Bookmark Topic Watch Topic
  • New Topic
programming forums Java Mobile Certification Databases Caching Books Engineering Micro Controllers OS Languages Paradigms IDEs Build Tools Frameworks Application Servers Open Source This Site Careers Other all forums
this forum made possible by our volunteer staff, including ...
Marshals:
  • Campbell Ritchie
  • Liutauras Vilda
  • Bear Bibeault
  • Paul Clapham
  • Jeanne Boyarsky
Sheriffs:
  • Devaka Cooray
  • Junilu Lacar
  • Tim Cooke
Saloon Keepers:
  • Tim Moores
  • Ron McLeod
  • Tim Holloway
  • Claude Moore
  • Stephan van Hulst
Bartenders:
  • Winston Gutkowski
  • Carey Brown
  • Frits Walraven

Who could alter the Java Runtime in one particular area? (work remotely)  RSS feed

 
Ranch Hand
Posts: 44
  • Mark post as helpful
  • send pies
  • Report post to moderator
I am attempting to find someone who can program or help me in the right direction with
what I'm after.  

I'm after someone to reprogram, build and submit an altered

64 bit Windows version of Java SE 11 Stable version
(as downloads from the zip file at https://jdk.java.net/11/ )


that is altered so as not to have any more floating point arithmetic errors
(denromals and pronormals) in it any more,
so that the default behaviour is range accurate, and doesn't have any pronormal or denormal
values at the most RHS digit excent of the range.  Probably also in a way which
continue to use and be layed out in memory in optimised fashion.
This is, unless such a version of this release of the OpenJDK is already out there,
for 64 bit Windows.

The classic example for this problem is this very known example:

double a = 0.1;
double b = 0.1;
double x = a*b;
out.println();
out.println(x);
out.println(x == 0.01);

Brackets and arithmetic ordering have apparently been part of the problem.

+, -, *, /, %, ++x, --x, x++, x--, (, ),  +=, -=, *=, /=, %=.

Hexidecimal mode needs to be corrected, on double and float.
Bitshifting operators may be impacted, and needs to be consistent
with the storage of the value; in the end it might not need to be changed,
or not very much.

I would like the default changed to be range accurate,
and things left so that if the strictfp keyword is used,
denormals and pronormals come back again
(normally, as before), if this is not very harder.


The Global Location for anyone doing this, if they can, will realistically be
anywhere in the world, although it could be better (or not) if you are not
very much offset, either before or after, around
Australian Central Standard time, UTC + 9:30, GMT + 9:30.

I am probably looking for a more advanced programmer who knows
how to acheive all of the relevant changes described, and can rebuild
the runtime so that it will work flawlessly, again,
at all used and mutual interface points, with no
problems, at all the interface points for floating point
computations and submissions.

Given that this issue is tantamount to an error fix in the Java SE language,
I am seeking a volunteer or lower cost worker who thinks similarly abou
t floating point operator errors at the Java defaulting as I do.

We can exchange through this forums by posting messages and replying,
certainly since they probably won't let people post email addresses in
their posts, really, anyway.

The last thing is that the results will need to be published on a free account,
on something like sourceforge or github, for download availability,
to comply with Java's GPL.

I look forward to someone who can easily
produce or even knows more about this problem!
 
Bartender
Posts: 5641
56
Eclipse IDE Firefox Browser Java MySQL Database VI Editor Windows
  • Mark post as helpful
  • send pies
  • Report post to moderator
This looks like a rehash to your post of 7 months ago which I thought Campbell addressed sufficiently.
 
Zachary Mitchell
Ranch Hand
Posts: 44
  • Mark post as helpful
  • send pies
  • Report post to moderator
Almost.  Except this time, I'm trying to locate
a programmer either through this public forum
or another one.  

Could someone reply to me here, pointing
me in a specific, particular direction?
 
Marshal
Posts: 24195
54
Eclipse IDE Firefox Browser MySQL Database
  • Mark post as helpful
  • send pies
  • Report post to moderator
Well, we do have a "Jobs Offered" forum. That seems like the ideal place for this post, at least on this forum.

Other places to post your job offer... well, I don't know. But I'm sure others will have an opinion.
 
Marshal
Posts: 63454
207
  • Mark post as helpful
  • send pies
  • Report post to moderator

Paul Clapham wrote:Well, we do have a "Jobs Offered" forum. . . . .

In which case it is necessary for the title of the thread to include the location of the job, as it says on the rubric for that forum. “Distance working,” or, “Working from home,” are acceptable alternatives.. Please go back and edit the title of the thread to include that information. Please also tell users what sort of payment you are offering.
 
Campbell Ritchie
Marshal
Posts: 63454
207
  • Mark post as helpful
  • send pies
  • Report post to moderator
note that threads may be deleted from Jobs Offered altogether for lack of location.
 
Saloon Keeper
Posts: 9856
199
  • Mark post as helpful
  • send pies
  • Report post to moderator
I can fix this problem for you as long as performance isn't an issue, and you also don't have problems with unexpected results when irrational numbers are involved.

