Mercurial > hg > octave-nkf
annotate liboctave/operators/mx-inlines.cc @ 18098:6c706a83070f
Tweak cset 8e056300994b defining 1 var per line in liboctave.
author | Rik <rik@octave.org> |
---|---|
date | Thu, 05 Dec 2013 09:48:11 -0800 |
parents | 8e056300994b |
children | 8cc66f091584 |
rev | line source |
---|---|
3 | 1 /* |
2 | |
17744
d63878346099
maint: Update copyright notices for release.
John W. Eaton <jwe@octave.org>
parents:
15285
diff
changeset
|
3 Copyright (C) 1993-2013 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> |
10146
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
30 #include <memory> |
2804 | 31 |
5525 | 32 #include "quit.h" |
33 | |
1650 | 34 #include "oct-cmplx.h" |
8758
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
35 #include "oct-locbuf.h" |
8897 | 36 #include "oct-inttypes.h" |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
37 #include "Array.h" |
10146
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
38 #include "Array-util.h" |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
39 |
13004
d9d65c3017c3
Make bsxfun automatic for most binary operators.
Jordi Gutiérrez Hermoso <jordigh@gmail.com>
parents:
11586
diff
changeset
|
40 #include "bsxfun.h" |
d9d65c3017c3
Make bsxfun automatic for most binary operators.
Jordi Gutiérrez Hermoso <jordigh@gmail.com>
parents:
11586
diff
changeset
|
41 |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
42 // Provides some commonly repeated, basic loop templates. |
461 | 43 |
8380
dbe67764e628
fix & improve speed of diagonal matrix multiplication
Jaroslav Hajek <highegg@gmail.com>
parents:
7789
diff
changeset
|
44 template <class R, class S> |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
45 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
|
46 { 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
|
47 |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
48 #define DEFMXUNOP(F, OP) \ |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
49 template <class R, class X> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
50 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
|
51 { 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
|
52 |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
53 DEFMXUNOP (mx_inline_uminus, -) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
54 |
9607
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
55 #define DEFMXUNOPEQ(F, OP) \ |
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
56 template <class R> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
57 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
|
58 { 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
|
59 |
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
60 DEFMXUNOPEQ (mx_inline_uminus2, -) |
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
61 |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
62 #define DEFMXUNBOOLOP(F, OP) \ |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
63 template <class X> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
64 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
|
65 { 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
|
66 |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
67 DEFMXUNBOOLOP (mx_inline_iszero, ==) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
68 DEFMXUNBOOLOP (mx_inline_notzero, !=) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
69 |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
70 #define DEFMXBINOP(F, OP) \ |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
71 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
|
72 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
|
73 { 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
|
74 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
|
75 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
|
76 { 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
|
77 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
|
78 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
|
79 { 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
|
80 |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
81 DEFMXBINOP (mx_inline_add, +) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
82 DEFMXBINOP (mx_inline_sub, -) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
83 DEFMXBINOP (mx_inline_mul, *) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
84 DEFMXBINOP (mx_inline_div, /) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
85 |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
86 #define DEFMXBINOPEQ(F, OP) \ |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
87 template <class R, class X> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
88 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
|
89 { 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
|
90 template <class R, class X> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
91 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
|
92 { 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
|
93 |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
94 DEFMXBINOPEQ (mx_inline_add2, +=) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
95 DEFMXBINOPEQ (mx_inline_sub2, -=) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
96 DEFMXBINOPEQ (mx_inline_mul2, *=) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
97 DEFMXBINOPEQ (mx_inline_div2, /=) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
98 |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
99 #define DEFMXCMPOP(F, OP) \ |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
100 template <class X, class Y> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
101 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
|
102 { 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
|
103 template <class X, class Y> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
104 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
|
105 { 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
|
106 template <class X, class Y> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
107 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
|
108 { 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
|
109 |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
110 DEFMXCMPOP (mx_inline_lt, <) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
111 DEFMXCMPOP (mx_inline_le, <=) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
112 DEFMXCMPOP (mx_inline_gt, >) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
113 DEFMXCMPOP (mx_inline_ge, >=) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
114 DEFMXCMPOP (mx_inline_eq, ==) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
115 DEFMXCMPOP (mx_inline_ne, !=) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
116 |
9553
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
117 // Convert to logical value, for logical op purposes. |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
118 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
|
119 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
|
120 { 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
|
121 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
|
122 { return x.value (); } |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
123 |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
124 template <class X> |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
125 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
|
126 { |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
127 for (size_t i = 0; i < n; i++) |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
128 r[i] = ! logical_value (x[i]); |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
129 } |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
130 |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
131 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
|
132 { |
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
133 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
|
134 } |
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
135 |
9553
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
136 #define DEFMXBOOLOP(F, NOT1, OP, NOT2) \ |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
137 template <class X, class Y> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
138 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
|
139 { \ |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
140 for (size_t i = 0; i < n; i++) \ |
9553
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
141 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
|
142 } \ |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
143 template <class X, class Y> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
144 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
|
145 { \ |
9553
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
146 const bool yy = (NOT2 logical_value (y)); \ |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
147 for (size_t i = 0; i < n; i++) \ |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
148 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
|
149 } \ |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
150 template <class X, class Y> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
151 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
|
152 { \ |
9553
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
153 const bool xx = (NOT1 logical_value (x)); \ |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
154 for (size_t i = 0; i < n; i++) \ |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
155 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
|
156 } |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
157 |
9553
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
158 DEFMXBOOLOP (mx_inline_and, , &, ) |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
159 DEFMXBOOLOP (mx_inline_or, , |, ) |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
160 DEFMXBOOLOP (mx_inline_not_and, !, &, ) |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
161 DEFMXBOOLOP (mx_inline_not_or, !, |, ) |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
162 DEFMXBOOLOP (mx_inline_and_not, , &, !) |
0c72d9284087
further bool ops tweaks
Jaroslav Hajek <highegg@gmail.com>
parents:
9550
diff
changeset
|
163 DEFMXBOOLOP (mx_inline_or_not, , |, !) |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
164 |
9612
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
165 #define DEFMXBOOLOPEQ(F, OP) \ |
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
166 template <class X> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
167 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
|
168 { \ |
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
169 for (size_t i = 0; i < n; i++) \ |
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
170 r[i] OP logical_value (x[i]); \ |
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
171 } \ |
13139
aa4a23337a0f
Enable BSX in-place for missing assignment operators
Jordi Gutiérrez Hermoso <jordigh@octave.org>
parents:
13005
diff
changeset
|
172 template <class X> \ |
aa4a23337a0f
Enable BSX in-place for missing assignment operators
Jordi Gutiérrez Hermoso <jordigh@octave.org>
parents:
13005
diff
changeset
|
173 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
