@johnregehr Never mind, I think I'm wrong. I think, indeed, you cannot do the chunking optimization in strict ISO C, in general.
@ch3root @sevenps @johnregehr @BRIAN_____ At -Os it makes a bunch of calls to memcpy1...
-
-
@RichFelker@sevenps@johnregehr@BRIAN_____ A separate function is not essential here, you can manually inline memcpy1.Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
@RichFelker@ch3root@sevenps@johnregehr@BRIAN_____ Can use __attribute__((always_inline)) to work around that with GCC and Clang but... -
@RichFelker@ch3root@sevenps@johnregehr@BRIAN_____ ... that would defeat the whole goal of sticking to ISO C, can do better otherwise. -
@RichFelker@ch3root@sevenps@johnregehr@BRIAN_____ It's silly that -O1 doesn't still doesn't do very conservative cost-based inlining... -
@RichFelker@ch3root@sevenps@johnregehr@BRIAN_____ ... and the same thing applies to -Os. Due to this, -O2 often generates smaller code. -
@RichFelker@ch3root@sevenps@johnregehr@BRIAN_____ For -Os, the problem is just that their inlining heuristic is way too conservative.
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.