lkml.org 
[lkml]   [2009]   [Apr]   [23]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [PATCH 02/22] Do not sanity check order in the fast path
From
Date
On Wed, 2009-04-22 at 18:34 -0700, Dave Hansen wrote:
> I'll also go and see what the actual .text size
> changes are from this patch both for alloc_pages() and
> alloc_pages_node() separately to make sure what we're dealing with
> here. Does this check even *exist* in the optimized code very
> often?

While this isn't definitive by any means, I did get some interesting
results. Pulling the check out of alloc_pages() had no effect at *all*
on text size because I'm trying with CONFIG_NUMA=n.

$ size i386-T41-laptop.{0,1}/vmlinux
text data bss dec hex filename
4348625 286560 860160 5495345 53da31 i386-T41-laptop.0/vmlinux
4348625 286560 860160 5495345 53da31 i386-T41-laptop.1/vmlinux

We get a slightly different when pulling the check out of
alloc_pages_node():

$ size i386-T41-laptop.{1,2}/vmlinux
text data bss dec hex filename
4348625 286560 860160 5495345 53da31 i386-T41-laptop.1/vmlinux
4348601 286560 860160 5495321 53da19 i386-T41-laptop.2/vmlinux

$ bloat-o-meter i386-T41-laptop.1/vmlinux i386-T41-laptop.2/vmlinux
add/remove: 0/0 grow/shrink: 9/7 up/down: 78/-107 (-29)
function old new delta
__get_user_pages 717 751 +34
st_read 1936 1944 +8
shmem_truncate_range 1660 1667 +7
pci_create_slot 410 417 +7
sg_build_indirect 449 455 +6
n_tty_read 1336 1342 +6
find_vma_prepare 103 108 +5
as_update_iohist 617 621 +4
ntfs_readdir 3426 3427 +1
enlarge_buffer 343 341 -2
__get_free_pages 36 33 -3
dma_generic_alloc_coherent 207 202 -5
mempool_alloc_pages 33 17 -16
futex_lock_pi 2120 2104 -16
kallsyms_lookup_name 102 82 -20
cache_alloc_refill 1171 1126 -45

I'm going to retry this with a NUMA config.

-- Dave



\
 
 \ /
  Last update: 2009-04-23 21:29    [W:0.220 / U:0.220 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site