Basic Concepts

Tensors

A tensor is a multidimensional array with: - Shape: Dimensions of the tensor - Data Type: float32, float64, int32, int64, uint8, bool - Device: CPU, CUDA, XPU, NPU, TPU - Data: Actual values stored in memory

Data Types

• TENSR_FLOAT32 - 32-bit floating point
• TENSR_FLOAT64 - 64-bit floating point
• TENSR_INT32 - 32-bit integer
• TENSR_INT64 - 64-bit integer
• TENSR_UINT8 - 8-bit unsigned integer
• TENSR_BOOL - Boolean

Devices

• TENSR_CPU - CPU execution
• TENSR_CUDA - NVIDIA GPU
• TENSR_XPU - Intel XPU
• TENSR_NPU - Neural Processing Unit
• TENSR_TPU - Tensor Processing Unit

Memory Management

Always free tensors after use:

Tensor* t = tensr_zeros(shape, 2, TENSR_FLOAT32, TENSR_CPU);
/* Use tensor */
tensr_free(t);