This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [wwwdocs] [committed] updated the vectorizer webpage
- From: Gerald Pfeifer <gerald at pfeifer dot com>
- To: Dorit Nuzman <DORIT at il dot ibm dot com>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Sun, 7 Oct 2007 15:34:55 +0200 (CEST)
- Subject: Re: [wwwdocs] [committed] updated the vectorizer webpage
- References: <OFCFE6127E.EA3AA683-ONC2257359.002F3B65-C2257359.002F5AB2@il.ibm.com>
On Mon, 17 Sep 2007, Dorit Nuzman wrote:
> committed the following update to the vectorizer webpage.
While reading this, I made some minor formatting and markup tweaks.
Also, I removed <meta http-equiv="Content-Type"...> which would create
duplicate entries on our production site. If you are using a validator
such as validator.w3.org, you can still specify this manually. (And in
general we may want to review how to handle preprocessing of our pages
to make this simpler; cf. the discussion FX started.)
Gerald
Index: vectorization.html
===================================================================
RCS file: /cvs/gcc/wwwdocs/htdocs/projects/tree-ssa/vectorization.html,v
retrieving revision 1.22
diff -u -3 -p -r1.22 vectorization.html
--- vectorization.html 18 Sep 2007 11:37:05 -0000 1.22
+++ vectorization.html 7 Oct 2007 13:30:52 -0000
@@ -1,7 +1,5 @@
<html xmlns="http://www.w3.org/1999/xhtml">
<head>
- <meta http-equiv="Content-Type" content="text/html" />
-
<title>Auto-vectorization in GCC</title>
</head>
@@ -37,20 +35,23 @@
<h2><a name="news">Latest News</a></h2>
<dl>
<dt>2007-09-17</dt>
- <dd>
- <ol>
- <li>-ftree-vectorize is going to be turned on under -O3. </li>
- <li>Cost-model tweaks and x86_64 specific costs committed to mainline (2007-09-10). </li>
- <li>Vectorization that exploits intra-iteration parallelism (ala SLP)
- was committed to mainline (2007-09-09). </li>
- <li>-freassociative-math can be used instead of -ffast-math to enable
- vectorization of reductions of floats (2007-09-04). </li>
- <li>Initial support for vectorization of outer-loops (doubly nested loops)
- was committed to mainline (2007-08-19).</li>
- <li>Run-time dependence testing using loop-versioning was committed
- to mainline (2007-08-16). </li>
- </ol>
- </dd>
+ <dd>
+ <ol>
+ <li><code>-ftree-vectorize</code> is going to be turned on under
+ <code>-O3</code>.</li>
+ <li>Cost-model tweaks and x86_64 specific costs committed to mainline
+ (2007-09-10).</li>
+ <li>Vectorization that exploits intra-iteration parallelism (ala SLP)
+ was committed to mainline (2007-09-09).</li>
+ <li><code>-freassociative-math</code> can be used instead of
+ <code>-ffast-math</code> to enable vectorization of reductions of
+ floats (2007-09-04).</li>
+ <li>Initial support for vectorization of outer-loops (doubly nested
+ loops) was committed to mainline (2007-08-19).</li>
+ <li>Run-time dependence testing using loop-versioning was committed
+ to mainline (2007-08-16).</li>
+ </ol>
+ </dd>
</dl>
<dl>
@@ -66,23 +67,24 @@
<dd>Mainline updates:
<ol>
<li>SPU specific costs for the cost model committed to mainline (2007-07-12).
- Tuning for other platforms (PPC, X86) ongoing. </li>
- <li>Initial cost model implementation committed to mainline (2007-06-08). </li>
+ Tuning for other platforms (PPC, x86) ongoing.</li>
+ <li>Initial cost model implementation committed to mainline
+ (2007-06-08).</li>
<li>Vectorization of fp/integer conversions of different sizes (e.g. float/short)
committed to mainline (2007-05-17).</li>
- <li>Data-refs analysis was rewritten and improved (2007-05-13). </li>
+ <li>Data-refs analysis was rewritten and improved (2007-05-13).</li>
</ol>
</dd>
<dd>Autovect-branch updates:
<ol>
<li>Outer-loop vectorization was enhanced to support aligned and unaligned
memory references in the inner-loop, using the optimized realignment
- scheme when possible. </li>
+ scheme when possible.</li>
<li>Vectorization that exploits intra-iteration parallelism (ala SLP)
was added to the vectorizer (that so far exploited only inter-iteration
- parallelism). </li>
+ parallelism).</li>
<li>The vectorizer cost model was extended to support the above two new vectorization
- features (outer-loop and "SLP"). </li>
+ features (outer-loop and "SLP").</li>
</ol>
</dd>
</dl>
@@ -93,7 +95,8 @@
<ol>
<li>Vectorization of fp/integer conversions of different sizes (e.g. float/short)
is soon to be committed to mainline.</li>
- <li>Initial cost model implementation is soon to be committed to mainline. </li>
+ <li>Initial cost model implementation is soon to be committed to
+ mainline.</li>
</ol>
</dd>
</dl>