Board index » kylix » Re: DElphi/Kylix Vs. Java ??

Re: DElphi/Kylix Vs. Java ??


2003-12-10 02:22:56 PM
kylix0
E>box state. PC - P5 with 512Mb of memory, for Windows and for Linux
E>equally slow.
Of course P4.
 
 

Re:Re: DElphi/Kylix Vs. Java ??

Ender wrote:
Quote
p>No, it is called you are attempting to run a program with less memory
p>than is recommended.

:-)))))) Interesting point of view. You trying to say that speed of
:program
depends from meeting memory requirements instead of actual memory usage?
CBX perfectly fit into 384Mb. It take much less than 384Mb of memory, and
it does not swap. So - memory is not the cause of slowdown.

How do you know it is running fine within 384 meg? Are you taking into
account JIT/JVM GC operations? I woud suspect you are right, however, I
would like to know how you know this.
Quote
Did you tried to install CBX? How you feel after speed of installer from
Zero G? Clicking the check box take about second for switching check box
state. PC - P5 with 512Mb of memory, for Windows and for Linux equally
slow.

Zero G installer is slow, period.
Quote
BuilderX contain many more functions and libraries than Delphi or Kylix?

It will. I do not know if Borland left the options open for the plug in or
not (for the GUI Builder). I do not know (but I doubt it), that Builder X
is fully optimized now, since it is a more or less a preview, not a final
product.
However consider:
a Full De{*word*81} and compile environment for Borland C++ for Windows, Intel
32 bit Compilers, Microsoft Visual C++ Compilers, Sun C++ Forte Compilers,
Metroworks C++ Compilers, GNU C++ compilers. Last time I looked, Delphi had
only one compiler, and Linux only one compiler. <G>Of course you know this
will take some overhead to supply this.
Add to this CVS integration, Rational Clear Case Integration, and Visual
Source Safe Integration. Also add Intel's VTune Integration.
Add the UML Modeler integration. Refactoring integration, XML Spy, Corba
Development and Debug test environment, JavaDoc integration, etc.
None of what I listed is in Delphi, except Corba, which does not support a
full debug environment for Corba (Delphi was using a COM wrapper for
Corba). I do not however, know about Delphi 7 and 8, since I have not
looked at either.
Again, I do not know what GUI hooks are there or not. I do not proclaim to
be an expert with BuilderX, I have only breifly looked at it. I do use
JBuilder however, everyday, and the list for the JB IDE is much more
extensive than anything Delphi has. Since they are sharing a common IDE,
and again, Builder X is a first release (without the GUI builder), I doubt
it has been optimised yet. I do know that JB 7 is not as fast as JB X, nor
is JB 9. JB 9 was the first to use JDK 1.4.x, whereas the rest of the
versions were built using 1.2 (JB 3, JB 3.5) or 1.3.x (4-7).
Borland recommends a minimum of 512 meg and recommends 768 meg of RAM. Since
you are shy at least 128 meg, I do not now how it would run on your
machine. I can run JB X on 256 meg of Ram, but it runs quite a bit
different from 256 meg to 512 meg or more of RAM. Speed difference is
pretty substantial.
 

Re:Re: DElphi/Kylix Vs. Java ??

Ender wrote:
Quote
RP>programs in the first place is a grave error. Blaming Java for the
RP>insufficiency of either buyer or developer is not a good argument.

Please correct me if i'm misunderstood. You saying that when program
working slow because of memory and/or CPU overhead of RTL (comparing to
other programs) then buyer or developer of program is guilty, not
developer of that RTL or entrie concept?
Correct. The concept leads to a design which needs a certain amount of
resources. Using tools to create applications propagates that. Buying it
without ascertaining that you have enough resources does not give a buyer
the right to complain that the tool or the concept is a failure. In the
same vein, choosing a tool and developing an application that will not run
properly in the environment it is intended too is a Bad Thing(TM).
Quote
RP>Again, same argument. That the JVM has overhead, yes it has, and it
RP>has its uses. That is what comes with the language. And that is what
RP>designers must take into account. Yes, it might require a bit of
RP>tuning, but the opportunities to do so are there, contratry to the
RP>possibilities a native application has in that respect.

Native apps does not use more memory than they really needed.
Is that so?
Quote
Why C++ Builder X does not use native widgets for my Linux? And main
question what is native widgets for Linux? Qt, GTK? It does not look
similar to any Qt or GTK app i have in my system installed.
From that POV, there isn't such a thing as a native widget. Actually, X has
a number of native widgets: Xaw Xaw3d, etcetera, etcetera, and Qt and GTK
are layers on top of that.
Quote
Do your cool engineers able to create Java program that will run faster,
eat less memory and be stable than native program that doing same things?
Of course we assume that both programs require equal resources for design
and testing and be developed equally qualified programmers.
I don't know. What I do know is that it is far more difficult to create the
same functionality in the same time in let's say C++ or OP, mainly because
of what is available and the extra checks and balances in Java and last but
not least the way Java does memory management.
--
Ruurd
 

{smallsort}

Re:Re: DElphi/Kylix Vs. Java ??

