Mercurial > hg > octave-nkf
annotate liboctave/operators/mx-inlines.cc @ 20830:b65888ec820e draft default tip gccjit
dmalcom gcc jit import
author | Stefan Mahr <dac922@gmx.de> |
---|---|
date | Fri, 27 Feb 2015 16:59:36 +0100 |
parents | a9574e3c6e9e |
children |
rev | line source |
---|---|
3 | 1 /* |
2 | |
19898
4197fc428c7d
maint: Update copyright notices for 2015.
John W. Eaton <jwe@octave.org>
parents:
19788
diff
changeset
|
3 Copyright (C) 1993-2015 John W. Eaton |
8777
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
4 Copyright (C) 2009 Jaroslav Hajek |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
5 Copyright (C) 2009 VZLU Prague |
3 | 6 |
7 This file is part of Octave. | |
8 | |
9 Octave is free software; you can redistribute it and/or modify it | |
10 under the terms of the GNU General Public License as published by the | |
7016 | 11 Free Software Foundation; either version 3 of the License, or (at your |
12 option) any later version. | |
3 | 13 |
14 Octave is distributed in the hope that it will be useful, but WITHOUT | |
15 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
16 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License | |
17 for more details. | |
18 | |
19 You should have received a copy of the GNU General Public License | |
7016 | 20 along with Octave; see the file COPYING. If not, see |
21 <http://www.gnu.org/licenses/>. | |
3 | 22 |
23 */ | |
24 | |
2828 | 25 #if !defined (octave_mx_inlines_h) |
26 #define octave_mx_inlines_h 1 | |
2804 | 27 |
28 #include <cstddef> | |
8650
a1ae2aae903e
abs,real,imag,conj: use code from mx-inlines rather than the generic map
Jaroslav Hajek <highegg@gmail.com>
parents:
8380
diff
changeset
|
29 #include <cmath> |
19576
af41e41ad28e
replace oct-mem.h inline indirections by standard function calls.
Kai T. Ohlhus <k.ohlhus@gmail.com>
parents:
18890
diff
changeset
|
30 #include <cstring> |
10146
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
31 #include <memory> |
2804 | 32 |
5525 | 33 #include "quit.h" |
34 | |
1650 | 35 #include "oct-cmplx.h" |
8758
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
36 #include "oct-locbuf.h" |
8897 | 37 #include "oct-inttypes.h" |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
38 #include "Array.h" |
10146
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
39 #include "Array-util.h" |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
40 |
13004
d9d65c3017c3
Make bsxfun automatic for most binary operators.
Jordi Gutiérrez Hermoso <jordigh@gmail.com>
parents:
11586
diff
changeset
|
41 #include "bsxfun.h" |
d9d65c3017c3
Make bsxfun automatic for most binary operators.
Jordi Gutiérrez Hermoso <jordigh@gmail.com>
parents:
11586
diff
changeset
|
42 |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
43 // Provides some commonly repeated, basic loop templates. |
461 | 44 |
8380
dbe67764e628
fix & improve speed of diagonal matrix multiplication
Jaroslav Hajek <highegg@gmail.com>
parents:
7789
diff
changeset
|
45 template <class R, class S> |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
46 inline void mx_inline_fill (size_t n, R *r, S s) throw () |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
47 { for (size_t i = 0; i < n; i++) r[i] = s; } |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
48 |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
49 #define DEFMXUNOP(F, OP) \ |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
50 template <class R, class X> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
51 inline void F (size_t n, R *r, const X *x) throw () \ |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
52 { for (size_t i = 0; i < n; i++) r[i] = OP x[i]; } |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
53 |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
54 DEFMXUNOP (mx_inline_uminus, -) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
55 |
9607
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
56 #define DEFMXUNOPEQ(F, OP) \ |
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
57 template <class R> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
58 inline void F (size_t n, R *r) throw () \ |
9607
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
59 { for (size_t i = 0; i < n; i++) r[i] = OP r[i]; } |
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
60 |
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
61 DEFMXUNOPEQ (mx_inline_uminus2, -) |
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
62 |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
63 #define DEFMXUNBOOLOP(F, OP) \ |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
64 template <class X> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
65 inline void F (size_t n, bool *r, const X *x) throw () \ |
14846
460a3c6d8bf1
maint: Use Octave coding convention for cuddled parenthis in function calls with empty argument lists.
Rik <octave@nomad.inbox5.com>
parents:
14138
diff
changeset
|
66 { const X zero = X (); for (size_t i = 0; i < n; i++) r[i] = x[i] OP zero; } |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
67 |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
68 DEFMXUNBOOLOP (mx_inline_iszero, ==) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
69 DEFMXUNBOOLOP (mx_inline_notzero, !=) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
70 |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
71 #define DEFMXBINOP(F, OP) \ |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
72 template <class R, class X, class Y> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
73 inline void F (size_t n, R *r, const X *x, const Y *y) throw () \ |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
74 { for (size_t i = 0; i < n; i++) r[i] = x[i] OP y[i]; } \ |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
75 template <class R, class X, class Y> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
76 inline void F (size_t n, R *r, const X *x, Y y) throw () \ |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
77 { for (size_t i = 0; i < n; i++) r[i] = x[i] OP y; } \ |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
78 template <class R, class X, class Y> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
79 inline void F (size_t n, R *r, X x, const Y *y) throw () \ |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
80 { for (size_t i = 0; i < n; i++) r[i] = x OP y[i]; } |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
81 |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
82 DEFMXBINOP (mx_inline_add, +) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
83 DEFMXBINOP (mx_inline_sub, -) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
84 DEFMXBINOP (mx_inline_mul, *) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
85 DEFMXBINOP (mx_inline_div, /) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
86 |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
87 #define DEFMXBINOPEQ(F, OP) \ |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
88 template <class R, class X> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
89 inline void F (size_t n, R *r, const X *x) throw () \ |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
90 { for (size_t i = 0; i < n; i++) r[i] OP x[i]; } \ |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
91 template <class R, class X> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
92 inline void F (size_t n, R *r, X x) throw () \ |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
93 { for (size_t i = 0; i < n; i++) r[i] OP x; } |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
94 |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
95 DEFMXBINOPEQ (mx_inline_add2, +=) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
96 DEFMXBINOPEQ (mx_inline_sub2, -=) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
97 DEFMXBINOPEQ (mx_inline_mul2, *=) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
98 DEFMXBINOPEQ (mx_inline_div2, /=) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
99 |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
100 #define DEFMXCMPOP(F, OP) \ |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
101 template <class X, class Y> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
102 inline void F (size_t n, bool *r, const X *x, const Y *y) throw () \ |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
103 { for (size_t i = 0; i < n; i++) r[i] = x[i] OP y[i]; } \ |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
104 template <class X, class Y> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
105 inline void F (size_t n, bool *r, const X *x, Y y) throw () \ |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
106 { for (size_t i = 0; i < n; i++) r[i] = x[i] OP y; } \ |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
107 template <class X, class Y> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
108 inline void F (size_t n, bool *r, X x, const Y *y) throw () \ |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
109 { for (size_t i = 0; i < n; i++) r[i] = x OP y[i]; } |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
110 |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
111 DEFMXCMPOP (mx_inline_lt, <) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
112 DEFMXCMPOP (mx_inline_le, <=) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
113 DEFMXCMPOP (mx_inline_gt, >) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
114 DEFMXCMPOP (mx_inline_ge, >=) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
115 DEFMXCMPOP (mx_inline_eq, ==) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
116 DEFMXCMPOP (mx_inline_ne, !=) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
117 |
9553
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
118 // Convert to logical value, for logical op purposes. |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
119 template <class T> inline bool logical_value (T x) { return x; } |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
120 template <class T> inline bool logical_value (const std::complex<T>& x) |
10482
2645a6b1027b
fix typo in last patch
Jaroslav Hajek <highegg@gmail.com>
parents:
10481
diff
changeset
|
121 { return x.real () != 0 || x.imag () != 0; } |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
122 template <class T> inline bool logical_value (const octave_int<T>& x) |
9553
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
123 { return x.value (); } |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
124 |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
125 template <class X> |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
126 void mx_inline_not (size_t n, bool *r, const X* x) throw () |
9553
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
127 { |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
128 for (size_t i = 0; i < n; i++) |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
129 r[i] = ! logical_value (x[i]); |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
130 } |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
131 |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
132 inline void mx_inline_not2 (size_t n, bool *r) throw () |
9607
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
133 { |
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
134 for (size_t i = 0; i < n; i++) r[i] = ! r[i]; |
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
135 } |
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
136 |
9553
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
137 #define DEFMXBOOLOP(F, NOT1, OP, NOT2) \ |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
138 template <class X, class Y> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
139 inline void F (size_t n, bool *r, const X *x, const Y *y) throw () \ |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
140 { \ |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
141 for (size_t i = 0; i < n; i++) \ |
9553
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
142 r[i] = (NOT1 logical_value (x[i])) OP (NOT2 logical_value (y[i])); \ |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
143 } \ |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
144 template <class X, class Y> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
145 inline void F (size_t n, bool *r, const X *x, Y y) throw () \ |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
146 { \ |
9553
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
147 const bool yy = (NOT2 logical_value (y)); \ |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
148 for (size_t i = 0; i < n; i++) \ |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
149 r[i] = (NOT1 logical_value (x[i])) OP yy; \ |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
150 } \ |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
151 template <class X, class Y> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
152 inline void F (size_t n, bool *r, X x, const Y *y) throw () \ |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
153 { \ |
9553
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
154 const bool xx = (NOT1 logical_value (x)); \ |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
