Both the article and cperciva use the algorithmic structure of CRC; but the article's trick still runs front-to-back. You need a "combine chunks" operation if you want to outsource part of the work to another processor.

It exists as well, but isn't mentioned in the article. The article just cares about single process speed-up.

The other "merge CRCs" solution starts out resembling:

  /* Return the CRC-64 of two sequential blocks, where crc1 is the CRC-64 of the
   first block, crc2 is the CRC-64 of the second block, and len2 is the length
   of the second block. */
  uint64_t crc64_combine(uint64_t crc1, uint64_t crc2, uintmax_t len2)