Mailinglist Archive: opensuse (4570 mails)

< Previous Next >
Re: [SLE] Clearing up the FUD on CLI/Mono
  • From: Anders Johansson <andjoh@xxxxxxxxxx>
  • Date: Sun, 13 Nov 2005 00:59:32 +0000 (UTC)
  • Message-id: <200511130201.28611.andjoh@xxxxxxxxxx>
On Sunday 13 November 2005 01:34, Randall R Schulz wrote:
> I'd think you'd not be given to folklore about program efficiency.
> Static compilation to native code as gcj (or conventional C++) does it
> is _not_ the best way to optimize performance. Java's so-called JIT
> (it's really on-demand) native code compilation has optimization
> opportunities not available when only static program analysis is
> possible.

Well, gcc has more to offer than just static compilation, although the new
stuff isn't really ready yet. But the profiling that you can do in later
versions of gcc has a lot of the same characteristics as JIT

I do know that a lot of the comments about java efficiency are exaggerated,
but it does still have overhead when you run it in the JVM

> For all practial purposes and for a sizeable majority of programs, Java
> is every bit as fast as C++ and what's more, the opportunities for
> optimization of C++ are more played-out than are those for Java, which
> is still improving its performance characteristics.

Well, that's a little backwards, since if it's possible to do in a java system
it's certainly possible to do for C/C++, and you'd still have the overhead of
the JVM

I just think the gcj option to compile a native executable is pretty cool


< Previous Next >
Follow Ups