Quote
Kylix has been dommed since day one. The LInux world is not interested in
investing in technology by spending money.
They have different priorities and goals (this doesn't mean they are bad.
Just different.).

I agreed with the rest of your post but Kylix wasnt doomed because the
linux world doesnt want to spend money. We plonked down $25,000 for
an enterprise license for oracle on red hat. We have spent tons of
money on linux products.
The reason kylix was doomed is because exactly 0% of linux developers
want to use object pascal and the VCL. If they had come out with a
compliant C++ compiler with a real C++ framework 5 years ago they
would have sold tons of it. Whether its too later for CBuilderX
remains to be seen. I dont think so.
They also gotta come to grips with their pricing reality. They may
get away with charging $5,000 for JBuilder architect but nobody is
forking out that much cash for a C++ product. I've said this a
million times but put all of the dbExpress drivers, single user source
control, wxDesigner, some light modeling into the pro sku for $1,000
and you've got yourself a winner.
 

Re:Re: DElphi/Kylix Vs. Java ??

Quote
MS will not hand the MS CLR/CLI to anyone. Sun has. That is the point.

Sun hasn't handed anything to anybody. Show me proof.
Quote
that is not controlled by Sun alone. It is controlled rather by the JCP. I
The standard in controlled by a democratic process but the product is
owned wholly by Sun.
 

Re:Re: DElphi/Kylix Vs. Java ??

Quote
Huh? No one can invent a 100% compatible MS.NET period!! It is impossible,
since MS allows NO ONE but MS to determine the direction and future for
NET. That is the point. IBM cannot create a 100% compatible NET for AS400,
AIX, OS390, etc. Neither can Mono, Blackdown, Oracle, Ensignia, Nokia, etc.

Your confusing ownership with future direction. I repeat Sun owns the
rights to EVERYTHING and the can and i believe will start charging for
it when the shit hits the fan for them financially.
As I understand it because m$ turned over C# and the CLR to the EMCA
you a free to clean room both and m$ can never come after you. Sun
has not done this.
 

Re:Re: DElphi/Kylix Vs. Java ??

Quote
>1. Insufficient performance.

Unfounded claim, Ender. Insufficient for what? Scientific calculations
involving huge matrices? Well, in that case, use Fortran.

Startup time is ridiculous. Aside from Swing though once is starts up
and eats 30MB of ram, its does run fast.
Quote
>2. Memory requirements for VM.

Tunable parameter, thank you.

Please explain what you mean by tunable because every app I have ever
written uses minimally 30MB of ram. Plus you have to run a seperate
VM for every process. Can you imagine if the 50 apps I am running on
my desktop right now were running in 50 different 30MB VM?
Is there a way using the JDK to not JIT code and cut the memory usage
down?
Quote
>4. Usage of not native GUI widgets.

That's nonsense. Swing uses native widgets in the end. If you want to use
non-native ones, no problem, if you want native ones, no problem.

Swing does not use native widgets and its slow as heck. SWT uses
native controls.
 

Re:Re: DElphi/Kylix Vs. Java ??

sorry i run CBuilderx and ArgoUML on my athlon 2000 with 768 MB of ram
and they still run like a dog. I can't understand why people
apologize for java's gross misuse of memory. Visual Studio net 2004
flies on this machine.
 

Re:Re: DElphi/Kylix Vs. Java ??

Quote
I don't know. What I do know is that it is far more difficult to create the
same functionality in the same time in let's say C++ or OP, mainly because
of what is available and the extra checks and balances in Java and last but
not least the way Java does memory management.

I'd agree with OP but not with C++. Memory management is very easy to
deal with in modern C++. I hardly ever have to manually call delete
anymore. I'm not sure what the rest of the checks and balances are.
 

Re:Re: DElphi/Kylix Vs. Java ??

Mike Margerum wrote:
Quote
>>1. Insufficient performance.
>
>Unfounded claim, Ender. Insufficient for what? Scientific calculations
>involving huge matrices? Well, in that case, use Fortran.
>
Startup time is ridiculous. Aside from Swing though once is starts up
and eats 30MB of ram, its does run fast.

What JDK version are you using? This is not true with 1.4.2,
Quote
>>2. Memory requirements for VM.
>
>Tunable parameter, thank you.
>
Please explain what you mean by tunable because every app I have ever
written uses minimally 30MB of ram. Plus you have to run a seperate
VM for every process. Can you imagine if the 50 apps I am running on
my desktop right now were running in 50 different 30MB VM?