|
174 { 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
|
175 |
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
176 DEFMXBOOLOPEQ (mx_inline_and2, &=) |
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
177 DEFMXBOOLOPEQ (mx_inline_or2, |=) |
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
178 |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
179 template <class T> |
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
180 inline bool |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
181 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
|
182 { |
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
183 for (size_t i = 0; i < n; i++) |
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
184 { |
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
185 if (xisnan (x[i])) |
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
186 return true; |
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
187 } |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
188 |
9814
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
189 return false; |
8380
dbe67764e628
fix & improve speed of diagonal matrix multiplication
Jaroslav Hajek <highegg@gmail.com>
parents:
7789
diff
changeset
|
190 } |
dbe67764e628
fix & improve speed of diagonal matrix multiplication
Jaroslav Hajek <highegg@gmail.com>
parents:
7789
diff
changeset
|
191 |
10900
b64803a8be4e
optimize element-wise sparse-dense multiplication and division
Jaroslav Hajek <highegg@gmail.com>
parents:
10643
diff
changeset
|
192 template <class T> |
b64803a8be4e
optimize element-wise sparse-dense multiplication and division
Jaroslav Hajek <highegg@gmail.com>
parents:
10643
diff
changeset
|
193 inline bool |
b64803a8be4e
optimize element-wise sparse-dense multiplication and division
Jaroslav Hajek <highegg@gmail.com>
parents:
10643
diff
changeset
|
194 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
|
195 { |
b64803a8be4e
optimize element-wise sparse-dense multiplication and division
Jaroslav Hajek <highegg@gmail.com>
parents:
10643
diff
changeset
|
196 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
|
197 { |
b64803a8be4e
optimize element-wise sparse-dense multiplication and division
Jaroslav Hajek <highegg@gmail.com>
parents:
10643
diff
changeset
|
198 if (! xfinite (x[i])) |
b64803a8be4e
optimize element-wise sparse-dense multiplication and division
Jaroslav Hajek <highegg@gmail.com>
parents:
10643
diff
changeset
|
199 return false; |
b64803a8be4e
optimize element-wise sparse-dense multiplication and division
Jaroslav Hajek <highegg@gmail.com>
parents:
10643
diff
changeset
|
200 } |
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 return true; |
b64803a8be4e
optimize element-wise sparse-dense multiplication and division
Jaroslav Hajek <highegg@gmail.com>
parents:
10643
diff
changeset
|
203 } |
b64803a8be4e
optimize element-wise sparse-dense multiplication and division
Jaroslav Hajek <highegg@gmail.com>
parents:
10643
diff
changeset
|
204 |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
205 template <class T> |
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
206 inline bool |
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
207 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
|
208 { |
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
209 for (size_t i = 0; i < n; i++) |
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
210 { |
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
211 if (x[i] < 0) |
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
212 return true; |
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
213 } |
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 return false; |
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
216 } |
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
217 |
13756
6dfebfa334cb
allow negative data log plots with OpenGL+FLTK graphics (bug #34232)
John W. Eaton <jwe@octave.org>
parents:
13139
diff
changeset
|
218 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
|
219 inline bool |
6dfebfa334cb
allow negative data log plots with OpenGL+FLTK graphics (bug #34232)
John W. Eaton <jwe@octave.org>
parents:
13139
diff
changeset
|
220 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
|
221 { |
6dfebfa334cb
allow negative data log plots with OpenGL+FLTK graphics (bug #34232)
John W. Eaton <jwe@octave.org>
parents:
13139
diff
changeset
|
222 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
|
223 { |
6dfebfa334cb
allow negative data log plots with OpenGL+FLTK graphics (bug #34232)
John W. Eaton <jwe@octave.org>
parents:
13139
diff
changeset
|
224 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
|
225 return true; |
6dfebfa334cb
allow negative data log plots with OpenGL+FLTK graphics (bug #34232)
John W. Eaton <jwe@octave.org>
parents:
13139
diff
changeset
|
226 } |
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 return false; |
6dfebfa334cb
allow negative data log plots with OpenGL+FLTK graphics (bug #34232)
John W. Eaton <jwe@octave.org>
parents:
13139
diff
changeset
|
229 } |
6dfebfa334cb
allow negative data log plots with OpenGL+FLTK graphics (bug #34232)
John W. Eaton <jwe@octave.org>
parents:
13139
diff
changeset
|
230 |
9814
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
231 template<class T> |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
232 inline bool |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
233 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
|
234 { |
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
235 for (size_t i = 0; i < n; i++) |
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
236 { |
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
237 if (x[i].imag () != 0) |
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
238 return false; |
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
239 } |
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 return true; |
2b29f3472e20
add a couple of useful loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9800
diff
changeset
|
242 } |
2811 | 243 |
9800
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
244 #define DEFMXMAPPER(F, FUN) \ |
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
245 template <class T> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
246 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
|
247 { 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
|
248 |
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
249 template<class T> |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
250 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
|
251 { 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
|
252 template<class T> |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
253 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
|
254 { 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
|
255 |
9743
26abff55f6fe
optimize bsxfun for common built-in operations
Jaroslav Hajek <highegg@gmail.com>
parents:
9721
diff
changeset
|
256 // Pairwise minimums/maximums |
9800
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
257 #define DEFMXMAPPER2(F, FUN) \ |
9743
26abff55f6fe
optimize bsxfun for common built-in operations
Jaroslav Hajek <highegg@gmail.com>
parents:
9721
diff
changeset
|
258 template <class T> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
259 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
|
260 { 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
|
261 template <class T> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
262 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
|
263 { 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
|
264 template <class T> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
265 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
|
266 { for (size_t i = 0; i < n; i++) r[i] = FUN (x, y[i]); } |
3 | 267 |
9800
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
268 DEFMXMAPPER2 (mx_inline_xmin, xmin) |
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
269 DEFMXMAPPER2 (mx_inline_xmax, xmax) |
2811 | 270 |
10146
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
271 // Specialize array-scalar max/min |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
272 #define DEFMINMAXSPEC(T, F, OP) \ |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
273 template <> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
274 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
|
275 { \ |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
276 if (xisnan (y)) \ |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
277 std::memcpy (r, x, n * sizeof (T)); \ |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
278 else \ |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
279 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
|
280 } \ |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
281 template <> \ |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
282 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
|
283 { \ |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
284 if (xisnan (x)) \ |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
285 std::memcpy (r, y, n * sizeof (T)); \ |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
286 else \ |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
287 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
|
288 } |
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
289 |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
290 DEFMINMAXSPEC (double, mx_inline_xmin, <=) |
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
291 DEFMINMAXSPEC (double, mx_inline_xmax, >=) |
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
292 DEFMINMAXSPEC (float, mx_inline_xmin, <=) |
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
293 DEFMINMAXSPEC (float, mx_inline_xmax, >=) |
10146
9597eea7fa36
inline xmin/xmax & optimize special cases
Jaroslav Hajek <highegg@gmail.com>
parents:
9827
diff
changeset
|
294 |
9827
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
295 // Pairwise power |
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
296 #define DEFMXMAPPER2X(F, FUN) \ |
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
297 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
|
298 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
|
299 { 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
|
300 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
|
301 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
|
302 { 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
|
303 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
|
304 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
|
305 { 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
|
306 |
13005
4061106b1c4b
Enable automatic bsxfun for power operators
Jordi Gutiérrez Hermoso <jordigh@gmail.com>
parents:
13004
diff
changeset
|
307 // 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
|
308 // arguments provided. |
4061106b1c4b
Enable automatic bsxfun for power operators
Jordi Gutiérrez Hermoso <jordigh@gmail.com>
parents:
13004
diff
changeset
|
309 using std::pow; |
4061106b1c4b
Enable automatic bsxfun for power operators
Jordi Gutiérrez Hermoso <jordigh@gmail.com>
parents:
13004
diff
changeset
|
310 DEFMXMAPPER2X (mx_inline_pow, pow) |
9827
c15a5ed0da58
optimize bsxfun (@power, ...)
