一个计算文件blake2s信息摘要的并行算法实现

blake2sp_file.c

/* A really fast parallel blake2s hash for big file.
 *
 * Usage: Add this into blake2sp-ref.c
 */
#include <sys/stat.h>
int blake2sp_file( const char *filename, uint8_t *out, uint8_t outlen )
{
  memset(out, 0, outlen);
  static const size_t buf_len = 8*1024;
  struct stat st;
  if(stat(filename, &st) != 0 || !S_ISREG(st.st_mode)) return -1; // file not exist or not regular file
  if(st.st_size == 0) return 0; // empty file
  const size_t parallel_size = (st.st_size + PARALLELISM_DEGREE - 1) / PARALLELISM_DEGREE;

  uint8_t hash[PARALLELISM_DEGREE][BLAKE2S_OUTBYTES];
  blake2s_state S[PARALLELISM_DEGREE][1];
  blake2s_state FS[1];

  for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
    if( blake2sp_init_leaf( S[i], outlen, 0, i ) < 0 ) return -1;

  S[PARALLELISM_DEGREE - 1]->last_node = 1; // mark last node

#if defined(_OPENMP)
  #pragma omp parallel shared(S,hash), num_threads(PARALLELISM_DEGREE)
#else

  for( size_t id__ = 0; id__ < PARALLELISM_DEGREE; ++id__ )
#endif
  {
#if defined(_OPENMP)
    size_t id__ = omp_get_thread_num();
#endif
    FILE *file = fopen(filename, "rb");
    if(file)
    {
      size_t read_pos = id__ * parallel_size;
      size_t read_len = (id__ == PARALLELISM_DEGREE-1) ? (st.st_size - read_pos) : (parallel_size);
      uint8_t *buf = malloc(buf_len);

      fseek(file, read_pos, SEEK_SET);
      while( read_len >= buf_len )
      {
        fread(buf, buf_len, 1, file);
        blake2s_update( S[id__], buf, buf_len);
        read_len -= buf_len;
      }

      if(read_len > 0)
      {
        fread(buf, read_len, 1, file);
        blake2s_update( S[id__], buf, read_len);
      }

      blake2s_final( S[id__], hash[id__], BLAKE2S_OUTBYTES );
      free(buf);
      fclose(file);
    }
  }

  if( blake2sp_init_root( FS, outlen, 0 ) < 0 ) return -1;
  FS->last_node = 1;
  for( size_t i = 0; i < PARALLELISM_DEGREE; ++i )
    blake2s_update( FS, hash[i], BLAKE2S_OUTBYTES );

  return blake2s_final( FS, out, outlen );
}