I charge €120,- per hour.
 
Zachary Mitchell
Ranch Hand
Posts: 44
  • Mark post as helpful
  • send pies
  • Report post to moderator
To Stephan van Hulst,

where the RHS of a double or float operation is being concerned,
there would have to be some extra tidy up logic,
and some extra information to be held.

Apart from this sort of thing, what do you mean "while performance is not an issue"?

I am hoping that someone like you can do the changes that I seek
on the default operations of double and float, as well as Double and Float.

While the correct range tidy up for the operation might being to use
another 64 bit series of registers (a bit of extra memory that
it will likely have to resort to)  How on earth could
there be much of an extra performance penalt involved?

-Could you describe to me what you mean by this "performance" statement of yours?

-How much payment applicable time do you estimate for this task?

-I look upon fixing this problem as an error correction, and not
really some enhancement request.  Is it possible to lower your costs
because of this?

-How do you organise electronic funds transfers for payment?
What do you have set up?  Are you able to, when it comes to it,
privately share the information for a Bank funds transfer?


(please correspond with me for now on these forums,
but if the time comes, you would be emailing be to
poweruserm@live.com.au).
 
Zachary Mitchell
Ranch Hand
Posts: 44
  • Mark post as helpful
  • send pies
  • Report post to moderator
By the way, I understand and agree with irrational to rational number range truncation, which I agree with,
at the data level. I look forward to your response on this forum thread!
 
Campbell Ritchie
Marshal
Posts: 63454
207
  • Mark post as helpful
  • send pies
  • Report post to moderator
Please remember what I said about locations.
 
Zachary Mitchell
Ranch Hand
Posts: 44
  • Mark post as helpful
  • send pies
  • Report post to moderator
I have been hoping to use internet communications and uploading
to exchanging the relevant information.

My location coincides somewhere within the CST region
of the nation of Australia.  Does the forum administrator think that
I should be more specific, or for the sake of some confidentiality
should I leave it at that?  Can the administrator reply if I should be more specific?

Otherwise, I await Stephan van Hulst's replies!
 
Zachary Mitchell
Ranch Hand
Posts: 44
  • Mark post as helpful
  • send pies
  • Report post to moderator
Again, Stephan van Hulst, I at least
think that what I want, for floating point operations,
is more inline with version 1.1, before they changed it
for version 1.2.

I am excited for your reply!
 
Zachary Mitchell
Ranch Hand
Posts: 44
  • Mark post as helpful
  • send pies
  • Report post to moderator
It would also, lately, be required that

java.lang.StrictMath

be repaired.

I can mention it all later when you reply!
And by the way, what Timezone on earth are you in?
 
Saloon Keeper
Posts: 5288
143
  • Mark post as helpful
  • send pies
  • Report post to moderator
What kind of code do you intend to run on this JVM? I could imagine a number of existing libraries malfunctioning due to the altered underlying API.
 
Carey Brown
Bartender
Posts: 5641
56
Eclipse IDE Firefox Browser Java MySQL Database VI Editor Windows
  • Mark post as helpful
  • send pies
  • Report post to moderator

Zachary Mitchell wrote:It would also, lately, be required that
java.lang.StrictMath
be repaired.?


Just because it doesn't behave the way you'd like doesn't mean that it's broken. Read up on the IEEE specs and you'll start to get a feel for all the design decisions and trade-offs that went into the spec. And for additional info see "Numerical Recipes in C", there's a good bit in the front of the book about how to avoid underflows and overflows. Usually the issue is a poorly designed algorithm (mathematically speaking). IANAM
 
Zachary Mitchell
Ranch Hand
Posts: 44
  • Mark post as helpful
  • send pies
  • Report post to moderator
Well, any Java machine code really, that would be compatible on Windows 64 bit.

In terms of preconditions and postconditions, I wouldn't have thought
repairing the last decimal place of floating point, for all source code created,
and any consistent library invoked, would course any problems.

Because the means of correction of the final most RHS decimal place, in java code,
can't be calculated any other way in such a Java program, unless it used
BigDecimal and/or BigInteger in an internally consistent way,
or uses StrictMath (or Math, which only listens to StrictMath), which
I amasking for repairs to, in exactly the same way as the java
floating point.

Without seeming silly, with everything used or programmed in such
a new, altered, JVM, as all that, what other problem cold there, hereby, possibly be?
 
Zachary Mitchell
Ranch Hand
Posts: 44
  • Mark post as helpful
  • send pies
  • Report post to moderator
That such an altered JVM will not operate
previous code consistently is the case, and in fact something I aim to disregard
with my aims here.
 
Zachary Mitchell
Ranch Hand
Posts: 44
  • Mark post as helpful
  • send pies
  • Report post to moderator
By which I mean programmed software and libraries in java,
after the fact of the compiler and either virtual machine, surely?
 
