Skip to content

fix: align S[...] notation with address formula in Tile Layout section #141

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
xush6528 wants to merge 1 commit into
base: main
Choose a base branch
from
Open

fix: align S[...] notation with address formula in Tile Layout section #141

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 2 additions & 2 deletions chapter_data_layout/index.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -84,11 +84,11 @@ So far we have described layouts for whole tensors. GPU kernels, however, rarely
entire matrix at once; they work on smaller tiles, which are loaded, transformed, and computed on by
different parts of the hardware. The good news is that tiling asks for nothing new. It is still
just a layout, only now written with a few more dimensions. Cut an 8×8 matrix into 2×4 tiles and we
get a 4-D layout, with coordinates `(tile_row, row_in_tile, tile_col, col_in_tile)` and strides
get a 4-D layout, with coordinates `(tile_row, tile_col, row_in_tile, col_in_tile)` and strides
chosen so that each tile stays contiguous:

```text
S[(4, 2, 2, 4) : (16, 4, 8, 1)]
S[(4, 2, 2, 4) : (16, 8, 4, 1)]
```

A logical `(i, j)` first becomes `(i//2, j//4, i%2, j%4)` and then runs through the strides. What is
Expand Down