Gcc builtin review: strcpy, stpcpy, strcat, stpcat?

[OndÅej BÃlka <neleai at seznam dot cz>]

I will omit expansion to memcpy and flaws from previous thread.

First problem is that gcc in

int foo (char *c, char *c2){
  stpcpy (c, c2);
}

Replaces it with strcpy. One could argue that opposite way to replace
strcpy with stpcpy is faster.

Reason is register pressure. Strcpy needs extra register to save return
value while stpcpy has return value already in register used for writing
terminating zero. 

Also to decrease instruction cache pressure you should always call
stpcpy and newer strcpy. You can't do it other way as you would need
expensive strlen. A pressure is problem as strcpy tend to be relatively rarely
used.

A possible advantage of strcpy is that gcc4.9 started to exploit that it
returns x saving register spill in caller. I don't know how is that
useful.

With stpcpy and strcat there are more optimizations with higher payoff.

You should expoit that stpcpy returns end of string so gcc should
replace

strcpy (x,y)
return strlen (x)

with 

return stpcpy (x,y) - x

Then there is famous problem of quadratic performance of strcat loop.
That is something that gcc could eliminate with tracking end of string.

Also there was discussion on libc-alpha about adding stpcat. That would
also save some end of string calculations.

Otherwise strcat look ok modulo memcpy related problems.