doesn't yield any new cells. In theory this could happen if all
cells allocated with NEWCELL are either in use or conservatively
marked and all cluster spine cells are conservatively marked.
(Thanks to Dirk.)
scm_default_min_yield_1, scm_default_init_heap_size_2,
scm_default_min_yield_2, scm_default_max_segment_size): New global
variables. Can be customized by the application before booting
Guile. (We might want to be able to control these parameters
dynamically through the "options interface" in the future, but
note that that is additional functionality. Here we're giving
default values which the environment variables can override.)
the yucky, ugly and nasty conditional compilation back. sorry,
but it was either that or requiring GAWK to build Guile.
removed the code that implemented the old GC scheme.
* init.c (scm_boot_guile_1): removed the code conditioned on
!GUILE_NEW_GC_SCHEME.
* __scm.h: (GUILE_NEW_GC_SCHEME): removed.
* gc.c (scm_protect_object, scm_unprotect_object): change the
implementation to more efficient (at least in the time complexity
sense). the calls should now also be thread-safe -- I suspect
that people expect them to be. (thanks to Han-Wen Nienhuys)
removed the code that implemented the old GC scheme.
* hashtab.c (scm_hash_fn_create_handle_x): add missing
SCM_REALLOW_INTS before return.
(scm_before_gc_c_hook, scm_before_mark_c_hook,
scm_before_sweep_c_hook, scm_after_sweep_c_hook,
scm_after_gc_c_hook): New C level hooks.
(scm_after_gc_hook): New Scheme level hook.
(scm_gc_sweep): Moved scanning of weak vectors to weaks.c.
(scm_igc): Moved initialization of scm_weak_vectors and the call
to scm_guardian_gc_init to respective module.
(scm_mark_weak_vector_spines): Moved to weaks.c;
Call to scm_guardian_zombify moved to guardians.c;
Run scm_before_gc_c_hook, scm_before_sweep_c_hook,
scm_after_gc_c_hook at appropriate places.
(scm_init_gc): Initialize scm_after_gc_hook.
and we don't want that.
(INCLUDES): Removed all -I options except for the root source
directory and the root build directory.
* *.*: Change includes so that they always use the "prefixes"
libguile/, qt/, guile-readline/, or libltdl/.
(Thanks to Tim Mooney.)
made SCM values into pointers. This turned an arithmetic
computation of the address of the vcell into a pointer-arithmetic
one, thereby screwing up marking of structs.
Bugfix 2: Removed incompletely introduced loop variable `j' used
when protecting the tail array of a struct.
prediction.
(SCM_MAX): New macro.
(scm_freelist_t): New field: collected_1. Previous amount of
collected cells.
(gc_sweep_freelist_finish): Trigger based on two last values of
freelist->collected to avoid unnecessary allocation due to
temporary peaks.
trigger adjustmeant: Take yield (freed cells) for all freelists
into account.
(SCM_INIT_HEAP_SIZE_1, SCM_GC_TRIGGER_1, SCM_GC_TRIGGER_2): Tuned
to 50000 cell heap with 45% trigger.
(scm_gc_cells_collected): Reintroduced.
(SCM_HEAP_SIZE): New macro.
(scm_gc_sweep): Reintroduced correct computation of
scm_cells_allocated.
(scm_freelist_t): Corrected commentary for field `cluster_size':
Clustersize counts objects, not cells; New member
`clusters_allocated'.
* gc.h: (typedef struct scm_freelist_t) remove from here.
* gc.c: (CELL_UP, CELL_DN) make these macros take additional
parameter (the span).
(CLUSTER_SIZE_IN_BYTES, ALIGNMENT_SLACK) new macros.
(typedef struct scm_freelist_t) move here from gc.h, it had no
business being externally visible.
(typedef struct scm_heap_seg_data_t) renamed from
scm_heap_seg_data, to be style-compliant.
(scm_mark_locations) if the possible pointer points to a
multy-cell, check that it's properly aligned.
(init_heap_seg) alighn multy-cells properly, work with the
assumption that the segment size divides cleanly by cluster size
(so that there's no spill).
(round_to_cluster_size) new function.
(alloc_some_heap, make_initial_segment) use round_to_cluster_size
to satisfy the new init_heap_seg invariant.
