f4ccfa0f71
sets: moving unique/conjoin/conjoin-at to compiler.utilities.
2017-02-07 18:54:24 -08:00
dc3ee923ca
compiler.cfg: add peephole opt to never untag the input of ##bit-count. fixes #1764
2017-01-04 07:29:08 -08:00
4d5b83c3f5
compiler.cfg.builder.alien.tests: fix test that broke windows
2016-12-01 19:52:02 +01:00
75b6b12d0e
compiler.cfg.*,compiler.tree.*: more docs and tests
2016-11-30 16:36:25 +01:00
79dcce8f39
compiler.cfg.debugger: also print the instruction number
...
it's useful to see when debugging some compiler passes
2016-11-23 18:56:03 +01:00
b89a305ef7
compiler.cfg.linear-scan.assignment: correct representation used when spilling
...
Previously the rep-of word was used. But as the vregs representation can
change, you need to instead use the representation it uas as the spill
point.
2016-11-15 03:49:46 +01:00
b1ac5ae7ca
compiler.cfg.linear-scan: more tests
2016-11-15 03:49:46 +01:00
ffef75f8a7
compiler.cfg.*: more docs
2016-11-15 03:49:46 +01:00
4d9367249b
compiler.cfg.linear-scan.*: more tests
2016-11-15 03:49:46 +01:00
7053c72aa1
compiler.cfg.linear-scan.allocation.spilling: <linked-hash>
2016-11-15 03:49:46 +01:00
eeeb65d7d8
compiler.cfg.stacks-docs: fix a typo
2016-10-30 14:18:53 -07:00
a102f7ad5d
compiler.cfg.*: bunch of doc updates
2016-10-08 07:08:00 +02:00
ab7e077b1d
Docs: fixing docs & comments to work with new stack clearing
2016-09-12 02:11:51 +02:00
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
2f71e91dce
compiler.cfg.stacks.*: rename trace-stack-state2 -> trace-stack-state
2016-09-09 08:06:02 +02:00
9decb6a91e
compiler.cfg.*: better way to store the block height
...
It is stored as a height-state instance on the height slot of the
basic-block. It will make better analysis possible because you see how
much the height increased or decreased in the block.
2016-09-08 04:01:25 +02:00
670d2c344e
compiler.cfg.stacks.*: new word local-loc>global, replacing untranslate-loc
2016-09-08 04:01:25 +02:00
6f06b51443
compiler.cfg.stacks.local: simplified code for getting kill locations
2016-09-08 04:01:25 +02:00
c640e3b8c1
compiler.cfg.*: changing height-state from a 2-el array to a tuple
...
This way, it is a little easier to see what is going on in the local
analysis phase.
2016-09-08 04:01:25 +02:00
2edda80b5e
compiler.cfg.*: removing the adjust-d word
...
inc-stack can be used in emit-call-block if we check if the block is a
kill block in end-local-analysis
2016-09-05 12:12:01 +02:00
ee5c28f470
compiler.cfg.stacks.*: removing the adjust word which wasn't useful
2016-09-05 10:37:29 +02:00
367bff6339
compiler.cfg.builder: simplification of the shuffling logic
2016-09-04 06:22:54 +02:00
af7e4903a2
compiler.cfg.debugger: print -> write to not add stray newlines
2016-09-01 09:59:15 +02:00
26d53f9019
compiler.cfg.*: compiler doc updates
2016-09-01 02:16:30 +02:00
f7542e95a9
compiler.cfg.stacks.finalize: only insert blocks if they aren't empty
...
Previously, a lot of empty blocks were added too.
2016-09-01 01:54:11 +02:00
c26379ea7e
compiler.cfg.debugger: better code for outputting using the formatting vocab
2016-08-30 04:04:21 +02:00
b82ea14af0
sets,compiler.cfg.stacks.global: putting the refine word in sets
2016-08-30 02:34:37 +02:00
6c88577ee7
compiler.cfg.*: replacing the peek/replace/kill-sets with slots
...
So instead of storing the info in variables, it is stored in slots on
the basic-block tuple which is much nicer.
2016-08-29 12:07:47 +02:00
e8e1811542
compiler.cfg.stacks.height: removing vocab
...
It was so small so better to put the words it contained in the vocabs
where they are used.
2016-08-26 11:06:17 +02:00
f0834e7b36
compiler.prettyprint: new vocab for prettyprinting compiler types
...
The require-when ensures that the prettyprint methods are loaded when
both the compiler and prettyprint is loaded.
2016-08-15 15:16:41 +02:00
a36c0cb5ac
compiler.*: moving all words only relevant for testing to compiler.test
2016-08-12 15:13:51 +02:00
2a19e00ff3
compiler.cfg.test-words: vocab is unused so we can remove it
2016-08-12 14:06:19 +02:00
435a4e29b9
compiler.cfg.stacks.local: using compiler.cfg.debugger to ensure consistent prettyprint output
2016-08-12 13:55:10 +02:00
ddc5ece757
alien.*: frontend varargs support! #1677
...
alien-invoke gets an extra parameter indicating if the call uses varargs
or not. In 99.9% of the cases, the parameter should be f, but is t if
varargs are indicated. E.g
: do-printf ( fmt d -- st ) int f "printf" { c-string double } t alien-invoke ;
2016-08-10 21:34:37 +02:00
7ab3ebfdd5
compiler.cfg.builder.alien: simpler caller-stack-frame word
2016-08-10 21:34:37 +02:00
24a02a1c8f
compiler.*: Backend implementation of varargs
...
It is turned off by default. Support for using it coming soon. :)
2016-08-08 12:02:19 +02:00
a79309680c
compiler.*, stack-checker.*: fixes to make the tests work on 32bit too
2016-08-05 16:44:19 +02:00
1b08ef3593
compiler.cfg.builder.*: fix for tests that broke on windows
2016-08-05 01:01:25 +02:00
e67745aaba
stack-checker.*: removes the in-d and out-d slots from the alien-node-params tuple
...
That data is already on the #alien-node tuple so it doesn't need to be
stored twice.
2016-08-04 20:30:58 +02:00
5794ca6834
compiler.cfg.builder.alien.tests: fix for the test case i broke
2016-08-02 01:29:19 +02:00
af62d33ba9
compiler.cfg.*: lots of new tests
2016-08-01 21:52:06 +02:00
2e332f73b5
compiler.*: new docs
2016-08-01 21:52:05 +02:00
e359f3fca4
compiler.cfg.builder.alien.params: refactors the next-reg-param and reg-class-full? words
2016-08-01 21:52:05 +02:00
8aef16bed2
compiler.cfg.builder.alien.*: merge of with-param-regs and with-param-regs*
2016-08-01 21:52:05 +02:00
2fd5654473
Add EOL at EOF for all authors.txt files for consistency
...
In some instances replaced CR/LF with LF.
2016-07-30 09:25:30 -07:00
75a3e191b2
compiler.*: a bunch of random doc additions
2016-06-21 23:15:20 +02:00
2fe26444aa
compiler.cfg.builder.*: docs and tests
2016-05-24 17:22:38 +02:00
249483fc4f
compiler.cfg.value-numbering.*: new docs
2016-05-24 17:21:27 +02:00
e60d45a169
Docs: fixing help-lint failures
2016-05-20 13:35:08 +02:00
27f9613694
compiler.cfg.value-numbering.comparisons: some duplicated words that can
...
be removed
2016-05-19 16:15:36 +02:00
John Benediktsson
Jon Harper
Björn Lindqvist
Alexander Iljin