CyRK.cy¶

Submodules¶

CyRK.cy.common module¶

class CyRK.cy.common.CyrkErrorCodes(*values)¶

Bases: IntEnum

ARGUMENT_ERROR = -10¶

ARGUMENT_NOT_SET = -9¶

ATTRIBUTE_ERROR = -7¶

BAD_CONFIG_DATA = -22¶

BAD_INITIAL_STEP_SIZE = -80¶

BOUNDS_ERROR = -8¶

CONVERGED = 20¶

CYSOLVER_INITIALIZATION_ERROR = -5¶

DENSE_BUILD_FAILED = -56¶

DENSE_OUTPUT_NOT_SAVED = -21¶

ERROR_IMPORTING_PYTHON_MODULE = -70¶

EVENT_SETUP_FAILED = -60¶

EVENT_TERMINATED = 2¶

GENERAL_ERROR = -1¶

INCOMPATIBLE_INPUT = -6¶

INITIALIZING = 10¶

INTEGRATION_NOT_SUCCESSFUL = -57¶

JACOBIAN_IS_SINGULAR = -90¶

LSODA_INTERNAL_ERROR = -92¶

MAX_ITERATIONS_HIT = -51¶

MAX_STEPS_SYSARCH_EXCEEDED = -53¶

MAX_STEPS_USER_EXCEEDED = -52¶

MEMORY_ALLOCATION_ERROR = -40¶

NEWTON_CONVERGENCE_ERROR = -91¶

NO_ERROR = 0¶

NUMBER_OF_EQUATIONS_IS_ZERO = -50¶

OPTIMIZE_CONVERGENCE_ERROR = -31¶

OPTIMIZE_SIGN_ERROR = -30¶

OTHER_ERROR = -99¶

PROPERTY_NOT_SET = -2¶

PYTHON_FUNCTION_ERROR = -71¶

SETUP_NOT_CALLED = -20¶

STEP_SIZE_ERROR_ACCEPTANCE = -55¶

STEP_SIZE_ERROR_SPACING = -54¶

SUCCESSFUL_INTEGRATION = 1¶

UNINITIALIZED_CLASS = -4¶

UNSET_ERROR_CODE = -100¶

UNSUPPORTED_UNKNOWN_MODEL = -3¶

VECTOR_SIZE_EXCEEDS_LIMITS = -41¶

CyRK.cy.common.get_error_message(error_code)¶

Return the message for a CyRK status or error code.

Parameters:: error_code (int) – A CyrkErrorCodes member or its integer value.
Returns:: The code’s message, or a generic “unknown error code” message if CyRK defines none for it.
Return type:: str

CyRK.cy.cysolver_api module¶

class CyRK.cy.cysolver_api.ODEMethod(*values)¶

Bases: IntEnum

BASE_METHOD = 1¶

BDF = 6¶

DOP853 = 5¶

LSODA = 7¶

NO_METHOD_SET = 0¶

RADAU = 8¶

RK23 = 3¶

RK45 = 4¶

RK_BASE_METHOD = 2¶

TSIT5 = 9¶

VERN7 = 10¶

VERN8 = 11¶

class CyRK.cy.cysolver_api.WrapCySolverResult¶

Bases: object

Wrapper for the C++ class CySolverResult defined in “cysolution.cpp”

call(t)¶: Call the dense output interpolater and return y

call_vectorize(t_view)¶: Call the dense output interpolater and return y

error_code¶

event_terminate_index¶

event_terminated¶

finalize()¶

integration_method¶: Returns a string stating what integration method was used for this solution.

message¶

method¶: Alias for self.integration_method.

num_dy¶

num_events¶

num_y¶

print_diagnostics()¶

size¶

solve()¶

status¶

status_message¶

steps_taken¶

success¶

t¶

t_events¶

y¶

y_events¶

CyRK.cy.cysolver_test module¶

CyRK.cy.cysolver_test.cy_extra_output_tester()¶

CyRK.cy.cysolver_test.cytester(diffeq_number, t_span=None, y0=None, args=None, method='rk45', expected_size=0, max_num_steps=0, max_ram_MB=2000, dense_output=False, use_events=False, t_eval=None, rtol=0.001, atol=1e-06, rtol_array=None, atol_array=None, max_step=inf, first_step=0.0, solution_reuse=None, force_retain_solver=False, repeats=1, use_jacobian=False)¶

CyRK.cy.events module¶

CyRK.cy.events_test module¶

CyRK.cy.events_test.build_event_wrapper_test()¶

CyRK.cy.events_test.run_cysolver_with_events(use_dense, t_eval, use_termination, capture_extra)¶

CyRK.cy.helpers module¶

CyRK.cy.parallel_test module¶

Checks that CySolver can be driven from several C++ threads at once.

Every job solves the same Lotka-Volterra problem through c_cysolve_batch, and the results are compared with the first job’s, so any interference between threads shows up as a mismatch. Each runner returns its wall-clock time so the tests can compare thread counts. This module is also the worked example for CyRK.cy.parallel.

CyRK.cy.parallel_test.run_dense_output_threads_test(num_threads=4, num_repeats=100)¶

Solve a Lorenz problem with extra outputs once, then read its dense output from num_threads C++ threads.

Every thread compares each read, extra outputs included, with a serial read at the same time. A dense read only reads the shared solution, so no read may differ. Returns the number of reads that did.

CyRK.cy.parallel_test.run_parallel_benchmark(num_threads, num_runs, t_end)¶

Time num_runs identical Lotka-Volterra solves to t_end on num_threads threads; returns the ms taken.

Use it to find where threading starts to pay off on your machine. The per-job time grows with t_end (the solution is periodic, so the step count is proportional to it) on top of the solver’s fixed setup cost.

CyRK.cy.parallel_test.run_parallel_common_args_test(num_threads=2)¶: Solve a batch where every job reads the same argument vector, spread over num_threads C++ threads.

CyRK.cy.parallel_test.run_parallel_test(num_threads=2)¶: Solve a batch where each job has its own argument vector, spread over num_threads C++ threads.

CyRK.cy.pyhelpers module¶

CyRK.cy.pyhelpers.find_ode_method_int(ode_method_name: str)[source]¶

CyRK.cy.pysolver module¶

class CyRK.cy.pysolver.PySolver¶

Bases: WrapCySolverResult

call(t)¶: Call the dense output interpolater and return y

call_vectorize(t_view)¶: Call the dense output interpolater and return y

set_problem_parameters(py_diffeq, time_span, y0, method='RK45', t_eval=None, dense_output=False, events=None, args=None, expected_size=0, num_extra=0, first_step=0.0, max_step=inf, rtol=0.001, atol=1e-06, max_num_steps=0, max_ram_MB=2000, pass_dy_as_arg=False, force_retain_solver=True, min_step=0.0, lband=None, uband=None)¶

set_pydiffeq(diffeq_func, args, num_y, num_dy, pass_dy_as_arg=False)¶

solve()¶

CyRK.cy.pysolver.pysolve_ivp(py_diffeq, time_span, y0, method='RK45', t_eval=None, dense_output=False, events=None, args=None, expected_size=0, num_extra=0, first_step=0.0, max_step=inf, rtol=0.001, atol=1e-06, max_num_steps=0, max_ram_MB=2000, pass_dy_as_arg=False, solution_reuse=None, force_retain_solver=True, min_step=0.0, lband=None, uband=None)¶