Re: [PATCH net-next] net: lan966x: Improve the CPU TX bitrate.

[Horatiu Vultur <horatiu.vultur@microchip.com>]

The 03/08/2022 16:40, Jakub Kicinski wrote:
> 
> On Tue, 8 Mar 2022 23:30:00 +0100 Horatiu Vultur wrote:
> > > >  static int lan966x_port_inj_ready(struct lan966x *lan966x, u8 grp)
> > > >  {
> > > > -     u32 val;
> > > > +     unsigned long time = jiffies + usecs_to_jiffies(READL_TIMEOUT_US);
> > > > +     int ret = 0;
> > > >
> > > > -     return readx_poll_timeout_atomic(lan966x_port_inj_status, lan966x, val,
> > > > -                                      QS_INJ_STATUS_FIFO_RDY_GET(val) & BIT(grp),
> > > > -                                      READL_SLEEP_US, READL_TIMEOUT_US);
> > > > +     while (!(lan_rd(lan966x, QS_INJ_STATUS) &
> > > > +              QS_INJ_STATUS_FIFO_RDY_SET(BIT(grp)))) {
> > > > +             if (time_after(jiffies, time)) {
> > > > +                     ret = -ETIMEDOUT;
> > > > +                     break;
> > > > +             }
> > >
> > > Did you try setting READL_SLEEP_US to 0? readx_poll_timeout_atomic()
> > > explicitly supports that.
> >
> > I have tried but it didn't improve. It was the same as before.
> 
> Huh, is ktime_get() super expensive on that platform?

Hm.. it looks like. Just adding ktime_get() before the while loop, then
the performance drops like before.
I am using SOC_LAN966 which has an ARMv7 CPU. So I am not sure how
expensive is ktime_get().

> jiffies vs ktime seems to be the main difference?

-- 
/Horatiu