Re: svn commit: r33043 - trunk/subversion/libsvn_client

Philip Martin <philip_at_codematters.co.uk> writes:

> I'd be surprised if the "performance gain" is measurable, even without
> optimisation. Once you add optimisation any gain might well disapear
> altogether, have you looked at the generated code? As for

Take svn_diff_mem_string_diff for example. Using gcc 4.1.2 and -O2 I
get the same generated code

0x00007f1f53ebd50f <svn_diff_mem_string_diff+111>: callq 0x7f1f53eb7210 <svn_diff_diff_at_plt>
0x00007f1f53ebd514 <svn_diff_mem_string_diff+116>: mov 0xa0(%rsp),%rbx
0x00007f1f53ebd51c <svn_diff_mem_string_diff+124>: mov 0xa8(%rsp),%r12
0x00007f1f53ebd524 <svn_diff_mem_string_diff+132>: mov 0xb0(%rsp),%r13
0x00007f1f53ebd52c <svn_diff_mem_string_diff+140>: mov 0xb8(%rsp),%r14
0x00007f1f53ebd534 <svn_diff_mem_string_diff+148>: mov 0xc0(%rsp),%r15
0x00007f1f53ebd53c <svn_diff_mem_string_diff+156>: add $0xc8,%rsp
0x00007f1f53ebd543 <svn_diff_mem_string_diff+163>: retq

both before and after your change. Any performance gain is an illusion.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe_at_subversion.tigris.org
For additional commands, e-mail: dev-help_at_subversion.tigris.org
Received on 2008-09-13 21:27:58 CEST