[NTLK] NEWT/0 set-aref bug
simonbell at me.com
Wed Dec 17 05:53:10 EST 2014
> On 16 Dec 2014, at 22:13, Matthias Melcher <mm at matthiasm.com> wrote:
> technical question first: Simon, what's the status of your awesome work? Did you manage to port the code over to Intel, or better yet, make it processor agnostic?
My Newton framework is the core of NCX so yes, it is little-endian. Should still be big-endian too if compiled for the right target but I don’t do regression testing so I can’t be sure. Currently I am working on a 64-bit version. That means 64-bit Refs, which solves the 2010 time-in-seconds problem. You can take a sneak preview of the 64-bit NCX 2.1 at <http://newtonresearch.org/connection/NCX21.zip> It is built for Mac OS X Yosemite. Still a bit flaky; frequently crashes on startup but if you get past that stage seems to be okay. (Until you inspect lots of soup entries and fill the VBO pages which are not GC’d properly yet.)
Also working on a yet-another-NTK app which I use as a NewtonScript playground. The NewtonScript bytecode I posted before is copied from that. I had this mad idea of updating the NewtonScript compiler using LLVM to target both bytecode and native code. It’s serendipity that Matej discovered the set-aref anomaly in NEWT/0 just when I was in a position to test my own version.
More information about the NewtonTalk