is_head_of_free_region slowing down swsusp

is_head_of_free_region slowing down swsusp

[Pavel Machek]

Hi!

is_head_of_free_region with count_and_copy_zone results results in
pretty nasty O(number_of_free_regions^2) behaviour, and some users see
cpu spending 40 seconds there :-(.

Actually count_and_copy_zone would probably be happy with
"is_free_page()".

I asked Lukas (who is seeing this problem) to kill locking from
is_head_of_free_region [we are running singlethreaded at this point,
so it should be ok]. Do you have any other ideas?
								Pavel
-- 
People were complaining that M$ turns users into beta-testers...
...jr ghea gurz vagb qrirybcref, naq gurl frrz gb yvxr vg gung jnl!

Re: is_head_of_free_region slowing down swsusp

[Nigel Cunningham]

Hi Pavel.

On Fri, 2004-08-13 at 08:23, Pavel Machek wrote:
> is_head_of_free_region with count_and_copy_zone results results in
> pretty nasty O(number_of_free_regions^2) behaviour, and some users see
> cpu spending 40 seconds there :-(.
> 
> Actually count_and_copy_zone would probably be happy with
> "is_free_page()".

Take a look at my implementation. I do a one-time pass through the slow
path, building a bitmap of free pages. is_head_of_free_region is then
simply a O(1) loop through the bitmap.

Nigel
-- 
Nigel Cunningham
Christian Reformed Church of Tuggeranong
PO Box 1004, Tuggeranong, ACT 2901

Many today claim to be tolerant. But true tolerance can cope with others
being intolerant.

Re: is_head_of_free_region slowing down swsusp

[Pavel Machek]

Hi!

> > is_head_of_free_region with count_and_copy_zone results results in
> > pretty nasty O(number_of_free_regions^2) behaviour, and some users see
> > cpu spending 40 seconds there :-(.
> > 
> > Actually count_and_copy_zone would probably be happy with
> > "is_free_page()".
> 
> Take a look at my implementation. I do a one-time pass through the slow
> path, building a bitmap of free pages. is_head_of_free_region is then
> simply a O(1) loop through the bitmap.

I've seen that solution (thanks)... I'd like to do something simpler.

									Pavel
-- 
People were complaining that M$ turns users into beta-testers...
...jr ghea gurz vagb qrirybcref, naq gurl frrz gb yvxr vg gung jnl!

Re: is_head_of_free_region slowing down swsusp

[Nigel Cunningham]

Hi.

On Fri, 2004-08-13 at 09:22, Pavel Machek wrote:
> > Take a look at my implementation. I do a one-time pass through the slow
> > path, building a bitmap of free pages. is_head_of_free_region is then
> > simply a O(1) loop through the bitmap.
> 
> I've seen that solution (thanks)... I'd like to do something simpler.

Umm. Okay. I didn't think it was complicated.

Nigel
-- 
Nigel Cunningham
Christian Reformed Church of Tuggeranong
PO Box 1004, Tuggeranong, ACT 2901

Many today claim to be tolerant. But true tolerance can cope with others
being intolerant.

Re: is_head_of_free_region slowing down swsusp

[Pavel Machek]

Hi!

> is_head_of_free_region with count_and_copy_zone results results in
> pretty nasty O(number_of_free_regions^2) behaviour, and some users see
> cpu spending 40 seconds there :-(.
> 
> Actually count_and_copy_zone would probably be happy with
> "is_free_page()".
> 
> I asked Lukas (who is seeing this problem) to kill locking from
> is_head_of_free_region [we are running singlethreaded at this point,
> so it should be ok]. Do you have any other ideas?

Thanks to Lukas Horalek for some timing info....

I now reproduced in on my machine. updatedb is enough to trigger
40seconds of copying. Killing locking does not help :-(.

								Pavel
-- 
People were complaining that M$ turns users into beta-testers...
...jr ghea gurz vagb qrirybcref, naq gurl frrz gb yvxr vg gung jnl!