Jaroslav Hajek <highegg@gmail.com>
parents:
9814
diff
changeset
|
311 |
9800
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
312 // 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
|
313 // inlining. |
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
314 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
|
315 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
|
316 { 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
|
317 |
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
318 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
|
319 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
|
320 { for (size_t i = 0; i < n; i++) r[i] = fun (x[i]); } |
3 | 321 |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
322 // 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
|
323 // a pointer, to allow the compiler reduce number of instances. |
3 | 324 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
325 template <class R, class X> |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
326 inline Array<R> |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
327 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
|
328 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
|
329 { |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
330 Array<R> r (x.dims ()); |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
331 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
|
332 return r; |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
333 } |
2811 | 334 |
9800
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
335 // Shortcuts for applying mx_inline_map. |
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
336 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
337 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
|
338 inline Array<R> |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
339 do_mx_unary_map (const Array<X>& x) |
9800
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
340 { |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
341 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
|
342 } |
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
343 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
344 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
|
345 inline Array<R> |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
346 do_mx_unary_map (const Array<X>& x) |
9800
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
347 { |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
348 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
|
349 } |
ef4c4186cb47
improve some mx_inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
9766
diff
changeset
|
350 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
351 template <class R> |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
352 inline Array<R>& |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
353 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
|
354 void (*op) (size_t, R *) throw ()) |
9607
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
355 { |
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
356 op (r.numel (), r.fortran_vec ()); |
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
357 return r; |
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
358 } |
1be3c73ed7b5
reuse temporary arrays in nested expressions
Jaroslav Hajek <highegg@gmail.com>
parents:
9578
diff
changeset
|
359 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
360 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
|
361 inline Array<R> |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
362 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
|
363 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
|
364 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
|
365 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
|
366 const char *opname) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
367 { |
18084
8e056300994b
Follow coding convention of defining and initializing only 1 variable per line in liboctave.
Rik <rik@octave.org>
parents:
17769
diff
changeset
|
368 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
|
369 dim_vector dy = y.dims (); |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
370 if (dx == dy) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
371 { |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
372 Array<R> r (dx); |
9612
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
373 op (r.length (), r.fortran_vec (), x.data (), y.data ()); |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
374 return r; |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
375 } |
14056
c3d401562410
allow warning (or error) for automatic bsxfun
John W. Eaton <jwe@octave.org>
parents:
13756
diff
changeset
|
376 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
|
377 { |
d9d65c3017c3
Make bsxfun automatic for most binary operators.
Jordi Gutiérrez Hermoso <jordigh@gmail.com>
parents:
11586
diff
changeset
|
378 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
|
379 } |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
380 else |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
381 { |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
382 gripe_nonconformant (opname, dx, dy); |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
383 return Array<R> (); |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
384 } |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
385 } |
3 | 386 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
387 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
|
388 inline Array<R> |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
389 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
|
390 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
|
391 { |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
392 Array<R> r (x.dims ()); |
9612
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
393 op (r.length (), r.fortran_vec (), x.data (), y); |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
394 return r; |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
395 } |
3 | 396 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
397 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
|
398 inline Array<R> |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
399 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
|
400 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
|
401 { |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
402 Array<R> r (y.dims ()); |
9612
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
403 op (r.length (), r.fortran_vec (), x, y.data ()); |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
404 return r; |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
405 } |
3 | 406 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
407 template <class R, class X> |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
408 inline Array<R>& |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
409 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
|
410 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
|
411 void (*op1) (size_t, R *, X) throw (), |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
412 const char *opname) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
413 { |
18084
8e056300994b
Follow coding convention of defining and initializing only 1 variable per line in liboctave.
