acea55c692
math.vectors: add v+- word which is accelerated by SSE3
2009-09-20 17:43:16 -05:00
47d8763340
More integer SIMD work
...
- move generated vocab support from specialized-arrays to vocabs.generated
- add fuzz testing to math.vectors.simd
- add alien type support for integer SIMD vectors
- SIMD: parsing word generates a SIMD type, instead of pre-generating them all in math.vectors.simd
2009-09-20 16:48:17 -05:00
01d2ef415a
get compiler tests loading
2009-09-16 09:20:47 -05:00
172219e931
compiler.tree.propagation: fix unary-op type functions with complex number inputs
2009-09-15 14:30:20 -07:00
32b95c2cdf
math: add unordered comparison operators u< u<= u> u>= which behave exactly like < <= > >= except no floating point exceptions are set if one or both inputs are NaNs; also add efficient intrinsic for unordered? predicate, and fix propagation type functions for abs, absq, and bitnot
2009-09-12 22:20:13 -05:00
dd661edf46
compiler.tree.propagation: convert /i of a positive integer by a power of two into a shift
2009-09-11 21:03:11 -05:00
7f2e2b1777
Specialized array overhaul
...
- Replace hand-written specialized-arrays.* subvocabularies with new system; instead of USE:ing specialized-arrays.T, do SPECIALIZED-ARRAY: T
- Ditto for specialized-vectors; use SPECIALIZED-VECTOR:
- io.mmap.functor: removed entirely, use <mapped-array> instead
- struct-arrays and struct-vectors have been removed because specialized arrays and vectors subsume them entirely
2009-09-09 22:33:34 -05:00
1e5bc3b376
inline alien-vector and set-alien-vector if SIMD is not available for a small speedup
2009-09-08 13:56:17 -05:00
bbca00e2ae
Fix conflicts
2009-09-07 23:51:25 -05:00
9f9b6bca01
compiler.tree.propagation: type check inputs to unsafe foldable words manually, so that stuff like [ "Hi" { } fixnum+fast ] doesn't crash in the compiler
2009-09-07 23:40:23 -05:00
a1ae209f81
compiler.tree.propagation.call-effect: stronger call( inlining; now can inline 'a [ b ] curry call(' where 'a' is literal, [ b ] doesn't infer, but [ a b ] does infer. This simplifies classes.struct:memory>struct
2009-09-07 17:45:03 -05:00
e36a0d7ef4
compiler: clean up code generation for alien boxing/unboxing a bit
2009-09-03 21:22:43 -05:00
4d5a4222b6
More SIMD work
...
- Rename SIMD types and register representations: <type>-<count> rather than <count><type>-array
- Make a functor to define 256-bit vector types, use it to define float-8 type
- Make SIMD instructions pure-insns so that they participate in value numbering
2009-09-03 20:58:56 -05:00
bf81cb4259
math.vectors.simd: split off intrinsics into a sub-vocabulary, to avoid loading most of the SIMD code on bootstrap
2009-09-03 03:43:43 -05:00
ae051e0c9a
compiler.tree.propagation: type function for clone had an issue, sometimes clone would get optimized out because of incorrect constant folding
2009-09-03 02:40:18 -05:00
ff8c70dbe0
Initial implementation of SSE vector intrinsics:
...
- cpu.architecture: add SSE vector representations
- compiler.cfg.intrinsics.alien: remove an attempt at optimization that value numbering handles now
- compiler.cfg.representations: support instructions where the representation is set in the 'rep' slot, and support conversions between single and double floats
- alien-float, set-alien-float now use the single float representation, and the conversion is implicit; this fixes a long-standing bug where a register could get clobbered because of how %set-alien-float was defined on x86
- math.vectors.specialization: add support for SIMD specialization (where the vector word's body is replaced by another quotation), also specialize the 'sum' word
- math.vectors.simd: 4float-array, 2double-array, 4double-array types, and specializers for the math.vectors words
2009-09-03 02:33:07 -05:00
2bb6293217
compiler: add fixnum-min/max intrinsics; ~10% speedup on benchmark.yuv-to-rgb
2009-08-28 19:02:59 -05:00
d957ae4e44
Performance improvements to make struct-arrays benchmark faster
...
