This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: malloc attributes and realloc


 > From: Robert Dewar <dewar@gnat.com>
 > 
 > Kaveh R. Ghazi wrote:
 > 
 > > Note many (most?) allocation in GCC occurs in other functions besides
 > > the few libiberty functions with the attribute.  So I think the
 > > optimization is pretty good to get as much as it does.
 > 
 > Pretty amazing I would say. If those figures are correct, then it is
 > indeed surprising that this helps so much, and perhaps means that this
 > thread is not 100% idle chatter :-)

I would encourage someone to verify my results.  I bootstrapped
mainline with --disable-checking --enable-languages=c --disable-nls
--enable-multilib=no on sparc-sun-solaris2.7 both with and without the
patch below.  Then I compiled combine.i with -S -g -O2 five times with
each compiler to get my figures.  I'd be interested to hear results
from other platforms.

		--Kaveh


diff -rup orig/egcc-CVS20040102/include/ansidecl.h egcc-CVS20040102/include/ansidecl.h
--- orig/egcc-CVS20040102/include/ansidecl.h	Tue Jun 17 10:05:22 2003
+++ egcc-CVS20040102/include/ansidecl.h	Fri Jan  2 23:13:12 2004
@@ -244,7 +244,7 @@ So instead we use the macro below and te
 
 /* Attribute __malloc__ on functions was valid as of gcc 2.96. */
 #ifndef ATTRIBUTE_MALLOC
-# if (GCC_VERSION >= 2096)
+# if 0
 #  define ATTRIBUTE_MALLOC __attribute__ ((__malloc__))
 # else
 #  define ATTRIBUTE_MALLOC


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]