View previous topic :: View next topic |
Author |
Message |
Spargeltarzan Guru
Joined: 23 Jul 2017 Posts: 317
|
Posted: Tue Jan 16, 2018 2:56 pm Post subject: 3 Times longer emerge time with new CPU i7-4770 |
|
|
Dear Community,
I previously used the Intel i7-4770K and changed it to i7-4770 for vt-d support (in the 4th series K allowed overclocked - what I am not interested in anyway - at the cost of vt-d)
I checked the cpu flags which are the same as with the i7-4770K and recompiled my whole system successfully.
However, emerges take 3 times longer than before, here some examples:
Code: |
Wed Dec 6 08:55:51 2017 >>> app-office/libreoffice-5.4.2.2
merge time: 1 hour, 13 minutes and 36 seconds.
Mon Jan 8 14:37:34 2018 >>> app-office/libreoffice-5.4.2.2
merge time: 3 hours, 40 minutes and 10 seconds.
--------------------------------------------------------------------------------------
Wed Dec 6 09:56:36 2017 >>> net-libs/webkit-gtk-2.18.3
merge time: 57 minutes and 40 seconds.
Mon Jan 8 10:44:58 2018 >>> net-libs/webkit-gtk-2.18.4
merge time: 3 hours, 38 minutes and 14 seconds.
--------------------------------------------------------------------------------------
Wed Dec 6 07:49:54 2017 >>> mail-client/thunderbird-52.4.0
merge time: 47 minutes and 57 seconds.
Mon Jan 8 07:59:47 2018 >>> mail-client/thunderbird-52.5.0
merge time: 1 hour, 15 minutes and 34 seconds.
--------------------------------------------------------------------------------------
Wed Dec 6 12:24:18 2017 >>> sys-devel/gcc-7.2.0
merge time: 32 minutes and 13 seconds.
Sun Jan 7 19:50:38 2018 >>> sys-devel/gcc-7.2.0
merge time: 1 hour, 26 minutes and 34 seconds.
|
What might be the issue here, any idea?
Thank you in advance! _________________ ___________________
Regards
Spargeltarzan
Notebook: Lenovo YOGA 900-13ISK: Gentoo stable amd64, GNOME systemd, KVM/QEMU
Desktop-PC: Intel Core i7-4770K, 8GB Ram, AMD Radeon R9 280X, ZFS Storage, GNOME openrc, Dantrell, Xen |
|
Back to top |
|
|
mike155 Advocate
Joined: 17 Sep 2010 Posts: 4438 Location: Frankfurt, Germany
|
Posted: Tue Jan 16, 2018 3:20 pm Post subject: |
|
|
That's a huge difference!
1) Check number of CPUs that are up and running
2) It could be power management. Use powertop to check frequencies and sleep states.
3) If you work in a VM and if you have already installed Meltdown patches: check for CPU flags 'pcid' and 'invpcid' |
|
Back to top |
|
|
toralf Developer
Joined: 01 Feb 2004 Posts: 3922 Location: Hamburg
|
Posted: Tue Jan 16, 2018 3:23 pm Post subject: |
|
|
in /etc/portage/make.conf ? |
|
Back to top |
|
|
Spargeltarzan Guru
Joined: 23 Jul 2017 Posts: 317
|
Posted: Tue Jan 16, 2018 3:36 pm Post subject: |
|
|
mike155,
1) less /proc/cpuinfo shows 8 CPU. Gnome system monitor too.
2) While emerging webkit-gtk the cores/threats are clocking very varying. Sometimes Nr. 7 100%, somtimes Nr. 5 100%, 3-4 cores almost idle (2-3 %), one core 7% (checked with gnome system monitor). In the Panel it shows me 3398 Mhz.
Powertop shows ~ 1400 awakings/s, 60-100 % cpu usage. Don't know how to see sleep states.
3) No VM, bare-metal installation.
toralf,
I realised in make.conf MAKEOPTS=-j1, but I believe in the statistics above I applied -j9 manually for emerge -e.
The -j1 was also with the i7-4770K in it, so every time I didn't specify -j it was with -j1 I guess. _________________ ___________________
Regards
Spargeltarzan
Notebook: Lenovo YOGA 900-13ISK: Gentoo stable amd64, GNOME systemd, KVM/QEMU
Desktop-PC: Intel Core i7-4770K, 8GB Ram, AMD Radeon R9 280X, ZFS Storage, GNOME openrc, Dantrell, Xen |
|
Back to top |
|
|
Hu Moderator
Joined: 06 Mar 2007 Posts: 21624
|
Posted: Wed Jan 17, 2018 3:14 am Post subject: |
|
|
Spargeltarzan wrote: | I realised in make.conf MAKEOPTS=-j1, but I believe in the statistics above I applied -j9 manually for emerge -e. | How did you manually apply this? There are two different levels of parallelization possible here. When you pass -jN on the emerge command line, that parallelizes the number of packages Portage can try to build in parallel (subject to the dependency tree, so in the extreme case, it can be ignored if every package to install requires the immediately preceding package). When you set -jN in MAKEOPTS (whether via make.conf or via an environment override on the emerge command line), you grant Make (and, to varying degrees, other build tools) permission to parallelize the steps of an individual build. Therefore, if you had MAKEOPTS=-j1 in make.conf, did not override that, and did pass -j9 on the emerge command line, Portage was permitted to build up to 9 packages in parallel, but every package would build its components serially. If you are examining the required time to completion for individual large packages, which seems to be the case according to your earlier posts, then you would see poor throughput on a per-package basis, even if the total system usage was high (due to each core working a unique package). |
|
Back to top |
|
|
chithanh Developer
Joined: 05 Aug 2006 Posts: 2158 Location: Berlin, Germany
|
Posted: Wed Jan 17, 2018 1:59 pm Post subject: |
|
|
If you do parallel merges (emerge -jN), this can mess up the build times as installation may be on hold until a related package finishes merging (besides things taking longer anyway).
Also, check CPU temperatures. Maybe there was a problem applying thermal paste or the cooler is not fastened properly. |
|
Back to top |
|
|
|
|
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
|
|