Sycl Academy LogoSycl Academy Logo
listShow Lessons List
  1. What is SYCL
  2. Enqueueing a Kernel
  3. Managing Data
  4. Handling Errors
  5. Device Discovery
  6. Data Parallelism
  7. Introduction to USM
  8. Using USM
  9. Asynchronous Execution
  10. Data and Dependencies
  11. In Order Queue
  12. Advanced Data Flow
  13. Multiple Devices
  14. ND Range Kernels
  15. Image Convolution
  16. Coalesced Global Memory
  17. Vectors
  18. Local Memory Tiling
  19. Further Optimisations
  20. Matrix Transpose
  21. More SYCL Features
  22. Functors
Lesson
Source
Solution
GitHub Iconedit

Local Memory Tiling

In this exercise you will learn how to cache global memory into local memory in tiles according to work-groups in order to compare the performance difference.

1.) Use local memory

Allocate local memory for the kernel function by creating a local accessor.

The local accessor is created by specifying a range which is the number of elements to allocate.

Note that local memory is allocated per work-group and each work-group can only access it's own local memory.

2.) Cache global memory access in local memory

Cache the global memory required for each work-group in local memory by reading from global accessor and then writing to the local accessor.

Remember to use the local id to know where in local memory to store the data, this can be queried by calling the nd_item member function get_local_id.

Compare the performance with local memory and without local memory.

Build and execution hints

For DevCloud via JupiterLab follow these instructions.

For DPC++: instructions.

For AdaptiveCpp: instructions.