Skip to contentSkip to conversation

b72fd1470c9: -41.7% perf-profile.cpu-cycles.get_page_from_freelist.__alloc_pages_nodemask.alloc_pages_current.__page_cache_alloc.pagecache_get_page

LKP Aaron Lu 2 participants 4 messages Jul 31, 2014

lore
Share & tools
lore

Aaron Lu

author
#0 lore raw

FYI, we noticed the below changes on

git://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git master
commit b72fd1470c9735f53485d089aa918dc327a86077 ("mm: rearrange zone fields into read-only, page alloc, statistics and page reclaim lines")

test case: lkp-st02/dd-write/5m-11HDD-JBOD-cfq-xfs-10dd

e28c951ff01a805 b72fd1470c9735f53485d089a --------------- -------------------------

      1.06 ~ 6%     -41.7%       0.62 ~ 3%  TOTAL perf-profile.cpu-cycles.get_page_from_freelist.__alloc_pages_nodemask.alloc_pages_current.__page_cache_alloc.pagecache_get_page
      1.34 ~ 2%     -19.8%       1.07 ~ 2%  TOTAL perf-profile.cpu-cycles.__block_write_begin.xfs_vm_write_begin.generic_perform_write.xfs_file_buffered_aio_write.xfs_file_write_iter
      1.19 ~ 5%     -12.1%       1.05 ~ 4%  TOTAL perf-profile.cpu-cycles.copy_from_user_atomic_iovec.iov_iter_copy_from_user_atomic.generic_perform_write.xfs_file_buffered_aio_write.xfs_file_write_iter
      2.78 ~ 1%     -16.3%       2.32 ~ 4%  TOTAL perf-profile.cpu-cycles.__clear_user.read_zero.read_zero.vfs_read.sys_read
  2.96e+09 ~ 4%      -5.2%  2.806e+09 ~ 0%  TOTAL perf-stat.cache-misses
  3.86e+12 ~ 5%      -5.2%  3.658e+12 ~ 1%  TOTAL perf-stat.ref-cycles

Legend:
~XX% - stddev percent
[+-]XX% - change percent

Disclaimer:
Results have been estimated based on internal Intel analysis and are provided for informational purposes only. Any difference in system hardware or software design or configuration may affect actual performance.

Thanks,
Aaron

echo 1 > /sys/kernel/debug/tracing/events/writeback/balance_dirty_pages/enable echo 1 > /sys/kernel/debug/tracing/events/writeback/bdi_dirty_ratelimit/enable echo 1 > /sys/kernel/debug/tracing/events/writeback/global_dirty_state/enable echo 1 > /sys/kernel/debug/tracing/events/writeback/writeback_single_inode/enable mkfs -t xfs /dev/sdb1
mkfs -t xfs /dev/sdj1
mkfs -t xfs /dev/sdh1
mkfs -t xfs /dev/sde1
mkfs -t xfs /dev/sdc1
mkfs -t xfs /dev/sdk1
mkfs -t xfs /dev/sdd1
mkfs -t xfs /dev/sdf1
mkfs -t xfs /dev/sdg1
mkfs -t xfs /dev/sdi1
mkfs -t xfs /dev/sdl1
mount -t xfs -o nobarrier,inode64 /dev/sdb1 /fs/sdb1
mount -t xfs -o nobarrier,inode64 /dev/sdc1 /fs/sdc1
mount -t xfs -o nobarrier,inode64 /dev/sdd1 /fs/sdd1
mount -t xfs -o nobarrier,inode64 /dev/sde1 /fs/sde1
mount -t xfs -o nobarrier,inode64 /dev/sdf1 /fs/sdf1
mount -t xfs -o nobarrier,inode64 /dev/sdg1 /fs/sdg1
mount -t xfs -o nobarrier,inode64 /dev/sdh1 /fs/sdh1
mount -t xfs -o nobarrier,inode64 /dev/sdi1 /fs/sdi1
mount -t xfs -o nobarrier,inode64 /dev/sdj1 /fs/sdj1
mount -t xfs -o nobarrier,inode64 /dev/sdk1 /fs/sdk1
mount -t xfs -o nobarrier,inode64 /dev/sdl1 /fs/sdl1

