27f9613694
compiler.cfg.value-numbering.comparisons: some duplicated words that can
...
be removed
2016-05-19 16:15:36 +02:00
3810889f91
compiler.cfg.value-numbering.graph: docs and test
2016-05-19 16:10:06 +02:00
1eaa151a3c
compiler.cfg.value-numbering.comparisons: docs and tests
2016-05-19 14:04:23 +02:00
05de15bdc1
compiler.cfg.utilities: rewrite apply-passes as a macro
...
It makes for more detailed callstacks when profiling.
2016-05-18 23:49:03 +02:00
30451ebf5a
compiler.cfg.*: new slots ds-height and rs-height
...
They are used when recording the blocks stacks height instead of the
ds-heights and rs-heights variables. It makes the code a bit simpler.
2016-05-03 01:16:50 +02:00
b8e0d5e278
compiler.cfg.builder.*: the when* checks aren't needed
2016-04-30 09:28:30 +02:00
70e9834f2f
compiler.cfg.*: new system for emitting spill/reloads for gc calls ( #1471 )
...
GC calls were treated as sync points so all registers were spilled
around them. But if the gc call isn't triggered, it is unnecessary to
spill. This commit fixes that by handling ##call-gc specially and
putting ##spill and ##reloads in the same block as the gc call itself.
2016-04-22 19:21:15 +02:00
d5cb972a71
Docs: ui and compiler
2016-04-22 18:23:02 +02:00
720edcbd3b
style fixes - indenting slots and word bodies
2016-04-22 13:06:41 +02:00
92a85ebe86
compiler.*: more compiler docs
2016-04-22 12:57:54 +02:00
3e14ceafe2
compiler.*: more compiler tests
2016-04-22 12:44:25 +02:00
e58e1b9a65
compiler.cfg.ssa.interference.live-ranges: Docs
2016-04-22 12:33:39 +02:00
cf74a4dc0f
compiler.cfg.linear-scan.allocation.splitting: Docs
2016-04-22 12:16:46 +02:00
951c0586f0
compiler.cfg.linear-scan.allocation: removes no-free-registers?
2016-04-22 12:04:22 +02:00
331483ab98
classes.algebra: removing valid-classoid? (enforced in constructors).
2016-04-13 14:29:24 -07:00
e426e5c339
compiler.cfg.linear-scan.assignment: better vregs>regs
2016-04-11 01:17:13 +02:00
754f52d399
compiler.cfg.linear-scan.assignment: refactoring of assign-registers-in-block + tests
2016-04-11 01:17:13 +02:00
4bfcc86cc4
compiler.tests.stack-trace: use with-test-file.
2016-04-04 15:00:04 -07:00
e4ddd9f38e
compiler.cfg.linear-scan.allocation.spilling: fix $maybe docs.
2016-04-04 10:40:02 -07:00
2cc8b5e7de
compiler.cfg.linear-scan.allocation.spilling: fix stack effect
2016-04-03 20:56:30 +02:00
89e98e96a4
compiler.cfg.linear-scan.live-intervals: hairy- and clober-insn moved to
...
this vocab
2016-04-03 20:56:30 +02:00
1a71407112
compiler.cfg.linear-scan.assignment: refactoring + tests
...
better expire-old-intervals, activate-new-intervals and assign-gc-roots
2016-04-03 20:56:29 +02:00
86606213de
compiler.cfg.linear-scan.assignment: refactoring
...
Better definitions for expire-old-intervals, activate-new-intervals and
assign-registers-in-insn.
2016-04-03 20:56:29 +02:00
54d7b50d1b
compiler.cfg.linear-scan.live-intervals: better way to write insn>sync-point
2016-04-03 20:56:29 +02:00
3ef33708ae
compiler.cfg.linear-scan.assignment: dont think prepare-insn should be called twice here
2016-04-02 17:48:23 +02:00
026abb9c38
compiler.cfg.linear.scan.*: makes it so ranges and uses are guaranteed
...
to be vectors
Simplifies the code and makes it a little faster.
