#15889: Fix handling of mantissa rounding to respect ties round to even #16997
+107
−86
Conversation
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
TT-BrianLiu
requested review from
abhullar-tt,
pgkeller,
aliuTT,
tt-aho,
tt-dma,
tt-asaigal and
ubcheema
as code owners
TT-BrianLiu
force-pushed
the
bliu/main2
branch
from
ad64313 to
842747f
Compare
Comment on lines
+11
to
+13
| float float_input; | ||
| uint32_t expected_mantissa; | ||
| float expected_float_output; |
| // Set shared exponent as original float exponent (ie. skip logic for handling shared exponents) | ||
| auto shared_exp = uint32_input >> 23 & 0xFF; | ||
|
|
||
| auto output_mantissa = convert_u32_to_bfp<tt::DataFormat::Bfp8_b, false>(uint32_input, shared_exp, false); |
There was a problem hiding this comment.
what false means here, in both cases?
| auto output_mantissa = convert_u32_to_bfp<tt::DataFormat::Bfp8_b, false>(uint32_input, shared_exp, false); | ||
| EXPECT_EQ(output_mantissa, expected_mantissa); | ||
|
|
||
| uint32_t uint32_output = convert_bfp_to_u32(tt::DataFormat::Bfp8_b, output_mantissa, shared_exp, false); |
There was a problem hiding this comment.
wonder why one function is templated and the other accepts bfp type as argument
|
|
||
| void roundtrip_test_for_mantissa_rounding_with_bfp8( | ||
| float float_input, uint8_t expected_mantissa, float expected_float_output) { | ||
| auto uint32_input = *reinterpret_cast<const uint32_t*>(&float_input); |
There was a problem hiding this comment.
can we allow conversion functions to accept/return or float or uint32t?
ayerofieiev-tt
approved these changes
Jan 22, 2025
- Add tt_metal gtest to test convert_u32_to_bfp rounding logic for bfp8 - Remove unused functions in tt_metal/api/tt-metalium/bfloat8.hpp
TT-BrianLiu
force-pushed
the
bliu/main2
branch
from
842747f to
6828830
Compare
TT-BrianLiu
requested review from
dmakoviichuk-tt,
rfurko-tt,
cfjchu,
razorback3,
dongjin-na and
bbradelTT
as code owners
bbradelTT
approved these changes
Jan 23, 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.
Ticket
Link to Github Issue
Problem description
See issue for more context and examples. When we handle rounding of mantissa when going from float to block float, we always round up, which doesn't folow the "ties round to even" rule. This PR fixes that and adds lower c++ tests to test the different cases.
What's changed
Checklist