[pulseaudio-discuss] [PATCH] core: Fix a litte-endian bug in ARM svolume code

[Peter Meerwald]

> The old code seems to be written for big-endian code.
 
> NB: This one's really strange since this effectively means that our soft
> volumes on ARMv6 have been mostly broken (since almost everybody uses
> litte-endian). Review would be appreciated!

benchmark results on Cortex-A8 were horrible 
(http://lists.freedesktop.org/archives/pulseaudio-discuss/2012-February/012864.html):

runtime on beagle-xm, 800 MHz:

checking NEON volume_float32ne
NEON: 10223 usec.
ref: 46480 usec.
checking NEON volume_s16ne
NEON: 8484 usec.
ARM: 339272 usec.
ref: 20203 usec.

regards, p.

> ---
>  src/pulsecore/svolume_arm.c |   21 +++++++++++++++++++++
>  1 files changed, 21 insertions(+), 0 deletions(-)
> 
> diff --git a/src/pulsecore/svolume_arm.c b/src/pulsecore/svolume_arm.c
> index b5ccc67..f06a702 100644
> --- a/src/pulsecore/svolume_arm.c
> +++ b/src/pulsecore/svolume_arm.c
> @@ -75,13 +75,22 @@ static void pa_volume_s16ne_arm(int16_t *samples, const int32_t *volumes, unsign
>          " ldrd r2, [r6], #8               \n\t" /* 2 samples at a time */
>          " ldr  r0, [%0]                   \n\t"
>  
> +#ifdef WORDS_BIGENDIAN
>          " smulwt r2, r2, r0               \n\t"
>          " smulwb r3, r3, r0               \n\t"
> +#else
> +        " smulwb r2, r2, r0               \n\t"
> +        " smulwt r3, r3, r0               \n\t"
> +#endif
>  
>          " ssat r2, #16, r2                \n\t"
>          " ssat r3, #16, r3                \n\t"
>  
> +#ifdef WORDS_BIGENDIAN
>          " pkhbt r0, r3, r2, LSL #16       \n\t"
> +#else
> +        " pkhbt r0, r2, r3, LSL #16       \n\t"
> +#endif
>          " str  r0, [%0], #4               \n\t"
>  
>          MOD_INC()
> @@ -95,18 +104,30 @@ static void pa_volume_s16ne_arm(int16_t *samples, const int32_t *volumes, unsign
>          " ldrd r4, [r6], #8               \n\t"
>          " ldrd r0, [%0]                   \n\t"
>  
> +#ifdef WORDS_BIGENDIAN
>          " smulwt r2, r2, r0               \n\t"
>          " smulwb r3, r3, r0               \n\t"
>          " smulwt r4, r4, r1               \n\t"
>          " smulwb r5, r5, r1               \n\t"
> +#else
> +        " smulwb r2, r2, r0               \n\t"
> +        " smulwt r3, r3, r0               \n\t"
> +        " smulwb r4, r4, r1               \n\t"
> +        " smulwt r5, r5, r1               \n\t"
> +#endif
>  
>          " ssat r2, #16, r2                \n\t"
>          " ssat r3, #16, r3                \n\t"
>          " ssat r4, #16, r4                \n\t"
>          " ssat r5, #16, r5                \n\t"
>  
> +#ifdef WORDS_BIGENDIAN
>          " pkhbt r0, r3, r2, LSL #16       \n\t"
>          " pkhbt r1, r5, r4, LSL #16       \n\t"
> +#else
> +        " pkhbt r0, r2, r3, LSL #16       \n\t"
> +        " pkhbt r1, r4, r5, LSL #16       \n\t"
> +#endif
>          " strd  r0, [%0], #8              \n\t"
>  
>          MOD_INC()
> 

-- 

Peter Meerwald
+43-664-2444418 (mobile)