3ba79be651
Revert "add a %blend-vector intrinsic for v?"
...
This reverts commit 21e4b28b67
2009-09-30 23:40:37 -05:00
21e4b28b67
add a %blend-vector intrinsic for v?
2009-09-30 23:03:59 -05:00
1c8662ce4a
math.vectors.simd: add vbroadcast intrinsic, fix integer overflow issues
2009-09-29 22:58:20 -05:00
2b13245704
math.vectors.simd: add fast intrinsic for 'nth', replace broadcast primitive with shuffles
2009-09-29 04:48:11 -05:00
a6e8277b2c
math.vectors.simd: add vshuffle intrinsic
2009-09-28 23:12:13 -05:00
ba029a88ff
io.mmap: fix obsolete tests and get code to inline better
2009-09-28 03:18:27 -05:00
10c5fe5933
math.vectors.simd: add hlshift, hrshift (128-bit shift), vbitandn intrinsics
2009-09-28 02:17:46 -05:00
e5b94b11d7
Some fixes and cleanups in math.vectors
...
- Tighten up type inference for operations on complex float arrays
- Fix v. to have correct behavior with complex numbers
- Rename v<< and v>> to vlshift and vrshift to avoid clashing with v>> accessor
2009-09-24 06:58:33 -05:00
a4e1d5511e
Tweaks to reduce deployed image size
2009-09-24 06:24:43 -05:00
24039cb56a
math.vectors.simd: add v<< and v>> intrinsics for bitwise shifts on elements
2009-09-24 03:32:39 -05:00
abac963882
math.vectors.simd: new operations: vabs vsqrt vbitand vbitor vbitxor
2009-09-23 02:47:14 -05:00
fda8870848
Merge branch 'master' into integer-simd
2009-09-22 20:21:40 -05:00
f03b54287e
compiler.tree.propagation.branches: M: #phi propagate-around was unnecessary
2009-09-22 17:56:50 -05:00
2019c0a747
Merge branch 'master' of git://factorcode.org/git/factor into constraints
2009-09-22 16:09:33 -05:00
198fdc82ab
Fixing failing unit tests in compiler.tree.propagation due to constraints
2009-09-22 16:01:14 -05:00
e6cddf1fa7
Merge branch 'master' of git://factorcode.org/git/factor into integer-simd
2009-09-21 17:58:24 -05:00
85c916fedc
Rename specific-method to method-for-class, rename (effective-method) to method-for-object, and make both much faster
2009-09-21 17:42:20 -05:00
66871995c9
math.vectors.simd: add saturated arithmetic operations
2009-09-20 23:16:02 -05:00
78c949b9b7
math.vectors: add v+- word which is accelerated by SSE3
2009-09-20 17:43:16 -05:00
dfb43bd2ca
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
ac5ea1769b
get compiler tests loading
2009-09-16 09:20:47 -05:00
aeb6a01710
compiler.tree.propagation: fix unary-op type functions with complex number inputs
2009-09-15 14:30:20 -07:00
427bfb4ab8
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
10423e0d56
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
527db8995a
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
8eeeeb5c5b
inline alien-vector and set-alien-vector if SIMD is not available for a small speedup
2009-09-08 13:56:17 -05:00
17821626c3
Fix conflicts
2009-09-07 23:51:25 -05:00
b279c5751c
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
0b35d79aee
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
1f5193198b
compiler: clean up code generation for alien boxing/unboxing a bit
2009-09-03 21:22:43 -05:00
20dfbf7ac8
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
9cc705f6ba
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
11aadb74af
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
52b99c050e
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
f30aa5d20e
compiler: add fixnum-min/max intrinsics; ~10% speedup on benchmark.yuv-to-rgb
2009-08-28 19:02:59 -05:00
99bf9fadfb
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
5f33f7306f
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
4fe0257169
cpu.x86: use SQRTSD instruction for math.libm:fsqrt word
2009-08-25 23:22:15 -05:00
b57894a78c
compiler.tree.propagation: bitand custom inlining was wrong if the second input was a bignum
2009-08-20 03:47:07 -05:00
c9cc1fa6a9
Fix interval inference of abs, absq when input is a complex number
2009-08-19 16:06:37 -05:00
75137bafb1
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
b3693e3c3b
compiler.tree.propagation: remove method inlining heuristic
2009-08-17 22:29:05 -05:00
7c3824639e
move if-zero etc to math, remove 1-/1+ from math
2009-08-14 14:27:23 -05:00
d1ce837569
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
6f2170eb02
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
7956e63fc2
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
55d1b76ad7
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
638f5b6579
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
26531ddf8b
More accurate interval-mod and interval-rem
2009-08-08 22:01:12 -05:00
a2f52c1e41
compiler.tree.recursive: more accurate loop detection
2009-08-04 19:18:40 -05:00
Joe Groff
Slava Pestov
Daniel Ehrenberg
Slava Pestov
Doug Coleman