2016-04-02 17:48:23 +02:00
95c1b6bbe3
compiler.cfg.linear-scan.live-intervals: doc lint fixes
2016-04-01 23:42:00 +02:00
a60f122c5e
basis.compiler.cfg.linear-scan.*: fixing boostrap problems due to the
...
word rename
Sorry for the noise.
2016-04-01 20:16:13 +02:00
eb68ad6932
compiler.cfg.linear-scan.allocation.splitting: fix HINTS: specialization
...
for split-interval
2016-04-01 19:38:45 +02:00
52e78441f8
compiler.cfg.linear-scan.live-intervals: simplify uses-vregs*
2016-04-01 19:16:35 +02:00
76be7b54ed
compiler.cfg.linear-scan.*: removed start and end from live-interval-state
...
They weren't needed because start>> was always equal to ranges min and
end>> ranges max
2016-04-01 19:16:35 +02:00
46f781a56f
compiler.tests.stack-trace: remove temp-file usage.
2016-03-31 07:37:50 -07:00
41209a3a42
summary.txt: Delete empty lines and empty files.
2016-03-30 22:11:51 -07:00
391f8e433f
basis: Cleaning up tests using lists and IN: forms.
2016-03-30 20:04:14 -07:00
1721f9e314
fix some tests for the sets:set change.
2016-03-29 20:43:46 -07:00
71ef8a22c2
disambiguate namespaces:set and sets:set.
2016-03-29 17:14:42 -07:00
e7a42b1dc4
sets: rename unordered-set back to set.
2016-03-29 15:46:29 -07:00
3f18b71d07
cleanup some QUALIFIED: that are no longer needed.
2016-03-29 10:20:16 -07:00
ff99f69f02
compiler.cfg.save-contexts: fix $link markup
2016-03-27 23:47:08 +02:00
07aee7525f
compiler.cfg.save-contexts: ##save-context isn't need in front of ##call-gc
...
The reason is because the minor-gc primitive itself saves the
context. Not inlining context saves, saves 20 bytes/call site.
2016-03-27 19:21:50 +02:00
63ef600aa3
compiler.cfg.*: indentation fixes
2016-03-27 19:21:50 +02:00
d3bc2035a2
factor: remove rest of double paren words.
2016-03-25 03:13:27 -07:00
6e035c3640
compiler.tests: fix a few KERNEL-ERROR.
2016-03-22 15:22:15 -07:00
9cc73c9492
graphs: move to core and simplify.
...
1. remove add-vertex* and remove-vertex* that contradicted the documentation.
2. graphs use hash-sets of edges instead of hashtables of {edge,edge} pairs.
3. make graphs:closure work like classes:closure, use in classes.
2016-03-21 10:20:38 -07:00
bce3a88287
stack-checker.known-words: fix input class declaration for <string>
...
it should be { integer-array-capacity integer } ofcourse
2016-03-19 18:59:19 +01:00
6f9ff8813e
compiler.*, stack-checker.known-words: adds the integer-array-capacity
...
This commit adds the integer-array-capacity declaration on a few
words. It should make loopy code compile faster and fix #1339
2016-03-18 20:46:41 +01:00
be4484d708
compiler.tree.*: fixes the propagation and dead-code removal logic so it
...
becomes aware of integer-array-capacity
2016-03-18 20:39:28 +01:00
a3ce61f8da
compiler.tree.cleanup: more docs and a test
2016-03-18 19:21:31 +01:00
6576660069
compiler.tree.dead-code.simple: tests and more docs
2016-03-18 19:21:31 +01:00
707bc801a4
compiler.cfg.stacks.local: replace-sets values must be hash-sets ( #1507 )
...
The compiler hang was because { D: 3 D: 2 } was unequal to { D: 2 D: 3
}, so using a set should fix that.
