lanefu changed the topic of #armbian-rockchip to: Armbian - Linux for ARM development boards | Rockchip SoC | www.armbian.com | This channel is relayed to the equivalent Discord channel | this channel is logged
<DC-IRC> [Discord] <gin_kage> Interesting... I'm tempted to try using `system-uncached`, as that's what the blob itself is using.
<DC-IRC> [Discord] <gin_kage> Yeah, looks like `system-uncached` works just as well. Might even be a better option, since it wouldn't be dependent on the user setup (i.e. the combination of device tree, command line and kernel config), and is likely larger anyway.
<DC-IRC> [Discord] <lanefu> have any more luck tuning the rtl1825 NICS on the r5s?
<DC-IRC> [Discord] <lanefu> FYI I just got a iperf3 test accross vlan improvement by setting
<DC-IRC> [Discord] <lanefu> ```
<DC-IRC> [Discord] <lanefu> ethtool -K lan1 rx-gro-list on
<DC-IRC> [Discord] <lanefu> ethtool -K lan2 rx-gro-list on
<DC-IRC> [Discord] <lanefu> ```
<DC-IRC> [Discord] <pookyfan> Official R8125 drivers (built with RSS feature enabled) after setting RSS affinity to all cores give me pretty decent and stable 910+ Mbps, but I haven't tested mode than one pair of computers with `iperf`. I hope that when I will be testing two pairs of PCs with `iper`, I'll get around 2 Gbps, but I need to set up my network and bring more PCs so I can test that.
<DC-IRC> [Discord] <pookyfan> Official R8125 drivers (built with RSS feature enabled) after setting RSS affinity to all cores give me pretty decent and stable 910+ Mbps, but I haven't tested more than one pair of computers with `iperf`. I hope that when I will be testing two pairs of PCs with `iper`, I'll get around 2 Gbps, but I need to set up my network and bring more PCs so I can test that.
<DC-IRC> [Discord] <pookyfan> Official R8125 drivers (built with RSS feature enabled) after setting RSS affinity to all cores give me pretty decent and stable 910+ Mbps, but I haven't tested more than one pair of computers with `iperf`. I hope that when I will be testing two pairs of PCs with `iperf`, I'll get around 2 Gbps, but I need to set up my network and bring more PCs so I can test that.
<DC-IRC> [Discord] <q4a> I was talking with @m33ts4k0z and this is his output:
<DC-IRC> [Discord] <q4a> ```root@orangepi5-plus:~# ls /dev/dma_heap
<DC-IRC> [Discord] <q4a> reserved system
<DC-IRC> [Discord] <q4a> ```
<DC-IRC> [Discord] <q4a> So he has no `system-uncached`
<DC-IRC> [Discord] <gin_kage> That means his kernel is not up to date, missing some latest commits: https://github.com/armbian/linux-rockchip/commit/77a373a398cac692b40cf19446bd3fefa2001473
<DC-IRC> [Discord] <q4a> Hm.. I hope that he can share his result again. But there's another interesting thing:
<DC-IRC> [Discord] <q4a> `set(WSIALLOC_MEMORY_HEAP_NAME "linux,cma" `
<DC-IRC> [Discord] <q4a> There's 2 default names
<DC-IRC> [Discord] <q4a> Maybe we can set few names too
<DC-IRC> [Discord] <gin_kage> No, that's actually one name. You literally get a device named `/dev/dma_heap/linux,cma`, e.g. on A311D.
<DC-IRC> [Discord] <m33ts4k0z> Guys I used the compile script and chose vendor kernel. How can I pull the latest commit for the kernel?
<DC-IRC> [Discord] <Werner> Why do you think the compilation uses not the most recent commit?
<DC-IRC> [Discord] <q4a> Sorry, I don't like to build kernel and think that there is more smart guys that already did it)
<DC-IRC> [Discord] <q4a> My top is apps and system libs
<DC-IRC> [Discord] <Werner> Why do you think the compilation does not use the most recent commit?
lanefu has quit [Ping timeout: 265 seconds]
lanefu has joined #armbian-rockchip
<DC-IRC> [Discord] <gin_kage> It should use the latest, yes. Always does for me.
<DC-IRC> [Discord] <m33ts4k0z> because the output of ls /dev/dma_heap is not correct according to @gin_kage ? or?
<DC-IRC> [Discord] <gin_kage> When did you build that image/kernel?
<DC-IRC> [Discord] <m33ts4k0z> 2-3 days ago maybe
<DC-IRC> [Discord] <gin_kage> More curiously, how does the latest blob work on your machine without that heap. I mean, that would definitely prevent it from working.
<DC-IRC> [Discord] <m33ts4k0z> it does work with wayland. the vkcube-wayland works 😅
<DC-IRC> [Discord] <m33ts4k0z> btw i built that specific image yesterday after Werner telling me that i needed the mesa-vpu extension
<DC-IRC> [Discord] <gin_kage> Maybe it's not the very latest one... There was g24-04, that didn't require it.
<DC-IRC> [Discord] <m33ts4k0z> is the kernel being repulled and rebuild on every build? or is it cached? maybe i need to clean the kernel package?
<DC-IRC> [Discord] <gin_kage> It does use compile cache, yeah.
<DC-IRC> [Discord] <m33ts4k0z> can i pass a flag to force repull/rebuild somehow maybe=
<DC-IRC> [Discord] <m33ts4k0z> can i pass a flag to force repull/rebuild somehow maybe?
<DC-IRC> [Discord] <m33ts4k0z> ` ./compile.sh ENABLE_EXTENSIONS=mesa-vpu kernel` maybe something like this?
<DC-IRC> [Discord] <monkablyat> no need to use mesa-vpu extension if you like to test the mali blob
<DC-IRC> [Discord] <m33ts4k0z> i tried to use the kernel command but doesnt seem to work 🤔
<DC-IRC> [Discord] <m33ts4k0z> It fetched the head and i think it re-built the kernel now. lets see
<DC-IRC> [Discord] <Werner> yes. `ARTIFACT_IGNORE_CACHE=yes` will force rebuild of everything. Though ususally after a commit to `main` new artifacts are generated automatically.
<DC-IRC> [Discord] <m33ts4k0z> Thanks 😊 It looks like the kernel rebuilt now. i will flash the image and run the command again to see
<DC-IRC> [Discord] <m33ts4k0z> yes now i have the correct output 👍
<DC-IRC> [Discord] <Werner> Added this particular situation here: https://github.com/armbian/documentation/pull/602/files