requirements-gpu.txt

# GPU Acceleration Dependencies for AFOptimizer
# ==============================================
#
# IMPORTANT: Only install ONE of the options below based on your hardware!
# GPU acceleration is OPTIONAL - the software works fine on CPU only.
#
# Installation Instructions:
# -------------------------
#
# 1. For NVIDIA GPU Users (Windows/Linux):
#    - First, check your CUDA version: nvidia-smi
#    - Then install the matching CuPy version:
#
#      For CUDA 11.x:  pip install cupy-cuda11x>=12.0.0
#      For CUDA 12.x:  pip install cupy-cuda12x>=13.0.0
#
#    Note: If you're not sure, try CUDA 12.x first (most common)
#
# 2. For AMD/Intel GPU Users (Windows/Linux):
#    - Install OpenCL drivers first (from GPU manufacturer)
#    - Then run: pip install pyopencl>=2023.1
#
# 3. For Mac Users (Apple Silicon M1/M2/M3):
#    - OpenCL is built-in, just run: pip install pyopencl>=2023.1
#
# 4. For Mac Users (Intel):
#    - Same as option 2 above
#
# Example Installation Commands:
# ------------------------------
#
# Windows with NVIDIA GPU (CUDA 12):
#   pip install cupy-cuda12x>=13.0.0
#
# Mac with Apple Silicon:
#   pip install pyopencl>=2023.1
#
# Linux with AMD GPU:
#   pip install pyopencl>=2023.1
#
# Verification:
# ------------
# After installation, run this to test GPU detection:
#   python test_gpu_detection.py
#
# If GPU is not detected, the software will automatically use CPU.
# This is completely fine - CPU processing still works great!

# Optional: CuPy for NVIDIA GPUs (uncomment one based on your CUDA version)
# cupy-cuda11x>=12.0.0    # For CUDA 11.x
# cupy-cuda12x>=13.0.0    # For CUDA 12.x

# Optional: PyOpenCL for AMD/Intel/Apple GPUs
# pyopencl>=2023.1