- improved optimization of ##unbox-any-c-ptr on ##box-displaced-alien; convert it to ##unbox-c-ptr where possible using class info stored in the ##bda instruction
- make fcos, fsin, etc inline again; everything in math.libm inline again, except for fsqrt which is an intrinsic
- convert min and max on floats to float-min and float-max
- make min and max not inline, so that the above can work
- struct-arrays: rice a bit so that more fixnums come up
2009-08-28 05:21:16 -05:00
ba0f3a9911
compiler.tree.propagation.transforms: don't fail to compile if 'at' called on something that's not an assoc
2009-08-27 18:57:56 -05:00
0df8aadce2
cpu.x86: use SQRTSD instruction for math.libm:fsqrt word
2009-08-25 23:22:15 -05:00
f01f7ad6eb
compiler.tree.propagation: bitand custom inlining was wrong if the second input was a bignum
2009-08-20 03:47:07 -05:00
2dc99ea05f
Fix interval inference of abs, absq when input is a complex number
2009-08-19 16:06:37 -05:00
829107902e
compiler.tree.propagation: improved interval inference for absq eliminates a conditional from math.vectors:distance. Type inference for rational math also a bit sharper now
2009-08-19 02:33:41 -05:00
3047d4a451
compiler.tree.propagation: remove method inlining heuristic
2009-08-17 22:29:05 -05:00
7c92ab1ea5
move if-zero etc to math, remove 1-/1+ from math
2009-08-14 14:27:23 -05:00
3f3d57032b
Delete empty unit tests files, remove 1- and 1+, reorder IN: lines in a lot of places, minor refactoring
2009-08-13 19:21:44 -05:00
4ecf5a904a
More accurate wrap-interval in compiler.tree.propagation.info fixes test regression; constructing an interval with endpoints at infinity now outputs full-interval
2009-08-12 03:25:53 -05:00
7e35723db0
compiler.tree.propagation: be more careful with intervals, ensuring that the inferred interval of a value is a subset of the value class's interval. This improves accuracy, for example [ >fixnum 1 + >fixnum most-positive-fixnum <= ] constant-folds to true
2009-08-10 01:16:49 -05:00
8a9c15ab0b
compiler.tree.escape-analysis: if the output of an #introduce node has an immutable tuple class type declaration, and it is not passed to any subroutine calls, or returned from the word, then unbox it. This speeds up vector arithmetic words on specialized arrays, because the specialized array is unboxed up-front, eliminating an indirection on every loop iteration
2009-08-09 16:29:21 -05:00
e400d80d8b
More accurate interval inference for mod, rem, and propagation can now infer intervals in the case where a value might be f. so, [ [ 127 bitand ] [ drop f ] if dup [ 0 >= ] [ not ] if ] now constant-folds down to 't'!
2009-08-08 23:03:45 -05:00
61ea749bb6
More accurate interval-mod and interval-rem
2009-08-08 22:01:12 -05:00
c3d60e5899
compiler.tree.recursive: more accurate loop detection
2009-08-04 19:18:40 -05:00
82c1106945
compiler.cfg.ssa.destruction: new implementation: simpler and more correct
2009-08-02 10:26:52 -05:00
e01fc93812
compiler.tree.propagation.call-effect: fix case where quotation inlining could enter an infinite loop: [ dup call( quot -- ) ] dup call( quot -- ) etc
2009-08-01 23:34:14 -05:00
501629cf75
compiler.tree.propagation.info: fix load error
2009-07-27 00:30:24 -05:00
fc41140720
compiler.cfg.utilities: move value-info utilities to compiler.tree.propagation.info
2009-07-26 21:10:33 -05:00
1d735bc32d
compiler.tree.propagation.transforms: fix shift transform
2009-07-17 00:02:57 -05:00
3fb4fc1bde
Improve code generation for shift word: add intrinsics for fixnum-shift-fast in the case where the shift count is not constant, transform 1 swap shift into a more overflow check with open-coded fast case, transform bitand into fixnum-bitand in more cases
2009-07-16 23:50:48 -05:00
062e33f8fb
fixing stupid bug in propagation
2009-07-16 00:43:54 -05:00
ee3e84a1f8
define-partial-eval framework in propagation pass makes it easy to add transforms; moving some transforms from stack checker to propagation, making them stronger
2009-07-16 00:34:50 -05:00
bb06facb01
new is inlined in the propagation pass when the class is known
2009-07-14 14:16:39 -05:00
47500fad06
