voxcity.simulator_gpu.init_taichi
=================================

.. py:module:: voxcity.simulator_gpu.init_taichi

.. autoapi-nested-parse::

   Taichi initialization for simulator_gpu.

   This module provides centralized Taichi initialization to ensure ti.init()
   is called before any Taichi fields or kernels are used.



Attributes
----------

.. autoapisummary::

   voxcity.simulator_gpu.init_taichi.taichi


Functions
---------

.. autoapisummary::

   voxcity.simulator_gpu.init_taichi.init_taichi
   voxcity.simulator_gpu.init_taichi.ensure_initialized
   voxcity.simulator_gpu.init_taichi.is_initialized
   voxcity.simulator_gpu.init_taichi.reset


Module Contents
---------------

.. py:function:: init_taichi(arch: Optional[str] = None, default_fp: type = ti.f32, default_ip: type = ti.i32, debug: bool = False, suppress_fp16_warnings: bool = True, honor_env: bool = True, **kwargs) -> bool

   Initialize Taichi runtime if not already initialized.

   This function is idempotent - calling it multiple times is safe.
   The first call will initialize Taichi, subsequent calls will be no-ops.

   :param arch: Architecture to use. Options:
                - None (default): Auto-detect best available (GPU preferred)
                - 'gpu': Use GPU (CUDA, Vulkan, Metal, etc.)
                - 'cuda': Use CUDA specifically
                - 'vulkan': Use Vulkan
                - 'metal': Use Metal (macOS)
                - 'cpu': Use CPU
   :param default_fp: Default floating point type (ti.f32 or ti.f64)
   :param default_ip: Default integer type (ti.i32 or ti.i64)
   :param debug: Enable debug mode for better error messages
   :param suppress_fp16_warnings: Suppress Taichi's fp16 precision loss warnings
                                  (default: True). These warnings occur when using fp16 intermediate
                                  buffers for memory bandwidth optimization.
   :param \*\*kwargs: Additional arguments passed to ti.init()

   :returns: True if initialization was performed, False if already initialized.


.. py:function:: ensure_initialized()

   Ensure Taichi is initialized with default settings.

   This is a convenience function for lazy initialization.
   Call this before any Taichi operations if you're not sure
   whether init_taichi() has been called.


.. py:function:: is_initialized() -> bool

   Check if Taichi has been initialized.


.. py:function:: reset()

   Reset initialization state.

   Note: This does NOT reset Taichi itself (which cannot be reset once initialized).
   This only resets the tracking flag for testing purposes.


.. py:data:: taichi