gro: __napi_gro_receive() optimizations
authorEric Dumazet <eric.dumazet@gmail.com>
Fri, 27 Aug 2010 05:03:08 +0000 (22:03 -0700)
committerDavid S. Miller <davem@davemloft.net>
Fri, 27 Aug 2010 05:03:08 +0000 (22:03 -0700)
commit40d0802b3eb47d57e2d57a5244a18cbbe9632e13
treefd45987286080e1f711c2448f6fbdc7d362a0241
parent250ad8f55c06eb866cfb57f8d3ea6ff961a7d1d7
gro: __napi_gro_receive() optimizations

compare_ether_header() can have a special implementation on 64 bit
arches if CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS is defined.

__napi_gro_receive() and vlan_gro_common() can avoid a conditional
branch to perform device match.

On x86_64, __napi_gro_receive() has now 38 instructions instead of 53

As gcc-4.4.3 still choose to not inline it, add inline keyword to this
performance critical function.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
include/linux/etherdevice.h
net/8021q/vlan_core.c
net/core/dev.c