-;;; When the dynamic usage increases beyond this amount, the system
-;;; notes that a garbage collection needs to occur by setting
-;;; *NEED-TO-COLLECT-GARBAGE* to T. It starts out as NIL meaning
-;;; nobody has figured out what it should be yet.
-;;;
-;;; FIXME: *GC-TRIGGER* seems to be denominated in bytes, not words.
-;;; And limiting it to INDEX is fairly reasonable in order to avoid
-;;; bignum arithmetic on every allocation, and to minimize the need
-;;; for thought about weird gotchas of the GC-control mechanism itself
-;;; consing as it operates. But as of sbcl-0.7.5, 512Mbytes of memory
-;;; costs $54.95 at Fry's in Dallas but cheap consumer 64-bit machines
-;;; are still over the horizon, so gratuitously limiting our heap size
-;;; to FIXNUM bytes seems fairly stupid. It'd be reasonable to
-;;; (1) allow arbitrary UNSIGNED-BYTE values of *GC-TRIGGER*, or
-;;; (2) redenominate this variable in words instead of bytes, postponing
-;;; the problem to heaps which exceed 50% of the machine's address
-;;; space, or even
-;;; (3) redemoninate this variable in CONS-sized two-word units,
-;;; allowing it to cover the entire memory space at the price of
-;;; possible loss of clarity.
-;;; (And whatever is done, it'd also be good to rename the variable so
-;;; that it's clear what unit it's denominated in.)
-(declaim (type (or index null) *gc-trigger*))
-(defvar *gc-trigger* nil)
-
-;;; When T, indicates that a GC should have happened but did not due to
-;;; *GC-INHIBIT*.
-(defvar *need-to-collect-garbage* nil) ; initialized in cold init