Messages in this thread | | | Subject | Re: [PATCH net-next 01/14] etherdevice: Add eth_<foo>_addr CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS code | From | Joe Perches <> | Date | Tue, 03 Mar 2015 11:41:09 -0800 |
| |
On Tue, 2015-03-03 at 14:27 -0500, David Miller wrote: > From: Joe Perches <joe@perches.com> > Date: Tue, 03 Mar 2015 11:07:51 -0800 > > On Tue, 2015-03-03 at 10:42 -0800, Eric Dumazet wrote: > >> On Tue, 2015-03-03 at 10:25 -0800, Joe Perches wrote: > >> > >> > At least for arm gcc 4.6.3, it emits different code > >> > for net/l2tp/l2tp_eth.o > >> > >> Then it looks like arm gcc or arm linux memset() should be improved. > > > > Perhaps you can take that up with the gcc folk. > > > > I think it appropriate to improve the actual emitted > > code for the compiler I use. > > In the long term, this is poor time spent. If you fixed GCC > everyone would benefit in the world, not just kernel builders. > > Furthermore, none of this crap is in the fast path of anything. > > I'm not applying this series, it's basis is not well founded > yet you keep trying to argue otherwise.
Until such time as the linux crosstools compilers are updated, (they seem stuck on 4.6.3 from 3 years ago) https://www.kernel.org/pub/tools/crosstool/ I think the series is a trivial, small improvement.
I believe that's the only "argument" I've made.
Your choice to apply it or not, but if the series isn't appropriate, likely the existence of both eth_zero_addr and eth_broadcast_addr is suspect too.
cheers, Joe
| |