Deepity
Ultra-fast, SIMD-accelerated Predictive Coding
Loading...
Searching...
No Matches
Activations.h File Reference

Defines the activation functions of a predictive coding model. More...

#include <deepity/utils/activations/ActivationDispatch.h>
#include <deepity/utils/activations/VectorMath.h>
#include <deepity/utils/activations/Relu.h>
#include <deepity/utils/activations/Gelu.h>
#include <deepity/utils/activations/Tanh.h>
#include <deepity/utils/activations/Sigmoid.h>
#include <deepity/utils/activations/Linear.h>

Go to the source code of this file.

Detailed Description

Defines the activation functions of a predictive coding model.

This is an umbrella header: the actual implementations live in deepity/utils/activations/ (ActivationDispatch.h for the ActivationType dispatch tables, then one header per activation family – Relu.h, Gelu.h, Tanh.h, Sigmoid.h, Linear.h – plus the currently-unused VectorMath.h). Split out so each activation's SIMD implementation is easier to find and review in isolation; existing #include <deepity/utils/Activations.h> call sites are unaffected.

#include <deepity/utils/Activations.h>
Deep::tanh(array, arraysize)
Activations.h
Defines the activation functions of a predictive coding model.
Deep::tanh
static void tanh(float *, size_t) noexcept
See Tanh.h.
Note
Separate implementations exist for AVX512F, AVX2, SSE, and naive. Memory does NOT need to be aligned.
Version
1.1
Date
2026-08-23
Author
Jack Rose