Re: [RFC PATCH] ali_workqueue: Adaptive lock integration on multi-socket/core platform

[Joseph Myers <joseph at codesourcery dot com>]

Please see the contribution checklist 
<https://sourceware.org/glibc/wiki/Contribution%20checklist>.  For 
example:

* FSF copyright assignment (with employer assignment / disclaimer as 
applicable) needed.  People are unlikely to look in detail at code without 
an assignment because it could cause problems if the assignment never 
appears and they wish to implement something similar in future.

* Please format code according to the GNU Coding Standards.

* New features need documentation in the user manual, and to be mentioned 
in the NEWS file.

* APIs need to be architecture-independent, in the absence of a clear 
justification for an architecture-specific API.

* A new interface is not useful without an installed header declaring it 
for users (this patch only has an internal header, not an installed one).

* New interfaces need testcases added to the glibc testsuite in the patch 
adding the interface.

* No "Contributed by" in new files.

* New symbol versions must be the version number of the first glibc 
release to have the feature.  For something added now that would be 
GLIBC_2.29.

* ABI test baselines (for all architectures) must be updated in any patch 
adding new interfaces.

* New code should not use __sync_*, and should have comments explicitly 
explaining the synchronization used (in terms of the C11 memory model) 
when using atomics.

-- 
Joseph S. Myers
joseph@codesourcery.com