Tunable parameters for the jre are done via command line arguments as per
here:
www-106.ibm.com/developerworks/ibm/library/i-tuning/
Why in the world would you do this? JVMs are forward compatible, not
backwards compatible, as long as they are in the same family. For instance:
1.2.x apps can be run on 1.4.x JVMs
1.3.x apps can be run on 1.4.x JVMs
1.1.x apps can be run on the newer JVMs as long as they are not attempting
to use any of the older Swing 1 widgets.
There are a few exceptions, but very few.
Quote
Is there a way using the JDK to not JIT code and cut the memory usage
down?

Sure, but you are speaking of major speed impediments.
java -nojit [params]
 

Re:Re: DElphi/Kylix Vs. Java ??

Mike Margerum wrote:
Quote
sorry i run CBuilderx and ArgoUML on my athlon 2000 with 768 MB of ram
and they still run like a dog. I can't understand why people
apologize for java's gross misuse of memory. Visual Studio net 2004
flies on this machine.
Strange, since I have less of a machine than you (Athlon 1800, 512 meg DDR
Ram), and I get good performance. I suggest looking at the JVM you are
using. I am using 1.4.2.
Also what Video drivers are you using?
 

Re:Re: DElphi/Kylix Vs. Java ??

Mike Margerum wrote:
Quote
>MS will not hand the MS CLR/CLI to anyone. Sun has. That is the point.
>
Sun hasn't handed anything to anybody. Show me proof.

>that is not controlled by Sun alone. It is controlled rather by the JCP. I
The standard in controlled by a democratic process but the product is
owned wholly by Sun.
Sure the copyrights and patents to Java are owned by Sun, as are the
license. Never argued that they are not.
The process however, as you correctly stated, as to what constitutes the
specifications of Java (J2SE, J2EE, J2ME) is a democratic process of all of
those involved.
Those incorrectly denigrating Java seem to think that because C# is ECMA,
that MS has given up all rights to C# and that the NET framework is also
ECMA. Nothing could be further from the truth.
 

Re:Re: DElphi/Kylix Vs. Java ??

Mike Margerum wrote:
Quote
Startup time is ridiculous. Aside from Swing though once is starts up
and eats 30MB of ram, its does run fast.
Tune it. Run java -X in a console and take measures.
Quote
Please explain what you mean by tunable because every app I have ever
written uses minimally 30MB of ram. Plus you have to run a seperate
VM for every process. Can you imagine if the 50 apps I am running on
my desktop right now were running in 50 different 30MB VM?
Possible. Don't start from the commmand line but from a 'starter'
application.
Quote
Is there a way using the JDK to not JIT code and cut the memory usage
down?
-nojit switch?
Quote
Swing does not use native widgets and its slow as heck. SWT uses
native controls.
Go and ascertain this fact yourself. More specifically, have a look at the
plaf packages and ascertain that key methods are native, then look up the
native code for them. SWT is thinner, therefore faster, but less competent
than Swing.
--
Ruurd
 

Re:Re: DElphi/Kylix Vs. Java ??

Mike Margerum wrote:
Quote
>I don't know. What I do know is that it is far more difficult to create
>the same functionality in the same time in let's say C++ or OP, mainly
>because of what is available and the extra checks and balances in Java and
>last but not least the way Java does memory management.
>
I'd agree with OP but not with C++. Memory management is very easy to
deal with in modern C++. I hardly ever have to manually call delete
anymore. I'm not sure what the rest of the checks and balances are.
Enforcing method signature for example. Compile errors if you do not throw
or catch an exception specified by a called method for example. Compile
errors if you happen to forget an '=' as in:
if (x = true) {}
to name a few examples.
--
Ruurd
 

Re:Re: DElphi/Kylix Vs. Java ??

"Mike Margerum" < XXXX@XXXXX.COM >wrote in message
[..]
Quote
I agreed with the rest of your post but Kylix wasnt doomed because the
linux world doesnt want to spend money. We plonked down $25,000 for
an enterprise license for oracle on red hat. We have spent tons of
money on linux products.
Well, don't take it literally. I meant "most of it".
Quote
The reason kylix was doomed is because exactly 0% of linux developers
want to use object pascal and the VCL. If they had come out with a
compliant C++ compiler with a real C++ framework 5 years ago they
would have sold tons of it.
Fair enough. I forgot that part. It is probabily more important than what I
said.
Quote
They also gotta come to grips with their pricing reality. They may
get away with charging $5,000 for JBuilder architect but nobody is
forking out that much cash for a C++ product. I've said this a
million times but put all of the dbExpress drivers, single user source
control, wxDesigner, some light modeling into the pro sku for $1,000
and you've got yourself a winner.
Agreed.