>>>>> "Sam" == Sam Steingold <[EMAIL PROTECTED]> writes:
Sam> so what do I do?
Sam> (this is quite reproducible.)
Not sure. The gencgc version for sparc is still rather new, so it might be
buggy. Can you try a sparc build with the original Cheney GC?
Unfortunately, this only has half the space because it needs 2 spaces
for the heap.
Sam> later I get this:
Sam> !!! CMUCL has run out of dynamic heap space. You can control heap size
Sam> !!! with the -dynamic-space-size commandline option.
Sam> *A2 gc_alloc_new_region failed, nbytes=8.
Sam> Generation Boxed Unboxed LB LUB Alloc Waste Trig WP GCs Mem-age
Sam> 0: 0 0 0 0 0 0 2000000 0 0 0.0000
Sam> 1: 0 0 0 0 0 0 2000000 0 0 0.0000
Sam> 2: 0 0 0 0 0 0 2000000 0 0 0.0000
Sam> 3: 70349 60598 0 0 1071942920 774904 961963104 0 1
0.8955
Sam> 4: 65327 60441 4226 1075 1072919168 798080 2000000 0 0
0.0000
Sam> 5: 0 0 0 0 0 0 2000000 0 0 0.0000
Sam> 6: 0 0 0 0 0 0 0 0 0 0.0000
Sam> Total bytes alloc=2144862088
Sam> I run cmucl with "-dynamic-space-size 2047MB" option.
I guess is a C int bug. The space size should be an unsigned.
Sam> I gather that I am out of luck here.
Well, as some other messages here have pointed out, gencgc tends to
hang onto garbage until a full GC happens. Perhaps you can put a (gc
:full t) in some places.
Or you can tweak some parameters of the GC. Unfortunately, no one
really understands what those parameters do. Or you can (temporarily)
turn of the generational part of the GC by setting the C variable
gencgc_oldest_gen_to_gc to 0.
Sorry, this isn't much help,
Ray
