**** BEGIN LOGGING AT Fri Nov 06 02:59:57 2020 Nov 06 10:42:58 morning Nov 06 10:47:33 ka6sox: https://github.com/shr-project/test-oe-build-time/commit/17b180fadd8b73292ef8a4f9c9b2578fe1fa9031 bonaire results are merged, it is significantly faster then it was before, but my old $500 builder with AMD 1600AF is still faster Nov 06 10:50:08 and it's about half the speed of dual-Xeon-E5-2670-8-channels (which has only 64G ram and spinning disks) Nov 06 10:56:53 e.g. qtwebengine.do_compile alone taske 1500s on 1600AF 16G ram with j12 and 3000s on bonaire with 157G and -j 20, either the HVM overhead is still significant, or something is wrong Nov 06 10:57:42 the memory bandwidth in some cases seems significantly slower than dual-Xeon-E5-2670-8-channels (the only other DDR3 server in my tests) Nov 06 10:58:53 Tofe: rebuilding pinephone, tissot, mido, rosy on jenkins first 2 with kernel fix Nov 06 10:59:11 who has mido or rosy? they might need the same fix for new binutils Nov 06 14:12:10 JaMa: I have Mido, Tofe Rosy Nov 06 15:41:16 then when you have some time, please check if it is able to load kernel modules Nov 06 15:52:04 JaMa, thanks for the update! Nov 06 16:10:41 JaMa, its free and not in your bedroom making noise :) Nov 06 16:28:10 ka6sox: true, I'm not complaining that it's slow, just pointing that something might still be broken and gently pushing over the board.. :) Nov 06 16:28:53 2014 chip vs 2017 chip too Nov 06 16:29:08 and the AMD interconnects vs Intel ones on the chips Nov 06 16:29:32 Xeon-E5-2670 is 2012 Nov 06 16:29:39 even worse! Nov 06 16:29:56 5yrs is 5 lifetimes in CPUs as we know Nov 06 16:30:26 Xeon-E5-2670 from 2012 with just 64G, running on spinning disks is twice as fast as bonaire Nov 06 16:33:12 interesting...let me look at one more potential tweak. Nov 06 16:35:01 JaMa, I'll have to add a local SSD :) Nov 06 16:37:52 not sure if it's caused by HDD as it was building completely in tmpfs Nov 06 17:18:59 JaMa: sure, will do rosy **** ENDING LOGGING AT Sat Nov 07 03:00:32 2020