Rik <rik@octave.org>
parents:
17769
diff
changeset
|
414 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
|
415 dim_vector dx = x.dims (); |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
416 if (dr == dx) |
13139
aa4a23337a0f
Enable BSX in-place for missing assignment operators
Jordi Gutiérrez Hermoso <jordigh@octave.org>
parents:
13005
diff
changeset
|
417 { |
aa4a23337a0f
Enable BSX in-place for missing assignment operators
Jordi Gutiérrez Hermoso <jordigh@octave.org>
parents:
13005
diff
changeset
|
418 op (r.length (), r.fortran_vec (), x.data ()); |
aa4a23337a0f
Enable BSX in-place for missing assignment operators
Jordi Gutiérrez Hermoso <jordigh@octave.org>
parents:
13005
diff
changeset
|
419 } |
14056
c3d401562410
allow warning (or error) for automatic bsxfun
John W. Eaton <jwe@octave.org>
parents:
13756
diff
changeset
|
420 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
|
421 { |
aa4a23337a0f
Enable BSX in-place for missing assignment operators
Jordi Gutiérrez Hermoso <jordigh@octave.org>
parents:
13005
diff
changeset
|
422 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
|
423 } |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
424 else |
9612
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
425 gripe_nonconformant (opname, dr, dx); |
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
426 return r; |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
427 } |
7789
82be108cc558
First attempt at single precision tyeps
David Bateman <dbateman@free.fr>
parents:
7017
diff
changeset
|
428 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
429 template <class R, class X> |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
430 inline Array<R>& |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
431 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
|
432 void (*op) (size_t, R *, X) throw ()) |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
433 { |
9612
66970dd627f6
further liboctave design improvements
Jaroslav Hajek <highegg@gmail.com>
parents:
9607
diff
changeset
|
434 op (r.length (), r.fortran_vec (), x); |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
435 return r; |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
436 } |
3 | 437 |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
438 template <class T1, class T2> |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
439 inline bool |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
440 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
|
441 { |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
442 for (size_t i = 0; i < n; i++) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
443 if (x[i] != y[i]) |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
444 return false; |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
445 return true; |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
446 } |
3 | 447 |
10481
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
448 template <class T> |
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
449 inline bool |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
450 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
|
451 bool (*op) (size_t, const T *) throw ()) |
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
452 { |
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
453 return op (a.numel (), a.data ()); |
e8811e5dd699
avoid exception throwing in mx-inline loops
Jaroslav Hajek <highegg@gmail.com>
parents:
10365
diff
changeset
|
454 } |
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 // 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
|
457 // magic to avoid underflows, which we don't need here. |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
458 template <class T> |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
459 inline T cabsq (const std::complex<T>& c) |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
460 { return c.real () * c.real () + c.imag () * c.imag (); } |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
461 |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
462 // default. works for integers and bool. |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
463 template <class T> |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
464 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
|
465 template <class T> |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
466 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
|
467 // for octave_ints |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
468 template <class T> |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
469 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
|
470 template <class T> |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
471 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
|
472 // 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
|
473 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
|
474 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
|
475 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
|
476 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
|
477 // Ditto for complex. |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
478 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
|
479 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
|
480 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
|
481 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
|
482 |
9550
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
483 #define OP_RED_SUM(ac, el) ac += el |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
484 #define OP_RED_PROD(ac, el) ac *= el |
3d6a9aea2aea
refactor binary & bool ops in liboctave
Jaroslav Hajek <highegg@gmail.com>
parents:
9513
diff
changeset
|
485 #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
|
486 #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
|
487 |
15018
3d8ace26c5b4
maint: Use Octave coding conventions for cuddled parentheses in liboctave/.
Rik <rik@octave.org>
parents:
14846
diff
changeset
|
488 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
|
489 { ac += el; } |
15018
3d8ace26c5b4
maint: Use Octave coding conventions for cuddled parentheses in liboctave/.
Rik <rik@octave.org>
parents:
14846
diff
changeset
|
490 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
|
491 { ac += el; } // FIXME: guaranteed? |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
492 template <class T> |
15018
3d8ace26c5b4
maint: Use Octave coding conventions for cuddled parentheses in liboctave/.
Rik <rik@octave.org>
parents:
14846
diff
changeset
|
493 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
|
494 { ac += el.double_value (); } |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
495 |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
496 // 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
|
497 #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
|
498 #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
|
499 |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
500 #define OP_RED_FCN(F, TSRC, TRES, OP, ZERO) \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
501 template <class T> \ |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
502 inline TRES \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
503 F (const TSRC* v, octave_idx_type n) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
504 { \ |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
505 TRES ac = ZERO; \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
506 for (octave_idx_type i = 0; i < n; i++) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
507 OP(ac, v[i]); \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
508 return ac; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
509 } |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
510 |
9721
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
511 #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
|
512 |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
513 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
|
514 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
|
515 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
|
516 OP_RED_FCN (mx_inline_prod, T, T, OP_RED_PROD, 1) |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
517 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
|
518 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
|
519 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
|
520 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
|
521 |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
522 |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
523 #define OP_RED_FCN2(F, TSRC, TRES, OP, ZERO) \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
524 template <class T> \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
525 inline void \ |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
526 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
|
527 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
528 for (octave_idx_type i = 0; i < m; i++) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
529 r[i] = ZERO; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
530 for (octave_idx_type j = 0; j < n; j++) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
531 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
532 for (octave_idx_type i = 0; i < m; i++) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
533 OP(r[i], v[i]); \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
534 v += m; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
535 } \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
536 } |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
537 |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
538 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
|
539 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
|
540 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
|
541 OP_RED_FCN2 (mx_inline_prod, T, T, OP_RED_PROD, 1) |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
542 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
|
543 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
|
544 |
10147
adc0143e9419
optimize any/all (x, 2) with small number of rows
Jaroslav Hajek <highegg@gmail.com>
parents:
10146
diff
changeset
|
545 #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
|
546 #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
|
547 |
adc0143e9419
optimize any/all (x, 2) with small number of rows
Jaroslav Hajek <highegg@gmail.com>
parents:
10146
diff
changeset
|
548 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
|
549 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
|
550 |
8758
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
551 // 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
|
552 // 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
|
553 // 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
|
554 |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
555 #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
|
556 template <class T> \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
557 inline void \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
558 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
|
559 { \ |
10147
adc0143e9419
optimize any/all (x, 2) with small number of rows
Jaroslav Hajek <highegg@gmail.com>
parents:
10146
diff
changeset
|
560 if (n <= 8) \ |
adc0143e9419
optimize any/all (x, 2) with small number of rows
Jaroslav Hajek <highegg@gmail.com>
parents:
10146
diff
changeset
|
561 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
|
562 \ |
8758
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
563 /* 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
|
564 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
|
565 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
|
566 octave_idx_type nact = m; \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
567 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
|
568 { \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
569 octave_idx_type k = 0; \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
570 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
|
571 { \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
572 octave_idx_type ia = iact[i]; \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
573 if (! PRED (v[ia])) \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
574 iact[k++] = ia; \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
575 } \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
576 nact = k; \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
577 v += m; \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
578 } \ |
83c9d60c3c47
implement short-circuiting row-reduction any/all algorithm
Jaroslav Hajek <highegg@gmail.com>
parents:
8756
diff
changeset
|
579 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
|
580 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
|
581 } |
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 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
|
584 OP_ROW_SHORT_CIRCUIT (mx_inline_all, xis_false, true) |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
585 |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
586 #define OP_RED_FCNN(F, TSRC, TRES) \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
587 template <class T> \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
588 inline void \ |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
589 F (const TSRC *v, TRES *r, octave_idx_type l, \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
590 octave_idx_type n, octave_idx_type u) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
591 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
592 if (l == 1) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
