> I would prefer that rte_memcpy be laid to rest and go away. > It never had a reason to be there.
DPDK still contains remains of hand crafted code dating back from a time where compilers (read: the oldest compilers supported by DPDK) didn't produce efficient code. Rte_memcpy() is a prime example of this. There are other examples, e.g. the use of Duff's Device directly in C source code. This code was written a decade ago, and modern compilers are now much better at producing efficient code. Testing would be able to verify the theory that modern compilers (read: the oldest compilers supported by DPDK) produce code that is as efficient as the hand crafted code. If so, the hand crafted code has outlived its purpose, and should be considered obsolete cruft and be completely removed. Until we are certain that removing it doesn't degrade performance, we should keep it where it makes a difference (in the fast path). Special thanks goes to Stephen for his efforts in cleaning up DPDK!