[QST] How does local_tile work? Could you provide an detailed explanation?

[ziyuhuang123]

auto gA = local_tile(mA, blk_shape, blk_coord, Step<_1, X,_1>{}); // (BLK_M,BLK_K,k)

I am learning this line in example code: https://github.com/NVIDIA/cutlass/blob/main/examples/cute/tutorial/sgemm_nt_1.cu

How we get this? By the way, I print it out, size is not BLK_M, BLK_K, k, but is BLK_M, BLK_K, K/BLK_K

Why??

[ziyuhuang123]

And what is step??? How we use it? Any example or doc? Thanks!!

[ziyuhuang123]

I see in another code, seems no need to use this Step?

  Tensor gA = local_tile(A, make_tile(Int<kTileM>{}, Int<kTileK>{}), make_coord(iy, _));
  Tensor gB = local_tile(B, make_tile(Int<kTileN>{}, Int<kTileK>{}), make_coord(ix, _));
  Tensor gC = local_tile(C, make_tile(Int<kTileM>{}, Int<kTileN>{}), make_coord(iy, ix));
  //  gA(kTileM, kTileK, num_tile_k)
  //  gB(kTileN, kTileK, num_tile_k)
  //  gC(kTileM, kTileN) 

[github-actions[bot]]

This issue has been labeled inactive-30d due to no recent activity in the past 30 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed. This issue will be labeled inactive-90d if there is no activity in the next 60 days.

[github-actions[bot]]

This issue has been labeled inactive-90d due to no recent activity in the past 90 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed.