I forgot that You may ask: what the hell is that HT/SMT/NICE thing?
About the project: > SMT/Hyperthreading, nice and scheduling policies
This is an experimental nrjQL branch with more responsiveness and realtime
behaviour:
- all this will be noticeable only when we have cpu equipped with HT
(HypeThread)
- It improves the performance of foreground applications at the expense of
backgound ones,
so your desktop/gui/applications will remain fast even if you run folding@home,
mprime, seti@home etc.,
but those background tasks will slow down more.
Read here, that's my topic with all the related links:
Kernel nrjQL 3.15.10-99 with better HyperThread and Realtime
http://mib.pianetalinux.org/forum/viewtopic.php?f=38&t=4446
Just a received feedback:
"Kernel with HT/SMT is working perfectly, much faster than the official scaled
kernel,
the performance is much better, taking 30% less time on the task of converting
video,
best gaming performance and processes in parallel and indexing faster..."
Il Venerdì 19 Settembre 2014 18:33, Nicolò Costanza <[email protected]> ha
scritto:
Observed and tested cases >>> HT/SMT/NICE (CONFIG_SMT_NICE=y)
Everything is working fine HOST & GUEST have the same feature as enabled!
a> Everything seems running fine if:
You run on Host PC (even different than omv2014.1) a kernel with HT/SMT/NICE
and
You run on it a Guest vm (it was tested vmware) with a kernel with HT/SMT/NICE
b> Everything seems running rather SLOW, only if:
You run on Host PC (even different than omv2014.1) a kernel W/O HT/SMT/NICE
and
You run on it a Guest vm (it was tested vmware) with a kernel with HT/SMT/NICE
The culprit seems to be > "ata_sff" running forever at 99%
To check and verify if your kernel are with or without, you can use this
command:
$ cat /boot/config-$(uname -r) | grep NICE
if the commad returns
CONFIG_SMT_NICE=y
you have it!
if the commad returns a null
you have NOT it
Considerations on VMWare:
We should consider that VMWare 10 (2013), will be soon replaced by the newer
VMWare 11 (2014)
with a full native support for the newest kernels and a newer hardware layer
virtualization,
so these troubles may be solved in the vmware 11 itself...
currently vmware needs external, not official patches, to allow it working with
current kernel
a preview:
http://www.vladan.fr/vmware-workstation-tech-preview-july-2014-shows-virtual-hardware-11-possibilities/
Joeghi's concerns:
The issue may be only for the reviewers that will try to test the new omv2014.1
from inside a vmware session,
if they have a normal plain kernel,
they may suffer a bad working, and write bad things about our new distro.
I think that every decision has its pros and cons... ;-)
thanks for Your attention,
bye, NicCo
---
Il Venerdì 19 Settembre 2014 14:55, Nicolò Costanza <[email protected]> ha
scritto:
I know it seems late, with the expected release on September 26...
Propose the doing of these tests now, but I as the kernel maintainer
I can not leave unheeded the documented feedback when I receive!
thanks, for your compehension (and testing...)bye, NicCo
Il Venerdì 19 Settembre 2014 14:45, Nicolò Costanza <[email protected]> ha
scritto:
Hi
"Joeghi" is testing different kernels and/with vmware guests
with kernel 3.15.10-1, he reports that > "ata_sff" is at 99%
and so all vmware vm guests are running very slow
with kernel 3.15.10-1.1, he reports that everything is fine
I don't know if this issue may affect only vmware linux vms,
or also other like vbox, or other fields of use or applications
from these rebuils, kernel ver/rel are 3.15.10-1.1
https://abf.rosalinux.ru/kernels_3_15x/Kernel-3.15.10-70-ONE/build_lists
you can install kernel
3.15.10-1.1 for helping us with this testing:
already tested with vmware,
this new release performs very well with omv2014.1 vms guests
these kernels have not ht/smt, probably this feature can harm vmare only,
or may harm other things, but we need to test really these things deeply!
Please, compare these new release 3.15.10-1.1 with the currently official
3.15.10-1
and send us your feedback ASAP!
thanks, for your help with testing
bye, NicCo
_______________________________________________
OM-Cooker mailing list
[email protected]
http://ml.openmandriva.org/listinfo.cgi/om-cooker-openmandriva.org