593 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
594 for (octave_idx_type i = 0; i < u; i++) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
595 { \ |
8756
d0755c9db5ed
implement fast logical sum (counting)
Jaroslav Hajek <highegg@gmail.com>
parents:
8751
diff
changeset
|
596 r[i] = F<T> (v, n); \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
597 v += n; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
598 } \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
599 } \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
600 else \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
601 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
602 for (octave_idx_type i = 0; i < u; i++) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
603 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
604 F (v, r, l, n); \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
605 v += l*n; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
606 r += l; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
607 } \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
608 } \ |
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 |
8743
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
611 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
|
612 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
|
613 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
|
614 OP_RED_FCNN (mx_inline_prod, T, T) |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
615 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
|
616 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
|
617 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
|
618 OP_RED_FCNN (mx_inline_all, T, bool) |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
619 |
8780
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
620 #define OP_CUM_FCN(F, TSRC, TRES, OP) \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
621 template <class T> \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
622 inline void \ |
8780
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
623 F (const TSRC *v, TRES *r, octave_idx_type n) \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
624 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
625 if (n) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
626 { \ |
8780
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
627 TRES t = r[0] = v[0]; \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
628 for (octave_idx_type i = 1; i < n; i++) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
629 r[i] = t = t OP v[i]; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
630 } \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
631 } |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
632 |
8780
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
633 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
|
634 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
|
635 OP_CUM_FCN (mx_inline_cumcount, bool, T, +) |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
636 |
8780
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
637 #define OP_CUM_FCN2(F, TSRC, TRES, OP) \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
638 template <class T> \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
639 inline void \ |
8780
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
640 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
|
641 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
642 if (n) \ |
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 for (octave_idx_type i = 0; i < m; i++) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
645 r[i] = v[i]; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
646 const T *r0 = r; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
647 for (octave_idx_type j = 1; j < n; j++) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
648 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
649 r += m; v += m; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
650 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
|
651 r[i] = r0[i] OP v[i]; \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
652 r0 += m; \ |
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 } \ |
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 |
8780
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
657 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
|
658 OP_CUM_FCN2 (mx_inline_cumprod, T, T, *) |
10643
9852264314d1
fix cumulative logical sum
Jaroslav Hajek <highegg@gmail.com>
parents:
10482
diff
changeset
|
659 OP_CUM_FCN2 (mx_inline_cumcount, bool, T, +) |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
660 |
8780
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
661 #define OP_CUM_FCNN(F, TSRC, TRES) \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
662 template <class T> \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
663 inline void \ |
8780
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
664 F (const TSRC *v, TRES *r, octave_idx_type l, \ |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
665 octave_idx_type n, octave_idx_type u) \ |
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 if (l == 1) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
668 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
669 for (octave_idx_type i = 0; i < u; i++) \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
670 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
671 F (v, r, n); \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
672 v += n; r += n; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
673 } \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
674 } \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
675 else \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
676 { \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
677 for (octave_idx_type i = 0; i < u; i++) \ |
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 F (v, r, l, n); \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
680 v += l*n; \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
681 r += l*n; \ |
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 } \ |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
684 } |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
685 |
8780
ea76466605ba
support native cumsum, gripe on overflow in sum/cumsum
Jaroslav Hajek <highegg@gmail.com>
parents:
8777
diff
changeset
|
686 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
|
687 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
|
688 OP_CUM_FCNN (mx_inline_cumcount, bool, T) |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
689 |
8751
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
690 #define OP_MINMAX_FCN(F, OP) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
691 template <class T> \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
692 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
|
693 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
694 if (! n) return; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
695 T tmp = v[0]; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
696 octave_idx_type i = 1; \ |
8776
d23c33ec6bd3
fix min/max behaviour with NaNs
Jaroslav Hajek <highegg@gmail.com>
parents:
8759
diff
changeset
|
697 if (xisnan (tmp)) \ |
d23c33ec6bd3
fix min/max behaviour with NaNs
Jaroslav Hajek <highegg@gmail.com>
parents:
8759
diff
changeset
|
698 { \ |
d23c33ec6bd3
fix min/max behaviour with NaNs
Jaroslav Hajek <highegg@gmail.com>
parents:
8759
diff
changeset
|
699 for (; i < n && xisnan (v[i]); i++) ; \ |
d23c33ec6bd3
fix min/max behaviour with NaNs
Jaroslav Hajek <highegg@gmail.com>
parents:
8759
diff
changeset
|
700 if (i < n) tmp = v[i]; \ |
d23c33ec6bd3
fix min/max behaviour with NaNs
Jaroslav Hajek <highegg@gmail.com>
parents:
8759
diff
changeset
|
701 } \ |
d23c33ec6bd3
fix min/max behaviour with NaNs
Jaroslav Hajek <highegg@gmail.com>
parents:
8759
diff
changeset
|
702 for (; i < n; i++) \ |
8751
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
703 if (v[i] OP tmp) tmp = v[i]; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
704 *r = tmp; \ |
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 template <class T> \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
707 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
|
708 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
709 if (! n) return; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
710 T tmp = v[0]; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
711 octave_idx_type tmpi = 0; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
712 octave_idx_type i = 1; \ |
8776
d23c33ec6bd3
fix min/max behaviour with NaNs
Jaroslav Hajek <highegg@gmail.com>
parents:
8759
diff
changeset
|
713 if (xisnan (tmp)) \ |
d23c33ec6bd3
fix min/max behaviour with NaNs
Jaroslav Hajek <highegg@gmail.com>
parents:
8759
diff
changeset
|
714 { \ |
d23c33ec6bd3
fix min/max behaviour with NaNs
Jaroslav Hajek <highegg@gmail.com>
parents:
8759
diff
changeset
|
715 for (; i < n && xisnan (v[i]); i++) ; \ |
d23c33ec6bd3
fix min/max behaviour with NaNs
Jaroslav Hajek <highegg@gmail.com>
parents:
8759
diff
changeset
|
716 if (i < n) { tmp = v[i]; tmpi = i; } \ |
d23c33ec6bd3
fix min/max behaviour with NaNs
Jaroslav Hajek <highegg@gmail.com>
parents:
8759
diff
changeset
|
717 } \ |
d23c33ec6bd3
fix min/max behaviour with NaNs
Jaroslav Hajek <highegg@gmail.com>
parents:
8759
diff
changeset
|
718 for (; i < n; i++) \ |
8751
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
719 if (v[i] OP tmp) { tmp = v[i]; tmpi = i; }\ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
720 *r = tmp; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
721 *ri = tmpi; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
722 } |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
723 |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
724 OP_MINMAX_FCN (mx_inline_min, <) |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
725 OP_MINMAX_FCN (mx_inline_max, >) |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
726 |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
727 // 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
|
728 // 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
|
729 // proceed to a faster code. |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
730 |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
731 #define OP_MINMAX_FCN2(F, OP) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
732 template <class T> \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
733 inline void \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
734 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
|
735 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
736 if (! n) return; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
737 bool nan = false; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
738 octave_idx_type j = 0; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
739 for (octave_idx_type i = 0; i < m; i++) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
740 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
741 r[i] = v[i]; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
742 if (xisnan (v[i])) nan = true; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
743 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
744 j++; v += m; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
745 while (nan && j < n) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
746 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
747 nan = false; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
748 for (octave_idx_type i = 0; i < m; i++) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
749 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
750 if (xisnan (v[i])) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
751 nan = true; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
752 else if (xisnan (r[i]) || v[i] OP r[i]) \ |
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 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
755 j++; v += m; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
756 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
757 while (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 for (octave_idx_type i = 0; i < m; i++) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
760 if (v[i] OP r[i]) r[i] = v[i]; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
761 j++; v += m; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
762 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
763 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
764 template <class T> \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
765 inline void \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
766 F (const T *v, T *r, octave_idx_type *ri, \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
767 octave_idx_type m, octave_idx_type n) \ |
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 if (! n) return; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
770 bool nan = false; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
771 octave_idx_type j = 0; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
772 for (octave_idx_type i = 0; i < m; i++) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
773 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
774 r[i] = v[i]; ri[i] = j; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
775 if (xisnan (v[i])) nan = true; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
776 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
777 j++; v += m; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
778 while (nan && j < n) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
779 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
780 nan = false; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
