Struct DenseKernel 

pub struct DenseKernel { /* private fields */ }

The dense simulation kernel - zero-allocation hot loop.

This kernel is designed for maximum throughput (>10M ticks/sec) by:

• Using zero-copy data access
• Operating on raw memory via AccessPolicy
• Avoiding allocations in the hot loop

Implementations

impl DenseKernel

pub fn new(config: DenseKernelConfig) -> Self

Create a new dense kernel with the given configuration.

pub fn run<S, F>( &self, data: &DataTable, initial_state: S, strategy: F, ) -> Result<DenseKernelResult<S>>

where F: FnMut(usize, &[*const f64], &mut S) -> i32,

Run simulation on a DataTable with a closure-based strategy.

Extracts f64 column pointers from the DataTable and iterates effective_start..config.end, calling the strategy each tick.

pub unsafe fn run_jit( &self, column_ptrs: &[*const f64], state_ptr: *mut u8, kernel: SimulationKernelFn, ) -> Result<DenseKernelResult<()>>

Run simulation with JIT-compiled kernel function.

This is the highest-performance path - bypasses SeriesCursor and calls the kernel directly with raw pointers.

Arguments

• column_ptrs - Slice of column data pointers
• state_ptr - Mutable pointer to state buffer (TypedObject)
• kernel - JIT-compiled kernel function

Returns

Result containing execution statistics (state is modified in-place).

Safety

The column_ptrs must point to valid f64 arrays with length >= self.config.end. The state_ptr must point to a valid TypedObject buffer.