Path: utzoo!attcan!uunet!cs.utexas.edu!usc!apple!amdahl!dgcad!gary
From: gary@dgcad.SV.DG.COM (Gary Bridgewater)
Newsgroups: comp.lang.c++
Subject: Re: Cfront inlining
Message-ID: <1505@proa.SV.DG.COM>
Date: 7 Apr 90 09:41:22 GMT
References: <117671@ti-csl.csc.ti.com>
Reply-To: gary@proa.SV.DG.COM ()
Organization: Data General SDD, Sunnyvale, CA
Lines: 47

In article <117671@ti-csl.csc.ti.com> NEATH@cgi.com writes:
>
>In article  <117072@ti-csl.csc.ti.com>  NEATH@cgi.com  (me!)  wrote about   the
>inlining policy of cfront 2.0 and the complextiy variable `imeasure'. When this
>reaches 13, cfront decides not to inline
>As such, that dosn't seem unreasonable: if you have 13 assignment statements in
>a function, you probably shouldn't be inlining it all over the place.
                            ^ why?

>There are, however, other circumstances  where the apparently somewhat arbitray
>complexity measure of 13 is an annoyance.  

like, whenever it is invoked...

>... To remedy this problem, it  appears  we must select from
>one of the following unattractive options:
>...
>   o Modify the cfront source code to allow a larger complexity measure so
>     that this particular case actually stays inline.

There you go.

>It would seem that the inline recommendation to the compiler is  a far trickier
>subject to deal with than some static complexity measure. 

Here's a bizarre thought - just let the programmer decide.  If, for example, the
programmer inserts an "inline" keyword then inline. If not, then don't.*  There
is always the outside chance the programmer did/did not use the "inline" keyword
for a reason. Call it a 50-50 chance, if you will, so, on average, it can't hurt.

After all, who suffers from the wrong decision?  If the general c++ community
doesn't understand inlining then let them post questions and get answers.  It
isn't the compiler that breaks the program.

>Why, for example, are  `for' loops disallowed in  inline functions, but `while'
>loops tolerated? 

Why is anything disallowed?  If it is currently beyond the coders to inline
certain things then document that as a deficiency, not a feature.

...
* Actually, the complier (C++ or the underlying cc) may be able to decide TO
inline in some cases - I don't want to restrict that feature other than wrt
debugging when the rule should be "what you code is what you get".
-- 
Gary Bridgewater, Data General Corporation, Sunnyvale California
gary@sv.dg.com or {amdahl,aeras,amdcad}!dgcad!gary
The impossible we understand right away - the obvious takes a little longer.