155 for (size_t i = 0; i < n; i++) \ |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
156 r[i] = xx OP (NOT2 logical_value (y[i])); \ |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
157 } |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
158 |
9553
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
159 DEFMXBOOLOP (mx_inline_and, , &, ) |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
160 DEFMXBOOLOP (mx_inline_or, , |, ) |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
161 DEFMXBOOLOP (mx_inline_not_and, !, &, ) |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
162 DEFMXBOOLOP (mx_inline_not_or, !, |, ) |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
163 DEFMXBOOLOP (mx_inline_and_not, , &, !) |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
164 DEFMXBOOLOP (mx_inline_or_not, , |, !) |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
165 |
9612
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
166 #define DEFMXBOOLOPEQ(F, OP) \ |
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
167 template <class X> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
168 inline void F (size_t n, bool *r, const X *x) throw () \ |
9612
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
169 { \ |
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
170 for (size_t i = 0; i < n; i++) \ |
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
171 r[i] OP logical_value (x[i]); \ |
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
172 } \ |
13139
aa4a23337a0f
Enable BSX in-place for missing assignment operators
Jordi Gutiérrez Hermoso <jordigh@octave.org>
parents:
13005
diff
changeset
|
173 template <class X> \ |
aa4a23337a0f
Enable BSX in-place for missing assignment operators
Jordi Gutiérrez Hermoso <jordigh@octave.org>
parents:
13005
diff
changeset
|
174 inline void F (size_t n, bool *r, X x) throw () \ |
aa4a23337a0f
Enable BSX in-place for missing assignment operators
Jordi Gutiérrez Hermoso <jordigh@octave.org>
parents:
13005
diff
changeset
|
175 { for (size_t i = 0; i < n; i++) r[i] OP x; } |
9612
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
176 |
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
177 DEFMXBOOLOPEQ (mx_inline_and2, &=) |
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
178 DEFMXBOOLOPEQ (mx_inline_or2, |=) |
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
179 |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
180 template <class T> |
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
181 inline bool |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
182 mx_inline_any_nan (size_t n, const T* x) throw () |
9814
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
183 { |
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
184 for (size_t i = 0; i < n; i++) |
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
185 { |
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
186 if (xisnan (x[i])) |
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
187 return true; |
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
188 } |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
189 |
9814
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
190 return false; |
8380
dbe67764e628
fix & improve speed of diagonal matrix multiplication
Jaroslav Hajek <highegg@gmail.com>
parents:
7789
diff
changeset
|
191 } |
dbe67764e628
fix & improve speed of diagonal matrix multiplication
Jaroslav Hajek <highegg@gmail.com>
parents:
7789
diff
changeset
|
192 |
10900
b64803a8be4e
optimize element-wise sparse-dense multiplication and division
Jaroslav Hajek <highegg@gmail.com>
parents:
10643
diff
changeset
|
193 template <class T> |
b64803a8be4e
optimize element-wise sparse-dense multiplication and division
Jaroslav Hajek <highegg@gmail.com>
parents:
10643
diff
changeset
|
194 inline bool |
b64803a8be4e
optimize element-wise sparse-dense multiplication and division
Jaroslav Hajek <highegg@gmail.com>
parents:
10643
diff
changeset
|
195 mx_inline_all_finite (size_t n, const T* x) throw () |
b64803a8be4e
optimize element-wise sparse-dense multiplication and division
Jaroslav Hajek <highegg@gmail.com>
parents:
10643
diff
changeset
|
196 { |
b64803a8be4e
optimize element-wise sparse-dense multiplication and division
Jaroslav Hajek <highegg@gmail.com>
parents:
10643
diff
changeset
|
197 for (size_t i = 0; i < n; i++) |
b64803a8be4e
optimize element-wise sparse-dense multiplication and division
Jaroslav Hajek <highegg@gmail.com>
parents:
10643
diff
changeset
|
198 { |
b64803a8be4e
optimize element-wise sparse-dense multiplication and division
Jaroslav Hajek <highegg@gmail.com>
parents:
10643
diff
changeset
|
199 if (! xfinite (x[i])) |
b64803a8be4e
optimize element-wise sparse-dense multiplication and division
Jaroslav Hajek <highegg@gmail.com>
parents:
10643
diff
changeset
|
200 return false; |
b64803a8be4e
optimize element-wise sparse-dense multiplication and division
Jaroslav Hajek <highegg@gmail.com>
parents:
10643
diff
changeset
|
201 } |
b64803a8be4e
optimize element-wise sparse-dense multiplication and division
Jaroslav Hajek <highegg@gmail.com>
parents:
10643
diff
changeset
|
202 |
b64803a8be4e
optimize element-wise sparse-dense multiplication and division
Jaroslav Hajek <highegg@gmail.com>
parents:
10643
diff
changeset
|
203 return true; |
b64803a8be4e
optimize element-wise sparse-dense multiplication and division
Jaroslav Hajek <highegg@gmail.com>
parents:
10643
diff
changeset
|
204 } |
b64803a8be4e
optimize element-wise sparse-dense multiplication and division
Jaroslav Hajek <highegg@gmail.com>
parents:
10643
diff
changeset
|
205 |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
206 template <class T> |
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
207 inline bool |
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
208 mx_inline_any_negative (size_t n, const T* x) throw () |
9827
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
209 { |
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
210 for (size_t i = 0; i < n; i++) |
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
211 { |
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
212 if (x[i] < 0) |
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
213 return true; |
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
214 } |
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
215 |
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
216 return false; |
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
217 } |
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
218 |
13756
6dfebfa334cb
allow negative data log plots with OpenGL+FLTK graphics (bug #34232)
John W. Eaton <jwe@octave.org>
parents:
13139
diff
changeset
|
219 template <class T> |
6dfebfa334cb
allow negative data log plots with OpenGL+FLTK graphics (bug #34232)
John W. Eaton <jwe@octave.org>
parents:
13139
diff
changeset
|
220 inline bool |
6dfebfa334cb
allow negative data log plots with OpenGL+FLTK graphics (bug #34232)
John W. Eaton <jwe@octave.org>
parents:
13139
diff
changeset
|
221 mx_inline_any_positive (size_t n, const T* x) throw () |
6dfebfa334cb
allow negative data log plots with OpenGL+FLTK graphics (bug #34232)
John W. Eaton <jwe@octave.org>
parents:
13139
diff
changeset
|
222 { |
6dfebfa334cb
allow negative data log plots with OpenGL+FLTK graphics (bug #34232)
John W. Eaton <jwe@octave.org>
parents:
13139
diff
changeset
|
223 for (size_t i = 0; i < n; i++) |
6dfebfa334cb
allow negative data log plots with OpenGL+FLTK graphics (bug #34232)
John W. Eaton <jwe@octave.org>
parents:
13139
diff
changeset
|
224 { |
6dfebfa334cb
allow negative data log plots with OpenGL+FLTK graphics (bug #34232)
John W. Eaton <jwe@octave.org>
parents:
13139
diff
changeset
|
225 if (x[i] > 0) |
6dfebfa334cb
allow negative data log plots with OpenGL+FLTK graphics (bug #34232)
John W. Eaton <jwe@octave.org>
parents:
13139
diff
changeset
|
226 return true; |
6dfebfa334cb
allow negative data log plots with OpenGL+FLTK graphics (bug #34232)
John W. Eaton <jwe@octave.org>
parents:
13139
diff
changeset
|
227 } |
6dfebfa334cb
allow negative data log plots with OpenGL+FLTK graphics (bug #34232)
John W. Eaton <jwe@octave.org>
parents:
13139
diff
changeset
|
228 |
6dfebfa334cb
allow negative data log plots with OpenGL+FLTK graphics (bug #34232)
John W. Eaton <jwe@octave.org>
parents:
13139
diff
changeset
|
229 return false; |
6dfebfa334cb
allow negative data log plots with OpenGL+FLTK graphics (bug #34232)
John W. Eaton <jwe@octave.org>
parents:
13139
diff
changeset
|
230 } |
6dfebfa334cb
allow negative data log plots with OpenGL+FLTK graphics (bug #34232)
John W. Eaton <jwe@octave.org>
parents:
13139
diff
changeset
|
231 |
9814
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
232 template<class T> |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
233 inline bool |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
234 mx_inline_all_real (size_t n, const std::complex<T>* x) throw () |
9814
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
235 { |
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
236 for (size_t i = 0; i < n; i++) |
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
237 { |
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
238 if (x[i].imag () != 0) |
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
239 return false; |
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
240 } |
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
241 |
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
242 return true; |
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
243 } |
2811 | 244 |
9800
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
245 #define DEFMXMAPPER(F, FUN) \ |
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
246 template <class T> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
247 inline void F (size_t n, T *r, const T *x) throw () \ |
9800
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
248 { for (size_t i = 0; i < n; i++) r[i] = FUN (x[i]); } |
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
249 |
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
250 template<class T> |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
251 inline void mx_inline_real (size_t n, T *r, const std::complex<T>* x) throw () |
9800
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
252 { for (size_t i = 0; i < n; i++) r[i] = x[i].real (); } |
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
253 template<class T> |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
254 inline void mx_inline_imag (size_t n, T *r, const std::complex<T>* x) throw () |
9800
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
255 { for (size_t i = 0; i < n; i++) r[i] = x[i].imag (); } |
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
256 |
9743
26abff55f6fe
optimize bsxfun for common built-in operations
Jaroslav Hajek <highegg@gmail.com>
parents:
9721
diff
changeset
|
257 // Pairwise minimums/maximums |
9800
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
258 #define DEFMXMAPPER2(F, FUN) \ |
9743
26abff55f6fe
optimize bsxfun for common built-in operations
Jaroslav Hajek <highegg@gmail.com>
parents:
9721
diff
changeset
|
259 template <class T> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
260 inline void F (size_t n, T *r, const T *x, const T *y) throw () \ |
9743
26abff55f6fe
optimize bsxfun for common built-in operations
Jaroslav Hajek <highegg@gmail.com>
parents:
9721
diff
changeset
|
261 { for (size_t i = 0; i < n; i++) r[i] = FUN (x[i], y[i]); } \ |
26abff55f6fe
optimize bsxfun for common built-in operations
Jaroslav Hajek <highegg@gmail.com>
parents:
9721
diff
changeset
|
262 template <class T> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
263 inline void F (size_t n, T *r, const T *x, T y) throw () \ |
9743
26abff55f6fe
optimize bsxfun for common built-in operations
Jaroslav Hajek <highegg@gmail.com>
parents:
9721
diff
changeset
|
264 { for (size_t i = 0; i < n; i++) r[i] = FUN (x[i], y); } \ |
26abff55f6fe
optimize bsxfun for common built-in operations
Jaroslav Hajek <highegg@gmail.com>
parents:
9721
diff
changeset
|
265 template <class T> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
266 inline void F (size_t n, T *r, T x, const T *y) throw () \ |
9743
26abff55f6fe
optimize bsxfun for common built-in operations
Jaroslav Hajek <highegg@gmail.com>
parents:
9721
diff
changeset
|
267 { for (size_t i = 0; i < n; i++) r[i] = FUN (x, y[i]); } |
3 | 268 |
9800
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
269 DEFMXMAPPER2 (mx_inline_xmin, xmin) |
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
270 DEFMXMAPPER2 (mx_inline_xmax, xmax) |
2811 | 271 |
10146
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
272 // Specialize array-scalar max/min |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
273 #define DEFMINMAXSPEC(T, F, OP) \ |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
274 template <> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
275 inline void F<T> (size_t n, T *r, const T *x, T y) throw () \ |
10146
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
276 { \ |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
277 if (xisnan (y)) \ |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