dd  if=/dev/zero of=/fs/sdb1/zero-1 status=noxfer &
dd  if=/dev/zero of=/fs/sdc1/zero-1 status=noxfer &
dd  if=/dev/zero of=/fs/sdd1/zero-1 status=noxfer &
dd  if=/dev/zero of=/fs/sde1/zero-1 status=noxfer &
dd  if=/dev/zero of=/fs/sdf1/zero-1 status=noxfer &
dd  if=/dev/zero of=/fs/sdg1/zero-1 status=noxfer &
dd  if=/dev/zero of=/fs/sdh1/zero-1 status=noxfer &
dd  if=/dev/zero of=/fs/sdi1/zero-1 status=noxfer &
dd  if=/dev/zero of=/fs/sdj1/zero-1 status=noxfer &
dd  if=/dev/zero of=/fs/sdk1/zero-1 status=noxfer &
dd  if=/dev/zero of=/fs/sdl1/zero-1 status=noxfer &
dd  if=/dev/zero of=/fs/sdb1/zero-2 status=noxfer &
dd  if=/dev/zero of=/fs/sdc1/zero-2 status=noxfer &
dd  if=/dev/zero of=/fs/sdd1/zero-2 status=noxfer &
dd  if=/dev/zero of=/fs/sde1/zero-2 status=noxfer &
dd  if=/dev/zero of=/fs/sdf1/zero-2 status=noxfer &
dd  if=/dev/zero of=/fs/sdg1/zero-2 status=noxfer &
dd  if=/dev/zero of=/fs/sdh1/zero-2 status=noxfer &
dd  if=/dev/zero of=/fs/sdi1/zero-2 status=noxfer &
dd  if=/dev/zero of=/fs/sdj1/zero-2 status=noxfer &
dd  if=/dev/zero of=/fs/sdk1/zero-2 status=noxfer &
dd  if=/dev/zero of=/fs/sdl1/zero-2 status=noxfer &
dd  if=/dev/zero of=/fs/sdb1/zero-3 status=noxfer &
dd  if=/dev/zero of=/fs/sdc1/zero-3 status=noxfer &
dd  if=/dev/zero of=/fs/sdd1/zero-3 status=noxfer &
dd  if=/dev/zero of=/fs/sde1/zero-3 status=noxfer &
dd  if=/dev/zero of=/fs/sdf1/zero-3 status=noxfer &
dd  if=/dev/zero of=/fs/sdg1/zero-3 status=noxfer &
dd  if=/dev/zero of=/fs/sdh1/zero-3 status=noxfer &
dd  if=/dev/zero of=/fs/sdi1/zero-3 status=noxfer &
dd  if=/dev/zero of=/fs/sdj1/zero-3 status=noxfer &
dd  if=/dev/zero of=/fs/sdk1/zero-3 status=noxfer &
dd  if=/dev/zero of=/fs/sdl1/zero-3 status=noxfer &
dd  if=/dev/zero of=/fs/sdb1/zero-4 status=noxfer &
dd  if=/dev/zero of=/fs/sdc1/zero-4 status=noxfer &
dd  if=/dev/zero of=/fs/sdd1/zero-4 status=noxfer &
dd  if=/dev/zero of=/fs/sde1/zero-4 status=noxfer &
dd  if=/dev/zero of=/fs/sdf1/zero-4 status=noxfer &
dd  if=/dev/zero of=/fs/sdg1/zero-4 status=noxfer &
dd  if=/dev/zero of=/fs/sdh1/zero-4 status=noxfer &
dd  if=/dev/zero of=/fs/sdi1/zero-4 status=noxfer &
dd  if=/dev/zero of=/fs/sdj1/zero-4 status=noxfer &
dd  if=/dev/zero of=/fs/sdk1/zero-4 status=noxfer &
dd  if=/dev/zero of=/fs/sdl1/zero-4 status=noxfer &
dd  if=/dev/zero of=/fs/sdb1/zero-5 status=noxfer &
dd  if=/dev/zero of=/fs/sdc1/zero-5 status=noxfer &
dd  if=/dev/zero of=/fs/sdd1/zero-5 status=noxfer &
dd  if=/dev/zero of=/fs/sde1/zero-5 status=noxfer &
dd  if=/dev/zero of=/fs/sdf1/zero-5 status=noxfer &
dd  if=/dev/zero of=/fs/sdg1/zero-5 status=noxfer &
dd  if=/dev/zero of=/fs/sdh1/zero-5 status=noxfer &
dd  if=/dev/zero of=/fs/sdi1/zero-5 status=noxfer &
dd  if=/dev/zero of=/fs/sdj1/zero-5 status=noxfer &
dd  if=/dev/zero of=/fs/sdk1/zero-5 status=noxfer &
dd  if=/dev/zero of=/fs/sdl1/zero-5 status=noxfer &
