mirrors/emacs
1
Fork 0
mirror of git://git.sv.gnu.org/emacs.git synced 2026-03-24 07:41:54 -07:00
Code Issues Projects Releases Wiki Activity
6607 commits 248 branches 194 tags 1.1 GiB
Commit graph

2618 commits

Author SHA1 Message Date
Gareth Rees
55305914da Fix the mortality of the world to avoid runaway feedback between the dynamic criterion and the mortality of the arena's top generation as described in <https://info.ravenbrook.com/mail/2018/07/13/19-53-23/0/>. 
Copied from Perforce
 Change: 194648
 2018-07-13 20:56:08 +01:00
Gareth Rees
99260e46d6 Generation's computed mortality has denominator of condemned memory (not new memory), so use the same denominator when predicting the mortality for a trace. 
Simpler to compute predicted mortality as casualties/condemned instead of 1 - survivors/condemned.

Copied from Perforce
 Change: 194634
 2018-07-13 12:07:54 +01:00
Gareth Rees
032bd714e0 Catch-up merge from master sources at changelevel 194628. 
Copied from Perforce
 Change: 194631
 2018-07-13 11:28:36 +01:00
Gareth Rees
1ef415715f Must get the new size of the generation before condemning it (afterwards, the condemned memory is accounted as "old", not "new"). 
Copied from Perforce
 Change: 194628
 2018-07-13 10:55:30 +01:00
Gareth Rees
b1e3bd9932 Improve describe output for generations. 
Must take SegBase(seg) before reclaiming the segment in case it gets destroyed.

Copied from Perforce
 Change: 194624
 2018-07-12 09:47:29 +01:00
Gareth Rees
21e6672e70 Reclaim the same generations that were condemned. 
Copied from Perforce
 Change: 194621
 2018-07-12 07:51:04 +01:00
Gareth Rees
90f5f01b84 Update common assertions in the manual. 
Convert capacity to bytes once on initialization rather than every time it is used.

Copied from Perforce
 Change: 194612
 2018-07-11 21:27:47 +01:00
Gareth Rees
c692e2b6e0 Represent the set of generations condemned for a trace. 
Eliminate duplication between policyCondemnChain and traceCondemnAll.
Compute better mortality estimate for collection of the world.

