Device Management

Supported Devices

• CPU: Standard CPU execution
• CUDA: NVIDIA GPU
• XPU: Intel XPU
• NPU: Neural Processing Unit
• TPU: Tensor Processing Unit

Device Selection

Tensor* t_cpu = tensr_zeros(shape, 2, TENSR_FLOAT32, TENSR_CPU);
Tensor* t_cuda = tensr_zeros(shape, 2, TENSR_FLOAT32, TENSR_CUDA);

Best Practices

• Use GPU for large tensors (>1000 elements)
• Transfer data in batches
• Synchronize after GPU operations
• Check device availability before use