Implementing CRC16 efficiently

Question

I got a program that I know will run only on very specific hardware, so I started using crc32 as supplied by the SSE4.2 instruction set, but now I need crc16 for a different purpose. Is there a way to coax the hardware to give me crc16 with the same performance as SSE's crc32?

Thanks

Answer 1

If you have the PCLMULQDQ instruction, you can use that to make a fast CRC. Though not quite as fast as the crc32 instruction.

Alternatively, a poor man's check value could be the low 16 bits of the result of crc32.