nmatrix 0.2.1 → 0.2.3
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +4 -4
- data/ext/nmatrix/data/data.cpp +9 -9
- data/ext/nmatrix/data/data.h +7 -8
- data/ext/nmatrix/data/ruby_object.h +1 -4
- data/ext/nmatrix/extconf.rb +9 -127
- data/ext/nmatrix/math.cpp +25 -25
- data/ext/nmatrix/math/asum.h +10 -31
- data/ext/nmatrix/math/cblas_templates_core.h +10 -10
- data/ext/nmatrix/math/getrf.h +2 -2
- data/ext/nmatrix/math/imax.h +12 -9
- data/ext/nmatrix/math/laswp.h +3 -3
- data/ext/nmatrix/math/long_dtype.h +16 -3
- data/ext/nmatrix/math/magnitude.h +54 -0
- data/ext/nmatrix/math/nrm2.h +19 -14
- data/ext/nmatrix/math/trsm.h +40 -36
- data/ext/nmatrix/math/util.h +14 -0
- data/ext/nmatrix/nmatrix.h +39 -1
- data/ext/nmatrix/ruby_nmatrix.c +45 -83
- data/ext/nmatrix/storage/common.h +9 -3
- data/ext/nmatrix/storage/dense/dense.cpp +4 -4
- data/ext/nmatrix/storage/list/list.cpp +2 -2
- data/ext/nmatrix/storage/yale/class.h +1 -1
- data/lib/nmatrix/blas.rb +103 -34
- data/lib/nmatrix/io/fortran_format.rb +8 -5
- data/lib/nmatrix/io/harwell_boeing.rb +11 -10
- data/lib/nmatrix/io/market.rb +9 -6
- data/lib/nmatrix/io/mat5_reader.rb +54 -29
- data/lib/nmatrix/io/mat_reader.rb +26 -14
- data/lib/nmatrix/io/point_cloud.rb +19 -11
- data/lib/nmatrix/math.rb +224 -5
- data/lib/nmatrix/mkmf.rb +103 -0
- data/lib/nmatrix/nmatrix.rb +20 -6
- data/lib/nmatrix/shortcuts.rb +415 -0
- data/lib/nmatrix/version.rb +1 -1
- data/spec/00_nmatrix_spec.rb +50 -1
- data/spec/02_slice_spec.rb +21 -21
- data/spec/blas_spec.rb +25 -3
- data/spec/math_spec.rb +233 -5
- data/spec/shortcuts_spec.rb +145 -5
- data/spec/spec_helper.rb +24 -1
- metadata +20 -4
checksums.yaml
CHANGED
@@ -1,7 +1,7 @@
|
|
1
1
|
---
|
2
2
|
SHA1:
|
3
|
-
metadata.gz:
|
4
|
-
data.tar.gz:
|
3
|
+
metadata.gz: 8dd60ade9441cfcdc2c70cd18a60571110c18145
|
4
|
+
data.tar.gz: 3ae0a112c051f510718a2d7afad97327bb6808c2
|
5
5
|
SHA512:
|
6
|
-
metadata.gz:
|
7
|
-
data.tar.gz:
|
6
|
+
metadata.gz: 6d9fcb8e0ba7ebb62cd30fa7aba1aafe0bcf1dba7ce7e77a12f3e249280212c668313c94b3e80ce40528829061ef0a6e8b140eda8d8f66175023133712393f72
|
7
|
+
data.tar.gz: ba0a384a026c56c01595b7de0b3a3a741bbe021d2b44e358a14c8f5d665c020262d0b400f6eade6a20ee91d65f90e47d730cc015430db5a0539ad4d684ceb6eb
|
data/ext/nmatrix/data/data.cpp
CHANGED
@@ -198,39 +198,39 @@ const nm::dtype_t Upcast[nm::NUM_DTYPES][nm::NUM_DTYPES] = {
|
|
198
198
|
void rubyval_to_cval(VALUE val, nm::dtype_t dtype, void* loc) {
|
199
199
|
using namespace nm;
|
200
200
|
switch (dtype) {
|
201
|
-
case BYTE:
|
201
|
+
case nm::BYTE:
|
202
202
|
*reinterpret_cast<uint8_t*>(loc) = static_cast<uint8_t>(RubyObject(val));
|
203
203
|
break;
|
204
204
|
|
205
|
-
case INT8:
|
205
|
+
