About highmem in 2.6

About highmem in 2.6

[Luis Miguel García]

Hi,

When I first installed 2.4, someone told me that if I had 1 gb ram it 
was better to not use highmem because those extra aditional mb was not 
worth the speed penalty of using the feature.

Sorry for my ignorance (and my sucking english) but must I enable 
highmem now with 2.6? or have it any speed penalty althought?

Thanks a lot!

Re: About highmem in 2.6

[Dave McCracken]

--On Wednesday, February 11, 2004 18:47:04 +0100 Luis Miguel García
<ktech@wanadoo.es> wrote:

> When I first installed 2.4, someone told me that if I had 1 gb ram it was
> better to not use highmem because those extra aditional mb was not worth
> the speed penalty of using the feature.
> 
> Sorry for my ignorance (and my sucking english) but must I enable highmem
> now with 2.6? or have it any speed penalty althought?

I don't know if anyone has actually measured the relative performance, but
I'd expect the answer to be the same as 2.4.  There is a small but
measurable performance penalty for enabling highmem which is higher than
the benefit of the extra 128 meg of memory you get when you have 1G.  If
you have more than 1G it's better to enable highmem.

Dave McCracken

Re: About highmem in 2.6

[Mark de Vries]

Dave McCracken wrote:
> --On Wednesday, February 11, 2004 18:47:04 +0100 Luis Miguel García
> <ktech@wanadoo.es> wrote:
> 
> 
>>When I first installed 2.4, someone told me that if I had 1 gb ram it was
>>better to not use highmem because those extra aditional mb was not worth
>>the speed penalty of using the feature.
>>
>>Sorry for my ignorance (and my sucking english) but must I enable highmem
>>now with 2.6? or have it any speed penalty althought?
> 
> 
> I don't know if anyone has actually measured the relative performance, but
> I'd expect the answer to be the same as 2.4.  There is a small but
> measurable performance penalty for enabling highmem which is higher than
> the benefit of the extra 128 meg of memory you get when you have 1G.  If
> you have more than 1G it's better to enable highmem.
> 

I've been using this patch for a while now on my box (with 1GB):
http://www.kernel.org/pub/linux/kernel/people/andrea/kernels/v2.4/2.4.23aa1/00_3.5G-address-space-5
(kernel is 'vanilla' otherwise)

This allows you to use your full 1GB w/out highmem support.... (2G/2G 
user/kernel addr space split, or something..)

Anything (potentially) wrong/bad about this patch??

Is there a simmilar patch for 2.6??

Rgds,
Mark.

pls. CC in reply, I'm not on the list....

Re: About highmem in 2.6

[Bongani Hlope]

[-- Attachment #1: Type: text/plain, Size: 1658 bytes --]

On Wed, 11 Feb 2004 20:13:10 +0100
Mark de Vries <m.devries@nl.tiscali.com> wrote:

> Dave McCracken wrote:
> > --On Wednesday, February 11, 2004 18:47:04 +0100 Luis Miguel García
> > <ktech@wanadoo.es> wrote:
> > 
> > 
> >>When I first installed 2.4, someone told me that if I had 1 gb ram it was
> >>better to not use highmem because those extra aditional mb was not worth
> >>the speed penalty of using the feature.
> >>
> >>Sorry for my ignorance (and my sucking english) but must I enable highmem
> >>now with 2.6? or have it any speed penalty althought?
> > 
> > 
> > I don't know if anyone has actually measured the relative performance, but
> > I'd expect the answer to be the same as 2.4.  There is a small but
> > measurable performance penalty for enabling highmem which is higher than
> > the benefit of the extra 128 meg of memory you get when you have 1G.  If
> > you have more than 1G it's better to enable highmem.
> > 
> 
> I've been using this patch for a while now on my box (with 1GB):
> http://www.kernel.org/pub/linux/kernel/people/andrea/kernels/v2.4/2.4.23aa1/00_3.5G-address-space-5
> (kernel is 'vanilla' otherwise)
> 
> This allows you to use your full 1GB w/out highmem support.... (2G/2G 
> user/kernel addr space split, or something..)
> 
> Anything (potentially) wrong/bad about this patch??
> 
> Is there a simmilar patch for 2.6??
> 

There is nothing wrong with that patch, the problem with Highmem support on x86 is that is uses an Intel hack to address the full 1Gb of memory, which make memory access a bit slower. The question is, does the 128Mb additional memory worth that penalty?

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: About highmem in 2.6

[Krzysztof Halasa]

Bongani Hlope <bonganilinux@mweb.co.za> writes:

> There is nothing wrong with that patch, the problem with Highmem support
> on x86 is that is uses an Intel hack to address the full 1Gb of memory,
> which make memory access a bit slower. The question is, does the 128Mb
> additional memory worth that penalty?

2GB/2GB split doesn't use any Intel hack nor highmem. In fact for
1 GB of RAM I use a little different split which covers the whole RAM
and gives more virtual RAM, something like 1.2/2.8 GB.
-- 
Krzysztof Halasa, B*FH

Re: About highmem in 2.6

[Martin J. Bligh]

> I've been using this patch for a while now on my box (with 1GB):
> http://www.kernel.org/pub/linux/kernel/people/andrea/kernels/v2.4/2.4.23aa1/00_3.5G-address-space-5
> (kernel is 'vanilla' otherwise)
> 
> This allows you to use your full 1GB w/out highmem support.... (2G/2G user/kernel addr space split, or something..)
> 
> Anything (potentially) wrong/bad about this patch??

The only problem is that you lose some virtual address space for the user,
which isn't a problem for most people.
 
> Is there a simmilar patch for 2.6??

ftp://ftp.kernel.org/pub/linux/kernel/people/mbligh/patches/2.6.2/2.6.2-mjb1/212-config_page_offset

You might have to tweak the config stuff a tiny bit to get the patch to apply 
if you don't have 4/4 split in.

M.

Re: About highmem in 2.6

[Brandon Low]

On Thu, 02/12/04 at 05:02:53 +0100, Krzysztof Halasa wrote:
> Bongani Hlope <bonganilinux@mweb.co.za> writes:
> 
> > There is nothing wrong with that patch, the problem with Highmem support
> > on x86 is that is uses an Intel hack to address the full 1Gb of memory,
> > which make memory access a bit slower. The question is, does the 128Mb
> > additional memory worth that penalty?
> 
> 2GB/2GB split doesn't use any Intel hack nor highmem. In fact for
> 1 GB of RAM I use a little different split which covers the whole RAM
> and gives more virtual RAM, something like 1.2/2.8 GB.

Any chance of me getting a copy of the patch which allows the finer
grained tuning of the kernel/user split?

Thanks.

Brandon Low
> -- 
> Krzysztof Halasa, B*FH
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

Re: About highmem in 2.6

[Krzysztof Halasa]

>> 2GB/2GB split doesn't use any Intel hack nor highmem. In fact for
>> 1 GB of RAM I use a little different split which covers the whole RAM
>> and gives more virtual RAM, something like 1.2/2.8 GB.

While there might a patch for it on the net (so one can configure it with
make *config), I just change 0xC0000000 in arch/i386/vmlinux.lds and in
include/asm-i386/page.h (__PAGE_OFFSET) to something like 0xB0000000.
-- 
Krzysztof Halasa, B*FH