[RFC PATCH 4/7] ath79: prepare for 1-port TP-Link EAP2x5 devices

[Sander Vanheule]

On Sat, 2020-07-18 at 13:03 +0200, mail at adrianschmutzler.de wrote:
> > I tried 80MHz and 65MHz, but that didn't result in proper reads.
> > 50MHz,
> > 55MHz, and 60MHz do appear to work on my eap245-v1, so I really
> > wouldn't
> > risk going above 50MHz.
> 
> If you've taken the 25 MHz somewhere reasonable, as you stated, I'd
> also be fine with that.
> 
> My motivation is mainly to prevent copy/pasting that value from other
> devices without thinking, ending up with slower read speed without a
> real reason.
> 
> So, either do a simple read test whether 50 MHz is faster, and take
> it if there is relevant improvement, or stick to the 25 Mhz from GPL
> sources anyway.

Using `time dd if=/dev/mtd3 of=/dev/null` I took a few data points:

    f (MHz) |    25  |    40  |    60
      T (s) |  7.22  |  5.28  |  4.37

Resulting in T = T_0 + S/f, with T_0 = 2.275 s and S = 123 s×MHz

At 50MHz, reading the entire firmware partition (mtd3) would take 4.7s.
This is 35% faster (-2.5s) than with a 25MHz SPI clock. Not
insignificant, but I must say I didn't notice any difference in boot time between 25MHz and 50MHz; that still took ~30s. So I would opt to stay with 25MHz.

Best,
Sander