Performance of highpte

Performance of highpte

[Martin J. Bligh]

Kernbench-2: (make -j N vmlinux, where N = 2 x num_cpus)
                        Elapsed        User      System         CPU
         2.5.59-mjb5       45.64      564.71      110.73     1479.50
 2.5.59-mjb5-highpte       46.38      565.32      118.35     1473.50

Hmmm. Looks like we need to dust off UKVA to me.

diffprofile:

3790 page_remove_rmap
3213 default_idle
1299 kmap_atomic
803 kmap_atomic_to_page
776 kmem_cache_free
676 __pte_chain_free
486 page_add_rmap
240 unmap_all_pages
225 kmem_cache_alloc
166 vm_enough_memory
132 do_generic_mapping_read
100 handle_mm_fault
82 buffered_rmqueue
79 __copy_from_user_ll
67 update_atime
66 kunmap_atomic
63 release_pages
58 filemap_nopage
55 file_move
51 generic_file_open
51 find_get_page
...
-52 dput
-61 do_schedule
-63 get_empty_filp
-74 do_page_fault
-96 vfs_read
-97 path_lookup
-121 fd_install
-159 pte_alloc_one
-260 .text.lock.file_table
-372 page_address

Performance of highpte

[Martin J. Bligh]

Hmmm. Looks like we need to dust off UKVA to me.

diffprofile:

3790 page_remove_rmap
3213 default_idle
1299 kmap_atomic
803 kmap_atomic_to_page
776 kmem_cache_free
676 __pte_chain_free
486 page_add_rmap
240 unmap_all_pages
225 kmem_cache_alloc
166 vm_enough_memory
132 do_generic_mapping_read
100 handle_mm_fault
82 buffered_rmqueue
79 __copy_from_user_ll
67 update_atime
66 kunmap_atomic
63 release_pages
58 filemap_nopage
55 file_move
51 generic_file_open
51 find_get_page
...
-52 dput
-61 do_schedule
-63 get_empty_filp
-74 do_page_fault
-96 vfs_read
-97 path_lookup
-121 fd_install
-159 pte_alloc_one
-260 .text.lock.file_table
-372 page_address

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/

Re: Performance of highpte

[Martin J. Bligh]

Odd. linux-mm helpfully stripped the results ... I'll try once more below,
if that doesn't work, try getting it off linux-kernel.


> Kernbench-2: (make -j N vmlinux, where N = 2 x num_cpus)
>                         Elapsed        User      System         CPU
>          2.5.59-mjb5       45.64      564.71      110.73     1479.50
>  2.5.59-mjb5-highpte       46.38      565.32      118.35     1473.50


--On Saturday, February 08, 2003 11:29:33 -0800 "Martin J. Bligh"
<mbligh@aracnet.com> wrote:

> Hmmm. Looks like we need to dust off UKVA to me.
> 
> diffprofile:
> 
> 3790 page_remove_rmap
> 3213 default_idle
> 1299 kmap_atomic
> 803 kmap_atomic_to_page
> 776 kmem_cache_free
> 676 __pte_chain_free
> 486 page_add_rmap
> 240 unmap_all_pages
> 225 kmem_cache_alloc
> 166 vm_enough_memory
> 132 do_generic_mapping_read
> 100 handle_mm_fault
> 82 buffered_rmqueue
> 79 __copy_from_user_ll
> 67 update_atime
> 66 kunmap_atomic
> 63 release_pages
> 58 filemap_nopage
> 55 file_move
> 51 generic_file_open
> 51 find_get_page
> ...
> -52 dput
> -61 do_schedule
> -63 get_empty_filp
> -74 do_page_fault
> -96 vfs_read
> -97 path_lookup
> -121 fd_install
> -159 pte_alloc_one
> -260 .text.lock.file_table
> -372 page_address
> 
> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to majordomo@kvack.org.  For more info on Linux MM,
> see: http://www.linux-mm.org/
> 
> 


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/

Re: Performance of highpte

[Benjamin LaHaise]

On Sat, Feb 08, 2003 at 11:33:06AM -0800, Martin J. Bligh wrote:
> Odd. linux-mm helpfully stripped the results ... I'll try once more below,
> if that doesn't work, try getting it off linux-kernel.

They weren't actually stripped, they somehow became part of the header.  
Hmmmm, now that I have a test message I can actually track down the 
problem.

		-ben
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/