dd  if=/dev/zero of=/fs/sdb1/zero-6 status=noxfer &
dd  if=/dev/zero of=/fs/sdc1/zero-6 status=noxfer &
dd  if=/dev/zero of=/fs/sdd1/zero-6 status=noxfer &
dd  if=/dev/zero of=/fs/sde1/zero-6 status=noxfer &
dd  if=/dev/zero of=/fs/sdf1/zero-6 status=noxfer &
dd  if=/dev/zero of=/fs/sdg1/zero-6 status=noxfer &
dd  if=/dev/zero of=/fs/sdh1/zero-6 status=noxfer &
dd  if=/dev/zero of=/fs/sdi1/zero-6 status=noxfer &
dd  if=/dev/zero of=/fs/sdj1/zero-6 status=noxfer &
dd  if=/dev/zero of=/fs/sdk1/zero-6 status=noxfer &
dd  if=/dev/zero of=/fs/sdl1/zero-6 status=noxfer &
dd  if=/dev/zero of=/fs/sdb1/zero-7 status=noxfer &
dd  if=/dev/zero of=/fs/sdc1/zero-7 status=noxfer &
dd  if=/dev/zero of=/fs/sdd1/zero-7 status=noxfer &
dd  if=/dev/zero of=/fs/sde1/zero-7 status=noxfer &
dd  if=/dev/zero of=/fs/sdf1/zero-7 status=noxfer &
dd  if=/dev/zero of=/fs/sdg1/zero-7 status=noxfer &
dd  if=/dev/zero of=/fs/sdh1/zero-7 status=noxfer &
dd  if=/dev/zero of=/fs/sdi1/zero-7 status=noxfer &
dd  if=/dev/zero of=/fs/sdj1/zero-7 status=noxfer &
dd  if=/dev/zero of=/fs/sdk1/zero-7 status=noxfer &
dd  if=/dev/zero of=/fs/sdl1/zero-7 status=noxfer &
dd  if=/dev/zero of=/fs/sdb1/zero-8 status=noxfer &
dd  if=/dev/zero of=/fs/sdc1/zero-8 status=noxfer &
dd  if=/dev/zero of=/fs/sdd1/zero-8 status=noxfer &
dd  if=/dev/zero of=/fs/sde1/zero-8 status=noxfer &
dd  if=/dev/zero of=/fs/sdf1/zero-8 status=noxfer &
dd  if=/dev/zero of=/fs/sdg1/zero-8 status=noxfer &
dd  if=/dev/zero of=/fs/sdh1/zero-8 status=noxfer &
dd  if=/dev/zero of=/fs/sdi1/zero-8 status=noxfer &
dd  if=/dev/zero of=/fs/sdj1/zero-8 status=noxfer &
dd  if=/dev/zero of=/fs/sdk1/zero-8 status=noxfer &
dd  if=/dev/zero of=/fs/sdl1/zero-8 status=noxfer &
dd  if=/dev/zero of=/fs/sdb1/zero-9 status=noxfer &
dd  if=/dev/zero of=/fs/sdc1/zero-9 status=noxfer &
dd  if=/dev/zero of=/fs/sdd1/zero-9 status=noxfer &
dd  if=/dev/zero of=/fs/sde1/zero-9 status=noxfer &
dd  if=/dev/zero of=/fs/sdf1/zero-9 status=noxfer &
dd  if=/dev/zero of=/fs/sdg1/zero-9 status=noxfer &
dd  if=/dev/zero of=/fs/sdh1/zero-9 status=noxfer &
dd  if=/dev/zero of=/fs/sdi1/zero-9 status=noxfer &
dd  if=/dev/zero of=/fs/sdj1/zero-9 status=noxfer &
dd  if=/dev/zero of=/fs/sdk1/zero-9 status=noxfer &
dd  if=/dev/zero of=/fs/sdl1/zero-9 status=noxfer &
dd  if=/dev/zero of=/fs/sdb1/zero-10 status=noxfer &
dd  if=/dev/zero of=/fs/sdc1/zero-10 status=noxfer &
dd  if=/dev/zero of=/fs/sdd1/zero-10 status=noxfer &
dd  if=/dev/zero of=/fs/sde1/zero-10 status=noxfer &
dd  if=/dev/zero of=/fs/sdf1/zero-10 status=noxfer &
dd  if=/dev/zero of=/fs/sdg1/zero-10 status=noxfer &
dd  if=/dev/zero of=/fs/sdh1/zero-10 status=noxfer &
dd  if=/dev/zero of=/fs/sdi1/zero-10 status=noxfer &
dd  if=/dev/zero of=/fs/sdj1/zero-10 status=noxfer &
dd  if=/dev/zero of=/fs/sdk1/zero-10 status=noxfer &
dd  if=/dev/zero of=/fs/sdl1/zero-10 status=noxfer &

