__builtin_memcpy_inline usage introduction.

[devnexen]

for now concealed in few places in zend. builtin producing only a load and store
operands and works with compile time lengths.

[devnexen]

to answer the 'what's for?' question, it is a kind of micro-optimization (unit tests finish ~1 sec faster). Granted it is not available on gcc but clang tough.

[arnaud-lb]

Is the added complexity worth it ? I would expect compilers to already inline memcpy when the cost/benefits is favorable to inlining.

[devnexen]

I would expect compilers to already inline memcpy when the cost/benefits is favorable to inlining.

Not necessarily, not passing any compiler flags but the default ones I get a little bump (tested with clang 14).
The complexity is relative, less impactful than the EXPECT macro nor need to care about arguments changes or anything.