call( and execute( inline known quotations/words in the propagation pass
2009-07-14 01:12:45 -05:00
e0d84eb3a2
compiler.tree.propagation: better length propagation
2009-07-09 02:28:30 -05:00
407377fc98
rename peek -> last and update all usages
2009-05-25 16:38:33 -05:00
a54c78007b
add a clamp word to math.order, use clamp word throughout libraries
2009-05-24 21:35:50 -05:00
66a9416473
Update documentation for stricter vocabulary search path semantics
2009-05-16 00:29:21 -05:00
fb6df472a2
Cleaning up USING: lists for new strict semantics
2009-05-14 16:54:16 -05:00
b67b617746
Fix compiler breakage if a GENERIC# has an incorrect stack effect declaration
2009-05-12 21:23:52 -05:00
466533d509
Fix overly-eager strength reduction for mod, and add a type function for >integer (reported by Joe Groff)
2009-05-07 12:32:06 -05:00
029d93a838
Make walker work better with call( and breakpoints which are nested inside combinators
2009-05-05 09:12:32 -05:00
f2ec59d658
Fix infinite loop when compiling a word containing a tuple literal with circular structure in it. This was triggered by call( inline caching in core-foundation.fsevents on Mac OS X
2009-05-01 09:36:53 -05:00
e5cdb7ac2d
call( fast-path now supports curry and compose
2009-04-30 21:08:29 -05:00
964fbd0a24
Fix test failures from ratio/complex built-in removal
2009-04-30 00:38:55 -05:00
fc4894fbdf
Replace ratio and complex built-in types with tuples defined in the library. This frees up two lo-tags, so move array and quotation over to these tags and update compiler for new tags
2009-04-30 00:27:35 -05:00
3dc9fdf9db
Fleshed out new dispatch code
2009-04-24 20:43:01 -05:00
057f75e9a1
Refactor compiler.tree.builder to fix various regressions
2009-04-21 23:02:00 -05:00
469c9ee21d
Debugging stack checking
2009-04-21 16:09:53 -05:00
05f3f9dcb9
Fixing unit tests for stack effect inference changes
2009-04-20 21:15:19 -05:00
8414693142
stack-checker: trust word declarations instead of recursively checking them
2009-04-20 18:44:45 -05:00
ec49307c88
Never inline default methods, and fix inlining of methods with hints
2009-04-20 03:27:18 -05:00
dba4c0d589
fixing compiler errors in basis
2009-04-17 12:46:04 -05:00
0a5b076c11
Make the walker infer
2009-04-16 23:14:11 -05:00
a4e62dfdba
Fix for math.parser syntax change
2009-04-13 14:47:39 -05:00
37e278ed02
Use [ ] [ ] map-reduce instead of unclip [ ] reduce
2009-04-06 17:50:54 -05:00
932d44cab8
Small speedup for code using H{ } clone and with-scope
2009-03-31 08:16:04 -05:00
be4fb1e7d9
Move call( and execute( to core
2009-03-16 20:11:36 -05:00
4fdb5d0557
Fix infinite fixed point iteration bug found by littledan; generalize-counter-interval wasn't called in all the right places
2009-03-12 17:30:24 -05:00
42224eb4e7
Propagation pass: add inlined node counts to total node count, so that the more we inline the less we are eager to inline more
2009-03-06 23:34:01 -06:00
03cb32ad9c
Make execute( faster, add execute-unsafe( and make effect tuple slots read only
2009-03-01 20:12:35 -06:00
78ce670101
Merge branch 'for-slava' of git://git.rfc1149.net/factor
2009-02-28 14:42:56 -06:00
bda8b2dda6
Better inlining for both-fixnums?
2009-02-26 14:11:26 -06:00
da2ebe21a6
Fix two overzealous [ ] removal in tests
2009-02-18 03:06:14 +01:00
087d931c36
Remove the tail argument from do/until/while
2009-02-18 02:41:14 +01:00
6b25e99470
Add summary for heaps more vocabs
2009-02-16 21:05:13 -06:00
b65b88364c
Updating lots of things to use call( -- )
2009-02-08 22:12:11 -06:00
3c408342ef
renaming: contain? -> any?, deep-contains? -> deep-any?, pad-left -> pad-head, pad-right -> pad-tail, trim-left -> trim-head, trim-right -> trim-tail
2009-01-29 22:19:07 -06:00
990513db60
use CONSTANT: in basis
2009-02-22 19:13:08 -06:00
bf5ff11720
Update usages of all-equal? now that its in the grouping vocab
2009-01-16 15:14:30 -06:00
e661c67189
refactor a word to use smart combinators
2009-01-08 23:09:38 -06:00
34792a9f23
Remove >r/r>
2008-12-17 19:17:37 -06:00
c5160d76e9
Merge qualified, alias, symbols, constants into core
2008-12-17 18:10:01 -06:00
60a1378a0a
Faster dynamic variable lookup: 6% speedup in benchmark.fib5
...
