kernel configuration, hpet (was notebook with lunar?)

Kok, Auke sofar at foo-projects.org
Thu Mar 13 18:38:51 CET 2008


Zbigniew Luszpinski wrote:
> Tuesday 11 of March 2008 22:42:28 Kok, Auke wrote:
>> Zbigniew Luszpinski wrote:
>>> Tuesday 11 of March 2008 17:04:24 Kok, Auke wrote:
>>>> Zbigniew Luszpinski wrote:
>>>>> [ ] HPET Timer Support
>>>> turn this ON. *always*
>>> My Asus A8N-VM CSM is too old to support this. This is not hardware
>>> limitation but bios acpi problem. Asus refused to add hpet entry to acpi
>>> table and reserved this feature for later boards. What a pity Linux needs
>>> hpet entry in acpi table to use hpet - otherwise I could have it.
>> there are patches that force-enable the HPET in this case, you might want
>> to look at those as HPET will significantly improve battery life on
>> laptops.
> 
> None of the patches I found support my chipset. :-(
> So I wrote new tiny one to enable hpet also for my chipset. :-)
> If you (or someone else here) uses hpet tell me if my output is correct and 
> does not miss anything:
> 
> $ dmesg | grep -i hpet
> Kernel command line: root=/dev/sda8 ro vga=773 video=vesafb:mtrr:4,ywrap 
> vt.default_utf8=0 hpet=force
> Force enabled HPET at base address 0xfed00000
> hpet clockevent registered
> Time: hpet clocksource has been installed.
> 
> $ grep -i hpet /proc/timer_list
> Clock Event Device: hpet
>  set_next_event: hpet_legacy_next_event
>  set_mode:       hpet_legacy_set_mode
> 
> Any other check for hpet I do not know?
> If the results will be OK I will make patch public on nvnews.net for further 
> public testing.
> 
> I'm not sure if hpet will help me on my single core CPU:
> "The HPET provides a stable time base on SMP systems, unlike the TSC, but it 
> is more expensive to access, as it is off-chip." source: CONFIG_HPET_TIMER 
> help from menuconfig.
> 
> So far do not see big difference. Will look how it behaves under load.

try running powertop and seeing if it still complains about HPET being missing.

>>>>> -*- 64 bit Memory and IO resources (EXPERIMENTAL)
>>>> I doubt you need this and it just makes things slower. probably better
>>>> to turn off.
>>> lspci -v tells me that many devices have 64bit+ caps (all are integrated
>>> in Nvidia C51 and MCP51 chipset). I do not remember that I turned it on
>>> because I usually keep away from experimental when I do not have to.
>>> Probably something other autoenabled it as requirement.
>> all this option does is use more bytes to access the same memory. Since you
>> are unlikely to have 4gb+ memory in your laptop, you don't need it and it
>> just wastes CPU cycles.
> 
> I can not disable it. This is dependency for:
>  [*] PAE (Physical Address Extension) Support
> 
> As for memory I have 4x256MiB==1GiB (quadbank :-D ). Of course I'm joking. 
> Quadbank is not supported/useful now. But some casual people say wow.

I don't think you need PAE anyway. you don't have 4gb of memory so that's not
going to be used anyway either.

>>>> looks good otherwise. :)
>>> Thanks for comments. When I replace Asus junk mobo with real hardware I
>>> will remember to enable hpet and upgrade CPU to dual/quad core too. :-)
>>> Only wait for apps using sse4/5.
>> lol, asus usually makes OK motherboards.
> 
> I do not think so. Asus saved cost by not providing heat spreader on south 
> bridge (other manufacturers have some iron on this very same south bridge).
> Southbridge is hot as hell. I got SATA errors few times.

lol, ok I will reconsider ASUS mobo's now :)

> Another issue is bios. Dirty and buggy as 2 year old sandwich.
> It took Asus 6 months to  clean up acpi tables from 1 bad character. Asus only 
> have done this fix under heavy, massive pressure from Linux users.
> google: A8N-VM acpi bad character
> A never fixed bug is in VGA vbios. Some VGA video modes in pure DOS are 
> broken. (the tech support from Asus told me to install widows driver to fix 
> pure DOS video mode problem. When I asked how Windows driver will cure pure 
> DOS video mode problem he answered Asus does not support DOS OS - and that 
> was all the help from Asus).
> Also the mentioned hpet is not present in acpi table so Linux without patching 
> will not see it. In Windows hpet will never be accessible. Asrock and Asus in 
> later models (Vista certified) add HPET enable switch in bios setup. And 
> those mobos use the same chipset I have and look identical (wow new feature 
> hpet).

windows doesn't use HPET at all afaik- the support is just missing in the OS. (I
think)

> Another not solved problem is with AsusUpdate app which freezes Windows when 
> is started. AsusUpdate also reports no file on server error when bios is up 
> to date. My mobo has 3 years so I doubt there will be any new updates. Asus 
> is the worst mobo maker I ever saw. (I could say more but the list is long 
> and contains some ugly things like hangs and freezeing in bioses less than 
> 1007 version). No more Asus. Never.
> 
> These bugs are not important for 99% of people but I use everything what 
> factory implements (except overclocking I do not like).
> 
>> it's more important for libraries to use sse+ stuff, those play an
>> extremely important role under linux. looks like that effort is being
>> worked on too right now.
> 
> It is strange why Intel does not implement 3dnow. They have agreement with AMD 
> on free exchanging CPU opcodes. 3dnow is still useful in older apps 
> (especially top classic games). Some present apps/libs also make use of 
> 3dnow.

not really strange at all. Intel was developing SSE2 when AMD came out with 3dnow
and a lot of that complements functionality but isn't exactly the same. Everyone
is better off if they're distinctly named differently and recognizeable.

AMD also did not choose to implement some SSE2/3 functionality since these engines
can be complex and might not actually be workable based on the CPU design. on top
of that they take die space and that is expensive, so every chip manufacturer will
decide for himself which is better for profit ($$!!) and let marketing choose how
to sell this to their customers, that's just how that works.

we're seeing the same thing now with SSE4 and AMD coming up with a subset of these
instructions and vice versa - it's like cars: some car makers advertise "hybrid!"
and others sell "All Wheel Drive!". Sure it would be nice if all cars had the same
features, but they would all be equally dull ;)


Auke


More information about the Lunar mailing list