278 std::memcpy (r, x, n * sizeof (T)); \ |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
279 else \ |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
280 for (size_t i = 0; i < n; i++) r[i] = (x[i] OP y) ? x[i] : y; \ |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
281 } \ |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
282 template <> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
283 inline void F<T> (size_t n, T *r, T x, const T *y) throw () \ |
10146
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
284 { \ |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
285 if (xisnan (x)) \ |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
286 std::memcpy (r, y, n * sizeof (T)); \ |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
287 else \ |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
288 for (size_t i = 0; i < n; i++) r[i] = (y[i] OP x) ? y[i] : x; \ |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
289 } |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
290 |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
291 DEFMINMAXSPEC (double, mx_inline_xmin, <=) |
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
292 DEFMINMAXSPEC (double, mx_inline_xmax, >=) |
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
293 DEFMINMAXSPEC (float, mx_inline_xmin, <=) |
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
294 DEFMINMAXSPEC (float, mx_inline_xmax, >=) |
10146
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
295 |
9827
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
296 // Pairwise power |
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
297 #define DEFMXMAPPER2X(F, FUN) \ |
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
298 template <class R, class X, class Y> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
299 inline void F (size_t n, R *r, const X *x, const Y *y) throw () \ |
9827
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
300 { for (size_t i = 0; i < n; i++) r[i] = FUN (x[i], y[i]); } \ |
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
301 template <class R, class X, class Y> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
302 inline void F (size_t n, R *r, const X *x, Y y) throw () \ |
9827
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
303 { for (size_t i = 0; i < n; i++) r[i] = FUN (x[i], y); } \ |
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
304 template <class R, class X, class Y> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
305 inline void F (size_t n, R *r, X x, const Y *y) throw () \ |
9827
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
306 { for (size_t i = 0; i < n; i++) r[i] = FUN (x, y[i]); } |
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
307 |
13005
4061106b1c4b
Enable automatic bsxfun for power operators
Jordi Gutiérrez Hermoso <jordigh@gmail.com>
parents:
13004
diff
changeset
|
308 // Let the compiler decide which pow to use, whichever best matches the |
4061106b1c4b
Enable automatic bsxfun for power operators
Jordi Gutiérrez Hermoso <jordigh@gmail.com>
parents:
13004
diff
changeset
|
309 // arguments provided. |
4061106b1c4b
Enable automatic bsxfun for power operators
Jordi Gutiérrez Hermoso <jordigh@gmail.com>
parents:
13004
diff
changeset
|
310 using std::pow; |
4061106b1c4b
Enable automatic bsxfun for power operators
Jordi Gutiérrez Hermoso <jordigh@gmail.com>
parents:
13004
diff
changeset
|
311 DEFMXMAPPER2X (mx_inline_pow, pow) |
9827
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
312 |
9800
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
313 // Arbitrary function appliers. The function is a template parameter to enable |
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
314 // inlining. |
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
315 template <class R, class X, R fun (X x)> |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
316 inline void mx_inline_map (size_t n, R *r, const X *x) throw () |
9800
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
317 { for (size_t i = 0; i < n; i++) r[i] = fun (x[i]); } |
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
318 |
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
319 template <class R, class X, R fun (const X& x)> |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
320 inline void mx_inline_map (size_t n, R *r, const X *x) throw () |
9800
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
321 { for (size_t i = 0; i < n; i++) r[i] = fun (x[i]); } |
3 | 322 |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
323 // Appliers. Since these call the operation just once, we pass it as |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
324 // a pointer, to allow the compiler reduce number of instances. |
3 | 325 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
326 template <class R, class X> |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
327 inline Array<R> |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
328 do_mx_unary_op (const Array<X>& x, |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
329 void (*op) (size_t, R *, const X *) throw ()) |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
330 { |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
331 Array<R> r (x.dims ()); |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
332 op (r.numel (), r.fortran_vec (), x.data ()); |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
333 return r; |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
334 } |
2811 | 335 |
9800
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
336 // Shortcuts for applying mx_inline_map. |
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
337 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
338 template <class R, class X, R fun (X)> |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
339 inline Array<R> |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
340 do_mx_unary_map (const Array<X>& x) |
9800
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
341 { |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
342 return do_mx_unary_op<R, X> (x, mx_inline_map<R, X, fun>); |
9800
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
343 } |
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
344 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
345 template <class R, class X, R fun (const X&)> |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
346 inline Array<R> |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
347 do_mx_unary_map (const Array<X>& x) |
9800
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
348 { |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
349 return do_mx_unary_op<R, X> (x, mx_inline_map<R, X, fun>); |
9800
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
350 } |
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
351 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
352 template <class R> |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
353 inline Array<R>& |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
354 do_mx_inplace_op (Array<R>& r, |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
355 void (*op) (size_t, R *) throw ()) |
9607
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
356 { |
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
357 op (r.numel (), r.fortran_vec ()); |
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
358 return r; |
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
359 } |
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
360 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
361 template <class R, class X, class Y> |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
362 inline Array<R> |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
363 do_mm_binary_op (const Array<X>& x, const Array<Y>& y, |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
364 void (*op) (size_t, R *, const X *, const Y *) throw (), |
13004
d9d65c3017c3
Make bsxfun automatic for most binary operators.
Jordi Gutiérrez Hermoso <jordigh@gmail.com>
parents:
11586
diff
changeset
|
365 void (*op1) (size_t, R *, X, const Y *) throw (), |
d9d65c3017c3
Make bsxfun automatic for most binary operators.
Jordi Gutiérrez Hermoso <jordigh@gmail.com>
parents:
11586
diff
changeset
|
366 void (*op2) (size_t, R *, const X *, Y) throw (), |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
367 const char *opname) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
368 { |
18084
8e056300994b
Follow coding convention of defining and initializing only 1 variable per line in liboctave.
Rik <rik@octave.org>
parents:
17769
diff
changeset
|
369 dim_vector dx = x.dims (); |
8e056300994b
Follow coding convention of defining and initializing only 1 variable per line in liboctave.
Rik <rik@octave.org>
parents:
17769
diff
changeset
|
370 dim_vector dy = y.dims (); |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
371 if (dx == dy) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
372 { |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
373 Array<R> r (dx); |
20442
a9574e3c6e9e
Deprecate Array::length() and Sparse::length() in favour of ::numel().
Carnë Draug <carandraug@octave.org>
parents:
20428
diff
changeset
|
374 op (r.numel (), r.fortran_vec (), x.data (), y.data ()); |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
375 return r; |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
376 } |
14056
c3d401562410
allow warning (or error) for automatic bsxfun
John W. Eaton <jwe@octave.org>
parents:
13756
diff
changeset
|
377 else if (is_valid_bsxfun (opname, dx, dy)) |
13004
d9d65c3017c3
Make bsxfun automatic for most binary operators.
Jordi Gutiérrez Hermoso <jordigh@gmail.com>
parents:
11586
diff
changeset
|
378 { |
d9d65c3017c3
Make bsxfun automatic for most binary operators.
Jordi Gutiérrez Hermoso <jordigh@gmail.com>
parents:
11586
diff
changeset
|
379 return do_bsxfun_op (x, y, op, op1, op2); |
d9d65c3017c3
Make bsxfun automatic for most binary operators.
Jordi Gutiérrez Hermoso <jordigh@gmail.com>
parents:
11586
diff
changeset
|
380 } |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
381 else |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
382 { |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
383 gripe_nonconformant (opname, dx, dy); |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
384 return Array<R> (); |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
385 } |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
386 } |
3 | 387 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
388 template <class R, class X, class Y> |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
389 inline Array<R> |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
390 do_ms_binary_op (const Array<X>& x, const Y& y, |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
391 void (*op) (size_t, R *, const X *, Y) throw ()) |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
392 { |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
393 Array<R> r (x.dims ()); |
20442
a9574e3c6e9e
Deprecate Array::length() and Sparse::length() in favour of ::numel().
Carnë Draug <carandraug@octave.org>
parents:
20428
diff
changeset
|
394 op (r.numel (), r.fortran_vec (), x.data (), y); |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
395 return r; |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
396 } |
3 | 397 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
398 template <class R, class X, class Y> |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
399 inline Array<R> |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
400 do_sm_binary_op (const X& x, const Array<Y>& y, |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
401 void (*op) (size_t, R *, X, const Y *) throw ()) |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
402 { |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
403 Array<R> r (y.dims ()); |
20442
a9574e3c6e9e
Deprecate Array::length() and Sparse::length() in favour of ::numel().
Carnë Draug <carandraug@octave.org>
parents:
20428
diff
changeset
|
404 op (r.numel (), r.fortran_vec (), x, y.data ()); |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
405 return r; |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
406 } |
3 | 407 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
408 template <class R, class X> |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
409 inline Array<R>& |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
410 do_mm_inplace_op (Array<R>& r, const Array<X>& x, |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
411 void (*op) (size_t, R *, const X *) throw (), |
13139
aa4a23337a0f
Enable BSX in-place for missing assignment operators
Jordi Gutiérrez Hermoso <jordigh@octave.org>
parents:
13005
diff
changeset
|
412 void (*op1) (size_t, R *, X) throw (), |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
413 const char *opname) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
414 { |
18084
8e056300994b
Follow coding convention of defining and initializing only 1 variable per line in liboctave.
Rik <rik@octave.org>
parents:
17769
diff
changeset
|
415 dim_vector dr = r.dims (); |
8e056300994b
Follow coding convention of defining and initializing only 1 variable per line in liboctave.
