gaspersic/triton
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

[DOCS][TUTORIALS] Fixed typo

Browse Source
This commit is contained in:
Philippe Tillet
2020-02-10 04:22:45 -05:00
committed by Philippe Tillet
parent 404dd18333
commit 69326ae6de
1 changed files with 3 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
docs/tutorials/custom-operation.rst
View File

@@ -23,7 +23,7 @@ Let us start with something simple, and see how Triton can be used to create a c
// bounds checking
bool check[TILE] = offset < N;
// write-back
*?(check)pz = *?(check)*px + *?(check)py;
*?(check)pz = *?(check)px + *?(check)py;
}
As you can see, arrays are first-class citizen in Triton. This has a number of important advantages that will be highlighted in the next tutorial. For now, let's keep it simple and see how to execute the above operation in PyTorch.
@@ -53,7 +53,7 @@ As you will see, a wrapper for the above Triton function can be created in just
// bounds checking
bool check[TILE] = offset < N;
// write-back
*?(check)pz = *?(check)*px + *?(check)py;
*?(check)pz = *?(check)px + *?(check)py;
}
"""
# create callable kernel for the source-code
@@ -99,4 +99,4 @@ Executing the above code will:
- Cache the resulting custom op
- Call the resulting custom op
In other words, the first program run will generate and cache a bunch of files in $HOME/.triton/cache, but subsequent runs should be just as fast as using a handwritten custom operation.
In other words, the first program run will generate and cache a bunch of files in $HOME/.triton/cache, but subsequent runs should be just as fast as using a handwritten custom operation.