likely/unlikely for if ()

[shawnl]

if (likely a < 0)
    foo;

If (unlikely b < 0)
    bar;

like in linux https://github.com/torvalds/linux/blob/master/include/linux/compiler.h#L39

[PavelVozenilek]

I'd proposed this in #489, with if+ and if-. More wordy or complicated syntax wouldn't be worth the trouble.

[shawnl]

if+ and if- is nice, especially as it corresponds to "closing" and "expanding" symbols. Closing

[BarabasGitHub]

I think I actually like the wordy ones better because it is in no way clear what if+ is supposed to do imo. Op di 7 aug. 2018 01:05 schreef Shawn Landden <[email protected]>:

…

if+ and if- is nice, especially as it corresponds to "closing" and "expanding" symbols. Closing — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#1342 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AL-sGicMvV22ad4qDV2jBsu85JyPL9m4ks5uOMu1gaJpZM4VxN--> .

[shawnl]

However "likely" and "unlikely" are in the wrong place here. They need to be the same as adding "comptime" to an expression

[PavelVozenilek]

@BarabasGitHub: if likely/unlikely functionality is expected to have measurable effect on performance, it should be used frequently. if+/if- doesn't clutter visually the code, unlike the alternatives. If they are used frequently, people will internalize their meaning quickly.

[shawnl]

yeah but likely/unlikely would work the same as comptime, which is nice