Copied from Perforce
 Change: 194611
 2018-07-11 21:14:43 +01:00
Gareth Rees
97fa28e3b7 There is now no difference between the sets of sources for compilers mv and pc, so eliminate the duplication. 
Copied from Perforce
 Change: 194602
 2018-07-11 17:09:07 +01:00
Gareth Rees
590850726a Merge branch/2014-10-26/sc into the master sources. 
Copied from Perforce
 Change: 194595
 2018-07-11 15:20:49 +01:00
Gareth Rees
13e3f5d1cd Assert that stackhot gives us an address that's warmer than the jump buffer, as suggested by nb. 
Copied from Perforce
 Change: 194594
 2018-07-11 15:02:19 +01:00
Gareth Rees
b4068ba48c Attribute_noinline is not needed by the smoke tests. 
Copied from Perforce
 Change: 194590
 2018-07-11 14:44:28 +01:00
Gareth Rees
99d0b6393c Avoid "warning c4312: 'type cast': conversion from 'unsigned int' to 'addr' of greater size" from microsoft visual c. 
Copied from Perforce
 Change: 194583
 2018-07-11 12:56:00 +01:00
Gareth Rees
1dc8b245f0 Ssixi3.c and ssixi6.c were deleted, so don't try to compile them. 
Copied from Perforce
 Change: 194580
 2018-07-11 11:08:56 +01:00
Gareth Rees
e5e8d034d9 Force stackhot not to be inlined, avoiding the problem described in <https://info.ravenbrook.com/mail/2018/07/11/09-48-49/0/> 
Copied from Perforce
 Change: 194577
 2018-07-11 10:57:28 +01:00
Gareth Rees
268178f4cf Merge branch 2016-03-27/cbs-tidy into the master sources. 
Copied from Perforce
 Change: 194572
 2018-07-10 15:31:25 +01:00
Gareth Rees
cf80504dbb Justify invalidation in rangefinish, as suggested by rb <https://info.ravenbrook.com/mail/2018/07/10/11-01-54/0/> 
Copied from Perforce
 Change: 194560
 2018-07-10 12:13:33 +01:00
Gareth Rees
1cd422b339 Update copyright as suggested by dl in review <https://info.ravenbrook.com/mail/2018/07/10/09-59-36/0/> 
Copied from Perforce
 Change: 194555
 2018-07-10 11:22:18 +01:00
Gareth Rees
db4c75c932 Add awl to gcbench. 
Copied from Perforce
 Change: 194552
 2018-07-10 10:06:19 +01:00
Gareth Rees
80b2832260 Need to capture the stack in mps_arena_roots_walk, as explained in design.mps.stack-scan.anal.entry-points. 
Copied from Perforce
 Change: 194497
 2018-07-07 21:30:00 +01:00
Gareth Rees
81fdc239e7 Fix issues identified in review by gdr <https://info.ravenbrook.com/mail/2018/07/07/15-54-49/0/> 
Copied from Perforce
 Change: 194494
 2018-07-07 19:23:58 +01:00
Gareth Rees
9893976d39 Fumbled the merge. 
Copied from Perforce
 Change: 194488
 2018-07-06 21:47:19 +01:00
Gareth Rees
317cc9a5a6 Catch-up merge from master sources at changelevel 194484 to branch/2014-10-26/sc. 
Copied from Perforce
 Change: 194487
 2018-07-06 21:42:35 +01:00
Gareth Rees
f1b880321b Don't make multiple calls to controlalloc when you could just make one and divvy it up yourself. 
Copied from Perforce
 Change: 194484
 2018-07-06 18:56:12 +01:00
Gareth Rees
bcc01c1fd8 No need for rvalue(&(...)) -- the result of the & operator is not an lvalue. 
Copied from Perforce
 Change: 194456
 2018-07-05 22:53:02 +01:00
Gareth Rees
ba03d79694 Catch-up merge from master sources at changelevel 194449 to branch/2018-06-18/mrgseg. 
Copied from Perforce
 Change: 194452
 2018-07-05 16:22:24 +01:00
Gareth Rees
c868d94400 Update mpscopyrightnotice, as required by §3.1.1 of https://info.ravenbrook.com/project/mps/master/procedure/version-create 
Copied from Perforce
 Change: 194446
 2018-07-05 14:55:06 +01:00
Gareth Rees
940c52fddd Merge branch/2016-03-01/mvff-control into the master sources. 
Copied from Perforce
 Change: 194442
 2018-07-05 13:37:01 +01:00
Gareth Rees
bb53551d8f Merge branch/2018-07-04/mvff-perf into the master sources. 
Copied from Perforce
 Change: 194439
 2018-07-05 12:54:16 +01:00
Gareth Rees
fc3d0e91b9 Restore descriptive version of comment as suggested by rb <https://info.ravenbrook.com/mail/2018/07/05/11-08-52/0/> 
Copied from Perforce
 Change: 194436
 2018-07-05 12:17:41 +01:00
Gareth Rees
b30964104a Rename "node" to "rangetree" as suggested in review by gdr <https://info.ravenbrook.com/mail/2018/07/05/10-17-35/0/> 
Get Xcode builds working.
Add rangetree.[ch] to the source code index.