case nm::INT8:
|
206
206
|
*reinterpret_cast<int8_t*>(loc) = static_cast<int8_t>(RubyObject(val));
|
207
207
|
break;
|
208
208
|
|
209
|
-
case INT16:
|
209
|
+
case nm::INT16:
|
210
210
|
*reinterpret_cast<int16_t*>(loc) = static_cast<int16_t>(RubyObject(val));
|
211
211
|
break;
|
212
212
|
|
213
|
-
case INT32:
|
213
|
+
case nm::INT32:
|
214
214
|
*reinterpret_cast<int32_t*>(loc) = static_cast<int32_t>(RubyObject(val));
|
215
215
|
break;
|
216
216
|
|
217
|
-
case INT64:
|
217
|
+
case nm::INT64:
|
218
218
|
*reinterpret_cast<int64_t*>(loc) = static_cast<int64_t>(RubyObject(val));
|
219
219
|
break;
|
220
220
|
|
221
|
-
case FLOAT32:
|
221
|
+
case nm::FLOAT32:
|
222
222
|
*reinterpret_cast<float32_t*>(loc) = static_cast<float32_t>(RubyObject(val));
|
223
223
|
break;
|
224
224
|
|
225
|
-
case FLOAT64:
|
225
|
+
case nm::FLOAT64:
|
226
226
|
*reinterpret_cast<float64_t*>(loc) = static_cast<float64_t>(RubyObject(val));
|
227
227
|
break;
|
228
228
|
|
229
|
-
case COMPLEX64:
|
229
|
+
case nm::COMPLEX64:
|
230
230
|
*reinterpret_cast<Complex64*>(loc) = RubyObject(val).to<Complex64>();
|
231
231
|
break;
|
232
232
|
|
233
|
-
case COMPLEX128:
|
233
|
+
case nm::COMPLEX128:
|
234
234
|
*reinterpret_cast<Complex128*>(loc) = RubyObject(val).to<Complex128>();
|
235
235
|
break;
|
236
236
|
|
data/ext/nmatrix/data/data.h
CHANGED
@@ -121,18 +121,17 @@ namespace nm {
|
|
121
121
|
|
122
122
|
template <typename Type>
|
123
123
|
Complex<Type>& Complex<Type>::operator=(const RubyObject& other) {
|
124
|
-
|
125
|
-
case T_COMPLEX:
|
124
|
+
if (RB_TYPE_P(other.rval, T_COMPLEX)) {
|
126
125
|
this->r = NUM2DBL(rb_funcall(other.rval, rb_intern("real"), 0));
|
127
126
|
this->i = NUM2DBL(rb_funcall(other.rval, rb_intern("imag"), 0));
|
128
|
-
|
129
|
-
|
130
|
-
|
131
|
-
|
127
|
+
}
|
128
|
+
else if (RB_TYPE_P(other.rval, T_FLOAT) ||
|
129
|
+
RB_TYPE_P(other.rval, T_FIXNUM) ||
|
130
|
+
RB_TYPE_P(other.rval, T_BIGNUM)) {
|
132
131
|
this->r = NUM2DBL(other.rval);
|
133
132
|
this->i = 0.0;
|
134
|
-
|
135
|
-
|
133
|
+
}
|
134
|
+
else {
|
136
135
|
rb_raise(rb_eTypeError, "not sure how to convert this type of VALUE to a complex");
|
137
136
|
}
|
138
137
|
return *this;
|
@@ -45,10 +45,7 @@
|
|
45
45
|
/*
|
46
46
|
* Macros
|
47
47
|
*/
|
48
|
-
#define NM_RUBYVAL_IS_NUMERIC(val) (FIXNUM_P(val) or (
|
49
|
-
#define NMATRIX_CHECK_TYPE(val) \
|
50
|
-
if (TYPE(val) != T_DATA || (RDATA(val)->dfree != (RUBY_DATA_FUNC)nm_delete && RDATA(val)->dfree != (RUBY_DATA_FUNC)nm_delete_ref)) \
|
51
|
-
rb_raise(rb_eTypeError, "Expected NMatrix on left-hand side of operation.");
|
48
|
+
#define NM_RUBYVAL_IS_NUMERIC(val) (FIXNUM_P(val) or RB_FLOAT_TYPE_P(val) or RB_TYPE_P(val, T_COMPLEX))