Zachary Mitchell
Ranch Hand
Posts: 44
  • Mark post as helpful
  • send pies
  • Report post to moderator

Just because it doesn't behave the way you'd like doesn't mean that it's broken. Read up on the IEEE specs and
you'll start to get a feel for all the design decisions and trade-offs that went into the spec. And for additional info see
"Numerical Recipes in C", there's a good bit in the front of the book about how to avoid underflows and overflows.
Usually the issue is a poorly designed algorithm (mathematically speaking). IANAM



-For the purposes of their range, float and double are far quicker, smaller in memory,
and far more organised in source code to program, read and change.  At any rate,
due to thinking around JEP 306

https://bugs.openjdk.java.net/browse/JDK-8175916

that may or may not happen, I still wish to pursue someone
who can consistently introduce the changes I seek to
the JVM most immediately useful to me.

I have more properly now re-published this in the jobs area under

https://coderanch.com/t/706892/careers/Altering-Java-Win-floating-point
 
Paul Clapham
Marshal
Posts: 24195
54
Eclipse IDE Firefox Browser MySQL Database
  • Mark post as helpful
  • send pies
  • Report post to moderator
I merged your stuff with the following thread. I hope that is okay by you.
 
Zachary Mitchell
Ranch Hand
Posts: 44
  • Mark post as helpful
  • send pies
  • Report post to moderator
I am attempting to find someone who can program some important
alterations to the JVM for my sake, and that of the public too.

I'm after someone to reprogram, build and submit an altered

64 bit Windows version of Java SE 11 Stable version
(as downloads from the zip file at https://jdk.java.net/11/ )

that is altered so as not to have any more floating point arithmetic errors
(denromals and pronormals) produced any more, so that the default
behaviour is range accurate, and doesn't have any pronormal or denormal
values at the most RHS digit excent of the range.  Probably, also in a way which
continue to use and be layed out in memory in optimised fashion.
This is, unless such a version of this release of the OpenJDK is already out there,
for 64 bit Windows, which I believe that such is not.

The types involved are float and double, Float and Double (autoboxing
and autounboxing).

The classic example for this problem is this very known example:

double a = 0.1;
double b = 0.1;
double x = a*b;
out.println();
out.println(x);
out.println(x == 0.01);


Brackets and arithmetic ordering have apparently been part of the problem.

+, -, *, /, %, ++x, --x, x++, x--, (, ),  +=, -=, *=, /=, %=.

Hexidecimal mode needs to be corrected, on double and float.
Bitshifting operators may be impacted, and needs to be consistent
with the storage of the value; in the end it might not need to be changed,
or not very much.

java.lang.StrictMath will need to be updated so that all its methods operate
with accurate value consideration for the final decimal place is all function
input values. If any of it contains the strictfp keyword, those will all
have to be removed.


I would like the default changed to be range accurate,
and things left so that if the strictfp keyword is used,
denormals and pronormals come back again
(normally, as before), if this is not very harder.

The Global Location for anyone doing this, if they can, will realistically be
anywhere in the world, although it could be better (or not) if you are not
very much offset, either before or after, around
Australian Central Standard time, UTC + 9:30, GMT + 9:30.

I am probably looking for a more advanced programmer who knows
how to acheive all of the relevant changes described, and can rebuild
the runtime so that it will work flawlessly, again,
at all used and mutual interface points, with no
problems, at all the interface points for floating point
computations and submissions.
The ability to accept currency converted payment
through paypal details would be totally ideal, instead of the more challenging
but possible bank transfer approach.

Given that this issue is tantamount to an error fix in the Java SE language,
I am seeking a volunteer or lower cost worker who thinks similarly abou
t floating point operator errors at the Java defaulting as I do.

We can exchange through this forums by posting messages and replying,
certainly since they probably won't let people post email addresses in
their posts, really, anyway.

The last thing is that the results will need to be published on
sourceforge, for download availability, to comply with Java's GPL.

Last that I learned, Stephan van Hulst
seemed willing to help me, and I look forward to his reply.

I look forward to someone who can easily
produce or even knows more about this problem!
 
Zachary Mitchell
Ranch Hand
Posts: 44
  • Mark post as helpful
  • send pies
  • Report post to moderator
I have been hoping to use internet communications and uploading
to exchanging the relevant information.

My location coincides somewhere within the CST region
of the nation of Australia.  Does the forum administrator think that
I should be more specific, or for the sake of some confidentiality
should I leave it at that?  Can the administrator reply if I should be more specific?
 
Zachary Mitchell
Ranch Hand
Posts: 44
  • Mark post as helpful
  • send pies
  • Report post to moderator
Stephan van Hulst, or any one else so fully
capable, reply or entertain address
my requests here?
 
Paul Clapham
Marshal
Posts: 24195
54
Eclipse IDE Firefox Browser MySQL Database
  • Mark post as helpful
  • send pies
  • Report post to moderator
Zachary, you have made your request. There's no need to repeat it every twenty minutes. So I'm locking this thread.
 
Let me tell you a story about a man named Jed. He made this tiny ad:
Become a Java guru with IntelliJ IDEA
https://www.jetbrains.com/idea/
    Bookmark Topic Watch Topic
  • New Topic
Boost this thread!