ryeppp 0.0.2 → 0.0.3
Sign up to get free protection for your applications and to get access to all the features.
- data/.gitignore +20 -0
- data/.rspec +2 -0
- data/Gemfile +4 -0
- data/LICENSE.txt +22 -0
- data/README.md +100 -0
- data/Rakefile +39 -0
- data/ext/templates/ryeppp.c.rb +518 -0
- data/lib/ryeppp.rb +6 -0
- data/lib/ryeppp/bench.rb +233 -0
- data/lib/ryeppp/version.rb +3 -0
- data/ryeppp.gemspec +29 -0
- data/spec/ryeppp_spec.rb +166 -0
- data/spec/spec_helper.rb +7 -0
- metadata +17 -2
data/.gitignore
ADDED
data/.rspec
ADDED
data/Gemfile
ADDED
data/LICENSE.txt
ADDED
@@ -0,0 +1,22 @@
|
|
1
|
+
Copyright (c) 2013 TODO: Write your name
|
2
|
+
|
3
|
+
MIT License
|
4
|
+
|
5
|
+
Permission is hereby granted, free of charge, to any person obtaining
|
6
|
+
a copy of this software and associated documentation files (the
|
7
|
+
"Software"), to deal in the Software without restriction, including
|
8
|
+
without limitation the rights to use, copy, modify, merge, publish,
|
9
|
+
distribute, sublicense, and/or sell copies of the Software, and to
|
10
|
+
permit persons to whom the Software is furnished to do so, subject to
|
11
|
+
the following conditions:
|
12
|
+
|
13
|
+
The above copyright notice and this permission notice shall be
|
14
|
+
included in all copies or substantial portions of the Software.
|
15
|
+
|
16
|
+
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
|
17
|
+
EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
|
18
|
+
MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
|
19
|
+
NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
|
20
|
+
LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
|
21
|
+
OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
|
22
|
+
WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
|
data/README.md
ADDED
@@ -0,0 +1,100 @@
|
|
1
|
+
# Ryeppp
|
2
|
+
|
3
|
+
This gem provides bindings to the [Yeppp!](http://www.yeppp.info/) library.
|
4
|
+
According to the documentation, "Yeppp! is a high-performance SIMD-optimized
|
5
|
+
mathematical library for x86, ARM, and MIPS processors on Windows, Android, Mac
|
6
|
+
OS X, and GNU/Linux systems."
|
7
|
+
|
8
|
+
## Installation
|
9
|
+
|
10
|
+
Add this line to your application's Gemfile:
|
11
|
+
|
12
|
+
gem 'ryeppp'
|
13
|
+
|
14
|
+
And then execute:
|
15
|
+
|
16
|
+
$ bundle
|
17
|
+
|
18
|
+
Or install it yourself as:
|
19
|
+
|
20
|
+
$ gem install ryeppp
|
21
|
+
|
22
|
+
You'll also need to [download](http://www.yeppp.info/downloads.html) the Yeppp!
|
23
|
+
library and unpack it in $HOME.
|
24
|
+
|
25
|
+
## Usage
|
26
|
+
|
27
|
+
The Ryeppp gem provides a single class `Ryeppp` with the following methods:
|
28
|
+
|
29
|
+
* `add_v64fv64f_v64f` - pairwise add 2 vectors of floats
|
30
|
+
* `add_v64sv64s_v64s` - pairwise add 2 vectors of integers
|
31
|
+
|
32
|
+
* `subtract_v64fv64f_v64f` - pairwise subtract 2 vectors of floats
|
33
|
+
* `subtract_v64sv64s_v64s` - pairwise subtract 2 vectors of integers
|
34
|
+
|
35
|
+
* `multiply_v64fs64f_v64f` - multiply a vector of floats by a float
|
36
|
+
* `multiply_v64sv64s_v64s` - multiply a vector of integers by an integer
|
37
|
+
* `multiply_v64fv64f_v64f` - pairwise multiply 2 vectors of floats
|
38
|
+
* `multiply_v64ss64s_v64s` - pairwise multiply 2 vectors of integers
|
39
|
+
|
40
|
+
* `dotproduct_v64fv64f_s64f` - compute the dot product of 2 vectors of floats