|
52
49
|
|
53
50
|
/*
|
54
51
|
* Classes and Functions
|
data/ext/nmatrix/extconf.rb
CHANGED
@@ -25,74 +25,18 @@
|
|
25
25
|
# This file checks for ATLAS and other necessary headers, and
|
26
26
|
# generates a Makefile for compiling NMatrix.
|
27
27
|
|
28
|
-
require "mkmf"
|
28
|
+
require File.expand_path("../../../lib/nmatrix/mkmf", __FILE__)
|
29
29
|
|
30
|
-
|
31
|
-
|
32
|
-
|
33
|
-
|
34
|
-
|
35
|
-
|
36
|
-
|
37
|
-
#include <ruby.h>
|
38
|
-
SRC
|
39
|
-
|
40
|
-
|
41
|
-
src << <<"SRC" unless header.nil?
|
42
|
-
#include <#{header}>
|
43
|
-
SRC
|
44
|
-
|
45
|
-
r = try_link(src + <<"SRC")
|
46
|
-
int main() { return 0; }
|
47
|
-
int t() { #{type} a; return 0; }
|
48
|
-
SRC
|
49
|
-
|
50
|
-
unless r
|
51
|
-
print "no\n"
|
52
|
-
return false
|
53
|
-
end
|
54
|
-
|
55
|
-
$defs.push(format("-DHAVE_%s", type.upcase))
|
56
|
-
|
57
|
-
print "yes\n"
|
58
|
-
|
59
|
-
return true
|
60
|
-
end
|
61
|
-
|
62
|
-
# Function derived from NArray's extconf.rb.
|
63
|
-
def create_conf_h(file) #:nodoc:
|
64
|
-
print "creating #{file}\n"
|
65
|
-
File.open(file, 'w') do |hfile|
|
66
|
-
header_guard = file.upcase.sub(/\s|\./, '_')
|
67
|
-
|
68
|
-
hfile.puts "#ifndef #{header_guard}"
|
69
|
-
hfile.puts "#define #{header_guard}"
|
70
|
-
hfile.puts
|
71
|
-
|
72
|
-
# FIXME: Find a better way to do this:
|
73
|
-
hfile.puts "#define RUBY_2 1" if RUBY_VERSION >= '2.0'
|
74
|
-
|
75
|
-
for line in $defs
|
76
|
-
line =~ /^-D(.*)/
|
77
|
-
hfile.printf "#define %s 1\n", $1
|
78
|
-
end
|
79
|
-
|
80
|
-
hfile.puts
|
81
|
-
hfile.puts "#endif"
|
82
|
-
end
|
83
|
-
end
|
84
|
-
|
85
|
-
if RUBY_VERSION < '1.9'
|
86
|
-
raise(NotImplementedError, "Sorry, you need at least Ruby 1.9!")
|
87
|
-
else
|
88
|
-
$INSTALLFILES = [['nmatrix.h', '$(archdir)'], ['nmatrix.hpp', '$(archdir)'], ['nmatrix_config.h', '$(archdir)'], ['nm_memory.h', '$(archdir)']]
|
89
|
-
if /cygwin|mingw/ =~ RUBY_PLATFORM
|
90
|
-
$INSTALLFILES << ['libnmatrix.a', '$(archdir)']
|
91
|
-
end
|
92
|
-
end
|
30
|
+
$INSTALLFILES = [
|
31
|
+
['nmatrix.h' , '$(archdir)'],
|
32
|
+
['nmatrix.hpp' , '$(archdir)'],
|
33
|
+
['nmatrix_config.h', '$(archdir)'],
|
34
|
+
['nm_memory.h' , '$(archdir)'],
|
35
|
+
['ruby_constants.h', '$(archdir)']
|
36
|
+
]
|
93
37
|
|
94
38
|
if /cygwin|mingw/ =~ RUBY_PLATFORM
|
95
|
-
|
39
|
+
$INSTALLFILES << ['libnmatrix.a', '$(archdir)']
|
96
40
|
end
|
97
41
|
|
98
42
|
$DEBUG = true
|
@@ -105,70 +49,8 @@ basenames = %w{nmatrix ruby_constants data/data util/io math util/sl_list storag
|
|
105
49
|
$objs = basenames.map { |b| "#{b}.o" }
|
106
50
|
$srcs = basenames.map { |b| "#{b}.cpp" }
|
107
51
|
|
108
|
-
def find_newer_gplusplus #:nodoc:
|
109
|
-
print "checking for apparent GNU g++ binary with C++0x/C++11 support... "
|
110
|
-
[9,8,7,6,5,4,3].each do |minor|
|
111
|
-
ver = "4.#{minor}"
|
112
|
-
gpp = "g++-#{ver}"
|
113
|
-
result = `which #{gpp}`
|
114
|
-
next if result.empty?