781 for (octave_idx_type i = 0; i < m; i++) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
782 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
783 if (xisnan (v[i])) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
784 nan = true; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
785 else if (xisnan (r[i]) || v[i] OP r[i]) \ |
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 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
788 j++; v += m; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
789 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
790 while (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 for (octave_idx_type i = 0; i < m; i++) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
793 if (v[i] OP r[i]) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
794 { r[i] = v[i]; ri[i] = j; } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
795 j++; v += m; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
796 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
797 } |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
798 |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
799 OP_MINMAX_FCN2 (mx_inline_min, <) |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
800 OP_MINMAX_FCN2 (mx_inline_max, >) |
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 #define OP_MINMAX_FCNN(F) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
803 template <class T> \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
804 inline void \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
805 F (const T *v, T *r, octave_idx_type l, \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
806 octave_idx_type n, octave_idx_type u) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
807 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
808 if (! n) return; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
809 if (l == 1) \ |
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 for (octave_idx_type i = 0; i < u; i++) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
812 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
813 F (v, r, n); \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
814 v += n; r++; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
815 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
816 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
817 else \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
818 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
819 for (octave_idx_type i = 0; i < u; i++) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
820 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
821 F (v, r, l, n); \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
822 v += l*n; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
823 r += l; \ |
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 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
826 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
827 template <class T> \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
828 inline void \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
829 F (const T *v, T *r, octave_idx_type *ri, \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
830 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
|
831 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
832 if (! n) return; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
833 if (l == 1) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
834 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
835 for (octave_idx_type i = 0; i < u; i++) \ |
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 F (v, r, ri, n); \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
838 v += n; r++; ri++; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
839 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
840 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
841 else \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
842 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
843 for (octave_idx_type i = 0; i < u; i++) \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
844 { \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
845 F (v, r, ri, l, n); \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
846 v += l*n; \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
847 r += l; ri += l; \ |
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 } \ |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
850 } |
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 OP_MINMAX_FCNN (mx_inline_min) |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
853 OP_MINMAX_FCNN (mx_inline_max) |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
854 |
8777
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
855 #define OP_CUMMINMAX_FCN(F, OP) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
856 template <class T> \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
857 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
|
858 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
859 if (! n) return; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
860 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
|
861 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
|
862 octave_idx_type j = 0; \ |
8777
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
863 if (xisnan (tmp)) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
864 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
865 for (; i < n && xisnan (v[i]); i++) ; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
866 for (; j < i; j++) r[j] = tmp; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
867 if (i < n) tmp = v[i]; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
868 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
869 for (; i < n; i++) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
870 if (v[i] OP tmp) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
871 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
872 for (; j < i; j++) r[j] = tmp; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
873 tmp = v[i]; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
874 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
875 for (; j < i; j++) r[j] = 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 template <class T> \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
878 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
|
879 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
880 if (! n) return; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
881 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
|
882 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
|
883 octave_idx_type j = 0; \ |
8777
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
884 if (xisnan (tmp)) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
885 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
886 for (; i < n && xisnan (v[i]); i++) ; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
887 for (; j < i; j++) { r[j] = tmp; ri[j] = tmpi; } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
888 if (i < n) { tmp = v[i]; tmpi = i; } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
889 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
890 for (; i < n; i++) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
891 if (v[i] OP tmp) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
892 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
893 for (; j < i; j++) { r[j] = tmp; ri[j] = tmpi; } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
894 tmp = v[i]; tmpi = i; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
895 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
896 for (; j < i; j++) { r[j] = tmp; ri[j] = tmpi; } \ |
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 |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
899 OP_CUMMINMAX_FCN (mx_inline_cummin, <) |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
900 OP_CUMMINMAX_FCN (mx_inline_cummax, >) |
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 // 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
|
903 // 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
|
904 // proceed to a faster code. |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
905 |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
906 #define OP_CUMMINMAX_FCN2(F, OP) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
907 template <class T> \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
908 inline void \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
909 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
|
910 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
911 if (! n) return; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
912 bool nan = false; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
913 const T *r0; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
914 octave_idx_type j = 0; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
915 for (octave_idx_type i = 0; i < m; i++) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
916 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
917 r[i] = v[i]; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
918 if (xisnan (v[i])) nan = true; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
919 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
920 j++; v += m; r0 = r; r += m; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
921 while (nan && j < 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 nan = false; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
924 for (octave_idx_type i = 0; i < m; i++) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
925 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
926 if (xisnan (v[i])) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
927 { r[i] = r0[i]; nan = true; } \ |
8949
e31d47f2c9bb
fixes to row-reducing cummin/cummax
Jaroslav Hajek <highegg@gmail.com>
parents:
8920
diff
changeset
|
928 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
|
929 r[i] = v[i]; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
930 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
931 j++; v += m; r0 = r; r += m; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
932 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
933 while (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 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
|
936 if (v[i] OP r0[i]) \ |
8777
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
937 r[i] = v[i]; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
938 else \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
939 r[i] = r0[i]; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
940 j++; v += m; r0 = r; r += m; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
941 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
942 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
943 template <class T> \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
944 inline void \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
945 F (const T *v, T *r, octave_idx_type *ri, \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
946 octave_idx_type m, octave_idx_type n) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
947 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
948 if (! n) return; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
949 bool nan = false; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
950 const T *r0; const octave_idx_type *r0i; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
951 octave_idx_type j = 0; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
952 for (octave_idx_type i = 0; i < m; i++) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
953 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
954 r[i] = v[i]; ri[i] = 0; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
955 if (xisnan (v[i])) nan = true; \ |
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 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
|
958 while (nan && j < n) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
959 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
960 nan = false; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
961 for (octave_idx_type i = 0; i < m; i++) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
962 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
963 if (xisnan (v[i])) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
964 { 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
|
965 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
|
966 { r[i] = v[i]; ri[i] = j; }\ |
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 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
|
969 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
970 while (j < n) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
971 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
972 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
|
973 if (v[i] OP r0[i]) \ |
8777
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
974 { r[i] = v[i]; ri[i] = j; } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
975 else \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
976 { r[i] = r0[i]; ri[i] = r0i[i]; } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
977 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
|
978 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
979 } |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
980 |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
981 OP_CUMMINMAX_FCN2 (mx_inline_cummin, <) |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
982 OP_CUMMINMAX_FCN2 (mx_inline_cummax, >) |
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 #define OP_CUMMINMAX_FCNN(F) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
985 template <class T> \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