sleep 298
killall -9 dd

#1 lore raw
Quoted 16 lines · Aaron Lu

FYI, we noticed the below changes on

git://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git master
commit b72fd1470c9735f53485d089aa918dc327a86077 ("mm: rearrange zone fields into read-only, page alloc, statistics and page reclaim lines")

test case: lkp-st02/dd-write/5m-11HDD-JBOD-cfq-xfs-10dd

e28c951ff01a805 b72fd1470c9735f53485d089a
--------------- -------------------------

      1.06 ~ 6%     -41.7%       0.62 ~ 3%  TOTAL perf-profile.cpu-cycles.get_page_from_freelist.__alloc_pages_nodemask.alloc_pages_current.__page_cache_alloc.pagecache_get_page
      1.34 ~ 2%     -19.8%       1.07 ~ 2%  TOTAL perf-profile.cpu-cycles.__block_write_begin.xfs_vm_write_begin.generic_perform_write.xfs_file_buffered_aio_write.xfs_file_write_iter
      1.19 ~ 5%     -12.1%       1.05 ~ 4%  TOTAL perf-profile.cpu-cycles.copy_from_user_atomic_iovec.iov_iter_copy_from_user_atomic.generic_perform_write.xfs_file_buffered_aio_write.xfs_file_write_iter
      2.78 ~ 1%     -16.3%       2.32 ~ 4%  TOTAL perf-profile.cpu-cycles.__clear_user.read_zero.read_zero.vfs_read.sys_read
  2.96e+09 ~ 4%      -5.2%  2.806e+09 ~ 0%  TOTAL perf-stat.cache-misses
  3.86e+12 ~ 5%      -5.2%  3.658e+12 ~ 1%  TOTAL perf-stat.ref-cycles
Quoted 4 lines

Legend:
~XX% - stddev percent
[+-]XX% - change percent

I'm not exactly sure what I'm reading here. I think it is reporting on cpu cycles and cache misses used in various kernel functions. It's not clear what the units are but it looks like percentages of overall cycles spent in the reported functions. That may or may not be good depending on whether there is a higher cost elsewhere pushing the percentages down but that detail is not in the report. It looks like this is reporting that fewer clock cycles are being spent and incurring fewer cache misses. What is the problem?

Signature · 2 lines
Mel Gorman
SUSE Labs
author
#2 lore raw
Quoted 22 lines · Mel Gorman
Quoted 16 lines · Aaron Lu

FYI, we noticed the below changes on

git://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git master
commit b72fd1470c9735f53485d089aa918dc327a86077 ("mm: rearrange zone fields into read-only, page alloc, statistics and page reclaim lines")

test case: lkp-st02/dd-write/5m-11HDD-JBOD-cfq-xfs-10dd

e28c951ff01a805 b72fd1470c9735f53485d089a
--------------- -------------------------