- If type of first input of equal? is known but not second, rewrite it as swap equal? and try to inline again
- Fix hints on methods to work better
- Put hints for word and fixnum keys on at* and set-at on hashtables
2008-12-17 14:57:24 -06:00
ec23584f48
Finish up O(1) byte array allocation
2008-12-09 18:17:04 -06:00
a90118da5d
Add inc-at word to core, and update some usages of at+ to use it instead
2008-12-09 16:54:48 -06:00
640b37cb70
More permissive
2008-12-08 19:14:18 -06:00
e4f8448eb1
Fix some problems with arithmetic type inference, exposed by recent changes to log2 word
...
- declared input type for bignum-shift was stricter than the runtime behavior, leading to bad propagation of type info if shift count was a bignum
- types inferred for type functions which used number-valued/integer-valued/real-valued were not always precise, eg bignum bignum bitxor => integer
- add interval-log2, type function for (log2)
- remove math-class-min, it was useless
2008-12-07 19:44:49 -06:00
145b635eb6
More optimization intended to reduce compile time. Another 10% speedup on compiling empty PEG parser
...
- new map-flat combinator replaces usages of 'map flatten' in compiler
- compiler.tree.def-use.simplified uses an explicit accumulator instead of flatten
- compiler.tree.tuple-unboxing uses an explicit accumulator instead of flatten
- fix inlining regression from last time: custom inlining results would sometimes be discarded
- compiler.tree's 3each and 3map combinators rewritten to not use flip
- rewrite math.partial-dispatch without locals (purely stylistic, no performance increase)
- hand-optimize flip for common arrays-of-arrays case
- don't run escape analysis and tuple unboxing if there are no allocations in the IR
2008-12-06 11:17:19 -06:00
a56d480aa6
Various optimizations leading to a 10% speedup on compiling empty EBNF parser:
...
- open-code getenv primitive
- inline tuple predicates in finalization
- faster partial dispatch
- faster built-in type predicates
- faster tuple predicates
- faster lo-tag dispatch
- compile V{ } clone and H{ } clone more efficiently
- add fixnum fast-path to =; avoid indirect branch if two fixnums not eq
- faster >alist on hashtables
2008-12-06 09:16:29 -06:00
044e2867d5
Teach compiler about string-nth range
2008-12-05 23:51:58 -06:00
aa838dbc2d
Fix compile errors
2008-12-05 09:04:16 -06:00
6c7005d588
Tweak inlining heuristic
2008-12-05 01:50:14 -06:00
fa6a2047f0
New inlining heuristic: number of usages within this word. Speeds up bootstrap by 10%
2008-12-04 07:05:59 -06:00
e5ed7447ed
Removing more >r/r> usages
2008-12-03 08:46:16 -06:00
d60c79c9bf
Expand instance? with literal class
2008-12-03 08:11:51 -06:00
402da00390
Fixing bootstrap with specialized arrays
2008-12-02 02:44:19 -06:00
e8a69a3aea
Fix conflict
2008-11-29 05:14:49 -06:00
86d45262dc
Add toutput ype propagation for #alien-invoke and #alien-indirect nodes
2008-11-29 03:47:38 -06:00
1fa819191d
Change inlining heuristic so that it behaves the same with dip and >r/r>
2008-11-24 05:46:26 -06:00
c678e6e362
Don't use combinators.short-circuit in compiler
2008-11-21 04:00:37 -06:00
79bffecc2e
Add type function for /mod
2008-11-21 01:54:56 -06:00
2fe364a7bb
Specialized arrays work in progress
2008-11-14 20:18:16 -06:00
f04b32ea02
Remove duplicated logic
2008-11-11 09:10:33 -06:00
5f4b247072
Propagation pass now uses a stack of hashtables for predicated constraints instead of cloning a hashtable. New strategy for recursive propagation; now converges with fewer iterations. ~15 sec bootstrap time improvement
2008-11-11 08:49:00 -06:00
43335d9c36
Intervals now use a special singleton for the full interval to eliminate FP math in the common case; fix a bug in value-info<=
2008-11-11 08:30:14 -06:00
bfd119e3b5
I/O ricing: various hints added so that UTF8 and ASCII fastpaths compile with less dispatch. 25% improvement on reverse-complement
2008-11-06 00:02:44 -06:00
cc879fa9b7
Tuple layouts are now arrays, instead of built-in types. The superclass
...
array is now part of the tuple layout object itself, and class hashcodes
are stored alongside class words there. This removes 2 indirections when
reading a superclass, and 3 when reading a superclass hashcode.