|
115
|
-
CONFIG['CXX'] = gpp
|
116
|
-
puts ver
|
117
|
-
return CONFIG['CXX']
|
118
|
-
end
|
119
|
-
false
|
120
|
-
end
|
121
|
-
|
122
|
-
def gplusplus_version
|
123
|
-
cxxvar = proc { |n| `#{CONFIG['CXX']} -E -dM - </dev/null | grep #{n}`.chomp.split(' ')[2] }
|
124
|
-
major = cxxvar.call('__GNUC__')
|
125
|
-
minor = cxxvar.call('__GNUC_MINOR__')
|
126
|
-
patch = cxxvar.call('__GNUC_PATCHLEVEL__')
|
127
|
-
|
128
|
-
raise("unable to determine g++ version (match to get version was nil)") if major.nil? || minor.nil? || patch.nil?
|
129
|
-
|
130
|
-
"#{major}.#{minor}.#{patch}"
|
131
|
-
end
|
132
|
-
|
133
|
-
|
134
|
-
if CONFIG['CXX'] == 'clang++'
|
135
|
-
$CXX_STANDARD = 'c++11'
|
136
|
-
|
137
|
-
else
|
138
|
-
version = gplusplus_version
|
139
|
-
if version < '4.3.0' && CONFIG['CXX'] == 'g++' # see if we can find a newer G++, unless it's been overridden by user
|
140
|
-
if !find_newer_gplusplus
|
141
|
-
raise("You need a version of g++ which supports -std=c++0x or -std=c++11. If you're on a Mac and using Homebrew, we recommend using mac-brew-gcc.sh to install a more recent g++.")
|
142
|
-
end
|
143
|
-
version = gplusplus_version
|
144
|
-
end
|
145
|
-
|
146
|
-
if version < '4.7.0'
|
147
|
-
$CXX_STANDARD = 'c++0x'
|
148
|
-
else
|
149
|
-
$CXX_STANDARD = 'c++11'
|
150
|
-
end
|
151
|
-
puts "using C++ standard... #{$CXX_STANDARD}"
|
152
|
-
puts "g++ reports version... " + `#{CONFIG['CXX']} --version|head -n 1|cut -f 3 -d " "`
|
153
|
-
end
|
154
|
-
|
155
52
|
#$libs += " -lprofiler "
|
156
53
|
|
157
|
-
# For release, these next two should both be changed to -O3.
|
158
|
-
$CFLAGS += " -O3 "
|
159
|
-
#$CFLAGS += " -static -O0 -g "
|
160
|
-
$CXXFLAGS += " -O3 -std=#{$CXX_STANDARD} " #-fmax-errors=10 -save-temps
|
161
|
-
#$CXXFLAGS += " -static -O0 -g -std=#{$CXX_STANDARD} "
|
162
|
-
|
163
|
-
CONFIG['warnflags'].gsub!('-Wshorten-64-to-32', '') # doesn't work except in Mac-patched gcc (4.2)
|
164
|
-
CONFIG['warnflags'].gsub!('-Wdeclaration-after-statement', '')
|
165
|
-
CONFIG['warnflags'].gsub!('-Wimplicit-function-declaration', '')
|
166
|
-
|
167
|
-
have_func("rb_array_const_ptr", "ruby.h")
|
168
|
-
have_macro("FIX_CONST_VALUE_PTR", "ruby.h")
|
169
|
-
have_macro("RARRAY_CONST_PTR", "ruby.h")
|
170
|
-
have_macro("RARRAY_AREF", "ruby.h")
|
171
|
-
|
172
54
|
create_conf_h("nmatrix_config.h")
|
173
55
|
create_makefile("nmatrix")
|
174
56
|
|
data/ext/nmatrix/math.cpp
CHANGED
@@ -9,8 +9,8 @@
|
|
9
9
|
//
|
10
10
|
// == Copyright Information
|
11
11
|
//
|
12
|
-
// SciRuby is Copyright (c) 2010 -
|
13
|
-
// NMatrix is Copyright (c) 2012 -
|
12
|
+
// SciRuby is Copyright (c) 2010 - present, Ruby Science Foundation
|
13
|
+
// NMatrix is Copyright (c) 2012 - present, John Woods and the Ruby Science Foundation
|
14
14
|
//
|
15
15
|
// Please see LICENSE.txt for additional copyright notices.
