Eric Sauvageau

The minified version of that fork returns an error here: ``` Uncaught SyntaxError: Unexpected identifier (at qrcode.min.js:1:16790) ``` However downloading the non-minified version and minifying it myself works. So, looks...

nettle building is broken. ``` make[7]: Entering directory '/home/merlin/amng.ax86/release/src/router/nettle' siv-cmac-aes256.o.d:168: *** missing separator. Stop. ```

> 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...

At a quick glance it looks about right. Earlier RT-AC68U used a 800 MHz bcm4708. Later revisions went 4709 (and then 4709c0). One thing to note about the 490x, this...

> Openssl does partially suffer from the "only handcrafted for 64-bit" problem. aesv8-armx.pl can generate Crypto Extension code for either AArch32 or AArch64, but sha512-armv4.pl has only Neon but not...

> I can't see anything from a Makefile configuration point of view - I think OpenSSL is already basically optimally configured w.r.t the assembly it does have. Ok, that's what...

Back in the day I ran benchmarks with Strongswan. bcmspu was faster than the ARM64 cyphers, one of the reasons being that bcmspu frees up the CPU to handle routing/NAT....

> Do you know what sort of ops that would have been performing? I was running an iperf test through an IPSEC tunnel between my desktop and an RT-AC86U. >...

Might be because you are doing it within the LAN (which would need to go through the NAT loopback - not sure IPSEC is able to do that). I do...

The current high-end HND models can reach around 350 Mbps of NAT throughput when HW acceleration is disabled or bypassed. It`s possible that IPSEC might bypass it (I don't know),...