numpy-minmax

What it is and what it does

numpy-minmax is a specialized library that computes both the minimum and maximum values of a NumPy array in a single pass using C code with AVX/AVX512 SIMD instructions. It targets the common case where you need both extrema but NumPy's separate amin and amax functions require two passes. The package is optimized for float32 arrays that are C-contiguous, F-contiguous, or 1D strided; it also handles contiguous int16 arrays. For other array layouts (strided arrays with ndim ≥ 2), it falls back to NumPy's standard functions.

The library depends on numpy and cffi and is distributed as compiled wheels for Python 3.10–3.14 across macOS (ARM64), Linux (x86_64, aarch64, musllinux variants), and Windows. It is actively maintained, with the latest release 49 days old, and carries no known security vulnerabilities. The project originated at Nomono, an audio AI startup, and is used for audio processing but applies to any float32 workload where both extrema are needed.

Use it for:

• Audio signal processing pipelines that need to normalize or clip audio samples by finding their dynamic range in one operation.
• Real-time data streaming where you compute min/max statistics on fixed-size buffers for monitoring or scaling.
• Scientific computing workflows that repeatedly compute both extrema on large float32 arrays and need lower latency.
• Batch image or signal processing where normalization to [0, 1] or [-1, 1] requires both bounds simultaneously.
• Performance-critical loops in numerical simulations where the overhead of two separate NumPy calls is measurable.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Finds the minimum and maximum values in a NumPy array using C and SIMD optimizations, delivering faster results than separate NumPy amin and amax calls.

Yes, if you have a workload that repeatedly computes both min and max on float32 NumPy arrays and latency matters. The package is actively maintained, has no security issues, carries a permissive MIT license, and offers a documented performance advantage. Install friction is moderate due to compilation, but wheels are available for common platforms and Python versions. Not worth installing if you rarely need both extrema together or if your arrays are primarily int16 or non-contiguous.

Install

numpy-minmax on PyPI

pip install numpy-minmax

uv add numpy-minmax

poetry add numpy-minmax

Installing numpy-minmax

Before you install

Medium install friction due to compiled C extensions with cffi and numpy dependencies. The package is actively maintained with a recent release and broad platform coverage (macOS, Linux, Windows across multiple architectures and Python versions 3.10–3.14).

License in practice

MIT License permits unrestricted use, modification, and distribution with minimal restrictions—suitable for both open-source and commercial projects.

Quickstart

pip install numpy-minmax

import numpy_minmax
import numpy as np

arr = np.arange(1337, dtype=np.float32)
min_val, max_val = numpy_minmax.minmax(arr)

Requires numpy and cffi as runtime dependencies; C compiler needed at install time.

Verify before relying

• Whether the 2.3x speedup claim holds across different CPU architectures and numpy versions beyond 1.24–1.26.
• Performance characteristics for int16 arrays and the exact conditions triggering fallback to numpy.amin/amax.
• Compatibility with non-standard array layouts or memory-mapped arrays.

Package facts

Evidence: numpy_minmax-0.6.0-cp310-cp310-macosx_11_0_arm64.whl; numpy_minmax-0.6.0-cp310-cp310-manylinux1_x86_64.manylinux_2_28_x86_64.manylinux_2_5_x86_64.whl; numpy_minmax-0.6.0-cp310-cp310-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl; numpy_minmax-0.6.0-cp310-cp310-musllinux_1_2_aarch64.whl; numpy_minmax-0.6.0-cp310-cp310-musllinux_1_2_x86_64.whl; numpy_minmax-0.6.0-cp310-cp310-win_amd64.whl; numpy_minmax-0.6.0-cp311-cp311-macosx_11_0_arm64.whl; numpy_minmax-0.6.0-cp311-cp311-manylinux1_x86_64.manylinux_2_28_x86_64.manylinux_2_5_x86_64.whl; numpy_minmax-0.6.0-cp311-cp311-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl; numpy_minmax-0.6.0-cp311-cp311-musllinux_1_2_aarch64.whl; numpy_minmax-0.6.0-cp311-cp311-musllinux_1_2_x86_64.whl; numpy_minmax-0.6.0-cp311-cp311-win_amd64.whl; numpy_minmax-0.6.0-cp312-cp312-macosx_11_0_arm64.whl; numpy_minmax-0.6.0-cp312-cp312-manylinux1_x86_64.manylinux_2_28_x86_64.manylinux_2_5_x86_64.whl; numpy_minmax-0.6.0-cp312-cp312-manylinux2014_aarch64.manylinux_2_17_aarch64.manylinux_2_28_aarch64.whl; numpy_minmax-0.6.0-cp312-cp312-musllinux_1_2_aarch64.whl; numpy_minmax-0.6.0-cp312-cp312-musllinux_1_2_x86_64.whl; numpy_minmax-0.6.0-cp312-cp312-win_amd64.whl; numpy_minmax-0.6.0-cp313-cp313-macosx_11_0_arm64.whl; numpy_minmax-0.6.0-cp313-cp313-manylinux1_x86_64.manylinux_2_28_x86_64.manylinux_2_5_x86_64.whl

Tags

More Libraries packages