Rik <rik@octave.org>
parents:
17769
diff
changeset
|
416 dim_vector dx = x.dims (); |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
417 if (dr == dx) |
13139
aa4a23337a0f
Enable BSX in-place for missing assignment operators
Jordi Gutiérrez Hermoso <jordigh@octave.org>
parents:
13005
diff
changeset
|
418 { |
20442
a9574e3c6e9e
Deprecate Array::length() and Sparse::length() in favour of ::numel().
Carnë Draug <carandraug@octave.org>
parents:
20428
diff
changeset
|
419 op (r.numel (), r.fortran_vec (), x.data ()); |
13139
aa4a23337a0f
Enable BSX in-place for missing assignment operators
Jordi Gutiérrez Hermoso <jordigh@octave.org>
parents:
13005
diff
changeset
|
420 } |
14056
c3d401562410
allow warning (or error) for automatic bsxfun
John W. Eaton <jwe@octave.org>
parents:
13756
diff
changeset
|
421 else if (is_valid_inplace_bsxfun (opname, dr, dx)) |
13139
aa4a23337a0f
Enable BSX in-place for missing assignment operators
Jordi Gutiérrez Hermoso <jordigh@octave.org>
parents:
13005
diff
changeset
|
422 { |
aa4a23337a0f
Enable BSX in-place for missing assignment operators
Jordi Gutiérrez Hermoso <jordigh@octave.org>
parents:
13005
diff
changeset
|
423 do_inplace_bsxfun_op (r, x, op, op1); |
aa4a23337a0f
Enable BSX in-place for missing assignment operators
Jordi Gutiérrez Hermoso <jordigh@octave.org>
parents:
13005
diff
changeset
|
424 } |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
425 else |
9612
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
426 gripe_nonconformant (opname, dr, dx); |
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
427 return r; |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
428 } |
7789
82be108cc558
First attempt at single precision tyeps
David Bateman <dbateman@free.fr>
parents:
7017
diff
changeset
|
429 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
430 template <class R, class X> |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
431 inline Array<R>& |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
432 do_ms_inplace_op (Array<R>& r, const X& x, |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
433 void (*op) (size_t, R *, X) throw ()) |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
434 { |
20442
a9574e3c6e9e
Deprecate Array::length() and Sparse::length() in favour of ::numel().
Carnë Draug <carandraug@octave.org>
parents:
20428
diff
changeset
|
435 op (r.numel (), r.fortran_vec (), x); |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
436 return r; |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
437 } |
3 | 438 |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
439 template <class T1, class T2> |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
440 inline bool |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
441 mx_inline_equal (size_t n, const T1 *x, const T2 *y) throw () |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
442 { |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
443 for (size_t i = 0; i < n; i++) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
444 if (x[i] != y[i]) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
445 return false; |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
446 return true; |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
447 } |
3 | 448 |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
449 template <class T> |
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
450 inline bool |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
451 do_mx_check (const Array<T>& a, |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
452 bool (*op) (size_t, const T *) throw ()) |
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
453 { |
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
454 return op (a.numel (), a.data ()); |
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
455 } |
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
456 |
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
457 // NOTE: we don't use std::norm because it typically does some heavyweight |
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
458 // magic to avoid underflows, which we don't need here. |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
459 template <class T> |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
460 inline T cabsq (const std::complex<T>& c) |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
461 { return c.real () * c.real () + c.imag () * c.imag (); } |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
462 |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
463 // default. works for integers and bool. |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
464 template <class T> |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
465 inline bool xis_true (T x) { return x; } |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
466 template <class T> |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
467 inline bool xis_false (T x) { return ! x; } |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
468 // for octave_ints |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
469 template <class T> |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
470 inline bool xis_true (const octave_int<T>& x) { return x.value (); } |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
471 template <class T> |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
472 inline bool xis_false (const octave_int<T>& x) { return ! x.value (); } |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
473 // for reals, we want to ignore NaNs. |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
474 inline bool xis_true (double x) { return ! xisnan (x) && x != 0.0; } |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
475 inline bool xis_false (double x) { return x == 0.0; } |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
476 inline bool xis_true (float x) { return ! xisnan (x) && x != 0.0f; } |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
477 inline bool xis_false (float x) { return x == 0.0f; } |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
478 // Ditto for complex. |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
479 inline bool xis_true (const Complex& x) { return ! xisnan (x) && x != 0.0; } |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
480 inline bool xis_false (const Complex& x) { return x == 0.0; } |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
481 inline bool xis_true (const FloatComplex& x) { return ! xisnan (x) && x != 0.0f; } |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
482 inline bool xis_false (const FloatComplex& x) { return x == 0.0f; } |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
483 |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
484 #define OP_RED_SUM(ac, el) ac += el |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
485 #define OP_RED_PROD(ac, el) ac *= el |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
486 #define OP_RED_SUMSQ(ac, el) ac += el*el |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
487 #define OP_RED_SUMSQC(ac, el) ac += cabsq (el) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
488 |
18890
8cc66f091584
Add "native" option to prod() (bug #40349).
Rik <rik@octave.org>
parents:
18098
diff
changeset
|
489 inline void op_dble_prod (double& ac, float el) |
8cc66f091584
Add "native" option to prod() (bug #40349).
Rik <rik@octave.org>
parents:
18098
diff
changeset
|
490 { ac *= el; } |
8cc66f091584
Add "native" option to prod() (bug #40349).
Rik <rik@octave.org>
parents:
18098
diff
changeset
|
491 inline void op_dble_prod (Complex& ac, const FloatComplex& el) |
8cc66f091584
Add "native" option to prod() (bug #40349).
Rik <rik@octave.org>
parents:
18098
diff
changeset
|
492 { ac *= el; } // FIXME: guaranteed? |
8cc66f091584
Add "native" option to prod() (bug #40349).
Rik <rik@octave.org>
parents:
18098
diff
changeset
|
493 template <class T> |
8cc66f091584
Add "native" option to prod() (bug #40349).
Rik <rik@octave.org>
parents:
18098
diff
changeset
|
494 inline void op_dble_prod (double& ac, const octave_int<T>& el) |
8cc66f091584
Add "native" option to prod() (bug #40349).
Rik <rik@octave.org>
parents:
18098
diff
changeset
|
495 { ac *= el.double_value (); } |
8cc66f091584
Add "native" option to prod() (bug #40349).
Rik <rik@octave.org>
parents:
18098
diff
changeset
|
496 |
15018
3d8ace26c5b4
maint: Use Octave coding conventions for cuddled parentheses in liboctave/.
Rik <rik@octave.org>
parents:
14846
diff
changeset
|
497 inline void op_dble_sum (double& ac, float el) |
9721
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
498 { ac += el; } |
15018
3d8ace26c5b4
maint: Use Octave coding conventions for cuddled parentheses in liboctave/.
Rik <rik@octave.org>
parents:
14846
diff
changeset
|
499 inline void op_dble_sum (Complex& ac, const FloatComplex& el) |
9721
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
500 { ac += el; } // FIXME: guaranteed? |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
501 template <class T> |
15018
3d8ace26c5b4
maint: Use Octave coding conventions for cuddled parentheses in liboctave/.
Rik <rik@octave.org>
parents:
14846
diff
changeset
|
502 inline void op_dble_sum (double& ac, const octave_int<T>& el) |
9721
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
503 { ac += el.double_value (); } |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
504 |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
505 // The following two implement a simple short-circuiting. |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
506 #define OP_RED_ANYC(ac, el) if (xis_true (el)) { ac = true; break; } else continue |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
507 #define OP_RED_ALLC(ac, el) if (xis_false (el)) { ac = false; break; } else continue |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
508 |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
509 #define OP_RED_FCN(F, TSRC, TRES, OP, ZERO) \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
510 template <class T> \ |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
511 inline TRES \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
512 F (const TSRC* v, octave_idx_type n) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
513 { \ |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
514 TRES ac = ZERO; \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
515 for (octave_idx_type i = 0; i < n; i++) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
516 OP(ac, v[i]); \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
517 return ac; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
518 } |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
519 |
9721
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
520 #define PROMOTE_DOUBLE(T) typename subst_template_param<std::complex, T, double>::type |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
521 |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
522 OP_RED_FCN (mx_inline_sum, T, T, OP_RED_SUM, 0) |
9721
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
523 OP_RED_FCN (mx_inline_dsum, T, PROMOTE_DOUBLE(T), op_dble_sum, 0.0) |
8756
d0755c9db5ed
implement fast logical sum (counting)
Jaroslav Hajek <highegg@gmail.com>
parents:
8751
diff
changeset
|
524 OP_RED_FCN (mx_inline_count, bool, T, OP_RED_SUM, 0) |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
525 OP_RED_FCN (mx_inline_prod, T, T, OP_RED_PROD, 1) |
18890
8cc66f091584
Add "native" option to prod() (bug #40349).
Rik <rik@octave.org>
parents:
18098
diff
changeset
|
526 OP_RED_FCN (mx_inline_dprod, T, PROMOTE_DOUBLE(T), op_dble_prod, 1) |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
527 OP_RED_FCN (mx_inline_sumsq, T, T, OP_RED_SUMSQ, 0) |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
528 OP_RED_FCN (mx_inline_sumsq, std::complex<T>, T, OP_RED_SUMSQC, 0) |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
529 OP_RED_FCN (mx_inline_any, T, bool, OP_RED_ANYC, false) |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
530 OP_RED_FCN (mx_inline_all, T, bool, OP_RED_ALLC, true) |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
531 |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
532 |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
533 #define OP_RED_FCN2(F, TSRC, TRES, OP, ZERO) \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
534 template <class T> \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
535 inline void \ |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
536 F (const TSRC* v, TRES *r, octave_idx_type m, octave_idx_type n) \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
537 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
538 for (octave_idx_type i = 0; i < m; i++) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
539 r[i] = ZERO; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
540 for (octave_idx_type j = 0; j < n; j++) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
541 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
542 for (octave_idx_type i = 0; i < m; i++) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
543 OP(r[i], v[i]); \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
544 v += m; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
545 } \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
546 } |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
547 |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
548 OP_RED_FCN2 (mx_inline_sum, T, T, OP_RED_SUM, 0) |
9721
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
549 OP_RED_FCN2 (mx_inline_dsum, T, PROMOTE_DOUBLE(T), op_dble_sum, 0.0) |
8756
d0755c9db5ed
implement fast logical sum (counting)
Jaroslav Hajek <highegg@gmail.com>
parents:
8751
diff
changeset
|
550 OP_RED_FCN2 (mx_inline_count, bool, T, OP_RED_SUM, 0) |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
551 OP_RED_FCN2 (mx_inline_prod, T, T, OP_RED_PROD, 1) |
18890
8cc66f091584
Add "native" option to prod() (bug #40349).
