Re: [The Java Posse] Re: Magic float kills Java?

On Tue, Feb 15, 2011 at 3:27 AM, Casper Bang wrote: > On Feb 15, 10:40 am, Carl Jokl wrote: >> Is the x87 the most accurate general purpose floating point unit on >> the market? > > This is the way I understand it: > > - Floating point data cannot be represented accurately as base-2 > (divided in

[The Java Posse] Re: Magic float kills Java?

In summary it is irony rather than sarcasm. Therefore I can proceed to laugh: Ha ha ha. -- You received this message because you are subscribed to the Google Groups "The Java Posse" group. To post to this group, send email to javaposse@googlegroups.com. To unsubscribe from this group, send emai

[The Java Posse] Re: Magic float kills Java?

On Feb 15, 10:40 am, Carl Jokl wrote: > Is the x87 the most accurate general purpose floating point unit on > the market? This is the way I understand it: - Floating point data cannot be represented accurately as base-2 (divided in significand and exponent). - IEEE 754 specifies floating point s

Re: [The Java Posse] Re: Magic float kills Java?

I highlighted that comment more as an example of the poor bastards trying to get some help from the compiler guys than for its accurate portrayal of '87 performance. Myself, I never expect the FPU to produce accurate data, but I'd sure like it to be repeatable... But on the other hand, the reason o

[The Java Posse] Re: Magic float kills Java?

Is the x87 the most accurate general purpose floating point unit on the market? I honestly don't know either way because I don't know about the relative strengths of different processor FPUs. Perhaps not important but it would help me interpret the line "wanting accuracy out of the most accurate ge

Re: [The Java Posse] Re: Magic float kills Java?

That GCC thread is hilarious! Watching hundreds of poor humans try and convince a few nerds that something clearly faulty is actually a bug. "I'd like to welcome the newest members of the bug 323community, where all x87 floating point errors in gcc

[The Java Posse] Re: Magic float kills Java?

I ran the Java patch on Friday against several different installed JREs, a couple Java 6, a 1.5 and a 1.4.2. It worked fine except for one case that rt.jar couldn't be put back in place because it was in use. The patch only works on Oracle-supplied VMs and fails early when run on the Mac VM. It mak

Re: [The Java Posse] Re: Magic float kills Java?

found this linked off the php version of the bug: http://www.exploringbinary.com/why-volatile-fixes-the-2-2250738585072011e-308-bug/ which would lead me to believe that the bug is actually a call from byte-code dow

Re: [The Java Posse] Re: Magic float kills Java?

You can represent the whole Brisbane bunch :) As for the bug, I believe it's in an underlying C library that both PHP and Java use internally to parse floats, which is why it doesn't happen on all VMs. Somebody (here?) mentioned that it doesn't affect Kaffe, for example. -Josh On 14 February 201

[The Java Posse] Re: Magic float kills Java?

It could be either. But my guess is Java. Is is "just a parser" and nothing fundamental that a JIT would be doing. > In other news are you coming to the roundup? you're a great presence on the > google group, would be good to meet you in person. Thanks dude. I'v been wanting to go. When the de

Re: [The Java Posse] Re: Magic float kills Java?

thanks Christian! So would you say then the bug is in the Java code called by Double.parse()? from the nature of the advisory I thought it was a bug in some native code running in the JVM... In other news are you coming to the roundup? you're a great presence on the google group, would be good to

[The Java Posse] Re: Magic float kills Java?

It's real. I tried it (before I read the reports that described it) http://twitpic.com/3vwtcq Certain rare cases cause the parsing method to get stuck in a loop. I'v never written a double parser but I can imagine how it happens. Perhaps it's a while(value > 0) loop that expects the value will al