SinCosLUT.jl

Fast SIMD sine/cosine by register-resident table lookup. Each sin/cos is a single hardware byte/word/dword permute over a table held entirely in vector registers — no multiplies, no memory gathers, and (unlike floating-point range reduction) no input range limit: the phase wraps for free via a bit-mask.

It is a deliberately low-precision, very high-throughput technique for GNSS carrier generation and bit-wise software correlators. Pick the output element type (Int8 / Int16 / Int32) to trade precision for speed. On a Zen 5 core it generates a carrier at ~17 ps/element (AVX-512) — several times faster than the polynomial alternatives (see the benchmarks).

Installation

using Pkg
Pkg.add("SinCosLUT")

Quick start

using SinCosLUT

tbl  = SinCosTable(Int8; steps = 64)          # 64-entry Int8 table
sins = zeros(Int8, 4096); coss = zeros(Int8, 4096)

# Drift-free carrier at 0.002 cycles/sample (= frequency / sampling_frequency):
generate_carrier!(sins, coss, tbl, 0.002)

# ...or straight from a frequency and sampling frequency:
generate_carrier!(sins, coss, tbl; frequency = 1000, sampling_frequency = 2e6)

Documentation

Full, always-runnable documentation lives at JuliaGNSS.github.io/SinCosLUT.jl/dev (every example is executed and checked when the docs are built):

• Usage guide — carrier generation, frequency forms, initial phase, amplitude precision, backends, arbitrary-index lookup, and the 1-bit (hard-limited) carrier.
• Fused, array-free generation — the allocation-free carrier_engine for fusing sin/cos straight into a correlation loop.
• Accuracy & drift-free phase — what the quantisation buys you and why the integer NCO phase never drifts.
• Benchmarks — measured throughput and accuracy vs the polynomial packages.
• API reference.

Choosing a package

Three JuliaGNSS packages compute fast SIMD sin/cos at different points on the speed/accuracy curve:

accuracy needed	use
≤ ~5-bit (very high throughput)	SinCosLUT (this package)
~6–13-bit, integer output	FixedPointSinCosApproximations.jl
float-grade, 12–24-bit	FastSinCos.jl

Note — the SIMD lookup runs on AVX-512 (all types), AVX2 (Int8), and NEON (Int8 and Int16); other type/CPU combinations fall back to a correct scalar path. SVE2 is not supported. See the usage guide.