349f99661d
compiler.*: Remove the scrubbing part of the GC maps
...
Instead of generating GC maps which describe which stack locations that
are uninitialized, we emit ##clear instructions for those locations in
front of ##call-gc instructions. This makes the context root scanning
much simpler because the GC can assume that all stack slots are
initialized. It also removes the compiler.cfg.stacks.vacant pass and
seem to reduce the image size slightly because many fewer GC maps needs
to be emitted.
2016-09-11 20:34:44 +02:00
640759ffed
compiler.cfg.*: new step in finalization: clear-uninitialized
...
this step is supposed to analyze the cfg, find where there are peeks
that can cause stack underflow, and insert replace-imm instructions so
that there are no uninitialized stack locations if they do. could fix #1187
2015-01-22 12:17:20 -08:00
163fef12f9
compiler.cfg.*: detached the normalize-height and schedule-instructions passes
...
those stages cause uninitialized "holes" in the stacks which in turn
causes gc errors if stack underflows occur
2015-01-22 12:17:20 -08:00
c65877104e
compiler: cleanup usings.
2014-12-13 16:10:21 -08:00
068ddd6c7b
compiler: cleanup cfg passes to have stack effect ( cfg -- )
2014-12-11 12:48:43 -08:00
e961a03837
compiler.cfg.stacks.vacant: compiler pass replacing uninitialized for more accurate stack maps
...
This pass uses a better algorithm for keeping track of all stack
manipulations which means that the stack maps it assigns are much more
precise.
2014-09-08 14:54:17 -07:00
f225bf7f55
compiler.cfg.finalization: reenable scheduling
...
Prolog safepoint appears to work with it enabled
2011-11-23 11:11:27 -08:00
6498bc65ce
compiler: emit safepoints in optimized prologs
2011-11-23 11:11:25 -08:00
c677c35de4
compiler.cfg: fix major facepalm with write barrier elimination
2010-09-05 21:39:45 -07:00
5fc9aa05b0
Revert "compiler.cfg: change linear ordering to place GC call blocks at the end"
...
This reverts commit 544a288192
2010-07-27 23:58:41 -04:00
8adde5360a
compiler.cfg: change linear ordering to place GC call blocks at the end
2010-07-27 13:00:28 -04:00
48e96ef032
compiler.cfg.scheduling: update to support multiple-output instructions
2010-07-15 17:38:34 -04:00
e27adb2830
compiler: re-architect low-level optimizer to allow more than one output value per instruction
2010-07-13 07:40:14 -04:00
a55c8ee671
FFI rewrite part 6: deconcatenatize
2010-07-02 15:44:12 -04:00
e86f434f26
Add GC maps to ##box, ##box-long-long, ##alien-invoke, ##alien-indirect and ##call-gc; remove ##gc-map instruction
2010-06-14 19:39:46 -04:00
4f66732c36
Fixing scheduling for compiler changes
2010-05-04 09:48:16 -05:00
b02208ceb2
Some fixes to merge scheduling
2010-05-03 17:54:46 -05:00
277705345f
Merge branch 'master' of git://factorcode.org/git/factor into s3
...
Conflicts:
basis/compiler/cfg/finalization/finalization.factor
2010-05-03 17:29:03 -05:00
4352902bb6
compiler.cfg.empty-blocks: remove unused pass
2010-05-03 18:23:54 -04:00
5509604ffe
Merge branch 'master' into s3
...
Conflicts:
basis/compiler/cfg/optimizer/optimizer.factor
2010-05-03 17:19:28 -05:00
f5c5d8b44c
compiler: remove flat machine representation and generate code directly from the CFG
2010-05-03 17:34:32 -04:00
43f269e4eb
Register allocation now uses SSA properties to coalesce values with different representations
2010-05-03 17:34:20 -04:00
9b34a4a054
compiler.cfg: remove unused 'reps' slot from compiler.cfg, and re-organize things in preparation for SSA register allocation
2010-05-03 17:34:18 -04:00
Björn Lindqvist
John Benediktsson
Joe Groff
Slava Pestov
Slava Pestov
Daniel Ehrenberg