-
Notifications
You must be signed in to change notification settings - Fork 61
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Datatype precision for compute kernel
Summary: Adding precision for compute kernels with an option `--comp-data-type`. This enables comm and comput kernels to have different datatype precisions. ## Background: 1) In the previous code, the output of the MatMul is always with FP32. However, in the workloads, the mm uses different precision type {bfloat16, fp32 etc}. 2) In the previous code, the input datatype supplied to compute functions were also converted to FP32. This diff fixes this by using the comp-data-type option. Test case: N5625782 that returned FP32 Reviewed By: kingchc Differential Revision: D59994979 fbshipit-source-id: 7a834b37125f5c2b2b7ffd86520790e66b4754dd
- Loading branch information
1 parent
7c820e5
commit 7b19f58
Showing
3 changed files
with
35 additions
and
15 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters