| |
| |
| |
|
Page: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
Comments:
<0> dude
<1> havent really looked too closely at it
<0> and that function would be totally radical if it were inline
<2> blah, the compiler will inline it if it feels like it
<3> no, i don't think so
<2> it's out of our hands, Jeremy
<3> i don't think anything is inlined w/o 'inline'
<3> the opposite is true, though
<3> even 'inline' functions might easily be not inlines
<0> yes
<2> actually int increment(int x) { return x+1; } would be inlined
<2> etc
<0> it may not inline it
<2> bet?
<3> after i check
<2> of course
<3> nope, it doesn't
<4> does the compiler tag functions that only do simple arithmetic with predefined values as worthy of inlining?
<4> doesn't even need to be predefined values, actually
<3> just as i said abover
<4> that makes it much simpler though
<4> what are the criteria for inlining automatically i wonder
<4> with say gcc
<3> there s no automatic inlining. period.
<3> (without explicit 'inline' keyword.)
<3> maybe in other languages
<3> not in C
<0> in C++ there is
<0> it can
<0> can be annoying
<0> in debugging
<0> :P
<2> eMishGLX, was inlined with /O1 (optimize for size)
<0> but the "let the compiler decide for you" comment is overly used
<3> obviously the answer is compiler-dependent
<2> this majestic code: http://rafb.net/paste/results/ETke6U24.html
<0> yes
<2> msvc ignores "inline" per se, also "__inline"
<2> yesterday i got it to finally inline with "__forceinline"
<5> optimize for size is slooow
<3> yes, with -Os gcc does auto-inline
<2> not necessarily, smaller size == better chance to fit in cache
<3> but not with -O
<3> and not with -O2
<2> zid`, I read some folklore that Microsoft compiles everything for size
<3> which is what i normally use
<3> it's weird that -O2 doesn't auto-inline
<3> i just knew it doesn't
<3> i never use -Os
<3> but what chance is for auto-inlining if function body is not in .h ?
<3> and if you put func body into .h, then you use 'static inline'
<2> eMishGLX, funnily, if the function body is in the cpp, and *even* in an already compiled .obj, link time code generation may still pull it out and inline it, if it would suit the purpose
<2> (MS)
<2> so the linker is a pretty solid compiler :)
<6> can scanf() get strings with spaces?
<3> Symm: MSVC can do this ?
<3> you sure ?
<3> honestly, i don't believe
<3> this is totally not linker job
<3> and it's never dumb*** cut-and-paste of the code. it requires intelligent code rewriting wrt registers use
<2> eMishGLX, yeah. google ltcg or link time code generation
<2> it's a VC 7.1 and up feature
<2> indeed, requires a lot of smarts
<2> BUT
<2> the .obj code is not pure binary
<2> it's some intermediate language
<2> so it's easier on the linker to finally compile it
<4> haha
<3> i wonder whether gcc whether picks this up
<4> why the *hell* is the linker doing compilatio
<4> n
<4> that's crazy
<3> *ever
<3> "The /LTCG option tells the linker to call the compiler and perform whole program optimization."
<3> so it's not really linker
Return to
#c
or
Go to some related
logs:
#bsd
#fedora
#nhl
#gentoo
#beginner
VIADSE
japanese toilettes
the whole livery line bow down like this with the big money
#beginner
gentoo cron 3am
|
|