Tile size
Tile size specifies the dimensions or number of elements in a tile, a portion of data processed as a unit.
Grokking GPU Programming
Tile size is the number of rows and columns in a tile used by a tiled computation.
- How it is used
- Configures the square chunks used to reuse matrix data in shared memory during tiled matrix multiplication.
- How deep it goes
- Defines tile size and relates it to thread-block matching and hardware limits, without comparing performance across tile sizes.
- Examples
- Uses 32 × 32 tiles, mismatched 16 × 16 blocks, and 64- and 128-element tiles to illustrate loading errors and limits.
- What you can do after
- Configure a one-element-per-thread tiled kernel by matching block dimensions to tile size and checking thread and shared-memory limits.
GPU Programming with Triton
Tile size is the configuration specifying the dimensions or element count of a tile processed by a kernel.
- How it is used
- Controls the amount of work assigned to a tile and can keep a softmax reduction local to one program instance.
- How deep it goes
- Explains a softmax sizing constraint and its resource tradeoff, but gives no general manual tuning procedure.
- Examples
- Uses a one-dimensional vector softmax with tile size at least N and one program instance per vector.
- What you can do after
- Configure a Triton softmax tile to cover its full reduction axis, trading global-memory staging for on-chip resource use.
Also covered in
What to know first
- GPU memoryGPU memory is memory available to a graphics processing unit for storing data and state used in graphics or computation.
- matrix-matrix multiplicationMatrix multiplication is an operation on compatible matrices in which each entry of the product is the dot product of a row of the first matrix and a column of the second.
- shared memoryShared memory is a memory region accessible to multiple processes, threads, or other concurrent execution contexts for sharing data.
Related concepts
- tileconfigures · A tile is a bounded block of elements from a larger array, matrix, or tensor that is treated as a unit of computation.
- tiled matrix multiplicationconfigures · Tiled matrix multiplication divides matrix operands into blocks and computes the product using those blocks to improve data reuse.
- thread layoutconfigures · A thread layout is the dimensional arrangement of threads within a CUDA thread block, such as a one-dimensional line, two-dimensional rectangle, or three-dimensional box.
- program instanceconfigures · A program instance is Triton's unit of work corresponding to one block of input elements in a compiled kernel launch.
- thread limitconfigures · A thread limit is the maximum number of threads that can be assigned to one CUDA block.
- arithmetic intensityproduces · Arithmetic intensity is the ratio of computational operations performed to bytes of data transferred by a computation.
Questions about tile size
What is tile size?
Tile size specifies the dimensions or number of elements in a tile, a portion of data processed as a unit.
Which Manning books explain tile size?
Grokking GPU Programming and GPU Programming with Triton.
What should I know before learning tile size?
Manning books that teach tile size usually introduce GPU memory, matrix-matrix multiplication and shared memory first.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.

