Tile Operations#
This guide shows how to use CUDA tile operations to apply a computation to
tensor regions. T.cuda.tile.apply repeats a specified PTX instruction, such
as a vector load or matrix multiplication. T.cuda.tile.ewise applies a
scalar TIRx function to register elements. You choose the operand views,
instruction variants, and synchronization required by the kernel.
Start with Tile Operands and Execution for a load, compute, and store example and an explanation of how operand axes divide the work across threads. The instruction reference lists each family’s parameters, operand layouts, and alignment requirements.
CUDA Instruction Guides#
The guides below show how to use these interfaces for data movement, arithmetic, asynchronous transfers, and matrix instructions. Asynchronous examples include the completion protocol required by their instruction family.