|Register pressure and inlining linuxkaffee_@_gmx.net (Stephan Ceram) (2008-11-08)|
|Re: Register pressure and inlining email@example.com (Jeremy Wright) (2008-11-09)|
|RE: Register pressure and inlining firstname.lastname@example.org (Murugesh) (2008-11-10)|
|Re: Register pressure and inlining email@example.com (Alex L.K) (2008-11-12)|
|Re: Register pressure and inlining firstname.lastname@example.org (Armel) (2008-11-12)|
|Re: Register pressure and inlining email@example.com (Ray Dillinger) (2008-11-12)|
|Re: Register pressure and inlining firstname.lastname@example.org (Christopher Glaeser) (2008-11-12)|
|Re: Register pressure and inlining email@example.com (Bob Morgan) (2008-12-17)|
|From:||Bob Morgan <firstname.lastname@example.org>|
|Date:||Wed, 17 Dec 2008 00:11:26 GMT|
|References:||08-11-036 08-11-045 08-11-047 08-11-051|
|Keywords:||registers, architecture, optimize|
|Posted-Date:||18 Dec 2008 17:24:51 EST|
On Wed, 12 Nov 2008 08:45:33 -0800, "Christopher Glaeser"
>> Is there any example that inlining acctually causes performance loss?
>Expanded code may not fit in small instruction cache.
The language semantics can sometimes lead to poorer performance with
inlining. The classical example is the DAXPY function in Fortran. When
not compiled inline, DAXPY can take advantage of Fortran language
rules concerning non-overlapping of dummy and global arguments. When
compiled inline the overlap may be difficult to identify with
dependence analysis, so performance may be hurt.
Return to the
Search the comp.compilers archives again.