comparison dv.c @ 1875:45a1592dadca libavcodec

* moving some of the commonly used bit reading/writing functions from common.c -> common.h so that they can be inlined. + performace gain ~1% (measured with DV decoding) + code bloat 0.05% Looks like a win-win solution.

1874:fab737b59790

    DCTELEM *mb;
    uint8_t partial_bit_count;
    uint32_t partial_bit_buffer; /* we can't use uint16_t here */
} EncBlockInfo;

static always_inline int dv_bits_left(PutBitContext* s)
{
    return (s->buf_end - s->buf) * 8 - 
           ((s->buf_ptr - s->buf) * 8 + 32 - (int64_t)s->bit_left);
}

static always_inline void dv_encode_ac(EncBlockInfo* bi, PutBitContext* pb_pool, 
                                       int pb_size)
{
    int run;
    int bits_left;

    uint32_t vlc = bi->partial_bit_buffer;
    
    bi->partial_bit_count = bi->partial_bit_buffer = 0;
vlc_loop:
       /* Find suitable storage space */
       for (; size > (bits_left = dv_bits_left(pb)); pb++) {
          if (bits_left) {
              size -= bits_left;
	      put_bits(pb, bits_left, vlc >> size);
	      vlc = vlc & ((1<<size)-1);
	  }

1875:45a1592dadca

    DCTELEM *mb;
    uint8_t partial_bit_count;
    uint32_t partial_bit_buffer; /* we can't use uint16_t here */
} EncBlockInfo;

static always_inline void dv_encode_ac(EncBlockInfo* bi, PutBitContext* pb_pool, 
                                       int pb_size)
{
    int run;
    int bits_left;

    uint32_t vlc = bi->partial_bit_buffer;
    
    bi->partial_bit_count = bi->partial_bit_buffer = 0;
vlc_loop:
       /* Find suitable storage space */
       for (; size > (bits_left = put_bits_left(pb)); pb++) {
          if (bits_left) {
              size -= bits_left;
	      put_bits(pb, bits_left, vlc >> size);
	      vlc = vlc & ((1<<size)-1);
	  }