On Sat, Nov 28, 2015 at 12:03 PM, Ganesh Ajjanagadde
<gajjanaga...@gmail.com> wrote:
> On systems having cbrt, there is no reason to use the slow pow function.
>
> Sample benchmark (x86-64, Haswell, GNU/Linux):
> new:
> 5124920 decicycles in cbrt_tableinit, 1 runs, 0 skips
>
> old:
> 12321680 decicycles in cbrt_tableinit, 1 runs, 0 skips
>
> Reviewed-by: Ronald S. Bultje <rsbul...@gmail.com>
> Signed-off-by: Ganesh Ajjanagadde <gajjanaga...@gmail.com>
> ---
> libavcodec/cbrt_tablegen.h | 5 ++---
> libavcodec/cbrt_tablegen_template.c | 1 +
> 2 files changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/libavcodec/cbrt_tablegen.h b/libavcodec/cbrt_tablegen.h
> index 27a3e3a..59b5a1d 100644
> --- a/libavcodec/cbrt_tablegen.h
> +++ b/libavcodec/cbrt_tablegen.h
> @@ -29,7 +29,7 @@
> #include "libavcodec/aac_defines.h"
>
> #if USE_FIXED
> -#define CBRT(x) (int)floor((x).f * 8192 + 0.5)
> +#define CBRT(x) lrint((x).f * 8192)
> #else
> #define CBRT(x) x.i
> #endif
> @@ -49,13 +49,12 @@ static av_cold void AAC_RENAME(cbrt_tableinit)(void)
> {
> if (!cbrt_tab[(1<<13) - 1]) {
> int i;
> - /* cbrtf() isn't available on all systems, so we use powf(). */
> for (i = 0; i < 1<<13; i++) {
> union {
> float f;
> uint32_t i;
> } f;
> - f.f = pow(i, 1.0 / 3.0) * i;
> + f.f = cbrt(i) * i;
> cbrt_tab[i] = CBRT(f);
> }
> }
> diff --git a/libavcodec/cbrt_tablegen_template.c
> b/libavcodec/cbrt_tablegen_template.c
> index 9dd2cf5..1d71d34 100644
> --- a/libavcodec/cbrt_tablegen_template.c
> +++ b/libavcodec/cbrt_tablegen_template.c
> @@ -23,6 +23,7 @@
> #include <stdlib.h>
> #define CONFIG_HARDCODED_TABLES 0
> #include "cbrt_tablegen.h"
> +#include "libavutil/tablegen.h"
> #include "tableprint.h"
>
> int main(void)
> --
> 2.6.2
>
pushed, identical stand-alone patch reviewed by Ronald.
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
