chenyu202...@gmail.com:
> From: chenyu <chenyu202...@gmail.com>
>
> Optimizing 160k code size by converting static array to dynamic malloc memory.
>
> Signed-off-by: chenyu <chenyu202...@gmail.com>
> ---
> libavcodec/mpegaudiodata.h | 4 ++--
> libavcodec/mpegaudiodec_common_tablegen.h | 10 ++++++++--
> 2 files changed, 10 insertions(+), 4 deletions(-)
>
> diff --git a/libavcodec/mpegaudiodata.h b/libavcodec/mpegaudiodata.h
> index 720c4bee64..6dfb74cd01 100644
> --- a/libavcodec/mpegaudiodata.h
> +++ b/libavcodec/mpegaudiodata.h
> @@ -50,8 +50,8 @@ extern const unsigned char * const ff_mpa_alloc_tables[5];
> extern const int8_t ff_table_4_3_exp [TABLE_4_3_SIZE];
> extern const uint32_t ff_table_4_3_value[TABLE_4_3_SIZE];
> #else
> -extern int8_t ff_table_4_3_exp [TABLE_4_3_SIZE];
> -extern uint32_t ff_table_4_3_value[TABLE_4_3_SIZE];
> +extern int8_t *ff_table_4_3_exp;
> +extern uint32_t *ff_table_4_3_value;
> #endif
>
> /* VLCs for decoding layer 3 huffman tables */
> diff --git a/libavcodec/mpegaudiodec_common_tablegen.h
> b/libavcodec/mpegaudiodec_common_tablegen.h
> index bf402c9d84..66e93df27f 100644
> --- a/libavcodec/mpegaudiodec_common_tablegen.h
> +++ b/libavcodec/mpegaudiodec_common_tablegen.h
> @@ -34,9 +34,10 @@
> #else
> #include <math.h>
> #include "libavutil/attributes.h"
> +#include "libavutil/mem.h"
>
> -int8_t ff_table_4_3_exp [TABLE_4_3_SIZE];
> -uint32_t ff_table_4_3_value[TABLE_4_3_SIZE];
> +int8_t *ff_table_4_3_exp;
> +uint32_t *ff_table_4_3_value;
>
> #define FRAC_BITS 23
> #define IMDCT_SCALAR 1.759
> @@ -51,6 +52,11 @@ static av_cold void mpegaudiodec_common_tableinit(void)
> };
> double pow43_val = 0;
>
> +#if !CONFIG_HARDCODED_TABLES
> + ff_table_4_3_exp = (int8_t *)av_calloc(TABLE_4_3_SIZE, sizeof(int8_t));
> + ff_table_4_3_value = (uint32_t *)av_calloc(TABLE_4_3_SIZE,
> sizeof(uint32_t));
> +#endif
> +
> for (int i = 1; i < TABLE_4_3_SIZE; i++) {
> double f, fm;
> int e, m;
This does not improve "code size" at all; after all you are actually
adding code (and an indirection) with your allocation. If one interprets
"code size" as "size of the binary" (instead of just .text), then this
patch still doesn't make sense, because the tables are currently in .bss
which does not increase the size of the binaries (that's at least the
case for common systems, is it different for you?).
And of course you are missing the error check (in fact, you have no way
to return an error from init functions like this).
- Andreas
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