2008-11-05 22:20:29 -06:00
06b99c31ee
Fix regression. If a parsing word called the compiler, it might compile a caller of a generic before the generic gets built, which would throw an error since the inferred effect of the generic might have less inputs than the combination's dispatch#
2008-11-03 20:59:48 -06:00
0cc0a41e57
Update for compiler.intrinsics removal
2008-10-20 20:40:36 -05:00
f092622fac
CFG IR is now pure SSA
2008-10-20 01:56:28 -05:00
c19f2257f4
Fix permission bits
2008-10-02 08:34:49 -05:00
f539406ee1
Fold class predicates applied to literals
2008-10-02 05:12:38 -05:00
1a865f89ac
Some optimizer change sfores for new codegen
2008-09-17 22:18:08 -05:00
01129fb9bd
Add unit test for locals performance regresion
2008-09-13 03:09:16 -05:00
10c68ebb21
New modular arithmetic optimization pass
2008-09-12 18:08:38 -05:00
54232f80ca
Adding identity optimization pass, tweak inlining heuristic
2008-09-12 08:18:44 -05:00
2bf532263d
Oops
2008-09-12 05:36:06 -05:00
40da49bef5
Perform loop detection before normalization, clean up normalization pass, more aggressive recursive return value propagation. Fixes regression on nsieve benchmark
2008-09-12 05:17:27 -05:00
cf80dd122a
Updating code for make and fry changes
2008-09-10 22:11:40 -05:00
8a921c791c
if-empty changes
2008-09-06 19:13:59 -05:00
e1716d92b3
Remove delegation slot
2008-09-03 03:46:56 -05:00
a4a00f6e40
Adding missing optimizations to finalization pass
2008-09-02 22:59:49 -05:00
110a5e5162
Change equality semantics
2008-09-02 02:02:05 -05:00
669e0f8f0a
Propagate slot types of literals
2008-09-01 18:25:21 -05:00
29e5ed2adb
Fixing deploy tool
2008-08-31 19:17:04 -05:00
f5fbd94a4c
New compiled crossref implementation
2008-08-31 01:34:00 -05:00
3cef7c9992
Smarter usage tracking system
2008-08-30 02:31:27 -05:00
1fcbdf9d52
Fix more compiler bugs
2008-08-29 00:26:47 -05:00
74dccc7fbf
Debugging compiler
2008-08-28 22:28:34 -05:00
ea2d02f248
Add some failing tests
2008-08-27 16:25:37 -05:00
8c46a3d56f
Only annotate certain nodes ith value info
2008-08-24 14:02:33 -05:00
8b855b2445
Help lint fixes
2008-08-22 23:00:35 -05:00
40e926609a
Fixing unit tests
2008-08-22 22:07:59 -05:00
2440fc1ceb
Fixing remaining issues
2008-08-22 17:38:23 -05:00
9aa6d8ae04
Compiler fixes
2008-08-22 15:30:57 -05:00
86f11713e3
Fixing some more bugs
2008-08-22 03:12:15 -05:00
aa9341e579
Enforce that usages of >r/r> must be balanced within a quotation
2008-08-18 20:49:03 -05:00
9b6fb70eba
Debugging optimizer
2008-08-18 15:47:49 -05:00
2fa08c75b9
Fix bug, better tests
2008-08-15 21:45:05 -05:00
10b75276ff
Fix optimization of clone, (clone)
2008-08-15 02:49:52 -05:00
e1987d4af9
Debugging new optimizer
2008-08-14 23:35:19 -05:00
a61e13f7be
More aggressive DCE
2008-08-13 23:52:49 -05:00
56d272df9e
Merge branch 'master' into new_optimizer
2008-08-12 03:36:44 -05:00
f2a5a30c6f
New front-end and high-level optimizer lands
2008-08-12 03:22:45 -05:00
Slava Pestov
Joe Groff
Slava Pestov
Doug Coleman
Daniel Ehrenberg
Doug Coleman
sheeple
Samuel Tardieu
Daniel Ehrenberg