Post B5gnBDb3qjh79Y4Hho by azonenberg@ioc.exchange
(DIR) More posts by azonenberg@ioc.exchange
(DIR) Post #B5gdw7OliHMvzx2kT2 by azonenberg@ioc.exchange
0 likes, 0 repeats
Just kicked off what should be a build of every supported platform other than MacOS. Here goes...
(DIR) Post #B5geuAvRjEeKJPxume by azonenberg@ioc.exchange
0 likes, 0 repeats
What's the point of having half a terabyte of donated RAM and a 32-core / 64-thread CPU if you don't actually use it?I think this is the most I've loaded the server since the upgrade. Five simultaneous builder instances running, three with NVIDIA GTX 1630s and two using llvmpipe (plus two in the queue waiting for an available GPU), each with 8 3rd gen Xeon Scalable vCPUs and 32GB RAM - so somewhere in between a c6i.2xlarge and c6i.4xlarge, plus a small GPU on 3 of the 5. On top of my normal lab services and virtual desktop workloads.If this works, I have to tie up a few loose ends on management and stats then I can start working on setting up a runner VM on the mac mini.
(DIR) Post #B5gfawDAJN0783dzxA by Thunderbolt93@chaos.social
0 likes, 0 repeats
@azonenberg "I paid for the whole CPU so I'm gonna use the whole CPU"
(DIR) Post #B5gfgUQsVOMuiCho5Q by azonenberg@ioc.exchange
0 likes, 0 repeats
@Thunderbolt93 Somebody else did, this was dumpster dived from scrapped/ewasted servers and sent my way to use for this exact purpose.But I still don't want to waste it lol
(DIR) Post #B5gfqRxqXMLnx7sLdQ by azonenberg@ioc.exchange
0 likes, 0 repeats
@Thunderbolt93 big tech's table scraps are better than what i had in production so i'm not gonna say no to free stuff (especially if it's the same generation as the lower speed/core count CPU I already had, so a drop-in upgrade)
(DIR) Post #B5ggIfWjhkU67oePL6 by azonenberg@ioc.exchange
0 likes, 0 repeats
Aaand clean build, no test or build failures and only expected warnings from known things I need to tidy up in the code but haven't got to yet.
(DIR) Post #B5gjgrJBqv2FTeXbWK by azonenberg@ioc.exchange
0 likes, 0 repeats
To recap, the goal here is to move the CI infrastructure for ngscopeclient off of GitHub Actions and into my lab for a couple of reasons...1) Better performance than the free tier Azure instances GitHub uses (the standard Linux /Windows Azure runners are 4 vCPU and 16GB RAM while mine are 8 vCPU and 32GB RAM)2) Access to real GPUs for more realistic testing3) Better visibility into the test environment for debugging when things go wrong, I can spin up an instance and not terminate/wipe it when the build finishes then poke around, attach gdb to a test binary, etc4) Longer term: Hardware-in-loop testing. The ThunderScope folks are going to be sending me a PCIe AIC format ThunderScope dedicated to testing and eventually I'm going to be mounting it in the VM server, passing it through to a runner instance, and actually running development builds of the ThunderScope kernel driver and TS.NET while having the development builds of ngscopeclient talk to it, acquire real waveforms from a signal generator, and run end to end tests from trigger to decoded protocol data.
(DIR) Post #B5gm2jryOIe2gWN5MG by f4grx@chaos.social
0 likes, 0 repeats
@azonenberg and 5) not github
(DIR) Post #B5gnBDb3qjh79Y4Hho by azonenberg@ioc.exchange
0 likes, 0 repeats
@f4grx moving code hosting and issue tracking is a separate discussion there are good arguments in favor but the complex submodule setup makes migration very nontrivial.Moving the CI only makes sense in the near term regardless of whether we leave GH outright in the future or not.
(DIR) Post #B5gv7vkWY4LoZ5RBa4 by f4grx@chaos.social
0 likes, 0 repeats
@azonenberg I understand that, of course.
(DIR) Post #B5hBpFUgc4FzWJhuhU by scisystems@nrw.social
0 likes, 0 repeats
Already ideas on the CI software stack?