Comments on: JDK 7 Twice as Fast* as JDK 6 for Arrays and Arithmetic http://lingpipe-blog.com/2009/03/30/jdk-7-twice-as-fast-as-jdk-6-for-arrays-and-arithmetic/ Natural Language Processing and Text Analytics Wed, 08 Feb 2012 17:47:08 +0000 hourly 1 http://wordpress.com/ By: Java 7: Evolution of Java – Revolution for the JVM « dblackherod http://lingpipe-blog.com/2009/03/30/jdk-7-twice-as-fast-as-jdk-6-for-arrays-and-arithmetic/#comment-15482 Sun, 07 Aug 2011 23:21:52 +0000 http://lingpipe-blog.com/?p=1085#comment-15482 [...] optimizations. We’ve already seen some very encouraging results in String performance, array performance, and a new concurrent garbage collector (G1). I suspect many people will find that their existing [...]

]]> By: Bob Carpenter http://lingpipe-blog.com/2009/03/30/jdk-7-twice-as-fast-as-jdk-6-for-arrays-and-arithmetic/#comment-15278 Sun, 10 Jul 2011 22:29:56 +0000 http://lingpipe-blog.com/?p=1085#comment-15278 Earlier 1.6 releases weren’t in synch with 1.7. So when I wrote this, the performance difference was easily reproducible. And both were faster than 1.5. Again, at the time I wrote this.

Later 1.6 versions incorporated the relevant 1.7 optimizations, which is great for those of us with conservatively version updating customers.

]]> By: Bernd http://lingpipe-blog.com/2009/03/30/jdk-7-twice-as-fast-as-jdk-6-for-arrays-and-arithmetic/#comment-15276 Sun, 10 Jul 2011 11:41:21 +0000 http://lingpipe-blog.com/?p=1085#comment-15276 SUN VM isn’t hybrid on Linux or Windows. That particular mode was only available on solaris x86. On those architectures the 64bit install was just a small upgrade to a larger 32bit package.

]]> By: Bernd http://lingpipe-blog.com/2009/03/30/jdk-7-twice-as-fast-as-jdk-6-for-arrays-and-arithmetic/#comment-15272 Sat, 09 Jul 2011 15:32:10 +0000 http://lingpipe-blog.com/?p=1085#comment-15272 I am quite sure, that the array bounds check supression for cases like for(int i=0;i<a.length;i++) was added in 1.5 and earlier. At least most of the early JIT/C comparisions mention this "trick".

And I really doubt much runtime speed improvements in 1.7 over 1.6, since the hotspot releases are nearly in sync.

]]> By: Java 7 What’s New, Performance Benchmark – 1.5, 1.6, 1.7 | Taranfx: Technology Blog http://lingpipe-blog.com/2009/03/30/jdk-7-twice-as-fast-as-jdk-6-for-arrays-and-arithmetic/#comment-5391 Sat, 29 Aug 2009 06:28:31 +0000 http://lingpipe-blog.com/?p=1085#comment-5391 [...] optimizations. We’ve already seen some very encouraging results in String performance, array performance, and a new concurrent garbage collector (G1). I suspect many people will find that their existing [...]

]]> By: Alex Miller - What Java 7 will mean to you http://lingpipe-blog.com/2009/03/30/jdk-7-twice-as-fast-as-jdk-6-for-arrays-and-arithmetic/#comment-4478 Sun, 12 Apr 2009 02:13:24 +0000 http://lingpipe-blog.com/?p=1085#comment-4478 [...] optimizations. We’ve already seen some very encouraging results in String performance, array performance, and a new concurrent garbage collector (G1). I suspect many people will find that their existing [...]

]]> By: Java 7 é mais rápido que a versão 6 « TekiMania http://lingpipe-blog.com/2009/03/30/jdk-7-twice-as-fast-as-jdk-6-for-arrays-and-arithmetic/#comment-4472 Fri, 10 Apr 2009 19:58:18 +0000 http://lingpipe-blog.com/?p=1085#comment-4472 [...] A notícia original pode ser vista em: http://lingpipe-blog.com/2009/03/30/jdk-7-twice-as-fast-as-jdk-6-for-arrays-and-arithmetic/ [...]

]]> By: Alex Miller - Weekly twitter links (April 6th) http://lingpipe-blog.com/2009/03/30/jdk-7-twice-as-fast-as-jdk-6-for-arrays-and-arithmetic/#comment-4437 Mon, 06 Apr 2009 15:25:34 +0000 http://lingpipe-blog.com/?p=1085#comment-4437 [...] JDK 7 Twice as Fast as JDK 6 for Arrays and Arithmetic (via @andreisavu) [...]

]]> By: Ismael Juma http://lingpipe-blog.com/2009/03/30/jdk-7-twice-as-fast-as-jdk-6-for-arrays-and-arithmetic/#comment-4436 Mon, 06 Apr 2009 09:08:18 +0000 http://lingpipe-blog.com/?p=1085#comment-4436 Foo,

Compressed pointers are not enabled by default. They are available for JDK 7 and JDK 6 Update 14 (the implementation in JDK 7 has a few improvements at the moment[1]). I would also try the benchmark with compressed pointers enabled as they can improve performance by quite a bit in cases where memory bandwidth is important[2].

Ismael

[1] http://blog.juma.me.uk/2009/04/03/load-unsigned-and-better-compressed-oops/
[2] http://blog.juma.me.uk/2008/10/14/32-bit-or-64-bit-jvm-how-about-a-hybrid/

]]> By: Ismael Juma http://lingpipe-blog.com/2009/03/30/jdk-7-twice-as-fast-as-jdk-6-for-arrays-and-arithmetic/#comment-4435 Mon, 06 Apr 2009 09:01:42 +0000 http://lingpipe-blog.com/?p=1085#comment-4435 Hi,

These days, one cannot compare “JDK 6″ to “JDK 7″ because different HotSpot versions are available in different JDK 6 builds. For example, JDK 6 Update 14 (still in early access) includes HS14 which was the version in the JDK 7 snapshots until a few months ago.

What kind of results do you get when you run your benchmark with JDK 6 Update 14[1]?

Ismael

[1] https://jdk6.dev.java.net/

]]>