Kernel and execution¶

This page describes the simulation kernel and how it implements the execution model. For how .sitar source is translated to C++, see Translation.

The simulation kernel¶

The kernel is a small set of C++ header files in core/ — approximately 1000 lines total, with no compiled library. Generated model code includes the headers directly, allowing the compiler to inline across kernel and model boundaries.

The central classes are:

Class	Role
`sitar::module`	Base class for all modules. Provides `run()`, `log`, `current_time`, `this_cycle`, `this_phase`.
`sitar::net<W,C>`	Fixed-capacity FIFO for `token<W>`. `push()` returns false when full; `pull()` returns false when empty.
`sitar::inport<W>` / `sitar::outport<W>`	Typed port wrappers that delegate to the connected `net`.
`sitar::token<N>`	Fixed-size byte payload container with metadata (ID, type, sender, receiver).
`sitar::time`	(cycle, phase) stored as a 64-bit integer; supports arithmetic and comparison.
`sitar::logger`	Thin `ostream` wrapper that prepends a (time, module-id) prefix on `endl`.

For a full description of the kernel classes, see Development Notes - Code Organization.

Execution semantics¶

The execution model — two phases per cycle, phase 0 for reads, phase 1 for writes, the convergence rule, and how wait and wait until interact with the phase counter — is covered in detail in Execution Model. The kernel implements exactly what that page describes: the simulation loop calls run(t) for each module at each phase, modules advance their behavior or return immediately, and the loop iterates until stop_simulation() is called.