feat(tests): add unittest for gemm. - Githubissues

TiledTensor / TiledCUDA

TiledCUDA is a highly efficient kernel template library designed to elevate CUDA C’s level of abstraction for processing tiles.

MIT License

148 stars 10 forks source link

feat(tests): add unittest for gemm. #86

Closed haruhi55 closed 3 months ago

haruhi55 commented 3 months ago

This test for GEMM involves:

[x] Loading a data tile from global memory to shared memory.
[x] Chunking shared memory using the TileIterator.
[x] Loading the data tile from shared memory to register.
[x] Computing GEMM using Tensor Core's WMMA.
[x] Storing the data tile from register to shared memory.
[x] Storing the data tile from shared memory to global memory.
[x] pass unittest.

This PR relies on the prior merge of https://github.com/TiledTensor/TiledCUDA/pull/83.