986 inline void \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
987 F (const T *v, T *r, octave_idx_type l, \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
988 octave_idx_type n, octave_idx_type u) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
989 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
990 if (! n) return; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
991 if (l == 1) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
992 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
993 for (octave_idx_type i = 0; i < u; i++) \ |
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 F (v, r, n); \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
996 v += n; r += n; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
997 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
998 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
999 else \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1000 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1001 for (octave_idx_type i = 0; i < u; i++) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1002 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1003 F (v, r, l, n); \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1004 v += l*n; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1005 r += l*n; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1006 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1007 } \ |
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 template <class T> \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1010 inline void \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1011 F (const T *v, T *r, octave_idx_type *ri, \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1012 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
|
1013 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1014 if (! n) return; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1015 if (l == 1) \ |
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, ri, n); \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1020 v += n; r += n; ri += n; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1021 } \ |
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 else \ |
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 for (octave_idx_type i = 0; i < u; i++) \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1026 { \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1027 F (v, r, ri, l, n); \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1028 v += l*n; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1029 r += l*n; ri += l*n; \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1030 } \ |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1031 } \ |
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 |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1034 OP_CUMMINMAX_FCNN (mx_inline_cummin) |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1035 OP_CUMMINMAX_FCNN (mx_inline_cummax) |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1036 |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1037 template <class T> |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1038 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
|
1039 octave_idx_type order) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1040 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1041 switch (order) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1042 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1043 case 1: |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1044 for (octave_idx_type i = 0; i < n-1; i++) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1045 r[i] = v[i+1] - v[i]; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1046 break; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1047 case 2: |
9702
9ecd35a606e3
avoid some warnings from g++
John W. Eaton <jwe@octave.org>
parents:
9612
diff
changeset
|
1048 if (n > 1) |
10314
07ebe522dac2
untabify liboctave C++ sources
John W. Eaton <jwe@octave.org>
parents:
10158
diff
changeset
|
1049 { |
07ebe522dac2
untabify liboctave C++ sources
John W. Eaton <jwe@octave.org>
parents:
10158
diff
changeset
|
1050 T lst = v[1] - v[0]; |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1051 for (octave_idx_type i = 0; i < n-2; i++) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1052 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1053 T dif = v[i+2] - v[i+1]; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1054 r[i] = dif - lst; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1055 lst = dif; |
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 } |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1058 break; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1059 default: |
17769
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1060 { |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1061 OCTAVE_LOCAL_BUFFER (T, buf, n-1); |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1062 |
17769
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1063 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
|
1064 buf[i] = v[i+1] - v[i]; |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1065 |
17769
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1066 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
|
1067 { |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1068 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
|
1069 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
|
1070 } |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1071 |
17769
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1072 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
|
1073 r[i] = buf[i]; |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1074 } |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1075 } |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1076 } |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1077 |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1078 template <class T> |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
1079 void mx_inline_diff (const T *v, T *r, |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1080 octave_idx_type m, octave_idx_type n, |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1081 octave_idx_type order) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1082 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1083 switch (order) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1084 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1085 case 1: |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1086 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
|
1087 r[i] = v[i+m] - v[i]; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1088 break; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1089 case 2: |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1090 for (octave_idx_type i = 0; i < n-2; i++) |
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 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
|
1093 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
|
1094 } |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1095 break; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1096 default: |
17769
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1097 { |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1098 OCTAVE_LOCAL_BUFFER (T, buf, n-1); |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1099 |
17769
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1100 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
|
1101 { |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1102 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
|
1103 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
|
1104 |
17769
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1105 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
|
1106 { |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1107 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
|
1108 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
|
1109 } |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1110 |
17769
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1111 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
|
1112 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
|
1113 } |
49a5a4be04a1
maint: Use GNU style coding conventions for code in liboctave/
Rik <rik@octave.org>
parents:
17744
diff
changeset
|
1114 } |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1115 } |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1116 } |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1117 |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1118 template <class T> |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1119 inline void |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1120 mx_inline_diff (const T *v, T *r, |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1121 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
|
1122 octave_idx_type order) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1123 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1124 if (! n) return; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1125 if (l == 1) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1126 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1127 for (octave_idx_type i = 0; i < u; i++) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1128 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1129 mx_inline_diff (v, r, n, order); |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1130 v += n; r += n-order; |
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 else |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1134 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1135 for (octave_idx_type i = 0; i < u; i++) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1136 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1137 mx_inline_diff (v, r, l, n, order); |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1138 v += l*n; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1139 r += l*(n-order); |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1140 } |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1141 } |
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 |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1144 // Assistant function |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1145 |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1146 inline void |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1147 get_extent_triplet (const dim_vector& dims, int& dim, |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1148 octave_idx_type& l, octave_idx_type& n, |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1149 octave_idx_type& u) |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1150 { |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1151 octave_idx_type ndims = dims.length (); |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1152 if (dim >= ndims) |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1153 { |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1154 l = dims.numel (); |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1155 n = 1; |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1156 u = 1; |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1157 } |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1158 else |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1159 { |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1160 if (dim < 0) |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1161 dim = dims.first_non_singleton (); |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1162 |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1163 // calculate extent triplet. |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1164 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
|
1165 for (octave_idx_type i = 0; i < dim; i++) |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1166 l *= dims (i); |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1167 for (octave_idx_type i = dim + 1; i < ndims; i++) |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1168 u *= dims (i); |
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 } |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1171 |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1172 // Appliers. |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1173 // 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
|
1174 // maybe it can be done without an explicit parameter? |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1175 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1176 template <class R, class T> |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1177 inline Array<R> |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1178 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
|
1179 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
|
1180 octave_idx_type, octave_idx_type)) |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1181 { |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1182 octave_idx_type l, n, u; |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1183 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
|
1184 // 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
|
1185 if (dims.length () == 2 && dims(0) == 0 && dims(1) == 0) |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
1186 dims (1) = 1; |
1bd918cfb6e2
reimplement any & all using the new reduction code
Jaroslav Hajek <highegg@gmail.com>
parents:
8736
diff
changeset
|
