News for hackers. Stuff that matters.

A comment on Dangerous Embedded C Coding Standard Rules (2011). Parent comment

Re:Dangerous Embedded C Coding Standard Rules (2011) by gioele on Monday December 30, 2013 @03:55PM (#6986889)
> #1: Avoiding shifting versus doing division: Nope. Crappy compilers and bad architectures will yield bad results. Know what you're doing.

Which compiler, among those still used in 2014, does not convert a division by 2 into a shift?

[ Parent | on HN ]
Re:Dangerous Embedded C Coding Standard Rules (2011) by gilgoomesh on Monday December 30, 2013 @06:05PM (#6987699)
It's not just division by 2. Most compilers will convert any division by a constant into either a shift (powers of 2) or a multiply (all other).

But there are lots of oddball compilers for CPUs, DSPs, and other embedded quasi processors you've never heard of. Many don't really optimise at all.

[ Parent | on HN ]
Re:Dangerous Embedded C Coding Standard Rules (2011) by angersock on Monday December 30, 2013 @04:06PM (#6986952)
Memory escapes me...is a shift guaranteed by the standard (and if so, which) to sign-extend or not sign-extend?
[ Parent | on HN ]
Re:Dangerous Embedded C Coding Standard Rules (2011) by kps on Monday December 30, 2013 @04:33PM (#6987115)
No, it isn't.

A right shift of a signed, negative value is implementation-defined (which means, something in particular happens, and the compiler or its documentation is supposed to tell you what).

A left shift of a signed value is undefined (which means, the compiler is allowed to launch the missiles, or fail to launch the missiles, whichever is worse for you) if a nonzero bit could conceivable get into the sign.

[ Parent | on HN ]
Re:Dangerous Embedded C Coding Standard Rules (2011) by kristoffer on Monday December 30, 2013 @04:31PM (#6987107)
A right shift of a signed integer is implementation defined.
[ Parent | on HN ]