kjbracey2

> I have no idea what you are doing here. I thought I'd managed to explain it to you in the previous discussion! Once more for luck - as the...

> And I've intended the color graph be a generic solution to all meters. I don't understand why you seem to think this isn't generic. It's totally generic. It's just...

> For what I've seen. the only benefit for `single_value` _as a class member_ so far would be to easily plot it in the graph display. Yes, just like the...

Does the ax88u build work for you, at least? That was the one tested - I admit to not compiling every other model. (So did I mess up the PR...

Clean build of rt-ax86u in src-rt-5.02p1axhnd.675x of the tip of the PR branch works for me - at least it's got as far as generating libnettle.a, and is still running.

> Have you been able to measure any performance improvement from these changes to confirm these are actually doing anything? Keep in mind that userspace is compiled as 32-bit armv7,...

Oh, one question - this is what I pieced together for the CPU possibilities there are. Have I got this right? RT-AC68U: 4709 (6.x.4708) 2x A9 --- RT-AC88U: 4709 (7.14.114.x)...

> is there something that could be optimized there? Surely, yes, but only if you're in the business of hand-coding ARM/Neon assembly. It's the upstream package that lacks aarch32+SHA256 crypto...

Oops, correction time. There is actually an aarch32 crypto extension sha256 implementation in there, I'd just missed it. It's tucked at the bottom of `sha256-armv4.pl`...

> bcm6750 dose not support hwcrypto > bcm6755/6756 bcm4906/4908/4912/4915 support hwcrypto > all of them support fpu(neon/vfp3) The 675x are all Cortex-A7 based, as far as I can see. I...