Rik <rik@octave.org>
parents:
18098
diff
changeset
|
552 OP_RED_FCN2 (mx_inline_dprod, T, PROMOTE_DOUBLE(T), op_dble_prod, 0.0) |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
553 OP_RED_FCN2 (mx_inline_sumsq, T, T, OP_RED_SUMSQ, 0) |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
554 OP_RED_FCN2 (mx_inline_sumsq, std::complex<T>, T, OP_RED_SUMSQC, 0) |
8758
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
555 |
10147
adc0143e9419
optimize any/all (x, 2) with small number of rows
Jaroslav Hajek <highegg@gmail.com>
parents:
10146
diff
changeset
|
556 #define OP_RED_ANYR(ac, el) ac |= xis_true (el) |
adc0143e9419
optimize any/all (x, 2) with small number of rows
Jaroslav Hajek <highegg@gmail.com>
parents:
10146
diff
changeset
|
557 #define OP_RED_ALLR(ac, el) ac &= xis_true (el) |
adc0143e9419
optimize any/all (x, 2) with small number of rows
Jaroslav Hajek <highegg@gmail.com>
parents:
10146
diff
changeset
|
558 |
adc0143e9419
optimize any/all (x, 2) with small number of rows
Jaroslav Hajek <highegg@gmail.com>
parents:
10146
diff
changeset
|
559 OP_RED_FCN2 (mx_inline_any_r, T, bool, OP_RED_ANYR, false) |
adc0143e9419
optimize any/all (x, 2) with small number of rows
Jaroslav Hajek <highegg@gmail.com>
parents:
10146
diff
changeset
|
560 OP_RED_FCN2 (mx_inline_all_r, T, bool, OP_RED_ALLR, true) |
adc0143e9419
optimize any/all (x, 2) with small number of rows
Jaroslav Hajek <highegg@gmail.com>
parents:
10146
diff
changeset
|
561 |
8758
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
562 // Using the general code for any/all would sacrifice short-circuiting. |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
563 // OTOH, going by rows would sacrifice cache-coherence. The following algorithm |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
564 // will achieve both, at the cost of a temporary octave_idx_type array. |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
565 |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
566 #define OP_ROW_SHORT_CIRCUIT(F, PRED, ZERO) \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
567 template <class T> \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
568 inline void \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
569 F (const T* v, bool *r, octave_idx_type m, octave_idx_type n) \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
570 { \ |
10147
adc0143e9419
optimize any/all (x, 2) with small number of rows
Jaroslav Hajek <highegg@gmail.com>
parents:
10146
diff
changeset
|
571 if (n <= 8) \ |
adc0143e9419
optimize any/all (x, 2) with small number of rows
Jaroslav Hajek <highegg@gmail.com>
parents:
10146
diff
changeset
|
572 return F ## _r (v, r, m, n); \ |
adc0143e9419
optimize any/all (x, 2) with small number of rows
Jaroslav Hajek <highegg@gmail.com>
parents:
10146
diff
changeset
|
573 \ |
8758
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
574 /* FIXME: it may be sub-optimal to allocate the buffer here. */ \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
575 OCTAVE_LOCAL_BUFFER (octave_idx_type, iact, m); \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
576 for (octave_idx_type i = 0; i < m; i++) iact[i] = i; \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
577 octave_idx_type nact = m; \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
578 for (octave_idx_type j = 0; j < n; j++) \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
579 { \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
580 octave_idx_type k = 0; \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
581 for (octave_idx_type i = 0; i < nact; i++) \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
582 { \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
583 octave_idx_type ia = iact[i]; \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
584 if (! PRED (v[ia])) \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
585 iact[k++] = ia; \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
586 } \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
587 nact = k; \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
588 v += m; \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
589 } \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
590 for (octave_idx_type i = 0; i < m; i++) r[i] = ! ZERO; \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
591 for (octave_idx_type i = 0; i < nact; i++) r[iact[i]] = ZERO; \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
592 } |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
593 |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
594 OP_ROW_SHORT_CIRCUIT (mx_inline_any, xis_true, false) |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
595 OP_ROW_SHORT_CIRCUIT (mx_inline_all, xis_false, true) |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
596 |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
597 #define OP_RED_FCNN(F, TSRC, TRES) \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
598 template <class T> \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
599 inline void \ |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
600 F (const TSRC *v, TRES *r, octave_idx_type l, \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
601 octave_idx_type n, octave_idx_type u) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
602 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
603 if (l == 1) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
604 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
605 for (octave_idx_type i = 0; i < u; i++) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
606 { \ |
8756
d0755c9db5ed
implement fast logical sum (counting)
Jaroslav Hajek <highegg@gmail.com>
parents:
8751
diff
changeset
|
607 r[i] = F<T> (v, n); \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
608 v += n; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
609 } \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
610 } \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
611 else \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
612 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
613 for (octave_idx_type i = 0; i < u; i++) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
614 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
615 F (v, r, l, n); \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
616 v += l*n; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
617 r += l; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
618 } \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
619 } \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
620 } |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
621 |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
622 OP_RED_FCNN (mx_inline_sum, T, T) |
9721
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
623 OP_RED_FCNN (mx_inline_dsum, T, PROMOTE_DOUBLE(T)) |
8756
d0755c9db5ed
implement fast logical sum (counting)
Jaroslav Hajek <highegg@gmail.com>
parents:
8751
diff
changeset
|
624 OP_RED_FCNN (mx_inline_count, bool, T) |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
625 OP_RED_FCNN (mx_inline_prod, T, T) |
18890
8cc66f091584
Add "native" option to prod() (bug #40349).
Rik <rik@octave.org>
parents:
18098
diff
changeset
|
626 OP_RED_FCNN (mx_inline_dprod, T, PROMOTE_DOUBLE(T)) |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
627 OP_RED_FCNN (mx_inline_sumsq, T, T) |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
628 OP_RED_FCNN (mx_inline_sumsq, std::complex<T>, T) |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
629 OP_RED_FCNN (mx_inline_any, T, bool) |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
630 OP_RED_FCNN (mx_inline_all, T, bool) |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
631 |
8780
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
632 #define OP_CUM_FCN(F, TSRC, TRES, OP) \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
633 template <class T> \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
634 inline void \ |
8780
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
635 F (const TSRC *v, TRES *r, octave_idx_type n) \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
636 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
637 if (n) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
638 { \ |
8780
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
639 TRES t = r[0] = v[0]; \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
640 for (octave_idx_type i = 1; i < n; i++) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
641 r[i] = t = t OP v[i]; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
642 } \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
643 } |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
644 |
8780
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
645 OP_CUM_FCN (mx_inline_cumsum, T, T, +) |
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
646 OP_CUM_FCN (mx_inline_cumprod, T, T, *) |
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
647 OP_CUM_FCN (mx_inline_cumcount, bool, T, +) |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
648 |
8780
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
649 #define OP_CUM_FCN2(F, TSRC, TRES, OP) \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
650 template <class T> \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
651 inline void \ |
8780
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
652 F (const TSRC *v, TRES *r, octave_idx_type m, octave_idx_type n) \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
653 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
654 if (n) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
655 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
656 for (octave_idx_type i = 0; i < m; i++) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
657 r[i] = v[i]; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
658 const T *r0 = r; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
659 for (octave_idx_type j = 1; j < n; j++) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
660 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
661 r += m; v += m; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
662 for (octave_idx_type i = 0; i < m; i++) \ |
8780
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
663 r[i] = r0[i] OP v[i]; \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
664 r0 += m; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
665 } \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
666 } \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
667 } |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
668 |
8780
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
669 OP_CUM_FCN2 (mx_inline_cumsum, T, T, +) |
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
670 OP_CUM_FCN2 (mx_inline_cumprod, T, T, *) |
10643
9852264314d1
fix cumulative logical sum
Jaroslav Hajek <highegg@gmail.com>
parents:
10482
diff
changeset
|
671 OP_CUM_FCN2 (mx_inline_cumcount, bool, T, +) |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
672 |
8780
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
673 #define OP_CUM_FCNN(F, TSRC, TRES) \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
674 template <class T> \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
675 inline void \ |
8780
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
676 F (const TSRC *v, TRES *r, octave_idx_type l, \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
677 octave_idx_type n, octave_idx_type u) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
678 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
679 if (l == 1) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
680 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
681 for (octave_idx_type i = 0; i < u; i++) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
682 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
683 F (v, r, n); \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
684 v += n; r += n; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
685 } \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
686 } \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
687 else \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
688 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
689 for (octave_idx_type i = 0; i < u; i++) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
690 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
691 F (v, r, l, n); \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
692 v += l*n; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
693 r += l*n; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
694 } \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
695 } \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
696 } |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
697 |
8780
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
698 OP_CUM_FCNN (mx_inline_cumsum, T, T) |
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
699 OP_CUM_FCNN (mx_inline_cumprod, T, T) |
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
700 OP_CUM_FCNN (mx_inline_cumcount, bool, T) |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
701 |
8751
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
702 #define OP_MINMAX_FCN(F, OP) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
703 template <class T> \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
704 void F (const T *v, T *r, octave_idx_type n) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
705 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
706 if (! n) return; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
707 T tmp = v[0]; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
708 octave_idx_type i = 1; \ |
8776
d23c33ec6bd3
fix min/max behaviour with NaNs
Jaroslav Hajek <highegg@gmail.com>
parents:
8759
diff
changeset
|
709 if (xisnan (tmp)) \ |
d23c33ec6bd3
fix min/max behaviour with NaNs
Jaroslav Hajek <highegg@gmail.com>
parents:
8759
diff
changeset
|
710 { \ |
d23c33ec6bd3
fix min/max behaviour with NaNs
Jaroslav Hajek <highegg@gmail.com>
parents:
8759
diff
changeset
|
711 for (; i < n && xisnan (v[i]); i++) ; \ |
d23c33ec6bd3
fix min/max behaviour with NaNs
Jaroslav Hajek <highegg@gmail.com>
parents:
8759
diff
changeset
|
712 if (i < n) tmp = v[i]; \ |
d23c33ec6bd3
fix min/max behaviour with NaNs
Jaroslav Hajek <highegg@gmail.com>
parents:
8759
diff
changeset
|
713 } \ |
d23c33ec6bd3
fix min/max behaviour with NaNs
Jaroslav Hajek <highegg@gmail.com>
parents:
8759
diff
changeset
|
714 for (; i < n; i++) \ |
8751
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
715 if (v[i] OP tmp) tmp = v[i]; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
716 *r = tmp; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
717 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
718 template <class T> \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
719 void F (const T *v, T *r, octave_idx_type *ri, octave_idx_type n) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
720 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
721 if (! n) return; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
722 T tmp = v[0]; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
723 octave_idx_type tmpi = 0; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
724 octave_idx_type i = 1; \ |
8776
d23c33ec6bd3
fix min/max behaviour with NaNs
Jaroslav Hajek <highegg@gmail.com>
parents:
8759
diff
changeset
|
725 if (xisnan (tmp)) \ |
d23c33ec6bd3
fix min/max behaviour with NaNs
Jaroslav Hajek <highegg@gmail.com>
parents:
8759
diff
changeset
|
726 { \ |
d23c33ec6bd3
fix min/max behaviour with NaNs
Jaroslav Hajek <highegg@gmail.com>
parents:
8759
diff
changeset
|
727 for (; i < n && xisnan (v[i]); i++) ; \ |
d23c33ec6bd3
fix min/max behaviour with NaNs
Jaroslav Hajek <highegg@gmail.com>
parents:
8759
diff
changeset
|
728 if (i < n) { tmp = v[i]; tmpi = i; } \ |
d23c33ec6bd3
fix min/max behaviour with NaNs
Jaroslav Hajek <highegg@gmail.com>
parents:
8759
diff
changeset
|
729 } \ |
d23c33ec6bd3
fix min/max behaviour with NaNs
Jaroslav Hajek <highegg@gmail.com>
parents:
8759
diff
changeset
|
730 for (; i < n; i++) \ |
8751
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
731 if (v[i] OP tmp) { tmp = v[i]; tmpi = i; }\ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
732 *r = tmp; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
