User Guide

• Programming Model
  • Recognizing available devices
  • Specifying offload target
  • Data sharing
  • Compute follows data
• Kernel Programming Basics
  • Writing Data Parallel Kernels
  • Synchronization Functions
  • Writing Device Functions
  • Supported Atomic Operations
  • Supported Address Space Qualifiers
  • Reduction on SYCL-supported Devices
  • Universal Functions
  • Supported Python Features inside numba_dpex.kernel
• Compiling and Offloading dpnp statements
  • Parallel Range
• Debugging with Intel® Distribution for GDB*
  • Requirements
  • Example of Intel® Distribution for GDB* usage
  • Features and Limitations
• Caching Mechanism in numba-dpex
  • Algorithm
  • Settings
• Configuration Options for numba-dpex
• Useful links