2016-03-16 16:32:50 +01:00
db9093dd5c
compiler.cfg.stacks.*: new docs
...
Mostly from making comments into doc definitions
2016-03-16 15:21:42 +01:00
f80513cd99
compiler.cfg.dataflow-analysis: removes the -analysis from the dfa token
...
It makes the FORWARD-ANALYSIS: and BACKWARD-ANALYSIS: functors less
magical if the generated singletons name is the same as in the syntax.
2016-03-16 14:40:58 +01:00
4a472d7882
compiler.cfg.builder.blocks: return the slots, they are absolutely
...
needed for bootstrap
2016-03-16 14:08:49 +01:00
f906f1c282
compiler.cfg.*: the make-kill-block word is not needed
2016-03-16 12:26:03 +01:00
5ca571e304
compiler.cfg.builder.blocks: some cleanup
2016-03-16 11:48:31 +01:00
3a9b297b3b
compiler.cfg.*: better stack effect for end-stack-analysis
2016-03-16 10:53:33 +01:00
d62240a0a2
compiler.codegen.relocation.tests: fix test so it works on windows too
2016-03-13 07:37:49 +01:00
db4c4665f3
Docs: doc updates
2016-03-13 07:37:20 +01:00
076abf04ce
compiler.*: new tests
2016-03-09 23:24:43 +01:00
b51bd79819
compiler.cfg.*: new doc files
2016-03-09 23:24:43 +01:00
442827ade9
compiler.cfg.intrinsics.fixnum: more docs and tests
2016-03-09 23:24:43 +01:00
b27634bff7
compiler.cfg.builder.blocks: removes assignments to basic-block variable
...
Now it isn't used at all in the cfg creation step.
2016-03-09 23:24:43 +01:00
9df955e199
compiler.cfg.*: big changes which removes the last basic-block uses
...
All intrinsic code generating words signatures are changed from ( node
-- ) to ( block node -- block' ) so the current block is now always
passed on the stack.
2016-03-09 23:24:42 +01:00
482998974b
compiler.cfg.*: more refactoring to remove basic-block get:s
...
Now almost all words pass around the current basic block on the
stack. Left is updating all intrinsics.
2016-03-09 23:24:42 +01:00
f9c6d7cc43
compiler.cfg.*: changes to remove basic-block variable
...
The emit-node generics signature is changed to ( block node -- block' )
so that it always returns the next block to operate on. Signature for a
lot of related words are changed similarily. Now there is only a few
basic-block usages left.
2016-03-09 23:24:42 +01:00
07adc2ecae
compiler.cfg.*: changed generic emit-node to take the current block as
...
stack parameter
The idea is to eventually completely remove the basic-block dynamic
variable from cfg construction.
2016-03-09 23:24:42 +01:00
5663d39a76
compiler.*: more docs
2016-02-28 01:01:38 +01:00
7247906bf8
vm: remove vm-field-offset and context-field-offset
...
Shorter to just write "vm offset-of" and "context offset-of"
2015-12-14 09:29:18 +01:00
3affad7681
docs: using $maybe.
2015-12-11 17:05:45 -08:00
24db8a5d7f
compiler.tests.callstack-overflow: disable test for os x 64, see #1478
2015-12-08 02:16:51 +01:00
a35895f222
compiler.cfg.intrinsics: fix doc and make tests work on 32bit
2015-12-01 13:43:45 +01:00
02c80423ba
compiler.cfg.intrinsics.slots: refactor + docs & tests
2015-11-28 01:55:12 +01:00
6a2c36f5ec
compiler.cfg.builder.blocks: help-lint fix
2015-11-23 16:32:01 +01:00
63fd4d25cf
compiler.cfg.*: more fixes to pass basic-block on the stack than in a
...
dynamic variable
2015-11-22 01:06:11 +01:00
1421779c9e
compiler.cfg.*: refactors words to take a block parameter instead of
...