|
41
|
+
|
42
|
+
* `min_v64f_s64f` - find the minimum in a vector of floats
|
43
|
+
* `min_v64s_s64s` - find the minimum in a vector of integers
|
44
|
+
|
45
|
+
* `max_v64f_s64f` - find the maximum in a vector of floats
|
46
|
+
* `max_v64s_s64s` - find the maximum in a vector of integers
|
47
|
+
|
48
|
+
* `min_v64fv64f_v64f` - find the pairwise minimum in 2 vectors of floats
|
49
|
+
|
50
|
+
* `max_v64fv64f_v64f` - find the pairwise maximum in 2 vectors of floats
|
51
|
+
|
52
|
+
* `min_v64fs64f_v64f` - find the pairwise minimum of a vector of floats and a
|
53
|
+
constant
|
54
|
+
|
55
|
+
* `max_v64fs64f_v64f` - find the pairwise maximum of a vector of floats and a
|
56
|
+
constant
|
57
|
+
|
58
|
+
* `negate_v64f_s64f` - negate a vector of floats
|
59
|
+
* `negate_v64s_s64s` - negate a vector of integers
|
60
|
+
|
61
|
+
* `sum_v64f_s64f` - compute the sum of a vector of floats
|
62
|
+
* `sumabs_v64f_s64f` - compute the sum of the absolute values of a vector of
|
63
|
+
floats
|
64
|
+
* `sumsquares_v64f_s64f` - compute the sum of the squares of the values of a
|
65
|
+
vector of floats
|
66
|
+
|
67
|
+
* `log_v64f_v64f` - compute the natural logarithm of the values of a vector of
|
68
|
+
floats
|
69
|
+
* `exp_v64f_v64f` - compute the base-_e_ exponent of the values of a vector of
|
70
|
+
floats
|
71
|
+
* `sin_v64f_v64f` - compute the sine of the values of a vector of floats
|
72
|
+
* `cos_v64f_v64f` - compute the cosine of the values of a vector of floats
|
73
|
+
* `tan_v64f_v64f` - compute the tangent of the values of a vector of floats
|
74
|
+
|
75
|
+
* `evaluatepolynomial_v64fv64f_v64f` - evaluate the polynomial with the given
|
76
|
+
float coefficients in standard form at the given _x_-values
|
77
|
+
|
78
|
+
## Benchmarks
|
79
|
+
|
80
|
+
Since Yeppp! depends on your processor, using the library may or may not
|
81
|
+
provide performance improvements, so it's important to benchmark your code on
|
82
|
+
your own processor. A few [benchmarks](http://www.yeppp.info/benchmarks.html)
|
83
|
+
are provided, and the [examples](http://docs.yeppp.info/c/examples.html) also
|
84
|
+
provide performance comparisons.
|
85
|
+
|
86
|
+
To run the benchmarks, simply run
|
87
|
+
|
88
|
+
$ ruby lib/ryeppp/bench.rb
|
89
|
+
|
90
|
+
## TODO
|
91
|
+
|
92
|
+
Add more benchmarks.
|
93
|
+
|
94
|
+
## Contributing
|
95
|
+
|
96
|
+
1. Fork it
|
97
|
+
2. Create your feature branch (`git checkout -b my-new-feature`)
|
98
|
+
3. Commit your changes (`git commit -am 'Add some feature'`)
|
99
|
+
4. Push to the branch (`git push origin my-new-feature`)
|
100
|
+
5. Create new Pull Request
|
data/Rakefile
ADDED
@@ -0,0 +1,39 @@
|
|
1
|
+
require "bundler/gem_tasks"
|
2
|
+
require "rake/extensiontask"
|
3
|
+
require "rspec/core/rake_task"
|
4
|
+
|
5
|
+
spec = Gem::Specification.load('ryeppp.gemspec')
|
6
|
+
Rake::ExtensionTask.new('ryeppp', spec)
|
7
|
+
|
8
|
+
desc "Ryeppp unit tests"
|
9
|
+
RSpec::Core::RakeTask.new(:spec) do |t|
|
10
|
+
t.pattern = "spec/*_spec.rb"
|
11
|
+
t.verbose = true
|
12
|
+
end
|
13
|
+
|
14
|
+
desc "Ryeppp c generation"
|
15
|
+
task :c_generation do
|
16
|
+
current_dir = __FILE__.split(/\//)[0..-2].join('/')
|
17
|
+
require File.join(current_dir, 'ext/templates/ryeppp.c.rb')
|
18
|
+
c_code = [
|
19
|
+
HEADERS,
|
20
|
+
PRIMARY,
|
21
|
+
FUNCS.call('Add'),
