Enable fp16 nonnative support for dynamic dispatch, make more ergonomic for static dispatch #200
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
r-devulap
reviewed
Apr 29, 2025
src/avx512-16bit-qsort.hpp
Outdated
| } | ||
| replace_inf_with_nan(arr, arrsize, nan_count, descending); | ||
| replace_inf_with_nan_fp16( |
There was a problem hiding this comment.
Why can't we continue to use replace_inf_with_nan?
There was a problem hiding this comment.
The only difference between the two functions seems to be 0x7c01 v/s 0xFFFF and 0xFFFF seems to fail tests.
There was a problem hiding this comment.
I got rid of replace_inf_with_nan_fp16 by modifying replace_inf_with_nan to use 0x7c01 instead of 0xFFFF.
| @@ -208,6 +208,9 @@ jobs: | |||
|
|
|||
| - name: Run test suite on SPR | |||
| run: sde -spr -- ./builddir/testexe | |||
| - name: Run ICL fp16 tests | |||
| # Note: This filters for the _Float16 tests based on the number assigned to it, which could change in the future | |||
| run: sde -icx -- ./builddir/testexe --gtest_filter="*/simdsort/2*" | |||
There was a problem hiding this comment.
The np-multiarray-tgl job does test the float16 portion of the code on a TGL, but this is fine too.
7547df8 to
2c39de4
Compare
r-devulap
approved these changes
Apr 30, 2025
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Outdated suggestions cannot be applied.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.
This patch enabled the non-avx512fp16 _Float16 sorting to be used by the dynamic dispatch logic, as well as integrating it better into the static dispatch logic.
It is vastly faster than scalar, but a fair bit slower then the dedicated avx512fp16 code.
Comparison to scalar
Comparison to AVX512_FP16