If i am not mistaken the /Oi flag signals the compiler to replace functions with intrinsic versions. Also at some point Casey talked about rewriting sinf himself instead of using the c runtime library's version of it.
My question is, if we were to do that and implement sinf ourselves, do we give up on /Oi optimization or is there a way to give a hint to compiler that the function we wrote can be replaced with intrinsic sin?
My question is, if we were to do that and implement sinf ourselves, do we give up on /Oi optimization or is there a way to give a hint to compiler that the function we wrote can be replaced with intrinsic sin?
Edited by Furkan
on
According to MSDN: http://msdn.microsoft.com/en-us/library/tzkfha43.aspx , you can additionally "use /fp:strict or /Za flags to to override generation of true intrinsic floating-point options".
I guess i've found my answer. Intrinsics are built-in to compilers. This means which functions are intrinsics is predefined so there is no language construct to hint the compiler that the function we wrote is an intrinsic.
For example; memcpy is an intrinsic funtion. If i were to write my version of it, say frkn_memcpy, i can't tell the compiler to make the same /Oi optimization it makes for memcpy. And if i name it memcpy instead of frkn_memcpy, it will cause a compiler error saying "intrinsic functions can't be defined" if i try to compile with intrinsic optimizations enabled.
Helpful reading
http://msdn.microsoft.com/en-us/library/26td21ds.aspx
For example; memcpy is an intrinsic funtion. If i were to write my version of it, say frkn_memcpy, i can't tell the compiler to make the same /Oi optimization it makes for memcpy. And if i name it memcpy instead of frkn_memcpy, it will cause a compiler error saying "intrinsic functions can't be defined" if i try to compile with intrinsic optimizations enabled.
Helpful reading
http://msdn.microsoft.com/en-us/library/26td21ds.aspx