77: output[column, row] = output[column, row] * output[column,column]
__pyx_t_9 = __pyx_v_column;
__pyx_t_14 = __pyx_v_row;
__pyx_t_13 = __pyx_v_column;
__pyx_t_8 = __pyx_v_column;
__pyx_t_20 = __pyx_v_column;
__pyx_t_19 = __pyx_v_row;
*((__pyx_t_8inverter_DTYPE_t *) ( /* dim=1 */ (( /* dim=0 */ (__pyx_v_output.data + __pyx_t_20 * __pyx_v_output.strides[0]) ) + __pyx_t_19 * __pyx_v_output.strides[1]) )) = ((*((__pyx_t_8inverter_DTYPE_t *) ( /* dim=1 */ (( /* dim=0 */ (__pyx_v_output.data + __pyx_t_9 * __pyx_v_output.strides[0]) ) + __pyx_t_14 * __pyx_v_output.strides[1]) ))) * (*((__pyx_t_8inverter_DTYPE_t *) ( /* dim=1 */ (( /* dim=0 */ (__pyx_v_output.data + __pyx_t_13 * __pyx_v_output.strides[0]) ) + __pyx_t_8 * __pyx_v_output.strides[1]) ))));
}
}
}
}
}
}
#if ((defined(__APPLE__) || defined(__OSX__)) && (defined(__GNUC__) && (__GNUC__ > 2 || (__GNUC__ == 2 && (__GNUC_MINOR__ > 95)))))
#undef likely
#undef unlikely
#define likely(x) __builtin_expect(!!(x), 1)
#define unlikely(x) __builtin_expect(!!(x), 0)
#endif
}