Fix aarch64/arm64 Linux wheel build failure due to missing libatomic linking #126
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
meta-cla
bot
added
the
CLA Signed
mergennachin
force-pushed
the
fix_tokenizers_aarch64
branch
from
523fe5f to
4d8a9a0
Compare
mergennachin
force-pushed
the
fix_tokenizers_aarch64
branch
from
4d8a9a0 to
089ff80
Compare
…linking Problem The tokenizers library fails to build on aarch64 Linux systems during CI with: /opt/rh/gcc-toolset-13/root/usr/libexec/gcc/aarch64-redhat-linux/13/ld: cannot find -latomic: No such file or directory Root Cause - On aarch64, certain atomic operations require explicit linking with libatomic - The sentencepiece dependency detects libatomic (Found atomic: /usr/lib64/libatomic.so.1) but incorrectly adds just the string "atomic" to link flags instead of the actual library path - This causes the linker to fail when building the Python extension Solution Added proper libatomic detection and linking for aarch64/arm64 systems in two places: 1. For the tokenizers static library 2. For the pytorch_tokenizers_cpp Python extension module The fix uses CMake's find_library to locate libatomic and explicitly links it when building on aarch64/arm64 architectures.
mergennachin
force-pushed
the
fix_tokenizers_aarch64
branch
from
089ff80 to
3e608b5
Compare
Hmm, |
swolchok
reviewed
Sep 11, 2025
| # Link with atomic library on aarch64/arm64 systems | ||
| # Some aarch64 systems require explicit linking with libatomic for certain atomic operations | ||
| if(CMAKE_SYSTEM_PROCESSOR MATCHES "aarch64|arm64") | ||
| find_library(ATOMIC_LIB NAMES atomic libatomic.so libatomic.so.1) |
There was a problem hiding this comment.
why do we have to duplicate 1) the find_library call 2) the gating? can't we just if(ATOMIC_LIB) since the previous time we look, on lines 97-105, is unconditional?
larryliu0820
approved these changes
Sep 11, 2025
|
Please address comments before merging |
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.
Problem
The tokenizers library fails to build on aarch64 Linux systems during CI with:
/opt/rh/gcc-toolset-13/root/usr/libexec/gcc/aarch64-redhat-linux/13/ld: cannot find -latomic: No such file or directory
Root Cause
Solution
Added proper libatomic detection and linking for aarch64/arm64 systems in two places:
The fix uses CMake's find_library to locate libatomic and explicitly links it when building on aarch64/arm64 architectures.