Enable optional PyTorch solver

[slarson]

Summary

• allow selecting a PyTorch-based backend
• expose new torchEnabled() config helper
• manage Python solver lifetime in owPhysicsFluidSimulator
• call into Python solver routines instead of OpenCL when selected
• return position/velocity from Python solver via new get_state

Testing

• ./test.sh

---

https://chatgpt.com/codex/tasks/task_b_6861baca2b58832a80a732106b93599b

[Copilot]

Pull Request Overview

This PR enables an optional PyTorch-based backend for the fluid simulator by adding a new Torch solver and associated configuration, managing its Python interpreter lifecycle, and updating simulation routines accordingly.

• Introduces a torchEnabled flag and related configuration in owConfigProperty.
• Implements initialization and reset routines for the PyTorch solver in owPhysicsFluidSimulator.
• Updates the Python pytorch_solver module with additional methods including a get_state routine.

Reviewed Changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
src/owPhysicsFluidSimulator.cpp	Adds PyTorch solver initialization, simulation step updates, and Python cleanup.
src/owConfigProperty.cpp	Introduces a new flag and command-line processing for torch-based simulation.
pytorch_solver.py	Refactors some functions and adds a get_state method for solver state retrieval.
inc/owPhysicsFluidSimulator.h	Declares new torchSolver and useTorchBackend members.
inc/owConfigProperty.h	Adds accessor and member for torchEnabled and useTorch flag.

[Copilot]

The initialization logic for the torchSolver is duplicated between the constructor and reset method. Consider refactoring this code into a helper function to improve maintainability.

[Copilot]

The series of PyObject_CallMethod calls in simulationStep do not check their return values. Consider verifying these calls to handle potential failures in the Python solver routines.

Suggested change

	PyObject_CallMethod(torchSolver, "run_hash_particles", nullptr);
	PyObject_CallMethod(torchSolver, "run_sort", nullptr);
	PyObject_CallMethod(torchSolver, "run_index", nullptr);
	PyObject_CallMethod(torchSolver, "run_index_post_pass", nullptr);
	PyObject_CallMethod(torchSolver, "run_find_neighbors", nullptr);
	PyObject_CallMethod(torchSolver, "run_compute_density", nullptr);
	PyObject_CallMethod(torchSolver, "run_compute_pressure", nullptr);
	PyObject_CallMethod(torchSolver, "run_compute_pressure_force_acceleration",
	nullptr);
	PyObject_CallMethod(torchSolver, "run_integrate", nullptr);
	PyObject *result = PyObject_CallMethod(torchSolver, "run_hash_particles", nullptr);
	if (!result) { PyErr_Print(); return -1; } Py_XDECREF(result);
	result = PyObject_CallMethod(torchSolver, "run_sort", nullptr);
	if (!result) { PyErr_Print(); return -1; } Py_XDECREF(result);
	result = PyObject_CallMethod(torchSolver, "run_index", nullptr);
	if (!result) { PyErr_Print(); return -1; } Py_XDECREF(result);
	result = PyObject_CallMethod(torchSolver, "run_index_post_pass", nullptr);
	if (!result) { PyErr_Print(); return -1; } Py_XDECREF(result);
	result = PyObject_CallMethod(torchSolver, "run_find_neighbors", nullptr);
	if (!result) { PyErr_Print(); return -1; } Py_XDECREF(result);
	result = PyObject_CallMethod(torchSolver, "run_compute_density", nullptr);
	if (!result) { PyErr_Print(); return -1; } Py_XDECREF(result);
	result = PyObject_CallMethod(torchSolver, "run_compute_pressure", nullptr);
	if (!result) { PyErr_Print(); return -1; } Py_XDECREF(result);
	result = PyObject_CallMethod(torchSolver, "run_compute_pressure_force_acceleration", nullptr);
	if (!result) { PyErr_Print(); return -1; } Py_XDECREF(result);
	result = PyObject_CallMethod(torchSolver, "run_integrate", nullptr);
	if (!result) { PyErr_Print(); return -1; } Py_XDECREF(result);