erts: Replace FP16 implementation with a new one #9735
Conversation
|
Please also update vendor.info to not include fp16 anymore. |
lucioleKi
force-pushed
the
isabell/erts/fp16_2
branch
from
76f37b4 to
cdbc277
Compare
|
FP16 was thought to exists as a vendor library in the tests of the SBOM generation With this removal, please remove the mention in these lines of fp16 otp/.github/scripts/otp-compliance.es Lines 1584 to 1585 in 7031f3d |
lucioleKi
force-pushed
the
isabell/erts/fp16_2
branch
2 times, most recently
from
7ea4276 to
ab58af8
Compare
CT Test Results 3 files 141 suites 49m 55s ⏱️ Results for commit f7c3e36. ♻️ This comment has been updated with latest results. To speed up review, make sure that you have read Contributing to Erlang/OTP and that all checks pass. See the TESTING and DEVELOPMENT HowTo guides for details about how to run test locally. Artifacts// Erlang/OTP Github Action Bot |
lucioleKi
added
the
testing
The new implementation has the same behavior as the previous one. The reason we are writing this new implementation is that the previous version is external and has an MIT license. If it gets a newer version, we need to upgrade it and keep it up to date. Because FP16 has a clear specification and is stand-alone, it is easier to just write our own implementation. The newer compilers already have native support for FP16, so this implementation is only relevant for older compilers. For this reason, the new implementation has not been tested for speed.
lucioleKi
force-pushed
the
isabell/erts/fp16_2
branch
from
ab58af8 to
f7c3e36
Compare
The new implementation has the same behavior as the previous one. The reason we are writing this new implementation is that the previous version is external and has an MIT license. If it gets a newer version, we need to upgrade it and keep it up to date. Because FP16 has a clear specification and is stand-alone, it is easier to just write our own implementation.
The newer compilers already have native support for FP16, so this implementation is only relevant for older compilers. For this reason, the new implementation has not been tested for speed.