WWWdocs updates
Gerald Pfeifer
gerald@pfeifer.com
Mon Jan 15 01:05:00 GMT 2007
On Tue, 9 Jan 2007, Jan Hubicka wrote:
> Thanks for review. I've comitted the changes now (sorry for delay - I
> had my hands full by IPA merge and cgraph cleanups last weeks).
> Changes to 4.3 news was non-trivial, so I am attaching what I ended up.
Thanks! I'd like to propose some minor changes on top of these which
I include below. Okay?
Gerald
Index: changes.html
===================================================================
RCS file: /cvs/gcc/wwwdocs/htdocs/gcc-4.3/changes.html,v
retrieving revision 1.19
diff -u -3 -p -r1.19 changes.html
--- changes.html 10 Jan 2007 12:17:46 -0000 1.19
+++ changes.html 15 Jan 2007 01:03:27 -0000
@@ -68,15 +68,15 @@
never reaches the object file.
</li>
- <li>Inliner heuristic is now aware of stack frame consumption. New
+ <li>The inliner heuristic is now aware of stack frame consumption. New
command line parameters <code>--param large-stack-frame</code>
and <code>--param large-stack-frame-growth</code>
- can now be used to limit stack frame size growth caused by inlining.</li>
+ can be used to limit stack frame size growth caused by inlining.</li>
- <li>During feedback directed optimizations, the expected size of block
- <code>memcpy</code>, <code>memset</code> and <code>bzero</code> block
+ <li>During feedback directed optimizations, the expected block size
+ the <code>memcpy</code>, <code>memset</code> and <code>bzero</code>
functions operate on is discovered and for cases of commonly used small
- block size, specialized inline code is generated.</li>
+ sizes, specialized inline code is generated.</li>
<li><code>__builtin_expect</code> no longer requires its argument to be
a compile time constant.</li>
@@ -180,18 +180,19 @@
<li>Tuning for AMD Geode processors is available via
<code>-mtune=geode</code> and <code>-march=geode</code>.</li>
<li>Code generation of block move (<code>memcpy</code>) and block set
- (<code>memset</code>) was rewritten. GCC can now pick best algorithm
- (loop, unrolled loop, instruction with <code>rep</code> prefix or
- a library call) based on the size of block being copied and CPU being
- optimized for. A new option <code>-minline-stringops-dynamically</code>
- has been added. With this option string operations of unknown size are
- expanded in such that small blocks are copied by in-line code, while for
- large block a library call is used. This results in faster code than
+ (<code>memset</code>) was rewritten. GCC can now pick the best
+ algorithm (loop, unrolled loop, instruction with <code>rep</code>
+ prefix or a library call) based on the size of the block being
+ copied and the CPU being optimized for. A new option
+ <code>-minline-stringops-dynamically</code> has been added. With
+ this option string operations of unknown size are expanded such
+ that small blocks are copied by in-line code, while for
+ large blocks a library call is used. This results in faster code than
<code>-minline-all-stringops</code> when the library implementation is
capable of using cache hierarchy hints. The heuristic choosing
the particular algorithm can be overwritten via
<code>-mstringop-strategy</code>. Newly also <code>memset</code> of
- other value than 0 is inlined.</li>
+ values different from 0 is inlined.</li>
<li>GCC no longer places the <code>cld</code> instruction before string
operations. Both i386 and x86-64 ABI documents mandate the direction
flag to be clear at the entry of a function. It is now invalid to
More information about the Gcc-patches
mailing list