[wwwdocs] IPA branch merge

Jan Hubicka jh@suse.cz
Fri Jan 26 12:28:00 GMT 2007


Hi,
this should update webpages for IPA branch merge.
Honza
Index: index.html
===================================================================
RCS file: /cvs/gcc/wwwdocs/htdocs/index.html,v
retrieving revision 1.593
diff -c -p -r1.593 index.html
*** index.html	15 Jan 2007 20:42:04 -0000	1.593
--- index.html	26 Jan 2007 12:26:16 -0000
*************** mission statement</a>.</p>
*** 45,50 ****
--- 45,59 ----
  
  <dl class="news">
  
+ <dt>January 25, 2007</dt>
+ <dd>Interprocedural optimization passes have been reorganized to operate on SSA
+     This enables more precise function analysis and also optimization while
+     inlining significantly improving performance of programs with high
+     abstraction penalty.
+     Code from <A HREF="http://gcc.gnu.org/wiki/ipa-branch">ipa-branch</A>
+     contributed by Jan Hubicka, SUSElabs and Razya Ladelsky, IBM Haifa,
+     was reviewed by Diego Novillo, Richard Guenther, Roger Sayle and Ian
+     Lance Taylor.</dd>
  <dt>January 8, 2007</dt>
  <dd>Andrew Haley and Tom Tromey of Red Hat merged the
      <code>gcj-eclipse</code> branch to svn trunk.  GCC now uses the
Index: gcc-4.3/changes.html
===================================================================
RCS file: /cvs/gcc/wwwdocs/htdocs/gcc-4.3/changes.html,v
retrieving revision 1.24
diff -c -p -r1.24 changes.html
*** gcc-4.3/changes.html	20 Jan 2007 14:29:20 -0000	1.24
--- gcc-4.3/changes.html	26 Jan 2007 12:26:16 -0000
***************
*** 82,87 ****
--- 82,110 ----
      <li><code>__builtin_expect</code> no longer requires its argument to be
      a compile time constant.</li>
  
+     <li>Inteprocedural optimizers was reorganized to work on functions in SSA
+ 	form.  This enable more precise and cheaper dataflow analysis and makes
+ 	writing interprocedural optimizations easier.  Following improvements
+ 	has been implemented on the top of this framework:
+ 	<ul>
+ 	 <li>Pre inline optimization: few local optimization passes are run
+ 	     before inliner (and other interprocedural passes) are executed.
+ 	     This significantly improves accuracy of code growth estimates used
+ 	     by inliner and reduce overall memory footprint for large
+ 	     compilation units.</li>
+ 	 <li>Early inlining (a simple bottom-up inliner pass inlining only
+ 	     functions whose body is smaller than expected call overhead) is
+ 	     now executed in
+ 	     parallel with early optimization inlining optimized function bodies
+ 	     into unoptimized function that is subsequently optimized by early
+ 	     optimizers.  This enables compiler to quickly eliminate abstraction
+ 	     overhead in C++ programs.</li>
+ 	 <li>Iterprocedural constant propagation now operate on SSA form
+ 	     increasing accuracy of the analysis.</li>
+ 	</ul>
+ 
+ 
+ 
    </ul>
  
    <ul>



More information about the Gcc-patches mailing list