Simon Brenner@f4grx @whitequark I would say the point is convenient arch-independent access to crc hardware - you should get the special instruction when the polynomial matches whatever your cpu has and hopefully a clmul based fallback before resorting to a table. Much nicer than a tangle of ifdefs for every architecture and the research needed to know every crc instruction and which crc variant it produces…
I guess if you were properly enthusiastic about crcs you’d already be maintaining specialized assembly for 14 architectures and every commonly used polynomial and this is not so interesting… but now the rest of us can get gcc (or gcc’s architecture maintainers) to do it for us?
All GNU social JP content and data are available under the