scicomp-compute-core 0.1.5

Unified computational primitives: NumPy/CuPy arrays, FFT, linear algebra

scicomp-compute-core

GPU-accelerated computation core for the Math-Physics-ML MCP system.

Overview

This package provides high-performance array operations and numerical computing utilities with optional GPU acceleration using CuPy. It includes:

• Array utilities - Unified interface for NumPy and CuPy arrays
• GPU acceleration - Seamless CPU/GPU array conversion
• Numerical operations - Optimized mathematical functions
• Linear algebra - Matrix operations and decompositions
• FFT operations - Fast Fourier transforms with GPU support

Installation

# CPU only
pip install scicomp-compute-core

# With GPU support (CUDA 12.x)
pip install scicomp-compute-core[gpu]

Quick Start

from compute_core.arrays import to_gpu, to_numpy
import numpy as np

# Create array on CPU
arr = np.array([1, 2, 3, 4, 5])

# Move to GPU (if available)
gpu_arr = to_gpu(arr)

# Perform GPU operations
result = gpu_arr * 2

# Move back to CPU
cpu_result = to_numpy(result)

Features

• Automatic device selection - Works with or without GPU
• Transparent acceleration - Same API for CPU and GPU arrays
• Type safety - Full type hints for all functions
• Performance - Optimized for numerical computing workflows

Part of Math-Physics-ML MCP System

This package is part of a larger system. See the full documentation for details on all available components.