oxifft-codegen

Version: 0.2.0
Status: ⚠️ Early Development — API may change

Procedural macro crate for OxiFFT codelet generation.

Overview

This crate replaces FFTW's OCaml-based genfft code generator with Rust procedural macros. It generates highly optimized FFT kernels (codelets) at compile time.

Features

Compile-time code generation: Zero runtime overhead (radix 2, 4, 8, 16)
SIMD-aware: Can generate SIMD-specific code patterns (infrastructure in place)
Symbolic optimization: Built-in symbolic expression DAG for optimization passes

Procedural Macros

gen_notw_codelet!(size) — Non-twiddle base case codelets
gen_twiddle_codelet!(radix) — Twiddle-factor codelets for multi-radix FFT
gen_simd_codelet!(size) — SIMD-optimized codelets (infrastructure, generation pending)
gen_dft_codelet!(size) — Convenience wrapper (aliases gen_notw_codelet!)

Codelet Types

Non-Twiddle Codelets (notw)

Base case FFT kernels that don't require twiddle factors. Used at the leaves of the FFT recursion.

use oxifft_codegen::gen_notw_codelet;

// Generates codelet_notw_8 function
gen_notw_codelet!(8);

Twiddle Codelets

Codelets that apply twiddle factors as part of the Cooley-Tukey recursion.

use oxifft_codegen::gen_twiddle_codelet;

// Generates codelet_twiddle_4
gen_twiddle_codelet!(4);

SIMD Codelets

SIMD-optimized codelets (infrastructure present, full generation pending).

use oxifft_codegen::gen_simd_codelet;

// Generates SIMD-optimized size-8 codelet
gen_simd_codelet!(8);

Convenience Macro

use oxifft_codegen::gen_dft_codelet;

// Currently aliases gen_notw_codelet!
gen_dft_codelet!(8);

Supported Sizes

Size	Non-Twiddle	Twiddle	SIMD
2	✓	✓	Planned
4	✓	✓	Planned
8	✓	✓	Planned
16	✓	✓	Planned
32	✓	Planned	Planned
64	✓	Planned	Planned

Code Generation Strategy

The codelet generator follows FFTW's approach:

Symbolic representation: Build a DAG of FFT operations
Optimization passes:
- Common subexpression elimination (CSE)
- Strength reduction (replace multiplications with additions where possible)
- Dead code elimination
Code emission: Generate Rust code with optimal instruction ordering

Implementation Notes

Uses proc-macro2, quote, and syn for macro implementation
Generated code is generic over the Float trait (f32/f64)
Follows FFTW's codelet naming conventions for compatibility
All sizes 2–64 have generation infrastructure, with 2–16 fully implemented
8 tests passing

oxifft-codegen 0.2.0