Rely on -cl-fp32-correctly-rounded-divide-sqrt for precise divide and sqrt
#5415
Conversation
This file contains hidden or 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
…nd sqrt Signed-off-by: Whitney Tsang <[email protected]>
whitneywhtsang
force-pushed
the
whitneywhtsang/cl-fp32-correctly-rounded-divide-sqrt
branch
from
bc74db7 to
7f660f5
Compare
chengjunlu
approved these changes
Nov 3, 2025
|
As a temp. workaround this OK. We will want to set the flag only on a specific operation to avoid changing precision for the entire function. @whitneywhtsang please add TODO and reference a new issue number to track the final solution. |
etiotto
reviewed
Nov 3, 2025
Created #5419. |
There is existing unit test for precise divide and sqrt: https://github.com/intel/intel-xpu-backend-for-triton/blob/main/python/test/unit/language/test_core.py#L1015 |
Signed-off-by: Whitney Tsang <[email protected]>
etiotto
approved these changes
Nov 3, 2025
whitneywhtsang
deleted the
whitneywhtsang/cl-fp32-correctly-rounded-divide-sqrt
branch
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
OpenCL supports the build option
-cl-fp32-correctly-rounded-divide-sqrt, which changes the requirement forOpFDivto be correctly rounded. The disadvantage is it is added at kernel level, so all divide and sqrt are precise, i.e., we cannot have both precise and approximate div in one kernel.