I am trying this on the October gengc build as I write this. It does indeed pass the half gigabyte / fixnum thresshold smoothly. It does indeed crash in (room t) when the bignums start flying.
-dk Raymond Toy wrote: > > Dick> I'm using the interim release > Dick> cmucl-2003-09-15-sparcv9-solaris8 . I'm running the image > Dick> as > > Dick> lisp -dynamic-space-size 1023 > > Dick> Below the asterisks is the program I'm running. > > Dick> Below the plus signs is the transscript. > > Dick> Note that the system seems to lose all interest in > Dick> triggering any automatic GCs past a certain point. One GC, > Dick> but only one GC, takes place after the thresshold is > 2^29. > Dick> Does anyone have any insights? > > I vaguely recall some issue with this, but I can't find the messages > that discussed this. > > However, I tried your code a very recent CVS build. It does still GC > after the threshold has exceeded most-positive-fixnum and actually > finishes. Can you try again, with, say, the October snapshot? Note > that I used a Sparc build with gencgc. (Check *features* for > :gencgc.) > > That's the good news. The bad news is that room crashes. :-) > > Type-error in KERNEL::OBJECT-NOT-FIXNUM-ERROR-HANDLER: > 597445728 is not of type FIXNUM > [Condition of type TYPE-ERROR] > > It's printing out a summary and obviously it assumes something is a > fixnum but isn't. > > Ray
