00338f62f3
factor: put inline on same line as ; for experimentation
2015-08-10 12:55:27 -05:00
53efceb0ad
math.integers, comment and simplify bignum/f
...
change the "while" that could only execute once to "when"
change the f/loop word name to "mantissa-and-guard" since it's what it
computes
change the check against 2^53 to be explicit
2015-07-26 12:33:56 -07:00
5424ad5586
math.integers, bignum/f, improve performance.
...
This changes avoids looping many times if the denominator is a power of
2. After this change, the implementation matches the linked sbcl
algorithm. This was probably a mistake done when porting the algorithm.
Basically, as an optimization, all trailing zeros are removed from the
base2 representation of the denominator to have smaller bignums to
divide. But the previous factor implementation didn't take this into
account when making the initial guess of the shift of the numerator to
obtain a result in the range [2^54-1,2^53]. The loop would then correct
the initial guess by a factor of 2 at each iteration, so it would run as
many iteration as the denominator base2 power reduction, instead of only
a few times. For pathological cases, the speed up is huge (10^4):
1 1000 2^ bignum/f
2015-07-26 12:33:55 -07:00
cc6d90a096
math.integers, fix bignum/f for results in ]0x1.0p-1022,0x0.4p-1022]
2015-07-26 12:33:55 -07:00
89e3cc511e
math: moving >fraction to math.
2015-07-19 16:57:45 -07:00
69ee8b9190
core, basis: Don't use the IN: foo.private feature and undocument it.
...
It only works well when you are loading code dynamically.
2015-06-30 08:51:07 -07:00
621b50a8e5
cpu: enable bit-test intrinsic for fixnum-bit?.
2015-06-17 19:11:10 -07:00
db9cfeeae2
math.integers: simplify fixnum-bit?.
2015-06-17 18:42:11 -07:00
ed786070e2
math.integers: fixnum-bit? should not use fixnum-shift-fast.
2015-06-17 08:22:00 -07:00
d60d098c11
math.integers: speedup to fixnum-bit?.
2015-06-07 11:08:57 -07:00
eeaa91d6c1
vm: using bignum>fixnum-strict in integer>fixnum-strict.
2014-06-07 09:46:05 -07:00
5487480126
core: cleanup USING lists.
2013-03-05 10:34:47 -08:00
6d679cf3a2
math.integers: the fixnum-log2 intrinsic didn't work due to inlining.
2013-02-27 17:08:36 -08:00
7ba77673a8
math.integers: faster fixnum-log2.
2013-02-27 16:01:34 -08:00
eaed0db99a
Revert "math: speed up integer>fixnum operations."
...
This reverts commit b467db9a9a
2012-09-16 07:46:22 -07:00
6ea9c0fc02
math: speed up integer>fixnum operations.
2012-09-15 08:52:06 -07:00
089bdb1268
math.integers: inline fixnum-log2.
2012-09-10 17:20:12 -07:00
20100d7853
math.integers: faster fixnum-bit?.
2012-08-30 22:40:22 -07:00
cd9dd9d752
core: Throw an error when assigning a bignum to a fixnum tuple slot if the bignum doesn't fit. Fixes #594 .
2012-08-03 14:59:59 -07:00
6994b0eed8
math.integers: Bignum hashcode* -- call bignum>fixnum to avoid a dispatch.
2012-07-23 09:38:12 -07:00
fb4e3ad9bc
compiler: Fix bitand on ratios, floats. Fix shift on ratios, floats. Add integer>fixnum. Fixes #500 .
2012-07-23 09:31:12 -07:00
98fc1e28bd
math.integers: make bignum/f round to even on tie
...
Fixes #372
2011-11-26 17:44:29 -08:00
90f34d092c
math: neg? needs to be defined for all reals
2011-11-26 15:37:58 -08:00
4e8a1d0bb4
math: add neg? word
...
Do the right thing with integers or floats
2011-11-23 19:51:05 -08:00
943596575a
use radix literals
2011-11-23 19:03:40 -08:00
bb4529bff2
math.integers: fix underflow and overflow behavior
2010-12-18 16:00:17 -08:00
f27080498d
Remove bignum>float VM primitive, and use bignum/f to implement >float on bignums instead, for a slight accuracy gain. Also, bignum/f now has a more efficient post-scaling algorithm to break the circular dependency on bignum>float
2010-11-24 22:41:15 -08:00
0c840a7965
math.integers: bignum/f rounding was wrong (reported by Joe Groff)
2010-11-13 16:02:12 -08:00
65dcd56b0b
math.integers: fix 0 0 bignum/f (reported by Joe Groff)
2010-11-12 19:49:18 -08:00
1c9a95122a
math.integers: remove a -rot usage
2010-02-10 15:44:49 +13:00
52f2ac2bb9
Tests for propagation additions; making fixnum-bit? inline
2010-01-25 20:15:17 -06:00
d3590ea210
Merge branch 'master' of git://factorcode.org/git/factor
...
Conflicts:
basis/compiler/tree/propagation/transforms/transforms.factor
2010-01-20 00:15:55 -06:00
aa1651032d
Adding compiler transforms in propagation
2010-01-20 00:10:49 -06:00
47a5e96547
Rework min and max so that behavior with floats and NaNs is consistent between generic arithmetic and open-coded float intrinsics
2010-01-13 22:20:16 +13:00
d4a207662e
might want to accurately divide large negative fixnums too
2009-10-31 12:10:29 -05:00
a7011fe087
use bignum/f to get accurate division of fixnums >= 2^53 on 64-bit platforms
2009-10-31 12:06:56 -05:00
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
2bb6293217
compiler: add fixnum-min/max intrinsics; ~10% speedup on benchmark.yuv-to-rgb
2009-08-28 19:02:59 -05:00
030b1b816c
Add inline declarations for various assorted methods
2009-08-17 22:32:21 -05:00
14ef1649d4
add if-zero/when-zero/unless-zero to core/ and update usages
2009-08-11 17:59:40 -05:00
0ad6d1fb7b
add a few usages of iota, remove most 1+ and 1- from core
2009-05-01 19:58:24 -05:00
0ffd43e2e3
1.0/0.0 => 1/0.
2009-04-13 19:48:08 -05:00
25a877e50b
Merge OneEyed's patch
2009-02-28 17:06:55 -06:00
087d931c36
Remove the tail argument from do/until/while
2009-02-18 02:41:14 +01:00
669548e62e
Clean up some usages of tuck, and add Joe's curried cleave/spread/apply combinators to kernel vocabulary
2009-02-02 13:43:54 -06:00
b154b21aaa
Add new until combinator, and a new do word which acts like a modifier: do while, do until for loops which iterate at least once
2008-12-08 23:37:18 -06:00
73b3cd6367
Use eq? instead of number= since we only ever have a fixnum here
2008-12-08 13:59:59 -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
8a8f0c925c
Use BSR instruction to implement fixnum-log2 intrinsic
2008-12-06 15:31:17 -06:00
a4d9cdfeb3
Refactor all usages of >r/r> in core to use dip, 2dip, 3dip
...
Non-optimizing compiler now special-cases dip, 2dip, 3dip following a
literal quotation: this allows us to break the dip/slip meta-circle
without explicit calls to >r/r>
2008-11-23 02:44:56 -06:00
Doug Coleman
Jon Harper
John Benediktsson
Joe Groff
Slava Pestov
Daniel Ehrenberg
Doug Coleman
Slava Pestov
Samuel Tardieu