kjbracey2

Thanks - the code there seems to suggest all 675x have Neon, which would make sense. And looking more closely at the Merlin kernel config, I can see it enabled...

There are still people getting confused by this: https://www.snbforums.com/threads/fix-this-bug-pls-big-file-size-copy-and-the-memory-getting-full-ver-386-4.76947/

I'm also preparing a patch to accelerate crc32_be. With them all done, we can get rid of about 32K of code and tables for the slice-by-8 software solution, which should...

Updated to Linux 4.21 version - it is about 35% faster on my RT-AX88U, despite upstream changelog suggesting it was slightly slower on A53. Original test time: 170 µs 4.20...

That somewhat surprises me. Do you know what sort of ops that would have been performing? The performance difference mainly depends on the block size - it appears that (from...

I'll try to reproduce those iperf/IPSEC tests. I'm moderately confident that the results might be reversed with 4 cores. There's another factor I need to investigate which is that these...

Well, the IPSec speed test looks like a whole project. Spent a couple of hours trying to figure out how to get a tunnel working, and not managed it so...

Got it working LAN side with Ubuntu - the trick is to install the `network-manager-strongswan` package, which adds a simple VPN to the GUI. So I can do iperf3 tests...

Hang on, doesn't the router have a 300Mbit/s or so performance limit when hardware acceleration isn't operational due to QoS? Is that what we're simply bumping into - our network...

> you'd have to monitor CPU usage to see if strongswan is capping the CPU, or something else (like IO) is. It was clear that all tests, whether using arm-ce...