|
16
16
|
//
|
@@ -134,6 +134,7 @@
|
|
134
134
|
#include "math/cblas_enums.h"
|
135
135
|
|
136
136
|
#include "data/data.h"
|
137
|
+
#include "math/magnitude.h"
|
137
138
|
#include "math/imax.h"
|
138
139
|
#include "math/scal.h"
|
139
140
|
#include "math/laswp.h"
|
@@ -237,11 +238,14 @@ namespace nm {
|
|
237
238
|
int col_index[M];
|
238
239
|
|
239
240
|
for (int k = 0;k < M; ++k) {
|
240
|
-
DType akk
|
241
|
+
typename MagnitudeDType<DType>::type akk;
|
242
|
+
akk = magnitude( matrix[k * (M + 1)] ); // diagonal element
|
243
|
+
|
241
244
|
int interchange = k;
|
242
245
|
|
243
246
|
for (int row = k + 1; row < M; ++row) {
|
244
|
-
DType big
|
247
|
+
typename MagnitudeDType<DType>::type big;
|
248
|
+
big = magnitude( matrix[M*row + k] ); // element below the temp pivot
|
245
249
|
|
246
250
|
if ( big > akk ) {
|
247
251
|
interchange = row;
|
@@ -694,16 +698,12 @@ static VALUE nm_cblas_rot(VALUE self, VALUE n, VALUE x, VALUE incx, VALUE y, VAL
|
|
694
698
|
static VALUE nm_cblas_nrm2(VALUE self, VALUE n, VALUE x, VALUE incx) {
|
695
699
|
|
696
700
|
static void (*ttable[nm::NUM_DTYPES])(const int N, const void* X, const int incX, void* sum) = {
|
697
|
-
/* nm::math::cblas_nrm2<uint8_t,uint8_t>,
|
698
|
-
nm::math::cblas_nrm2<int8_t,int8_t>,
|
699
|
-
nm::math::cblas_nrm2<int16_t,int16_t>,
|
700
|
-
nm::math::cblas_nrm2<int32_t,int32_t>, */
|
701
701
|
NULL, NULL, NULL, NULL, NULL, // no help for integers
|
702
|
-
nm::math::cblas_nrm2<float32_t
|
703
|
-
nm::math::cblas_nrm2<float64_t
|
704
|
-
nm::math::cblas_nrm2<
|
705
|
-
nm::math::cblas_nrm2<
|
706
|
-
nm::math::cblas_nrm2<nm::RubyObject
|
702
|
+
nm::math::cblas_nrm2<float32_t>,
|
703
|
+
nm::math::cblas_nrm2<float64_t>,
|
704
|
+
nm::math::cblas_nrm2<nm::Complex64>,
|
705
|
+
nm::math::cblas_nrm2<nm::Complex128>,
|
706
|
+
nm::math::cblas_nrm2<nm::RubyObject>
|
707
707
|
};
|
708
708
|
|
709
709
|
nm::dtype_t dtype = NM_DTYPE(x);
|
@@ -748,16 +748,16 @@ static VALUE nm_cblas_nrm2(VALUE self, VALUE n, VALUE x, VALUE incx) {
|
|
748
748
|
static VALUE nm_cblas_asum(VALUE self, VALUE n, VALUE x, VALUE incx) {
|
749
749
|
|
750
750
|
static void (*ttable[nm::NUM_DTYPES])(const int N, const void* X, const int incX, void* sum) = {