      1.06 ~ 6%     -41.7%       0.62 ~ 3%  TOTAL perf-profile.cpu-cycles.get_page_from_freelist.__alloc_pages_nodemask.alloc_pages_current.__page_cache_alloc.pagecache_get_page
      1.34 ~ 2%     -19.8%       1.07 ~ 2%  TOTAL perf-profile.cpu-cycles.__block_write_begin.xfs_vm_write_begin.generic_perform_write.xfs_file_buffered_aio_write.xfs_file_write_iter
      1.19 ~ 5%     -12.1%       1.05 ~ 4%  TOTAL perf-profile.cpu-cycles.copy_from_user_atomic_iovec.iov_iter_copy_from_user_atomic.generic_perform_write.xfs_file_buffered_aio_write.xfs_file_write_iter
      2.78 ~ 1%     -16.3%       2.32 ~ 4%  TOTAL perf-profile.cpu-cycles.__clear_user.read_zero.read_zero.vfs_read.sys_read
  2.96e+09 ~ 4%      -5.2%  2.806e+09 ~ 0%  TOTAL perf-stat.cache-misses
  3.86e+12 ~ 5%      -5.2%  3.658e+12 ~ 1%  TOTAL perf-stat.ref-cycles
Quoted 4 lines

Legend:
~XX% - stddev percent
[+-]XX% - change percent

Quoted 7 lines

I'm not exactly sure what I'm reading here. I think it is reporting on cpu cycles and cache misses used in various kernel functions. It's not clear what the units are but it looks like percentages of overall cycles spent in the reported functions. That may or may not be good depending on whether there is a higher cost elsewhere pushing the percentages down but that detail is not in the report. It looks like this is reporting that fewer clock cycles are being spent and incurring fewer cache misses. What is the problem?

LKP does not report problems only, it will also report commits that make things better :-)

Quoted 1 line

From the perf-stat.cache-misses, I think it is indicating your commit

does something for good.

Thanks,
Aaron

#3 lore raw
Quoted 31 lines · Aaron Lu
Quoted 22 lines · Mel Gorman
Quoted 16 lines · Aaron Lu

FYI, we noticed the below changes on

git://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git master
commit b72fd1470c9735f53485d089aa918dc327a86077 ("mm: rearrange zone fields into read-only, page alloc, statistics and page reclaim lines")

test case: lkp-st02/dd-write/5m-11HDD-JBOD-cfq-xfs-10dd

e28c951ff01a805 b72fd1470c9735f53485d089a
--------------- -------------------------

      1.06 ~ 6%     -41.7%       0.62 ~ 3%  TOTAL perf-profile.cpu-cycles.get_page_from_freelist.__alloc_pages_nodemask.alloc_pages_current.__page_cache_alloc.pagecache_get_page
      1.34 ~ 2%     -19.8%       1.07 ~ 2%  TOTAL perf-profile.cpu-cycles.__block_write_begin.xfs_vm_write_begin.generic_perform_write.xfs_file_buffered_aio_write.xfs_file_write_iter
      1.19 ~ 5%     -12.1%       1.05 ~ 4%  TOTAL perf-profile.cpu-cycles.copy_from_user_atomic_iovec.iov_iter_copy_from_user_atomic.generic_perform_write.xfs_file_buffered_aio_write.xfs_file_write_iter
      2.78 ~ 1%     -16.3%       2.32 ~ 4%  TOTAL perf-profile.cpu-cycles.__clear_user.read_zero.read_zero.vfs_read.sys_read
  2.96e+09 ~ 4%      -5.2%  2.806e+09 ~ 0%  TOTAL perf-stat.cache-misses
  3.86e+12 ~ 5%      -5.2%  3.658e+12 ~ 1%  TOTAL perf-stat.ref-cycles
Quoted 4 lines

Legend:
~XX% - stddev percent
[+-]XX% - change percent

Quoted 7 lines

I'm not exactly sure what I'm reading here. I think it is reporting on cpu cycles and cache misses used in various kernel functions. It's not clear what the units are but it looks like percentages of overall cycles spent in the reported functions. That may or may not be good depending on whether there is a higher cost elsewhere pushing the percentages down but that detail is not in the report. It looks like this is reporting that fewer clock cycles are being spent and incurring fewer cache misses. What is the problem?

Quoted 6 lines

LKP does not report problems only, it will also report commits that make things better :-)

From the perf-stat.cache-misses, I think it is indicating your commit does something for good.

Hooray! Thanks for the good news :D

Signature · 2 lines
Mel Gorman
SUSE Labs