...
> I just got a new OS (XP) that didn't include a JVM or jview. I was using
> jview to run my local Java applications but now have to switch to java.exe.
> What a difference!! and not for the better.
>>..Am I doing something wrong? If not, how do I reinstall jview?
>
> 1) Get a time machine..
> ..
The work I mainly use Java for is internal lab work. Java is great at
crunching numbers and I really need the speed and not so much the
portability. So what I just did to solve my speed problem was to install the
MSJVM off the net (msjavx86.exe) and then install an old JDK which is needed
to compile.
For the application I was running I noticed a speed difference of just over
10 times. That's not mice nuts. I'm not sure why Sun can't make their
java.exe perform at a level equivalent (or even close) to jview? I mean,
aren't they the ones that developed the language in the first place?
Thomas
Andrew Thompson - 31 Aug 2006 18:35 GMT
> >>..Am I doing something wrong? If not, how do I reinstall jview?
> >
[quoted text clipped - 11 lines]
> java.exe perform at a level equivalent (or even close) to jview? I mean,
> aren't they the ones that developed the language in the first place?
You have struck one of the area where the MSVM shines
brightly - it is fast at what (little) it does. Although I have
never seen a repeatable test that supported it was *ten*
times faster than the Sun JVM, when doing raw number
crunching, it usually beat the Sun VM by a factor
of at least 2-4 (twice to four times as fast), from memory.
I suspect that part of it is that MS used lower level
calls to the hardware, much lower and faster than what
they are prepared to publicly document.
If that is the case, they would always have the potential
to build a faster VM than Sun could.
Also, Sun seems to concentrate their efforts on optimising
other things, rather than raw number-crunching ability
(which is rarely the bottleneck).
(concedes) You might still be able to find an
'unauthorised' distribution of the MSVM on the net.
Someone was offering it some time ago after MS
withdrew it's JVM from IE. I have not checked
recently if the site is still active, and do not have the
URL handy. I suppose Google is your best bet there.
Andrew T.
Oliver Wong - 31 Aug 2006 19:38 GMT
>>>..Am I doing something wrong? If not, how do I reinstall jview?
>>
[quoted text clipped - 11 lines]
> java.exe perform at a level equivalent (or even close) to jview? I mean,
> aren't they the ones that developed the language in the first place?
This might be like asking why a Microsoft employee can complete a 100
meter *dash* in 15 seconds, whereas a Sun employee takes 20 seconds to
complete a 100 meter *hurdle*. The so-called Microsoft "JVM" doesn't
implement the Java standard, so the two virtual machines aren't actually
competing under the same conditions.
If you need pure number crunching abilities without regards to
portability, I heard FORTRAN is one of the best languages for that.
- Oliver
~Glynne - 31 Aug 2006 21:16 GMT
> >>..Am I doing something wrong? If not, how do I reinstall jview?
> >
[quoted text clipped - 13 lines]
>
> Thomas
Since you mention "number crunching" and lousy performance in your
post, you are probably running into issues with the terrible
performance of the trig functions in the newer JVMs.
Java 1.3 was the last release with good trig performance. The trig
functions were "fixed" in the 1.4 release to provide infinite precision
during argument reduction. Now you can get an accurate value for
sin(2000000000).....if you're willing to wait long enough.
If you need better performance you either have to
1) code your own routines in Java (e.g. from the Cephes library), or
2) use JNI to access the C math functions.
~Glynne
Arne Vajhøj - 01 Sep 2006 01:39 GMT
> The work I mainly use Java for is internal lab work. Java is great at
> crunching numbers and I really need the speed and not so much the
[quoted text clipped - 6 lines]
> java.exe perform at a level equivalent (or even close) to jview? I mean,
> aren't they the ones that developed the language in the first place?
This does not match my experience.
My experience is that SUN JVM passed MS JVM when SUN JVM
went from 1.3.1 to 1.4.1 (I have never tested 1.4.0).
But of course it depends on what the program do.
Arne