|
751
|
-
nm::math::cblas_asum<uint8_t
|
752
|
-
nm::math::cblas_asum<int8_t
|
753
|
-
nm::math::cblas_asum<int16_t
|
754
|
-
nm::math::cblas_asum<int32_t
|
755
|
-
nm::math::cblas_asum<int64_t
|
756
|
-
nm::math::cblas_asum<float32_t
|
757
|
-
nm::math::cblas_asum<float64_t
|
758
|
-
nm::math::cblas_asum<
|
759
|
-
nm::math::cblas_asum<
|
760
|
-
nm::math::cblas_asum<nm::RubyObject
|
751
|
+
nm::math::cblas_asum<uint8_t>,
|
752
|
+
nm::math::cblas_asum<int8_t>,
|
753
|
+
nm::math::cblas_asum<int16_t>,
|
754
|
+
nm::math::cblas_asum<int32_t>,
|
755
|
+
nm::math::cblas_asum<int64_t>,
|
756
|
+
nm::math::cblas_asum<float32_t>,
|
757
|
+
nm::math::cblas_asum<float64_t>,
|
758
|
+
nm::math::cblas_asum<nm::Complex64>,
|
759
|
+
nm::math::cblas_asum<nm::Complex128>,
|
760
|
+
nm::math::cblas_asum<nm::RubyObject>
|
761
761
|
};
|
762
762
|
|
763
763
|
nm::dtype_t dtype = NM_DTYPE(x);
|
@@ -998,7 +998,7 @@ static VALUE nm_clapack_getrs(VALUE self, VALUE order, VALUE trans, VALUE n, VAL
|
|
998
998
|
|
999
999
|
// Allocate the C version of the pivot index array
|
1000
1000
|
int* ipiv_;
|
1001
|
-
if (
|
1001
|
+
if (!RB_TYPE_P(ipiv, T_ARRAY)) {
|
1002
1002
|
rb_raise(rb_eArgError, "ipiv must be of type Array");
|
1003
1003
|
} else {
|
1004
1004
|
ipiv_ = NM_ALLOCA_N(int, RARRAY_LEN(ipiv));
|
@@ -1050,7 +1050,7 @@ static VALUE nm_clapack_laswp(VALUE self, VALUE n, VALUE a, VALUE lda, VALUE k1,
|
|
1050
1050
|
|
1051
1051
|
// Allocate the C version of the pivot index array
|
1052
1052
|
int* ipiv_;
|
1053
|
-
if (
|
1053
|
+
if (!RB_TYPE_P(ipiv, T_ARRAY)) {
|
1054
1054
|
rb_raise(rb_eArgError, "ipiv must be of type Array");
|
1055
1055
|
} else {
|
1056
1056
|
ipiv_ = NM_ALLOCA_N(int, RARRAY_LEN(ipiv));
|
data/ext/nmatrix/math/asum.h
CHANGED
@@ -9,8 +9,8 @@
|
|
9
9
|
//
|
10
10
|
// == Copyright Information
|
11
11
|
//
|
12
|
-
// SciRuby is Copyright (c) 2010 -
|
13
|
-
// NMatrix is Copyright (c) 2012 -
|
12
|
+
// SciRuby is Copyright (c) 2010 - present, Ruby Science Foundation
|
13
|
+
// NMatrix is Copyright (c) 2012 - present, John Woods and the Ruby Science Foundation
|
14
14
|
//
|
15
15
|
// Please see LICENSE.txt for additional copyright notices.
