John CarmackVerified account

@ID_AA_Carmack The title is even truer today if we remove the "Floating-Point" bit. cc @jdmoon10

@ID_AA_Carmack If you don't have exact reproducibility, you don't have one language. Different semantics = different languages. "Java on X"?

@ID_AA_Carmack exact reproducibility was made optional with strictfp

@ID_AA_Carmack I read the PDF filename and all I could think of was @nineinchnails' Hurt... as sung by a Java programmer.

@ID_AA_Carmack JNI is the escape valve, I suppose. That way you get none of the portability, but lots of additional complexity.

@ID_AA_Carmack surely some of this has been resolved in the last 15 years! :O

@ID_AA_Carmack that is pretty horrible, now turn to JS numeric types and be scared

@ID_AA_Carmack Oh God, Mr. Carmack, you just had to give me another paper to read. It's like I'm back in school again.

I wonder if Microsoft's CIL does any better @id_aa_carmack

@ID_AA_Carmack Thanks for the link! As professor says:"Unintended numerical anomalies in computer games become Features celebrated in BLOGS"

@ID_AA_Carmack I'm more interested to know how this fits in combination with GPGPU. What is intel/nVidia/AMD's default mode of aritmetics.

@ID_AA_Carmack thanks for sharing, I love their writing style ... Sharp wit!

@ID_AA_Carmack Is this still relevant today? Has any of this been rectified in the proceeding years?

@ID_AA_Carmack 94.99999999% are clueless about floating point.

@ID_AA_Carmack An interesting old paper, that speaks the truth ^^

@ID_AA_Carmack @0xabad1dea On reading the first slide, it seems that Sun adopted the "Sue J++ out of existence" solution instead. Java #1!

Java's floating point is very bad: “@ID_AA_Carmack: an interesting old paper: http://www.cs.berkeley.edu/~wkahan/JAVAhurt.pdf …”

"@ID_AA_Carmack: This is an interesting old paper to read: http://www.cs.berkeley.edu/~wkahan/JAVAhurt.pdf …"

@ID_AA_Carmack Im reading.