733 *ri = tmpi; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
734 } |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
735 |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
736 OP_MINMAX_FCN (mx_inline_min, <) |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
737 OP_MINMAX_FCN (mx_inline_max, >) |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
738 |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
739 // Row reductions will be slightly complicated. We will proceed with checks |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
740 // for NaNs until we detect that no row will yield a NaN, in which case we |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
741 // proceed to a faster code. |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
742 |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
743 #define OP_MINMAX_FCN2(F, OP) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
744 template <class T> \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
745 inline void \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
746 F (const T *v, T *r, octave_idx_type m, octave_idx_type n) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
747 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
748 if (! n) return; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
749 bool nan = false; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
750 octave_idx_type j = 0; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
751 for (octave_idx_type i = 0; i < m; i++) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
752 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
753 r[i] = v[i]; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
754 if (xisnan (v[i])) nan = true; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
755 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
756 j++; v += m; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
757 while (nan && j < n) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
758 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
759 nan = false; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
760 for (octave_idx_type i = 0; i < m; i++) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
761 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
762 if (xisnan (v[i])) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
763 nan = true; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
764 else if (xisnan (r[i]) || v[i] OP r[i]) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
765 r[i] = v[i]; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
766 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
767 j++; v += m; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
768 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
769 while (j < n) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
770 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
771 for (octave_idx_type i = 0; i < m; i++) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
772 if (v[i] OP r[i]) r[i] = v[i]; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
773 j++; v += m; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
774 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
775 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
776 template <class T> \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
777 inline void \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
778 F (const T *v, T *r, octave_idx_type *ri, \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
779 octave_idx_type m, octave_idx_type n) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
780 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
781 if (! n) return; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
782 bool nan = false; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
783 octave_idx_type j = 0; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
784 for (octave_idx_type i = 0; i < m; i++) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
785 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
786 r[i] = v[i]; ri[i] = j; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
787 if (xisnan (v[i])) nan = true; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
788 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
789 j++; v += m; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
790 while (nan && j < n) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
791 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
792 nan = false; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
793 for (octave_idx_type i = 0; i < m; i++) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
794 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
795 if (xisnan (v[i])) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
796 nan = true; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
797 else if (xisnan (r[i]) || v[i] OP r[i]) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
798 { r[i] = v[i]; ri[i] = j; } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
799 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
800 j++; v += m; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
801 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
802 while (j < n) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
803 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
804 for (octave_idx_type i = 0; i < m; i++) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
805 if (v[i] OP r[i]) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
806 { r[i] = v[i]; ri[i] = j; } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
807 j++; v += m; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
808 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
809 } |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
810 |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
811 OP_MINMAX_FCN2 (mx_inline_min, <) |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
812 OP_MINMAX_FCN2 (mx_inline_max, >) |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
813 |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
814 #define OP_MINMAX_FCNN(F) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
815 template <class T> \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
816 inline void \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
817 F (const T *v, T *r, octave_idx_type l, \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
818 octave_idx_type n, octave_idx_type u) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
819 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
820 if (! n) return; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
821 if (l == 1) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
822 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
823 for (octave_idx_type i = 0; i < u; i++) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
824 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
825 F (v, r, n); \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
826 v += n; r++; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
827 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
828 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
829 else \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
830 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
831 for (octave_idx_type i = 0; i < u; i++) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
832 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
833 F (v, r, l, n); \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
834 v += l*n; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
835 r += l; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
836 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
837 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
838 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
839 template <class T> \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
840 inline void \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
841 F (const T *v, T *r, octave_idx_type *ri, \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
842 octave_idx_type l, octave_idx_type n, octave_idx_type u) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
843 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
844 if (! n) return; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
845 if (l == 1) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
846 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
847 for (octave_idx_type i = 0; i < u; i++) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
848 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
849 F (v, r, ri, n); \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
850 v += n; r++; ri++; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
851 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
852 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
853 else \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
854 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
855 for (octave_idx_type i = 0; i < u; i++) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
856 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
857 F (v, r, ri, l, n); \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
858 v += l*n; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
859 r += l; ri += l; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
860 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
861 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
862 } |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
863 |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
864 OP_MINMAX_FCNN (mx_inline_min) |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
865 OP_MINMAX_FCNN (mx_inline_max) |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
866 |
8777
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
867 #define OP_CUMMINMAX_FCN(F, OP) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
868 template <class T> \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
869 void F (const T *v, T *r, octave_idx_type n) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
870 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
871 if (! n) return; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
872 T tmp = v[0]; \ |
18084
8e056300994b
Follow coding convention of defining and initializing only 1 variable per line in liboctave.
Rik <rik@octave.org>
parents:
17769
diff
changeset
|
873 octave_idx_type i = 1; \ |
8e056300994b
Follow coding convention of defining and initializing only 1 variable per line in liboctave.
Rik <rik@octave.org>
parents:
17769
diff
changeset
|
874 octave_idx_type j = 0; \ |
8777
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
875 if (xisnan (tmp)) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
876 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
877 for (; i < n && xisnan (v[i]); i++) ; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
878 for (; j < i; j++) r[j] = tmp; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
879 if (i < n) tmp = v[i]; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
880 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
881 for (; i < n; i++) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
882 if (v[i] OP tmp) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
883 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
884 for (; j < i; j++) r[j] = tmp; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
885 tmp = v[i]; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
886 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
887 for (; j < i; j++) r[j] = tmp; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
888 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
889 template <class T> \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
890 void F (const T *v, T *r, octave_idx_type *ri, octave_idx_type n) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
891 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
892 if (! n) return; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
893 T tmp = v[0]; octave_idx_type tmpi = 0; \ |
18084
8e056300994b
Follow coding convention of defining and initializing only 1 variable per line in liboctave.
Rik <rik@octave.org>
parents:
17769
diff
changeset
|
894 octave_idx_type i = 1; \ |
8e056300994b
Follow coding convention of defining and initializing only 1 variable per line in liboctave.
Rik <rik@octave.org>
parents:
17769
diff
changeset
|
895 octave_idx_type j = 0; \ |
8777
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
896 if (xisnan (tmp)) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
897 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
898 for (; i < n && xisnan (v[i]); i++) ; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
899 for (; j < i; j++) { r[j] = tmp; ri[j] = tmpi; } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
900 if (i < n) { tmp = v[i]; tmpi = i; } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
901 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
902 for (; i < n; i++) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
903 if (v[i] OP tmp) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
904 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
905 for (; j < i; j++) { r[j] = tmp; ri[j] = tmpi; } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
906 tmp = v[i]; tmpi = i; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
907 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
908 for (; j < i; j++) { r[j] = tmp; ri[j] = tmpi; } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
909 } |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
910 |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
911 OP_CUMMINMAX_FCN (mx_inline_cummin, <) |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
912 OP_CUMMINMAX_FCN (mx_inline_cummax, >) |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
913 |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
914 // Row reductions will be slightly complicated. We will proceed with checks |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
915 // for NaNs until we detect that no row will yield a NaN, in which case we |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
916 // proceed to a faster code. |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
917 |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
918 #define OP_CUMMINMAX_FCN2(F, OP) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
919 template <class T> \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
920 inline void \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
921 F (const T *v, T *r, octave_idx_type m, octave_idx_type n) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
922 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
923 if (! n) return; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
924 bool nan = false; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
925 const T *r0; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
926 octave_idx_type j = 0; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
927 for (octave_idx_type i = 0; i < m; i++) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
928 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
929 r[i] = v[i]; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
930 if (xisnan (v[i])) nan = true; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
931 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
932 j++; v += m; r0 = r; r += m; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
933 while (nan && j < n) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
934 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
935 nan = false; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
936 for (octave_idx_type i = 0; i < m; i++) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
937 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
938 if (xisnan (v[i])) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
939 { r[i] = r0[i]; nan = true; } \ |
8949
e31d47f2c9bb
fixes to row-reducing cummin/cummax
Jaroslav Hajek <highegg@gmail.com>
parents:
8920
diff
changeset
|
940 else if (xisnan (r0[i]) || v[i] OP r0[i]) \ |
8777
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
941 r[i] = v[i]; \ |
19788
6b09dd576521
Fix cummin/cummax operations on rows when NaN element present (bug #44009).
