Technical description of neuron_s as a standalone execution descriptor. More...

Detailed Description

Technical description of neuron_s as a standalone execution descriptor.

Author: Oscar Sotomayor (Titux)

This document describes neuron_s strictly as an isolated structural unit. The structure contains no execution logic and defines no control flow. It only stores the information required by external code to perform computation.

No neural network theory, training semantics, or framework-level behavior are described here.

Concept

neuron_s is a passive data descriptor representing a single computation frame.

It does not define:

evaluation order
scheduling
ownership of referenced data

All operational semantics are implemented externally.

The structure can be used independently for:

rapid prototyping
debugging
low-level experimentation

Structural View

neuron_s may be interpreted as a compact execution frame:

* ┌──────────────────────────────┐
* │          neuron_s            │
* ├──────────────────────────────┤
* │ inputs   : N                 │  ← number of operands
* │ bff_idx  : k                 │  ← memory selector
* │ fn       : f                 │  ← operation selector
* ├──────────────────────────────┤
* │ in   → [ ptr0 | ptr1 | … ]   │  ← input references
* │ w    → [  w0  |  w1  | … ]   │  ← coefficients
* ├──────────────────────────────┤
* │ b        : bias              │
* │ out      : result            │
* └──────────────────────────────┘
*

The diagram represents structural layout only. It does not imply execution order.

External Execution Flow

Execution semantics are defined entirely outside the structure.

External code is responsible for:

iterating over inputs
dereferencing values through in
combining them with w and b
applying the operation selected by fn
storing the result in out

neuron_s only provides the data required for this process.

Referential Connectivity

All inputs are accessed through pointers.

This enables:

shared data sources
direct neuron-to-neuron linking (out → in)
integration with external variables without copying

neuron_s does not own input memory.

Memory Selector Role

Within neuron_s, bff_idx acts purely as an execution context selector.

It does not define a buffer mechanism. Its interpretation is determined by surrounding execution logic.

In standalone scenarios, bff_idx may be ignored without affecting structural correctness.

Definition in file neuron_s.dox.