Copied from Perforce
 Change: 194434
 2018-07-05 12:06:47 +01:00
Gareth Rees
827f9676f2 Fix issues identified in review by gdr <https://info.ravenbrook.com/mail/2018/07/05/10-17-35/0/> 
Copied from Perforce
 Change: 194430
 2018-07-05 11:20:51 +01:00
Gareth Rees
3af7e264a3 Fix the check of the cbs overlapping insertion failure case: in order for the inserted block not to overlap it's right neighbour, its limit must be <= the base of the right neighbour (not the limit as previously written). this mistake was introduced accidentally in change 182344. 
Copied from Perforce
 Change: 194427
 2018-07-05 10:43:47 +01:00
Gareth Rees
73702aa7f1 No need for rvalue(&(...)) -- the result of the & operator is not an lvalue. 
Copied from Perforce
 Change: 194424
 2018-07-05 10:24:22 +01:00
Gareth Rees
ae3cc16e4b Function implementations call the macro implementations, to reduce the burden of keeping the functions and macros consistent. 
Copied from Perforce
 Change: 194418
 2018-07-05 09:28:01 +01:00
Gareth Rees
08f78b4b0e Merge branch/2018-07-03/mfs-tractp into the master sources. 
Copied from Perforce
 Change: 194414
 2018-07-04 17:13:53 +01:00
Gareth Rees
2564ffcb61 Avoid emitting poolalloc and poolfree events in the hot variety, so that they not affected asymmetrically by the poolfree macro. 
Copied from Perforce
 Change: 194410
 2018-07-04 16:51:13 +01:00
Gareth Rees
a737d39444 Merge branch 2018-06-27/job004056. 
Copied from Perforce
 Change: 194409
 2018-07-04 16:46:04 +01:00
Gareth Rees
17504d0242 Mfsextend now takes (base, limit) instead of (base, size) for consistency with the rest of the mps. 
Copied from Perforce
 Change: 194408
 2018-07-04 16:31:45 +01:00
Gareth Rees
6c3da83fb2 Avoid "dereferencing type-punned pointer might break strict-aliasing rules" warning from gcc. 
Copied from Perforce
 Change: 194403
 2018-07-04 15:56:51 +01:00
Gareth Rees
0cf1718af9 Improve performance of mvff in hot builds by: 
1. Using macros to bypass PoolFree and Land* generic functions.
2. Compiling out assertions on functions on the critical paths for manual allocation and freeing.
3. Taking advantage of the fact that splay->updateNode does not change during a splay operation.

Copied from Perforce
 Change: 194397
 2018-07-04 14:30:22 +01:00
Gareth Rees
97f8625195 Fix issues identified in review https://info.ravenbrook.com/mail/2018/07/03/10-10-11/0/ 
Copied from Perforce
 Change: 194382
 2018-07-03 13:04:21 +01:00
Gareth Rees
4946085f89 Integrate branch/2016-03-15/sunrise/...@190086,190087 to pick up elimination of tractp from mfs. 
Copied from Perforce
 Change: 194372
 2018-07-03 10:34:46 +01:00
Gareth Rees
33c55964a5 Fix issues identified in review by gdr. 
Copied from Perforce
 Change: 194339
 2018-07-02 09:05:03 +01:00
Gareth Rees
a215a7c4b9 Avoid warning "./poolmv.c:934:13: error: 'static' function 'mvcheck' declared in header file" from clang 5.0. see https://travis-ci.org/ravenbrook/mps/jobs/398247514 
Copied from Perforce
 Change: 194328
 2018-06-29 15:12:55 +01:00
Gareth Rees
5ee645763e Catch-up merge from master sources at changelevel 194314 to branch/2016-03-01/mvff-control. 
Copied from Perforce
 Change: 194323
 2018-06-29 14:28:33 +01:00
Gareth Rees
d1d282d351 Restore fixclosure together with explanation. 
Copied from Perforce
 Change: 194314
 2018-06-29 13:54:55 +01:00
Gareth Rees
556bb96d41 Catch-up merge from master sources at changelevel 194268 to branch/2018-06-18/job004070. 
Copied from Perforce
 Change: 194274
 2018-06-28 14:31:02 +01:00
Gareth Rees
d22656c8c4 Catch-up merge from the master sources at changelevel 194194 to branch/2017-03-04/seg-methods. 
Copied from Perforce
 Change: 194267
 2018-06-28 13:52:36 +01:00
Gareth Rees
c320c86231 Use initonceexecuteonce to get thread-safe initialization of the global locks. 
This means that the MPS no longer supports Windows XP, so update documentation accordingly.
Exercise the global locks in lockut.

Copied from Perforce
 Change: 194253
 2018-06-28 12:39:06 +01:00