[libav-devel] [PATCH 12/13] arm64: int32_to_float_fmul neon asm

Martin Storsjö martin at martin.st
Thu Dec 10 14:55:00 CET 2015


On Wed, 9 Dec 2015, Janne Grunau wrote:

> 3% faster dts decoding on a cortex-a57.
>
>                                 cortex-a57   cortex-a53
> int32_to_float_fmul_array8_c:    1270.9       4475.6
> int32_to_float_fmul_array8_neon:  328.6        569.2
> int32_to_float_fmul_scalar_c:     928.5       4119.6
> int32_to_float_fmul_scalar_neon:  309.1        524.1
> ---
> libavcodec/fmtconvert.h              |  1 +
> libavcodec/aarch64/fmtconvert_init.c | 43 ++++++++++++++++++++
> libavcodec/fmtconvert.c              | 11 ++++--
> libavcodec/aarch64/fmtconvert_neon.S | 76 ++++++++++++++++++++++++++++++++++++
> libavcodec/aarch64/Makefile          |  2 +
> 5 files changed, 130 insertions(+), 3 deletions(-)
> create mode 100644 libavcodec/aarch64/fmtconvert_init.c
> create mode 100644 libavcodec/aarch64/fmtconvert_neon.S

Ok

// Martin


More information about the libav-devel mailing list