#13953: Incorrectly handled bfloat16 -0.0 in ttnn.signbit #28
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
atatuzunerTT
force-pushed
the
atuzuner/signbit_neg_zero
branch
from
0c64f1a to
c2c04c4
Compare
rdjogoTT
approved these changes
Mar 3, 2025
atatuzunerTT
force-pushed
the
atuzuner/signbit_neg_zero
branch
from
c2c04c4 to
f0f5ca4
Compare
fvranicTT
pushed a commit
that referenced
this pull request
Mar 4, 2025
### Ticket [Link to Github Issue](tenstorrent/tt-metal#13953) ### Problem description `-0.0` Incorrectly handled as `0.0` in `ttnn.signbit`. bfloat16 data format should preserve `-0.0` values through the unpacker, math, packer pipeline. However, `-0.0` in bfloat16 format gets passed into the kernel as `0.0` and `signbit` reads this as a positive value. The detailed reasoning is as follows. When unpacking to the Dest register, bfloat16 values are first packed into Src registers and then moved to Dest. When moving these values from Src, `-0.0` is interpreted as `0.0` and src zero flags are set. This leads to `-0.0` values not being moved, but instead, zeroes being written into relevant locations in Dest. ### What's changed Disabled src zero flags for eltwise unary/sfpu operations. `-0.0` values are now moved to Dest instead of written with zeroes. ### Checklist - [x] [All post commit](https://github.com/tenstorrent/tt-metal/actions/runs/13592906337) CI passes - [x] [Blackhole Post commit](https://github.com/tenstorrent/tt-metal/actions/runs/13592910095) CI passes (if applicable)
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Outdated suggestions cannot be applied.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
-0.0Incorrectly handled as0.0inttnn.signbit. bfloat16 data format should preserve-0.0values through the unpacker, math, packer pipeline. However,-0.0in bfloat16 format gets passed into the kernel as0.0andsignbitreads this as a positive value.The detailed reasoning is as follows. When unpacking to the Dest register, bfloat16 values are first packed into Src registers and then moved to Dest. When moving these values from Src,
-0.0is interpreted as0.0and src zero flags are set. This leads to-0.0values not being moved, but instead, zeroes being written into relevant locations in Dest.What's changed
Disabled src zero flags for eltwise unary/sfpu operations.
-0.0values are now moved to Dest instead of written with zeroes.Checklist