1187 |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1188 get_extent_triplet (dims, dim, l, n, u); |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1189 |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1190 // Reduction operation reduces the array size. |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1191 if (dim < dims.length ()) dims(dim) = 1; |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1192 dims.chop_trailing_singletons (); |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1193 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1194 Array<R> ret (dims); |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1195 mx_red_op (src.data (), ret.fortran_vec (), l, n, u); |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1196 |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1197 return ret; |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1198 } |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1199 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1200 template <class R, class T> |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1201 inline Array<R> |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1202 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
|
1203 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
|
1204 octave_idx_type, octave_idx_type)) |
8736
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 octave_idx_type l, n, u; |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1207 dim_vector dims = src.dims (); |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1208 get_extent_triplet (dims, dim, l, n, u); |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1209 |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1210 // 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
|
1211 Array<R> ret (dims); |
8736
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1212 mx_cum_op (src.data (), ret.fortran_vec (), l, n, u); |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1213 |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1214 return ret; |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1215 } |
53b4fdeacc2e
improve reduction functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8650
diff
changeset
|
1216 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1217 template <class R> |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1218 inline Array<R> |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1219 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
|
1220 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
|
1221 octave_idx_type, octave_idx_type)) |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1222 { |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1223 octave_idx_type l, n, u; |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1224 dim_vector dims = src.dims (); |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1225 get_extent_triplet (dims, dim, l, n, u); |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1226 |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1227 // 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
|
1228 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
|
1229 dims.chop_trailing_singletons (); |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1230 |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1231 Array<R> ret (dims); |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1232 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
|
1233 |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1234 return ret; |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1235 } |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1236 |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1237 template <class R> |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1238 inline Array<R> |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1239 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
|
1240 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
|
1241 octave_idx_type, octave_idx_type, octave_idx_type)) |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1242 { |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1243 octave_idx_type l, n, u; |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1244 dim_vector dims = src.dims (); |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1245 get_extent_triplet (dims, dim, l, n, u); |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1246 |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1247 // If the dimension is zero, we don't do anything. |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1248 if (dim < dims.length () && dims(dim) != 0) dims(dim) = 1; |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1249 dims.chop_trailing_singletons (); |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1250 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1251 Array<R> ret (dims); |
8751
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1252 if (idx.dims () != dims) idx = Array<octave_idx_type> (dims); |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1253 |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1254 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
|
1255 l, n, u); |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1256 |
9f7ce4bf7650
optimize min/max functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8743
diff
changeset
|
1257 return ret; |
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 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1260 template <class R> |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1261 inline Array<R> |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1262 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
|
1263 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
|
1264 octave_idx_type, octave_idx_type)) |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1265 { |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1266 octave_idx_type l, n, u; |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1267 dim_vector dims = src.dims (); |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1268 get_extent_triplet (dims, dim, l, n, u); |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1269 |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1270 Array<R> ret (dims); |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1271 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
|
1272 |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1273 return ret; |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1274 } |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1275 |
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, Array<octave_idx_type>& idx, int dim, |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1279 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
|
1280 octave_idx_type, octave_idx_type, octave_idx_type)) |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1281 { |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1282 octave_idx_type l, n, u; |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1283 dim_vector dims = src.dims (); |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1284 get_extent_triplet (dims, dim, l, n, u); |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1285 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1286 Array<R> ret (dims); |
8777
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1287 if (idx.dims () != dims) idx = Array<octave_idx_type> (dims); |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1288 |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1289 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
|
1290 l, n, u); |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1291 |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1292 return ret; |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1293 } |
724c0f46d9d4
implement cummin/cummax functions
Jaroslav Hajek <highegg@gmail.com>
parents:
8776
diff
changeset
|
1294 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1295 template <class R> |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1296 inline Array<R> |
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1297 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
|
1298 void (*mx_diff_op) (const R *, R *, |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1299 octave_idx_type, octave_idx_type, octave_idx_type, |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1300 octave_idx_type)) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1301 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1302 octave_idx_type l, n, u; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1303 if (order <= 0) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1304 return src; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1305 |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1306 dim_vector dims = src.dims (); |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1307 |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1308 get_extent_triplet (dims, dim, l, n, u); |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1309 if (dim >= dims.length ()) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1310 dims.resize (dim+1, 1); |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1311 |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1312 if (dims(dim) <= order) |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1313 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1314 dims (dim) = 0; |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1315 return Array<R> (dims); |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1316 } |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1317 else |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1318 { |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1319 dims(dim) -= order; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1320 } |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1321 |
10362
b47ab50a6aa8
simplify appliers in mx-inlines.cc
Jaroslav Hajek <highegg@gmail.com>
parents:
10314
diff
changeset
|
1322 Array<R> ret (dims); |
9513
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1323 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
|
1324 |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1325 return ret; |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1326 } |
9f870f73ab7d
implement built-in diff
Jaroslav Hajek <highegg@gmail.com>
parents:
8949
diff
changeset
|
1327 |
9721
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1328 // 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
|
1329 // 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
|
1330 // Accurate Sum And Dot Product, |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1331 // 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
|
1332 |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1333 template <class T> |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
1334 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
|
1335 const T& x) |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1336 { |
18084
8e056300994b
Follow coding convention of defining and initializing only 1 variable per line in liboctave.
Rik <rik@octave.org>
parents:
17769
diff
changeset
|
1337 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
|
1338 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
|
1339 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
|
1340 s = s1; |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1341 e += e1; |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1342 } |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1343 |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1344 template <class T> |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1345 inline T |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
1346 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
|
1347 { |
18098
6c706a83070f
Tweak cset 8e056300994b defining 1 var per line in liboctave.
Rik <rik@octave.org>
parents:
18084
diff
changeset
|
1348 T s, e; |
6c706a83070f
Tweak cset 8e056300994b defining 1 var per line in liboctave.
Rik <rik@octave.org>
parents:
18084
diff
changeset
|
1349 s = e = 0; |
9721
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1350 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
|
1351 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
|
1352 |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1353 return s + e; |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1354 } |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1355 |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1356 template <class T> |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1357 inline void |
11586
12df7854fa7c
strip trailing whitespace from source files
John W. Eaton <jwe@octave.org>
parents:
11523
diff
changeset
|
1358 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
|
1359 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
|
1360 { |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1361 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
|
1362 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
|
1363 e[i] = r[i] = T (); |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1364 |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1365 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
|
1366 { |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1367 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
|
1368 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
|
1369 |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1370 v += m; |
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 |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1373 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
|
1374 r[i] += e[i]; |
192d94cff6c1
improve sum & implement the 'extra' option, refactor some code
Jaroslav Hajek <highegg@gmail.com>
parents:
9702
diff
changeset
|
1375 } |
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 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
|
1378 |
2804 | 1379 #endif |