| Message ID | 1517956050-32412-3-git-send-email-rosenp@gmail.com |
|---|---|
| State | Rejected |
| Headers | show |
| Series | [LEDE-DEV,1/3] ag71xx: Remove timestamp struct member and use trans_start instead | expand |
On Wed, Feb 7, 2018 at 1:24 AM, Karl Palsson <karlp@tweak.net.au> wrote: > > Rosen Penev <rosenp@gmail.com> wrote: >> Seems to perform slightly better. > > Seems? Sounds like super awesome reason to change things. I did measure a performance improvement but it's hard to gauge how much it really is as since iperf results are variable now. I would say 5mbps max. Also this change is similar to the ramips one.
On 2018-02-06 23:27, Rosen Penev wrote: > Seems to perform slightly better. > > Signed-off-by: Rosen Penev <rosenp@gmail.com> Please check the context of your changes more carefully. The change you're making is affecting the init only, so it's completely irrelevant for performance. It is also wrong, because napi_alloc_frag may only be called from NAPI poll context. - Felix
diff --git a/target/linux/ar71xx/files/drivers/net/ethernet/atheros/ag71xx/ag71xx_main.c b/target/linux/ar71xx/files/drivers/net/ethernet/atheros/ag71xx/ag71xx_main.c index 13f500b..8b8355a 100644 --- a/target/linux/ar71xx/files/drivers/net/ethernet/atheros/ag71xx/ag71xx_main.c +++ b/target/linux/ar71xx/files/drivers/net/ethernet/atheros/ag71xx/ag71xx_main.c @@ -229,7 +229,7 @@ static int ag71xx_ring_rx_init(struct ag71xx *ag) struct ag71xx_desc *desc = ag71xx_ring_desc(ring, i); if (!ag71xx_fill_rx_buf(ag, &ring->buf[i], offset, - netdev_alloc_frag)) { + napi_alloc_frag)) { ret = -ENOMEM; break; }
Seems to perform slightly better. Signed-off-by: Rosen Penev <rosenp@gmail.com> --- .../ar71xx/files/drivers/net/ethernet/atheros/ag71xx/ag71xx_main.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)