|
22
|
+
FUNCS.call('Subtract'),
|
23
|
+
FUNCS.call('Multiply'),
|
24
|
+
DOT_PRODUCT,
|
25
|
+
MIN_MAX,
|
26
|
+
PAIRWISE_MIN_MAX,
|
27
|
+
CONSTANT_MIN_MAX,
|
28
|
+
NEGATE,
|
29
|
+
SUMS,
|
30
|
+
MATHS,
|
31
|
+
POLYNOMIAL,
|
32
|
+
INITIALIZER
|
33
|
+
].join("\n\n")
|
34
|
+
File.open(File.join(current_dir, 'ext/ryeppp/ryeppp.c'), 'w') do |f|
|
35
|
+
f.write(c_code)
|
36
|
+
end
|
37
|
+
end
|
38
|
+
|
39
|
+
task :default => [:c_generation, :compile, :spec]
|
@@ -0,0 +1,518 @@
|
|
1
|
+
MAX_SIGNED_INTEGER = 2**63 - 1
|
2
|
+
MIN_SIGNED_INTEGER = -1 * MAX_SIGNED_INTEGER
|
3
|
+
|
4
|
+
class String
|
5
|
+
def pluralize
|
6
|
+
"#{self}#{self.size > 0 && self[-1] == 'e' ? 's' : 'es'}"
|
7
|
+
end
|
8
|
+
end
|
9
|
+
|
10
|
+
def allocate_yep64_typed_array_and_assert(var_names, len_var_name, opts={})
|
11
|
+
var_names = Array(var_names)
|
12
|
+
type = opts[:type] || '{{type}}'
|
13
|
+
assert = opts[:assert].nil? ? true : opts[:assert]
|
14
|
+
%{#{var_names.map{|vn| %{Yep64#{type} *yep_#{vn} = (Yep64#{type}*)calloc(#{len_var_name}, sizeof(Yep64#{type}));}}.join("\n")}
|
15
|
+
#{assert ? var_names.map{|vn| %{assert(yep_#{vn} != NULL);}}.join("\n") : nil}}
|
16
|
+
end
|
17
|
+
|
18
|
+
def deinitialize_yeppp
|
19
|
+
%{/* Deinitialize the Yeppp! library */
|
20
|
+
status = yepLibrary_Release();
|
21
|
+
assert(status == YepStatusOk);}
|
22
|
+
end
|
23
|
+
|
24
|
+
def initialize_yeppp
|
25
|
+
%{/* Initialize the Yeppp! library */
|
26
|
+
status = yepLibrary_Init();
|
27
|
+
assert(status == YepStatusOk);}
|
28
|
+
end
|
29
|
+
|
30
|
+
def guard_integer_input_size(var_name, iteration_var_name)
|
31
|
+
%{if (rb_funcall(#{var_name}_a[#{iteration_var_name}], '>', 1, #{MAX_SIGNED_INTEGER})) {
|
32
|
+
rb_raise(rb_eRangeError, "input was too large for 64-bit signed integer");
|
33
|
+
}
|
34
|
+
if (rb_funcall(#{var_name}_a[#{iteration_var_name}], '<', 1, #{MIN_SIGNED_INTEGER})) {
|
35
|
+
rb_raise(rb_eRangeError, "input was too small for 64-bit signed integer");
|
36
|
+
}}
|
37
|
+
end
|
38
|
+
|
39
|
+
def load_ruby_array_from_yeppp_array(var_name, iteration_var_name, len_var_name, type)
|
40
|
+
%{/* Load the Ruby Array */
|
41
|
+
new_ary = rb_ary_new2(#{len_var_name});
|
42
|
+
for (#{iteration_var_name}=0; #{iteration_var_name}<#{len_var_name}; #{iteration_var_name}++) {
|
43
|
+
rb_ary_push(new_ary, #{type == 'f' ? 'DBL' : 'LONG'}2NUM((#{type == 'f' ? 'double' : 'long'})yep_#{var_name}[#{iteration_var_name}]));
|
44
|
+
}}
|
45
|
+
end
|
46
|
+
def load_ruby_array_from_yeppp_array_parameterized(var_name, iteration_var_name, len_var_name)
|
47
|
+
%{/* Load the Ruby Array */
|
48
|
+
new_ary = rb_ary_new2(#{len_var_name});
|
49
|
+
for (#{iteration_var_name}=0; #{iteration_var_name}<#{len_var_name}; #{iteration_var_name}++) {
|
50
|
+
rb_ary_push(new_ary, {{ruby_type}}2NUM(({{c_type}})yep_#{var_name}[#{iteration_var_name}]));
|
51
|
+
}}
|
52
|
+
end
|
53
|
+
def load_ruby_array_into_yeppp_array(var_name, iteration_var_name, len_var_name, type, permitted_types, opts={})
|
54
|
+
pt = permitted_types.map do |t|
|
55
|
+
case t
|
56
|
+
when :float
|
57
|
+
"TYPE(#{var_name}_a[#{iteration_var_name}]) != T_FLOAT"
|
58
|
+
when :integer
|
59
|
+
"TYPE(#{var_name}_a[#{iteration_var_name}]) != T_FIXNUM && TYPE(#{var_name}_a[#{iteration_var_name}]) != T_BIGNUM"
|
60
|
+
else
|
61
|
+
raise "Invalid permitted_type: #{t}."