using the basic-block dynamic variable
the idea is to make the code easier to follow by limiting the use of
dynamic variables
2015-11-22 00:37:34 +01:00
e1b22e0af4
compiler.tree.dead-code.simple: small refactoring and docs
2015-11-22 00:37:33 +01:00
097b40ce41
compiler.tree.propagation.recursive: better counter-class
2015-11-22 00:37:33 +01:00
a3be724f5a
compiler.tree.propagation.info: better code for wrap-interval
2015-11-16 12:30:24 +01:00
7b18618b3e
compiler.cfg.linear-scan.allocation.state: fixes the the 64-bit-o-centric testcases
2015-11-04 13:54:14 +01:00
bb6fa7267c
compiler.cfg.linear-scan.allocation.state: help-lint fixes
2015-11-04 13:38:59 +01:00
f692200547
compiler.cfg.*: refactoring which removes spill-area-size and
...
spill-area-align from cfg
This makes the code for building the stack frame simpler.
2015-11-04 07:53:19 +01:00
cfe4fec574
compiler.cfg.linear-scan.*: make it so cfg is given as an argument to next-spill-slot
2015-11-04 07:53:19 +01:00
4890239c3c
compiler.tree.propagation.tests: #1370 specific test
2015-09-22 09:03:16 +02:00
804348287c
Docs: fix a bunch of help-lint warnings
2015-09-22 08:52:59 +02:00
c5b92cedcb
compiler.*: even more compiler docs
2015-09-22 08:52:59 +02:00
4db173cadb
compiler.*: random doc and tests updates
2015-09-22 08:52:59 +02:00
3e485652fe
compiler.tree.propagation.slots: looks like you can do without the
...
constructor-output-class word
2015-09-22 08:52:59 +02:00
09ad5bd617
compiler.tree.propagation.slots: docs & tests for the vocab
2015-09-22 08:52:59 +02:00
4ee88bdc3d
compiler.tree.propagation.info: better logic for comparing value-infos
...
Now the slots of the value infos only matter if the class of the first
info is at least as large as the class of the second info
2015-09-22 08:52:37 +02:00
bc0cdff2c6
compiler.tree.propagation.info: fixed value-info<= so that it handles
...
slots correctly
if a slot is defined for a value-info, then it can't be <= a value-info
without that slot defined
2015-09-22 08:52:37 +02:00
86d35982ac
compiler.cfg.linear-scan.allocation.splitting: better def for split-uses
...
it seems to be just as fast as the wordier one
2015-09-22 08:52:37 +02:00
ea67379751
compiler.tree.*: new tests
2015-09-22 08:52:37 +02:00
b9ce4910af
compiler.tree.*: docs
2015-09-22 08:52:36 +02:00
aaef9af863
compiler.cfg.*: random doc updates
2015-09-22 08:51:46 +02:00
a1153e6de1
compiler.tree.propagation.info: more docs and tests for the vocab
2015-09-22 08:51:46 +02:00
bb8b77419f
compiler.tree.propagation.nodes: the loop-nesting variable appears unused
2015-09-22 08:51:05 +02:00
44092bf78d
compiler.tree.propagation.*: better to use 2array than the "cuter" /\ word
2015-09-22 08:51:05 +02:00
9369f8216f
compiler.tree.propagation.constraints: docs for this vocab
2015-09-22 08:51:04 +02:00
508ddc5c0b
compiler.cfg.linear-scan.ranges: merge adjacent ranges when adding them
...
this way, the number of live interval ranges you have to deal with is
much smaller
2015-09-22 08:51:04 +02:00
89e261f673
compiler.cfg.linear-scan.ranges: remove the live-range definition and
...
use integer pairs instead
so instead of ranges being a sequence of live-ranges it is now a
sequence of integer pairs instead. That makes the
compiler.cfg.linear-scan.ranges simpler and easier to generalize
2015-09-22 08:51:04 +02:00
2943e7f246
compiler.cfg.linear-scan.ranges: new word ranges-endpoints and remove relevant-ranges
2015-09-22 08:51:04 +02:00
8c781639ed
compiler.cfg.linear-scan.ranges: move more range-related features into
...
