Eric Dumazet a écrit :
ffffffff8011db70 <pfn_to_page>: /* pfn_to_page total: 487244 0.8036 */
18501 0.0305 :ffffffff8011db70: mov 2965994(%rip),%ecx
# <memnode_shift>
83312 0.1374 :ffffffff8011db76: mov %rdi,%rax
2 3.3e-06 :ffffffff8011db79: push %rbp
:ffffffff8011db7a: shl $0xc,%rax
21880 0.0361 :ffffffff8011db7e: mov %rsp,%rbp
601 9.9e-04 :ffffffff8011db81: shr %cl,%rax
:ffffffff8011db84: movzbq
0xffffffff803f0d60(%rax),%rax
56462 0.0931 :ffffffff8011db8c: mov
0xffffffff804264e0(,%rax,8),%rcx
26 4.3e-05 :ffffffff8011db94: mov %rdi,%rax
1 1.6e-06 :ffffffff8011db97: sub 0x3340(%rcx),%rax
164872 0.2719 :ffffffff8011db9e: lea 0x0(,%rax,8),%rdx
43 7.1e-05 :ffffffff8011dba6: shl $0x6,%rax
:ffffffff8011dbaa: sub %rdx,%rax
73574 0.1214 :ffffffff8011dbad: add 0x3330(%rcx),%rax
38 6.3e-05 :ffffffff8011dbb4: leaveq
67932 0.1120 :ffffffff8011dbb5: retq
Some comments about pfn_to_page()
1) reorder struct pglist_data
-----------------------------
0x3340(%rcx) is node_start_pfn, and 0x3330(%rcx) is node_mem_map
These offsets show they may be in different cache lines.
Maybe we should change 'struct pglist_data' ordering so that those two fields
are in the same memory cache line ?
2) Place memnodeshift in front of memnodemap
---------------------------------------------
For most setups, only the first bytes in memnodemap[] are read.
And the linker usually puts memnodeshift AFTER memnodemap.
Two different cache lines are used.
I suggest the following change (seems better than playing vmlinux.lds games)
include/asm-x86_64/mmzone.h
struct memnode {
int shift;
u8 map[NODEMAPSIZE];
} ____cacheline_aligned;
extern struct memnode memnode;
#define memnode_shift memnode.shift
#define memnodemap memnode.map
Thank you
Eric
-
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
