18/06/2021 01:26, Honnappa Nagarahalli: > > On Tue, Jun 15, 2021 at 09:54:51PM -0500, Joyce Kong wrote: > > > Since C11 memory model is adopted in DPDK now[1], use GCC's atomic > > > builtins in test cases. > > > > as previously discussed these atomics are not "C11" they are direct use of > > gcc > > builtins. please don't incorporate C11 into the title of the patches or > > commit > > messages since it isn't. > > GCC supports 2 types of built-in atomics, > __atomic_xxx[1] and __sync_xxx [2]. > We need a way to distinguish between them. > We are using "C11" as [1] says they match C++11 memory model.
I agree it would be more correct to mention "compiler builtin" as it is not strictly the C11 API. > [1] https://gcc.gnu.org/onlinedocs/gcc/_005f_005fatomic-Builtins.html > [2] https://gcc.gnu.org/onlinedocs/gcc-4.1.1/gcc/Atomic-Builtins.html > > > > > please do not integrate a patch that directly uses gcc builtins and > > extensions > > please maintain abstractions under the rte_ namespace. > > This is just another wrapper which is not required > according to the current requirements we are working with. Yes such wrapper is not required *today*. We have 2 options: 1/ introduce a wrapper now to anticipate any future issue 2/ introduce a wrapper later when required Given we already use these builtins, we should not block this patchset. If it is decided to change the policy, then we'll replace the calls to the compiler builtins in all the codebase. > > specifically this patch substantially increases coupling to a single > > compiler > > implementation reducing portability. > > > > as previously requested, please establish at a minimum macros in the rte_ > > namespace for this. > > This needs to be discussed at the Techboard. I have CCed the Techboard. > The Techboard meets once in 2 weeks. The details are at [3]. > Next meeting is on 6/30 at 10am CST. Can you please attend and make your case? I agree to discuss options 1 or 2 in a techboard meeting. > Alternately, you can send an email to techbo...@dpdk.org > explaining your case and we will debate and make a decision. > > [3] https://core.dpdk.org/techboard/ > > > > > thanks. > > > > > [1] > > > https://www.dpdk.org/blog/2021/03/26/dpdk-adopts-the-c11-memory- > > model/
