Re: [avr-gcc-list] Wrong excution order in 4.1.1, but not 3.4.5, regression?

[Joerg Wunsch]

As Haase Bjoern (PT-BEU/EMT) wrote:

> >http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30908

> IMO, the situation is not so clear. Frequently, inlining functions
> is *more* code efficient because the calling function possibly does
> not need to add code for saving registers. Possibly this decision
> determines if the caller needs a frame pointer or not.  Whether or
> not to inline code is a difficult decision and I am convinced that
> it is generally useful that even -Os inlines (tiny!) functions.

Apparently, it needs to be tuned for the AVR target though, as the
costs for calling a function are over-estimated.

See also the discussion that started in the bug tracker.

-- 
cheers, J"org               .-.-.   --... ...--   -.. .  DL8DTL

http://www.sax.de/~joerg/                        NIC: JW11-RIPE
Never trust an operating system you don't have sources for. ;-)