Janne Grunau <[email protected]> writes:

> The change is splitted into two patches to ease reviewing. The first patch
> adds the infrastructure and copies the h264 asm and renames the functions.
> The second patch changes the functions to the rv40 variants. The
> modifications are just the straight forward implementations of the
> variable bias rv40 does instead of the rounded shift.
> ---8<---
> based on the h264 asm
>
> 3% faster overall for the first 30 seconds of bourne.rmvb
> 57.845s vs 59.627s (fastest out of 10 tries each) on pandaboard
> ---
>  libavcodec/arm/Makefile            |    3 +
>  libavcodec/arm/rv40dsp_init_neon.c |   42 +++++
>  libavcodec/arm/rv40dsp_neon.S      |  295 
> ++++++++++++++++++++++++++++++++++++
>  libavcodec/rv34dsp.h               |    1 +
>  libavcodec/rv40dsp.c               |    2 +
>  5 files changed, 343 insertions(+), 0 deletions(-)
>  create mode 100644 libavcodec/arm/rv40dsp_init_neon.c
>  create mode 100644 libavcodec/arm/rv40dsp_neon.S

I will review this next week if I don't have time during the weekend.

-- 
Måns Rullgård
[email protected]
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to