|
16
16
|
//
|
@@ -60,6 +60,8 @@
|
|
60
60
|
#define ASUM_H
|
61
61
|
|
62
62
|
|
63
|
+
#include "math/magnitude.h"
|
64
|
+
|
63
65
|
namespace nm { namespace math {
|
64
66
|
|
65
67
|
/*
|
@@ -73,44 +75,21 @@ namespace nm { namespace math {
|
|
73
75
|
* complex64 -> float or double
|
74
76
|
* complex128 -> double
|
75
77
|
*/
|
76
|
-
template <typename
|
77
|
-
inline
|
78
|
-
|
79
|
-
if ((N > 0) && (incX > 0)) {
|
80
|
-
for (int i = 0; i < N; ++i) {
|
81
|
-
sum += std::abs(X[i*incX]);
|
82
|
-
}
|
83
|
-
}
|
84
|
-
return sum;
|
85
|
-
}
|
86
|
-
|
87
|
-
|
88
|
-
template <>
|
89
|
-
inline float asum(const int N, const Complex64* X, const int incX) {
|
90
|
-
float sum = 0;
|
91
|
-
if ((N > 0) && (incX > 0)) {
|
92
|
-
for (int i = 0; i < N; ++i) {
|
93
|
-
sum += std::abs(X[i*incX].r) + std::abs(X[i*incX].i);
|
94
|
-
}
|
95
|
-
}
|
96
|
-
return sum;
|
97
|
-
}
|
98
|
-
|
99
|
-
template <>
|
100
|
-
inline double asum(const int N, const Complex128* X, const int incX) {
|
101
|
-
double sum = 0;
|
78
|
+
template <typename DType, typename MDType = typename MagnitudeDType<DType>::type>
|
79
|
+
inline MDType asum(const int N, const DType* X, const int incX) {
|
80
|
+
MDType sum = 0;
|
102
81
|
if ((N > 0) && (incX > 0)) {
|
103
82
|
for (int i = 0; i < N; ++i) {
|
104
|
-
sum +=
|
83
|
+
sum += magnitude(X[i*incX]);
|
105
84
|
}
|
106
85
|
}
|
107
86
|
return sum;
|
108
87
|
}
|
109
88
|
|
110
89
|
|
111
|
-
template <typename
|
90
|
+
template <typename DType, typename MDType = typename MagnitudeDType<DType>::type>
|
112
91
|
inline void cblas_asum(const int N, const void* X, const int incX, void* sum) {
|
113
|
-
*reinterpret_cast<
|
92
|
+
*reinterpret_cast<MDType*>( sum ) = asum<DType,MDType>( N, reinterpret_cast<const DType*>(X), incX );
|
114
93
|
}
|
115
94
|
|
116
95
|
|
@@ -107,9 +107,9 @@ inline void cblas_rot(const int N, void* X, const int incX, void* Y, const int i
|
|
107
107
|
* complex64 -> float or double
|
108
108
|
* complex128 -> double
|
109
109
|
*/
|
110
|
-
template <typename
|
111
|
-
inline
|
112
|
-
return nm::math::asum<
|
110
|
+
template <typename DType, typename MDType = typename MagnitudeDType<DType>::type>
|
111
|
+
inline MDType asum(const int N, const DType* X, const int incX) {
|
112
|
+
return nm::math::asum<DType,MDType>(N,X,incX);
|
113
113
|
}
|
114
114
|
|
115
115
|
|
@@ -134,9 +134,9 @@ inline double asum(const int N, const Complex128* X, const int incX) {
|
|
134
134
|
}
|
135
135
|
|
136
136
|
|
137
|
-
template <typename
|
137
|
+
template <typename DType, typename MDType = typename MagnitudeDType<DType>::type>
|
138
138
|
inline void cblas_asum(const int N, const void* X, const int incX, void* sum) {
|
139
|
-
*static_cast<
|
139
|
+
*static_cast<MDType*>( sum ) = asum<DType, MDType>( N, static_cast<const DType*>(X), incX );
|
140
140
|
}
|
141
141
|
|
142
142
|
/*
|
@@ -149,9 +149,9 @@ inline void cblas_asum(const int N, const void* X, const int incX, void* sum) {
|
|
149
149
|
* complex64 -> float or double
|
150
150
|
* complex128 -> double
|
151
151
|
*/
|
152
|
-
template <typename
|
153
|
-
inline
|
154
|
-
return nm::math::nrm2<
|
152
|
+
template <typename DType, typename MDType = typename MagnitudeDType<DType>::type>
|
153
|
+
inline MDType nrm2(const int N, const DType* X, const int incX) {
|
154
|
+
return nm::math::nrm2<DType,MDType>(N, X, incX);
|
155
155
|
}
|
156
156
|
|
157
157
|
|
@@ -175,9 +175,9 @@ inline double nrm2(const int N, const Complex128* X, const int incX) {
|
|
175
175
|
return cblas_dznrm2(N, X, incX);
|
176
176
|
}
|
177
177
|
|
178
|
-
template <typename
|
178
|
+
template <typename DType, typename MDType = typename MagnitudeDType<DType>::type>
|
179
179
|
inline void cblas_nrm2(const int N, const void* X, const int incX, void* result) {
|
180
|
-
*static_cast<
|
180
|
+
*static_cast<MDType*>( result ) = nrm2<DType, MDType>( N, static_cast<const DType*>(X), incX );
|
181
181
|
}
|
182
182
|
|
183
183
|
//imax
|