Lachlan Andrew <lachlanbis@gmail.com>
parents:
19576
diff
changeset
|
942 else \ |
6b09dd576521
Fix cummin/cummax operations on rows when NaN element present (bug #44009).
Lachlan Andrew <lachlanbis@gmail.com>
parents:
19576
diff
changeset
|
943 r[i] = r0[i]; \ |
8777
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
944 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
945 j++; v += m; r0 = r; r += m; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
946 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
947 while (j < n) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
948 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
949 for (octave_idx_type i = 0; i < m; i++) \ |
8949
e31d47f2c9bb
fixes to row-reducing cummin/cummax
Jaroslav Hajek <highegg@gmail.com>
parents:
8920
diff
changeset
|
950 if (v[i] OP r0[i]) \ |
8777
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
951 r[i] = v[i]; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
952 else \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
953 r[i] = r0[i]; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
954 j++; v += m; r0 = r; r += m; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
955 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
956 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
957 template <class T> \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
958 inline void \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
959 F (const T *v, T *r, octave_idx_type *ri, \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
960 octave_idx_type m, octave_idx_type n) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
961 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
962 if (! n) return; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
963 bool nan = false; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
964 const T *r0; const octave_idx_type *r0i; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
965 octave_idx_type j = 0; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
966 for (octave_idx_type i = 0; i < m; i++) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
967 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
968 r[i] = v[i]; ri[i] = 0; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
969 if (xisnan (v[i])) nan = true; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
970 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
971 j++; v += m; r0 = r; r += m; r0i = ri; ri += m; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
972 while (nan && j < n) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
973 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
974 nan = false; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
975 for (octave_idx_type i = 0; i < m; i++) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
976 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
977 if (xisnan (v[i])) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
978 { r[i] = r0[i]; ri[i] = r0i[i]; nan = true; } \ |
8949
e31d47f2c9bb
fixes to row-reducing cummin/cummax
Jaroslav Hajek <highegg@gmail.com>
parents:
8920
diff
changeset
|
979 else if (xisnan (r0[i]) || v[i] OP r0[i]) \ |
8777
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
980 { r[i] = v[i]; ri[i] = j; }\ |
19788
6b09dd576521
Fix cummin/cummax operations on rows when NaN element present (bug #44009).
Lachlan Andrew <lachlanbis@gmail.com>
parents:
19576
diff
changeset
|
981 else \ |
6b09dd576521
Fix cummin/cummax operations on rows when NaN element present (bug #44009).
Lachlan Andrew <lachlanbis@gmail.com>
parents:
19576
diff
changeset
|
982 { r[i] = r0[i]; ri[i] = r0i[i]; }\ |
8777
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
983 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
984 j++; v += m; r0 = r; r += m; r0i = ri; ri += m; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
985 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
986 while (j < n) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
987 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
988 for (octave_idx_type i = 0; i < m; i++) \ |
8949
e31d47f2c9bb
fixes to row-reducing cummin/cummax
Jaroslav Hajek <highegg@gmail.com>
parents:
8920
diff
changeset
|
989 if (v[i] OP r0[i]) \ |
8777
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
990 { r[i] = v[i]; ri[i] = j; } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
991 else \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
992 { r[i] = r0[i]; ri[i] = r0i[i]; } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
993 j++; v += m; r0 = r; r += m; r0i = ri; ri += m; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
994 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
995 } |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
996 |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
997 OP_CUMMINMAX_FCN2 (mx_inline_cummin, <) |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
998 OP_CUMMINMAX_FCN2 (mx_inline_cummax, >) |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
999 |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1000 #define OP_CUMMINMAX_FCNN(F) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1001 template <class T> \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1002 inline void \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1003 F (const T *v, T *r, octave_idx_type l, \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1004 octave_idx_type n, octave_idx_type u) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1005 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1006 if (! n) return; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1007 if (l == 1) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1008 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1009 for (octave_idx_type i = 0; i < u; i++) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1010 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1011 F (v, r, n); \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1012 v += n; r += n; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1013 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1014 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1015 else \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1016 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1017 for (octave_idx_type i = 0; i < u; i++) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1018 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1019 F (v, r, l, n); \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1020 v += l*n; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1021 r += l*n; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1022 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1023 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1024 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1025 template <class T> \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1026 inline void \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1027 F (const T *v, T *r, octave_idx_type *ri, \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1028 octave_idx_type l, octave_idx_type n, octave_idx_type u) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1029 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1030 if (! n) return; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1031 if (l == 1) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1032 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1033 for (octave_idx_type i = 0; i < u; i++) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1034 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1035 F (v, r, ri, n); \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1036 v += n; r += n; ri += n; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1037 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1038 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1039 else \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1040 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1041 for (octave_idx_type i = 0; i < u; i++) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1042 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1043 F (v, r, ri, l, n); \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1044 v += l*n; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1045 r += l*n; ri += l*n; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1046 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1047 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1048 } |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1049 |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1050 OP_CUMMINMAX_FCNN (mx_inline_cummin) |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1051 OP_CUMMINMAX_FCNN (mx_inline_cummax) |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1052 |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1053 template <class T> |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1054 void mx_inline_diff (const T *v, T *r, octave_idx_type n, |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1055 octave_idx_type order) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1056 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1057 switch (order) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1058 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1059 case 1: |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1060 for (octave_idx_type i = 0; i < n-1; i++) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1061 r[i] = v[i+1] - v[i]; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1062 break; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1063 case 2: |
9702
9ecd35a606e3
avoid some warnings from g++
John W. Eaton <jwe@octave.org>
parents:
9612
diff
changeset
|
1064 if (n > 1) |
10314
07ebe522dac2
untabify liboctave C++ sources
John W. Eaton <jwe@octave.org>
parents:
10158
diff
changeset
|
1065 { |
07ebe522dac2
untabify liboctave C++ sources
John W. Eaton <jwe@octave.org>
parents:
10158
diff
changeset
|
1066 T lst = v[1] - v[0]; |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1067 for (octave_idx_type i = 0; i < n-2; i++) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1068 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1069 T dif = v[i+2] - v[i+1]; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1070 r[i] = dif - lst; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1071 lst = dif; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1072 } |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1073 } |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1074 break; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1075 default: |
17769
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1076 { |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1077 OCTAVE_LOCAL_BUFFER (T, buf, n-1); |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1078 |
17769
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1079 for (octave_idx_type i = 0; i < n-1; i++) |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1080 buf[i] = v[i+1] - v[i]; |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1081 |
17769
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1082 for (octave_idx_type o = 2; o <= order; o++) |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1083 { |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1084 for (octave_idx_type i = 0; i < n-o; i++) |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1085 buf[i] = buf[i+1] - buf[i]; |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1086 } |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1087 |
17769
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1088 for (octave_idx_type i = 0; i < n-order; i++) |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1089 r[i] = buf[i]; |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1090 } |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1091 } |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1092 } |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1093 |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1094 template <class T> |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
1095 void mx_inline_diff (const T *v, T *r, |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1096 octave_idx_type m, octave_idx_type n, |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1097 octave_idx_type order) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1098 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1099 switch (order) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1100 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1101 case 1: |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1102 for (octave_idx_type i = 0; i < m*(n-1); i++) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1103 r[i] = v[i+m] - v[i]; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1104 break; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1105 case 2: |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1106 for (octave_idx_type i = 0; i < n-2; i++) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1107 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1108 for (octave_idx_type j = i*m; j < i*m+m; j++) |
15118
a4e94933fed3
Fix bug #37033 in diff ()
Carlo de Falco <cdf@users.sourceforge.net>
parents:
14138
diff
changeset
|
1109 r[j] = (v[j+m+m] - v[j+m]) - (v[j+m] - v[j]); |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1110 } |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1111 break; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1112 default: |
17769
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1113 { |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1114 OCTAVE_LOCAL_BUFFER (T, buf, n-1); |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1115 |
17769
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1116 for (octave_idx_type j = 0; j < m; j++) |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1117 { |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1118 for (octave_idx_type i = 0; i < n-1; i++) |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1119 buf[i] = v[i*m+j+m] - v[i*m+j]; |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1120 |
17769
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1121 for (octave_idx_type o = 2; o <= order; o++) |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1122 { |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1123 for (octave_idx_type i = 0; i < n-o; i++) |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1124 buf[i] = buf[i+1] - buf[i]; |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1125 } |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1126 |
17769
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1127 for (octave_idx_type i = 0; i < n-order; i++) |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1128 r[i*m+j] = buf[i]; |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1129 } |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1130 } |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1131 } |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1132 } |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1133 |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1134 template <class T> |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1135 inline void |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1136 mx_inline_diff (const T *v, T *r, |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1137 octave_idx_type l, octave_idx_type n, octave_idx_type u, |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1138 octave_idx_type order) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1139 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1140 if (! n) return; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1141 if (l == 1) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1142 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1143 for (octave_idx_type i = 0; i < u; i++) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1144 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1145 mx_inline_diff (v, r, n, order); |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1146 v += n; r += n-order; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1147 } |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1148 } |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1149 else |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1150 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1151 for (octave_idx_type i = 0; i < u; i++) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1152 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1153 mx_inline_diff (v, r, l, n, order); |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1154 v += l*n; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1155 r += l*(n-order); |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1156 } |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1157 } |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1158 } |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1159 |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1160 // Assistant function |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1161 |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1162 inline void |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1163 get_extent_triplet (const dim_vector& dims, int& dim, |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1164 octave_idx_type& l, octave_idx_type& n, |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1165 octave_idx_type& u) |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1166 { |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1167 octave_idx_type ndims = dims.length (); |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1168 if (dim >= ndims) |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1169 { |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1170 l = dims.numel (); |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1171 n = 1; |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1172 u = 1; |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1173 } |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1174 else |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1175 { |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1176 if (dim < 0) |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1177 dim = dims.first_non_singleton (); |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1178 |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1179 // calculate extent triplet. |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1180 l = 1, n = dims(dim), u = 1; |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
1181 for (octave_idx_type i = 0; i < dim; i++) |
20428
b2100e1659ac
maint: Use cuddled parentheses when indexing dimension_vectors.
