[MPlayer-cvslog] r26005 - in trunk/libvo: aclib_template.c osd_template.c

[Ivan Kalvachev]

On Feb 15, 2008 11:52 PM, uau <subversion at mplayerhq.hu> wrote:
> Author: uau
> Date: Fri Feb 15 22:52:34 2008
> New Revision: 26005
>
> Log:
> libvo: change asm syntax to use ASMALIGN and " # nop"
>
> Change ".balign 16\n\t" to ASMALIGN(4) and "/nop" to " # nop". The new
> version is what other code in MPlayer uses, and works with old
> assembler versions like that used on OS X.
>
>
> Modified:
>    trunk/libvo/aclib_template.c
>    trunk/libvo/osd_template.c
>
> Modified: trunk/libvo/aclib_template.c
> >                         "xor %%"REG_a", %%"REG_a"       \n\t"
> -                       ".balign 16             \n\t"
> +                       ASMALIGN(4)
>                         "1:                     \n\t"
[..]
> -                               ".balign 16             \n\t"
> +                               ASMALIGN(4)
>                                 "2:                     \n\t"

I think that recent benchmark  showed that aligning of inner loop with
NOPs doesn't improve the speed. In fact it actually only slows down
the function by the number of nops executed.

Saying, maybe it is good idea to remove them.
Michael, it is your code.