reduce size of task_struct on 64-bit machines
authorWilliam Cohen <wcohen@redhat.com>
Tue, 8 May 2007 07:23:41 +0000 (00:23 -0700)
committerLinus Torvalds <torvalds@woody.linux-foundation.org>
Tue, 8 May 2007 18:14:58 +0000 (11:14 -0700)
commit97dc32cdb1b53832801159d5f634b41aad9d0a23
tree438f59c84d3528de5f68583d312beceb1aa32659
parent4d7bf11d649c72621ca31b8ea12b9c94af380e63
reduce size of task_struct on 64-bit machines

This past week I was playing around with that pahole tool
(http://oops.ghostprotocols.net:81/acme/dwarves/) and looking at the size
of various struct in the kernel.  I was surprised by the size of the
task_struct on x86_64, approaching 4K.  I looked through the fields in
task_struct and found that a number of them were declared as "unsigned
long" rather than "unsigned int" despite them appearing okay as 32-bit
sized fields.  On x86_64 "unsigned long" ends up being 8 bytes in size and
forces 8 byte alignment.  Is there a reason there a reason they are
"unsigned long"?

The patch below drops the size of the struct from 3808 bytes (60 64-byte
cachelines) to 3760 bytes (59 64-byte cachelines).  A couple other fields
in the task struct take a signficant amount of space:

struct thread_struct       thread;               688
struct held_lock           held_locks[30];       1680

CONFIG_LOCKDEP is turned on in the .config

[akpm@linux-foundation.org: fix printk warnings]
Cc: <linux-arch@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/proc/array.c
include/linux/sched.h