svn commit: r45753 - head/en_US.ISO8859-1/htdocs/news/status

Thu Oct 9 01:44:21 UTC 2014

Author: wblock
Date: Thu Oct  9 01:44:20 2014
New Revision: 45753
URL: https://svnweb.freebsd.org/changeset/doc/45753

Log:
  Add libcsse report.

Modified:
  head/en_US.ISO8859-1/htdocs/news/status/report-2014-07-2014-09.xml

Modified: head/en_US.ISO8859-1/htdocs/news/status/report-2014-07-2014-09.xml
==============================================================================

--- head/en_US.ISO8859-1/htdocs/news/status/report-2014-07-2014-09.xml	Wed Oct  8 22:37:42 2014	(r45752)
+++ head/en_US.ISO8859-1/htdocs/news/status/report-2014-07-2014-09.xml	Thu Oct  9 01:44:20 2014	(r45753)
@@ -1019,4 +1019,47 @@
 	commit bits were taken in for safe keeping.</p>
     </body>
   </project>
+
+  <project cat='bin'>
+    <title>SSE variants of libc routines for amd64</title>
+
+    <contact>
+      <person>
+	<name>
+	  <given>John</given>
+	  <common>Baldwin</common>
+	</name>
+	<email>jhb at FreeBSD.org</email>
+      </person>
+    </contact>
+
+    <links>
+      <url href="http://trac.baldwin.cx:8080/freebsd/wiki/LibCSSE" />
+    </links>
+
+    <body>
+      <p>I have written SSE/AVX-optimized versions of a few libc
+	routines for amd64.  So far the list includes <tt>memcpy</tt>,
+	<tt>memset</tt>, and <tt>strlen</tt>.  For each routine I have
+	written a simple regression test as well as performed some
+	simple microbenchmarks on various AMD and Intel CPUs.</p>
+
+      <p>The simplest routine is <tt>strlen</tt> which appears to be a
+	general win in microbenchmarks.  <tt>memcpy</tt> and
+	<tt>memset</tt> have proven trickier as different variants can
+	behave quite differently on different CPUs.</p>
+
+      <p>At present, I do not yet have a patch relative to libc.  Once
+	I do, this will be suitable for more testing.  I'd like to see
+	some real-world benchmarks that show measurable improvement
+	before pushing any of this up into the tree.</p>
+    </body>
+
+    <help>
+      <task>
+	<p>Create a branch that holds a modified libc and is suitable
+	  for testing</p>
+      </task>
+    </help>
+  </project>
 </report>