this vocab
2015-09-22 08:51:04 +02:00
fb45bc278f
compiler.cfg.linear-scan.live-intervals: the reg-class slot can be
...
removed
it's always the same as vreg>> rep-of reg-class-of so no need to store
that data twice
2015-09-22 08:51:04 +02:00
68c6654763
compiler.cfg.linear-scan.*: move words for splitting and intersecting ranges to the compiler.cfg.linear-scan.ranges vocab
2015-09-22 08:51:04 +02:00
b8d47ed845
compiler.cfg.linear-scan.ranges: new vocab to contain all the range
...
related stuff from live intervals
2015-09-22 08:51:03 +02:00
d0aae5cc20
compiler.cfg.linearization: number-blocks can set the number>> accessor
...
instead of using the numbers dynamic variable
2015-09-22 08:51:03 +02:00
e477f6996f
Fix comments to be ! not #!.
2015-09-08 16:15:10 -07:00
bc7f9ee669
VM: always use undecorated names when loading ffi functions
...
For win32, Factor tries four different function names when loading
stdcall and fastcall functions, in case decorated names are used in the
dll. It seems to not be necessary because a dll meant for 3rd party use
will always export undecorated
names (http://blogs.msdn.com/b/oldnewthing/archive/2004/01/12/57833.aspx ).
2015-09-02 21:54:41 +02:00
9725169ca7
compiler.cfg.intrinsics.misc: use the vm-special-object-offset word
2015-09-02 21:54:36 +02:00
22887c5816
VM: get rid of the odd RT_EXCEPTION_HANDLER relocation
...
The address to the win32 relocation handler can be stored in a special
object and be accessed by jit-install-seh. Doing it that way is better
because you can remove a lot of special code related to that relocation
type.
2015-09-02 21:54:30 +02:00
85e45ae29e
compiler.constants: remove a bunch of constants that are unused
...
(or only used by the defunkt arm backend)
2015-09-02 21:54:24 +02:00
a89b0c2b30
Docs: reorg of the compiler.constants docs a little
2015-09-02 21:54:21 +02:00
883f65d0e4
VM: fix one callstack overflow problem by "unlocking" the callstacks
...
border pages
Also a new vocab compiler.tests.callstack-overflow which is supposed to
contain all tests for callstack overflow-related problems.
2015-08-14 20:19:58 -07:00
bdc4880b17
compiler.cfg.save-contexts: context-save-needed not needed because it's
...
the same as gc-map-insn
2015-08-14 20:19:51 -07:00
0216f9f6d1
namespaces: rename namestack* to (get-namestack) for consistency.
2015-08-13 17:53:07 -07:00
ceb75057da
change ERROR: words from throw-foo back to foo.
2015-08-13 16:13:05 -07:00
ce046b77c8
compiler: rename D to D: because it wants a token. Same for R -> R:
2015-08-13 15:23:10 -07:00
62686907fe
compiler: fix rename
2015-08-13 12:55:28 -07:00
c21a154056
core: callstack is a builtin type and a class name. rename the *stack words that grab the current stack get-foostack to avoid the callstack builtin conflict and for better symmetry with set-foostack.
2015-08-13 10:11:59 -07:00
b6be8685c3
basis: ERROR: changes.
2015-08-13 03:20:39 -07:00
225c28f23a
compiler.tree.propagation: Fix build error
2015-08-13 02:53:03 -07:00
e3ddd337e0
compiler, alien: fix some duplication in ERROR: definitions. hope this doesnt trigger more failing builds than it fixed...
2015-08-12 13:58:38 -05:00
e0d1b14e5c
factor: don't need FROM: namespaces => set or namespaces:set anymore
2015-08-12 11:25:57 -05:00
46587f8457
factor: arch -> arch-name. fix breakage caused by other patch.
