1cdd080445
os-unix: Actually call pthread_join() to avoid an infinite loop and 100% cpu usage. It happens randomly. Fixes #738 . Thanks to kkrev or frankfranks.
2012-10-26 10:36:01 -07:00
afdaff3059
vm: make bignum-bitnot faster.
2012-08-31 15:22:07 -07:00
42f4193939
vm: simplify a bit the fixnum_divmod code.
2012-08-23 08:38:18 -07:00
ed8496b583
vm: speed up primitive_fixnum_divmod by 10-12%.
2012-08-22 21:46:59 -07:00
b541986b4e
vm: use tabs instead of spaces in bignum_gcd.
2012-08-22 21:31:27 -07:00
aad70160f0
vm: Add more comments about functions that allocate.
2012-08-15 22:00:08 -07:00
000efd9bbb
vm: Add messages about things that can allocate. Fix a gc bug in the primitive for quotation-code.
2012-08-15 21:30:55 -07:00
66de3a85dd
errors.cpp: Previous fix was correct, except we shouldn't clear the data_roots before gc'ing. Laugh out loud. Fixes #615 .
2012-08-15 19:49:04 -07:00
7e39d7ef30
errors.cpp: Patch looked like it would work. It either exposes more gc bugs or is incorrect.
...
Revert "errors.cpp: general_error() throws away its args when it calls compact_gc() when compiled with DEBUG=1. Save the args as data_roots instead. Fixes #615 . See #620."
This reverts commit dc46297050
2012-08-15 17:25:14 -07:00
beb202b3de
errors.cpp: general_error() throws away its args when it calls compact_gc() when compiled with DEBUG=1. Save the args as data_roots instead. Fixes #615 . See #620 .
2012-08-15 16:36:36 -07:00
47b4ebc80d
bignum.cpp: Fix calls to reallot_array() -- bignums can get new addresses if they are not in the nursery or are somehow larger. Fix a couple more correctness issues.
2012-08-15 01:51:58 -07:00
91d8326b77
math.cpp: Fix theoretical gc bug.
2012-08-15 01:51:58 -07:00
39c3f10138
bignum: incorporate some fixes from @slavapestov.
2012-08-10 14:28:44 -07:00
c4f85efc9d
vm: make sure we zero out the end of the bignums.
2012-08-04 18:02:27 -07:00
61a6b5692f
vm: a bit faster not to trim on each iteration...
2012-08-04 09:13:05 -07:00
f7bb7c6a32
vm: additional fix in bignum_gcd.
2012-08-03 21:50:48 -07:00
37001708cd
vm: some fixes to fast bignum gcd.
2012-08-03 12:13:46 -07:00
5c8a2a4057
vm: fix updating size inside bignum_gcd loop.
2012-08-02 17:36:07 -07:00
30e34ca616
bignum: Fix bignum_gcd algorithm from overwriting the wrong bignum memory. Add GC_BIGNUM to prevent gc crashes. Fixes #599 .
2012-08-02 17:16:03 -07:00
426c449ee1
vm: adding more GC_BIGNUM to WIN64 path. See #599 .
2012-08-02 17:04:22 -07:00
106ecdb6dd
compaction: Less than is the new greater than. Fixes #558 .
2012-07-25 17:57:50 -07:00
1a9caec368
Config.macosx: first two digits of 'xcodebuild -version' number is enough; throw away rest of version number.
...
$ xcodebuild -version
Xcode 4.3.3
Build version 4E3002
Fixes #543 .
2012-06-22 15:23:07 -07:00
e9ec95ad4f
compaction: Inifnite headaches--cured! Two compact functions, one was not like the other.
2012-05-27 18:26:03 -07:00
c577a5468e
vm: Typo for SITE_CFLAGS. Fixes #491 .
2012-04-10 16:28:04 -07:00
5794ab46a0
vm: use euclid gcd on win64 until we find a better way to do the 128-bit math.
2012-04-06 11:42:59 -07:00
eb556bfb06
vm: use twodigit type for all variables (fixes 32-bit warning).
2012-04-06 09:36:25 -07:00
4a29dafcdf
vm/Config: GNUMakefile doesn't actually put the CFLAGS on the command, so use SITE_CFLAGS instead. Put -m32 in Config.x86.32, -m64 in Config.x86.64. Fixes #488 .
2012-04-05 10:26:50 -07:00
22c26ff3f5
vm: adding bignum_gcd primitive.
2012-04-05 09:17:35 -07:00
3f9b2ef47a
use clang when building w/ xcode >4.3; it's faster
2012-02-18 23:40:38 -08:00
b094d4d0ec
vm: actually code_block_pic is a fine frame type
2011-12-21 21:12:54 -08:00
7121e0d598
vm: also smoke-check callstack after pic update
2011-12-21 18:31:37 -08:00
0b72b811fc
vm: flush after print_obj so it's usable from gdb
2011-12-21 13:25:28 -08:00
9f137c66d4
vm: fashion brigade
2011-12-21 13:25:05 -08:00
442923b561
vm: walk stack on inline_cache_miss if PIC_DEBUG=1
...
Reveals inline cache stack coherency bug behind #460
2011-12-21 13:24:35 -08:00
b6f323450f
vm: load mega cache object before popping frame
...
Fixes #457 harder
2011-12-20 14:38:01 -08:00
f513894435
vm: unwind-native-frames is special
...
Fixes #457
2011-12-19 16:54:25 -08:00
842eb27785
vm: cleanups
2011-12-14 09:56:50 -08:00
1c7bd9a7b4
vm: don't scribble memory handling leaf signals
2011-12-14 09:56:50 -08:00
7c9332e966
cpu.x86.*: clean up signal handler prolog code
...
Also bump the signal handler frame sizes up again since it looks like we shrank them too much.
2011-12-14 09:56:50 -08:00
7844c66d6e
vm: factor out word_stack_frame_size
2011-12-14 09:56:49 -08:00
a1ee881435
vm: get rid of now-trivial innermost_stack_frame
2011-12-14 09:56:49 -08:00
079025559d
simplify function prologs
2011-12-14 09:56:49 -08:00
2f20ee6fb0
vm: fix types after rebase
2011-12-14 09:56:49 -08:00
604ceb957c
vm: replace remaining stack_frame-based logic
2011-12-14 09:56:49 -08:00
d74f194b07
vm: remove _reversed from callstack iterator names
...
Now that they're the only game in town we can give them the short names
2011-12-14 09:56:49 -08:00
f6a5f48da0
vm: remove iterate_callstack_object
2011-12-14 09:56:48 -08:00
a1958c67bd
vm: callstack_reversed for callstack_to_array
2011-12-14 09:56:48 -08:00
0af5fff27c
remove iterate_callstack now that it's unused
2011-12-14 09:56:48 -08:00
822a658cc3
vm: callstack_reversed in call_frame_scrubber
2011-12-14 09:56:48 -08:00
9236c05e60
vm: stage code block map fixup properly for GC
...
Don't update the map until the very last thing, and pass untranslated addresses to the iterator functors. Somewhat complicated by the fact that, for startup_fixup, the map is initialized with fixed-up values, so the fixup functor needs a flag indicating whether it operates with a fixed or unfixed code heap map.
2011-12-14 09:56:48 -08:00
Doug Coleman
John Benediktsson
Joe Groff