doppler-dsp 0.2.7

Dead-simple, ultra-fast signal processing in C

doppler

Dead-simple, ultra-fast digital signal processing.

doppler is a lean C99 signal processing library built for one goal: maximum throughput with minimum friction — from any language. The full DSP stack lives in one portable core with paper-thin Python bindings and a Rust FFI. No runtime surprises, no framework lock-in.

What's inside

• NCO — 32-bit phase accumulator, 2¹⁶-entry LUT, AVX-512 batch generation, FM ctrl port
• FIR filter — AVX-512 complex taps, CI8/CI16/CI32/CF32 input types
• FFT — 1D and 2D, selectable backend (FFTW or pocketfft)
• Polyphase resampler — continuously-variable rate, Kaiser-designed bank; DPMFS variant (608 B, L1-resident) for cache-sensitive pipelines
• Halfband decimator — dedicated 2:1 decimator exploiting halfband symmetry; 375 MSa/s at 60 dB attenuation
• SIMD arithmetic — SSE2/AVX2 complex multiply via dp_c16_mul
• Signal streaming — low-latency ZMQ transport (PUB/SUB, PUSH/PULL, REQ/REP)
• Circular buffers — double-mapped ring buffers for zero-copy, lock-free IPC (F32/F64/I16)
• Multi-language — clean C ABI; Python bindings (NCO, FFT, streaming, buffers) and Rust FFI

Benchmarks

Measured on AMD Ryzen AI 7 350, Release build (-O3 -march=native). Re-run any suite with make blazing then the binary listed below.

NCO (dp_nco_*)

block=1 048 576 samples × 200 iterations — ./build/c/bench_nco_c

Rank	Variant	MSa/s	Notes
🥇	u32	20 341	Raw phase only — store + add, no LUT
🥈	u32_ovf	3 028	Raw phase + carry bit (ADD + SETB)
🥉	cf32	2 094	Free-running IQ — AVX-512 16-wide gather
4	u32_ovf_ctrl	1 424	FM ctrl + raw phase + carry
5	u32_ctrl	1 095	FM ctrl + raw phase
6	cf32_ctrl	525	FM ctrl + IQ — LUT + ctrl overhead

FIR (dp_fir_execute_*)

taps=19 block=409 600 samples × 100 iterations — ./build/c/bench_fir_c

Input type	MSa/s	Notes
CI8	317	8-bit complex integer input
CF32	280	32-bit complex float input
CI16	235	16-bit complex integer input

Halfband decimator (dp_hbdecim_cf32)

block=65 536 samples × 400 iterations — ./build/c/bench_hbdecim_c

N taps	Attenuation	MSa/s (input)
10	40 dB	540
19	60 dB	375
37	80 dB	201
73	80 dB	105

The halfband structure halves the FIR multiply count by exploiting coefficient symmetry; the pure-delay branch costs one multiply instead of N.

FFT (dp_fft*)

FFTW backend, estimate plan, 1 thread, complex double — ./build/c/bench_fft_c

1D (fft1d_execute / fft1d_execute_inplace)

Size	Out-of-place MSa/s	In-place MSa/s
1 024	925	757
4 096	474	379
16 384	146	139

2D (fft2d_execute / fft2d_execute_inplace)

Size	Out-of-place MSa/s	In-place MSa/s
64 × 64	722	616
128 × 128	252	241
256 × 256	33	30

Ring buffer (dp_f32 / dp_f64)

Lock-free SPSC, 268 M samples, batch=4096 — ./build/c/bench_buffer_c

Type	MSa/s	GB/s
f32 (8 B/sample)	5 129	38.2
f64 (16 B/sample)	2 401	35.8

Quick example

C:

#include <doppler.h>
#include <dp/fft.h>
#include <complex.h>

size_t shape[] = {1024};
dp_fft_global_setup(shape, 1, -1, 1, "estimate", NULL);

double complex in[1024], out[1024];
/* ... fill in[] ... */
dp_fft1d_execute(in, out);

Python:

from doppler.fft import fft
import numpy as np

x = np.random.randn(1024) + 1j * np.random.randn(1024)
spectrum = fft(x)

Documentation

Full docs at doppler-dsp.github.io/doppler

Document	Contents
Quick Start	Build, install, run the examples (Docker quickstart included)
Build Guide	CMake options, platform notes, Python setup, Docker details
Overview	Architecture and full API reference
Examples	C and Python code examples - FFT, SIMD, streaming

| CLAUDE.md | Development notes and project context (for contributors) |

Build

make          # build (Linux/macOS; MSYS2 on Windows)
make test     # run CTest suite

Or with Python bindings:

make && uv sync

See Build Guide for platform-specific instructions and all CMake options.

Licensing

The doppler source code is MIT-licensed.

If built with FFTW support (default), the resulting binary links against FFTW, which is licensed under the GNU General Public License (GPL). In this case, the distributed binary is covered by the GPL.

If built with -DUSE_FFTW=OFF, the pocketfft backend is used instead. pocketfft is BSD-3-Clause-licensed (see POCKETFFT_LICENSE) which is compatible with the MIT-license and so the resulting binary remains MIT-licensed with BSD-3-Clause licensed FFT features.

See Build Guide for details and the installed LICENSE files.