2015-08-12 11:08:24 -05:00
28dcd0667a
factor: rename set to unordered-set.
2015-08-12 10:54:19 -05:00
330c7ba156
factor: Removing SLOT: nth/at/global in favor of using change: at for tuple slots named ``at`` and change-at for hashtables.
...
Redo many FROM: and QUALIFIED: and using lists.
2015-08-12 08:49:02 -05:00
00338f62f3
factor: put inline on same line as ; for experimentation
2015-08-10 12:55:27 -05:00
0bdc27053b
compiler: oops, assoc-empty? vs empty?. undo harvest-values here.
2015-08-08 12:27:36 -05:00
76cb13ac52
basis: use harvest-values
2015-08-08 12:24:22 -05:00
43fb73fae6
factor: Rename I[ ]I to [I I]. Fix places where we do "string"token without whitespace.
2015-08-06 17:05:12 -05:00
136d793c6e
compiler.tests: remove some tests that can't work now.
...
check-combination-effect raises an error at parse-time.
2015-08-05 18:01:16 -07:00
22b871d82e
compiler.cfg.utilities: make it so block>cfg initializes spill-area-align
2015-07-29 13:01:24 +02:00
38afea84bf
fix doc using.
2015-07-28 20:20:40 -07:00
4ded3c491c
fix some help-lint errors.
2015-07-28 20:06:59 -07:00
8a731681bf
compiler.cfg.*: tests against #1308
2015-07-28 17:58:29 -07:00
a8c0a25d87
compiler.cfg.ssa.destruction.coalescing: refactor and maybe a fix for #1308
...
the eliminatable-copy? word is the fix. the previous code only checked
that the registers had the same register class, but you also need to
check that they dont have the same representation size. because a copy
from double-rep -> double-2-rep is not eliminatable
2015-07-28 17:58:29 -07:00
b544c1b352
compiler.cfg.*: some extra unit tests for the compiler
2015-07-28 17:58:29 -07:00
820207c5b0
compiler.cfg.ssa.destruction.coalescing: simpler code for setting up the
...
initial leader-map and class-element-map
2015-07-28 17:58:28 -07:00
5e29aac0a6
Docs: various more compiler-related docs
2015-07-28 17:58:28 -07:00
aca610813d
compiler.cfg.representations.selection: use a union instead of a generic
...
to see if an insn is peephole-optimizable + test
2015-07-28 17:58:28 -07:00
44c9b2c8e8
compiler.tree.*: new docs
2015-07-28 17:58:28 -07:00
2c5f00d865
compiler.cfg.*: even more docs
2015-07-28 17:58:28 -07:00
e095b54644
compiler.tree.escape-analysis.*: new doc files
2015-07-28 17:58:28 -07:00
06560b24a6
compiler.tree.escape-analysis.allocations: stub doc file from comments
2015-07-28 17:58:27 -07:00
b4e4792cb6
compiler.cfg.*: a bunch of docs updates
2015-07-28 17:58:27 -07:00
cdf9a8535b
compiler.cfg.representations.selection: some initial docs from comments
2015-07-28 17:58:27 -07:00
31aae02916
compiler.cfg.linear-scan.live-intervals: simpler code for finding the
...
sync points in the cfg
2015-07-28 17:58:27 -07:00
78d5c0a743
compiler.cfg.instructions: ##read -> read-insn and ##write -> write-insn
2015-07-28 17:58:27 -07:00
81c7659360
compiler.cfg.instructions: rename ##allocation to allocation-insn,
...
because it's nicer if all instruction unions have the -insn suffix and
the ## prefix is reserved for tuples
2015-07-28 17:58:27 -07:00
1fcf96cada
factor: remove """ string syntax for now. there are HEREDOC:, STRING:, escaping \", and something else soon.
2015-07-25 22:59:56 -07:00
Björn Lindqvist
John Benediktsson
Doug Coleman