Why is a byte card faster than a bit card?

Question

I read the Garbage Collection Guide. He says that when you make a card table, they use a bytemap instead of a bitmap and say it is faster than a bitmap due to the high caching speed? But, as I know, the cache line is usually 64 bytes, if we make changes in the byte, the racing dispute still exists, another processor will invalidate the line anyway, it's the same as a bit card, can anyone to help?

Answer 1

Not sure if I understood the context correctly, but overall:

• bitmap access

  BYTE/WORD/..., / .

  , 8- , :

  BYTE map[];
  

  :

  readed_bit=(map[bit>>3]>>(bit&7))&1;
  

  :

  map[bit>>3]|=1<<(bit&7);
  

  :

  map[bit>>3]&=255^(1<<(bit&7));
  

  bit - , . , .

• BYTE

  readed_byte=map[byte];
  

  :

  map[byte]=1;
  

  :

  map[byte]=0;
  

  byte - BYTE, . , , .

, HW, , BYTE ... , , , ...