Rik <rik@octave.org>
parents:
19898
diff
changeset
|
1182 l *= dims(i); |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1183 for (octave_idx_type i = dim + 1; i < ndims; i++) |
20428
b2100e1659ac
maint: Use cuddled parentheses when indexing dimension_vectors.
Rik <rik@octave.org>
parents:
19898
diff
changeset
|
1184 u *= dims(i); |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1185 } |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1186 } |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1187 |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1188 // Appliers. |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1189 // FIXME: is this the best design? C++ gives a lot of options here... |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1190 // maybe it can be done without an explicit parameter? |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1191 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1192 template <class R, class T> |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1193 inline Array<R> |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1194 do_mx_red_op (const Array<T>& src, int dim, |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
1195 void (*mx_red_op) (const T *, R *, octave_idx_type, |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1196 octave_idx_type, octave_idx_type)) |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1197 { |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1198 octave_idx_type l, n, u; |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1199 dim_vector dims = src.dims (); |
15018
3d8ace26c5b4
maint: Use Octave coding conventions for cuddled parentheses in liboctave/.
Rik <rik@octave.org>
parents:
14846
diff
changeset
|
1200 // M*b inconsistency: sum ([]) = 0 etc. |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
1201 if (dims.length () == 2 && dims(0) == 0 && dims(1) == 0) |
20428
b2100e1659ac
maint: Use cuddled parentheses when indexing dimension_vectors.
Rik <rik@octave.org>
parents:
19898
diff
changeset
|
1202 dims(1) = 1; |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
1203 |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1204 get_extent_triplet (dims, dim, l, n, u); |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1205 |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1206 // Reduction operation reduces the array size. |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1207 if (dim < dims.length ()) dims(dim) = 1; |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1208 dims.chop_trailing_singletons (); |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1209 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1210 Array<R> ret (dims); |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1211 mx_red_op (src.data (), ret.fortran_vec (), l, n, u); |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1212 |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1213 return ret; |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1214 } |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1215 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1216 template <class R, class T> |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1217 inline Array<R> |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1218 do_mx_cum_op (const Array<T>& src, int dim, |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
1219 void (*mx_cum_op) (const T *, R *, octave_idx_type, |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1220 octave_idx_type, octave_idx_type)) |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1221 { |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1222 octave_idx_type l, n, u; |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1223 dim_vector dims = src.dims (); |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1224 get_extent_triplet (dims, dim, l, n, u); |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1225 |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1226 // Cumulative operation doesn't reduce the array size. |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1227 Array<R> ret (dims); |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1228 mx_cum_op (src.data (), ret.fortran_vec (), l, n, u); |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1229 |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1230 return ret; |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1231 } |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1232 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1233 template <class R> |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1234 inline Array<R> |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1235 do_mx_minmax_op (const Array<R>& src, int dim, |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
1236 void (*mx_minmax_op) (const R *, R *, octave_idx_type, |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1237 octave_idx_type, octave_idx_type)) |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1238 { |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1239 octave_idx_type l, n, u; |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1240 dim_vector dims = src.dims (); |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1241 get_extent_triplet (dims, dim, l, n, u); |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1242 |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1243 // If the dimension is zero, we don't do anything. |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1244 if (dim < dims.length () && dims(dim) != 0) dims(dim) = 1; |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1245 dims.chop_trailing_singletons (); |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1246 |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1247 Array<R> ret (dims); |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1248 mx_minmax_op (src.data (), ret.fortran_vec (), l, n, u); |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1249 |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1250 return ret; |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1251 } |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1252 |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1253 template <class R> |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1254 inline Array<R> |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1255 do_mx_minmax_op (const Array<R>& src, Array<octave_idx_type>& idx, int dim, |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1256 void (*mx_minmax_op) (const R *, R *, octave_idx_type *, |
8751
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1257 octave_idx_type, octave_idx_type, octave_idx_type)) |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1258 { |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1259 octave_idx_type l, n, u; |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1260 dim_vector dims = src.dims (); |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1261 get_extent_triplet (dims, dim, l, n, u); |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1262 |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1263 // If the dimension is zero, we don't do anything. |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1264 if (dim < dims.length () && dims(dim) != 0) dims(dim) = 1; |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1265 dims.chop_trailing_singletons (); |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1266 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1267 Array<R> ret (dims); |
8751
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1268 if (idx.dims () != dims) idx = Array<octave_idx_type> (dims); |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1269 |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1270 mx_minmax_op (src.data (), ret.fortran_vec (), idx.fortran_vec (), |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1271 l, n, u); |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1272 |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1273 return ret; |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1274 } |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1275 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1276 template <class R> |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1277 inline Array<R> |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1278 do_mx_cumminmax_op (const Array<R>& src, int dim, |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
1279 void (*mx_cumminmax_op) (const R *, R *, octave_idx_type, |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1280 octave_idx_type, octave_idx_type)) |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1281 { |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1282 octave_idx_type l, n, u; |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1283 dim_vector dims = src.dims (); |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1284 get_extent_triplet (dims, dim, l, n, u); |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1285 |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1286 Array<R> ret (dims); |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1287 mx_cumminmax_op (src.data (), ret.fortran_vec (), l, n, u); |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1288 |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1289 return ret; |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1290 } |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1291 |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1292 template <class R> |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1293 inline Array<R> |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1294 do_mx_cumminmax_op (const Array<R>& src, Array<octave_idx_type>& idx, int dim, |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1295 void (*mx_cumminmax_op) (const R *, R *, octave_idx_type *, |
8777
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1296 octave_idx_type, octave_idx_type, octave_idx_type)) |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1297 { |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1298 octave_idx_type l, n, u; |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1299 dim_vector dims = src.dims (); |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1300 get_extent_triplet (dims, dim, l, n, u); |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1301 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1302 Array<R> ret (dims); |
8777
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1303 if (idx.dims () != dims) idx = Array<octave_idx_type> (dims); |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1304 |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1305 mx_cumminmax_op (src.data (), ret.fortran_vec (), idx.fortran_vec (), |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1306 l, n, u); |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1307 |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1308 return ret; |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1309 } |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1310 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1311 template <class R> |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1312 inline Array<R> |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1313 do_mx_diff_op (const Array<R>& src, int dim, octave_idx_type order, |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1314 void (*mx_diff_op) (const R *, R *, |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1315 octave_idx_type, octave_idx_type, octave_idx_type, |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1316 octave_idx_type)) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1317 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1318 octave_idx_type l, n, u; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1319 if (order <= 0) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1320 return src; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1321 |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1322 dim_vector dims = src.dims (); |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1323 |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1324 get_extent_triplet (dims, dim, l, n, u); |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1325 if (dim >= dims.length ()) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1326 dims.resize (dim+1, 1); |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1327 |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1328 if (dims(dim) <= order) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1329 { |
20428
b2100e1659ac
maint: Use cuddled parentheses when indexing dimension_vectors.
Rik <rik@octave.org>
parents:
19898
diff
changeset
|
1330 dims(dim) = 0; |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1331 return Array<R> (dims); |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1332 } |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1333 else |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1334 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1335 dims(dim) -= order; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1336 } |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1337 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1338 Array<R> ret (dims); |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1339 mx_diff_op (src.data (), ret.fortran_vec (), l, n, u, order); |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1340 |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1341 return ret; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1342 } |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1343 |
9721
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1344 // Fast extra-precise summation. According to |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1345 // T. Ogita, S. M. Rump, S. Oishi: |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1346 // Accurate Sum And Dot Product, |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1347 // SIAM J. Sci. Computing, Vol. 26, 2005 |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1348 |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1349 template <class T> |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
1350 inline void twosum_accum (T& s, T& e, |
9721
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1351 const T& x) |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1352 { |
18084
8e056300994b
Follow coding convention of defining and initializing only 1 variable per line in liboctave.
Rik <rik@octave.org>
parents:
17769
diff
changeset
|
1353 T s1 = s + x; |
8e056300994b
Follow coding convention of defining and initializing only 1 variable per line in liboctave.
Rik <rik@octave.org>
parents:
17769
diff
changeset
|
1354 T t = s1 - s; |
8e056300994b
Follow coding convention of defining and initializing only 1 variable per line in liboctave.
Rik <rik@octave.org>
parents:
17769
diff
changeset
|
1355 T e1 = (s - (s1 - t)) + (x - t); |
9721
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1356 s = s1; |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1357 e += e1; |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1358 } |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1359 |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1360 template <class T> |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1361 inline T |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
1362 mx_inline_xsum (const T *v, octave_idx_type n) |
9721
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1363 { |
18098
6c706a83070f
Tweak cset 8e056300994b defining 1 var per line in liboctave.
Rik <rik@octave.org>
parents:
18084
diff
changeset
|
1364 T s, e; |
6c706a83070f
Tweak cset 8e056300994b defining 1 var per line in liboctave.
Rik <rik@octave.org>
parents:
18084
diff
changeset
|
1365 s = e = 0; |
9721
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1366 for (octave_idx_type i = 0; i < n; i++) |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1367 twosum_accum (s, e, v[i]); |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1368 |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1369 return s + e; |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1370 } |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1371 |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1372 template <class T> |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1373 inline void |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
1374 mx_inline_xsum (const T *v, T *r, |
9721
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1375 octave_idx_type m, octave_idx_type n) |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1376 { |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1377 OCTAVE_LOCAL_BUFFER (T, e, m); |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1378 for (octave_idx_type i = 0; i < m; i++) |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1379 e[i] = r[i] = T (); |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1380 |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1381 for (octave_idx_type j = 0; j < n; j++) |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1382 { |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1383 for (octave_idx_type i = 0; i < m; i++) |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1384 twosum_accum (r[i], e[i], v[i]); |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1385 |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1386 v += m; |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1387 } |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1388 |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1389 for (octave_idx_type i = 0; i < m; i++) |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1390 r[i] += e[i]; |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1391 } |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1392 |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1393 OP_RED_FCNN (mx_inline_xsum, T, T) |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1394 |
2804 | 1395 #endif |