|
62
|
+
end
|
63
|
+
end.join(' && ')
|
64
|
+
%{/* Load #{var_name}_a into yep_#{var_name}. */
|
65
|
+
for (#{iteration_var_name}=0; #{iteration_var_name}<#{len_var_name}; #{iteration_var_name}++) {
|
66
|
+
if (#{pt}) {
|
67
|
+
rb_raise(rb_eTypeError, "input was not all #{permitted_types.map(&:to_s).map(&:pluralize).join(' and ')}");
|
68
|
+
}
|
69
|
+
#{guard_integer_input_size(var_name, iteration_var_name)}
|
70
|
+
yep_#{var_name}[#{opts[:reverse] ? "#{len_var_name} - #{iteration_var_name} - 1" : iteration_var_name}] = (Yep64#{type})NUM2#{type == 'f' ? 'DBL' : 'LONG'}(#{var_name}_a[#{iteration_var_name}]);
|
71
|
+
}}
|
72
|
+
end
|
73
|
+
def load_ruby_array_into_yeppp_array_parameterized(var_name, iteration_var_name, len_var_name)
|
74
|
+
%{/* Load #{var_name}_a into yep_#{var_name}. */
|
75
|
+
for (#{iteration_var_name}=0; #{iteration_var_name}<#{len_var_name}; #{iteration_var_name}++) {
|
76
|
+
if (TYPE(#{var_name}_a[#{iteration_var_name}]) != {{ruby_klass}} && TYPE(#{var_name}_a[#{iteration_var_name}]) != {{alt_ruby_klass}}) {
|
77
|
+
rb_raise(rb_eTypeError, "input was not all {{ruby_klass_human}}");
|
78
|
+
}
|
79
|
+
#{guard_integer_input_size(var_name, iteration_var_name)}
|
80
|
+
yep_#{var_name}[#{iteration_var_name}] = (Yep64{{type}})NUM2{{ruby_type}}(#{var_name}_a[#{iteration_var_name}]);
|
81
|
+
}}
|
82
|
+
end
|
83
|
+
|
84
|
+
def release_array_memory(var_names)
|
85
|
+
var_names = Array(var_names)
|
86
|
+
%{/* Release the memory allocated for array#{var_names.size == 1 ? nil : 's'} */
|
87
|
+
#{var_names.map{|vn| %{free(yep_#{vn});}}.join("\n")}}
|
88
|
+
end
|
89
|
+
|
90
|
+
def typed_variants(s, opts={})
|
91
|
+
[
|
92
|
+
['s', 'long', 'LONG', 'T_FIXNUM', 'T_BIGNUM', 'integers'],
|
93
|
+
['f', 'double', 'DBL', 'T_FLOAT', 'T_FLOAT', 'floats']
|
94
|
+
].map do |(type, c_type, ruby_type, ruby_klass, alt_ruby_klass, ruby_klass_human)|
|
95
|
+
if !opts[:only_type] || opts[:only_type] == type
|
96
|
+
s.gsub(/{{type}}/, type)
|
97
|
+
.gsub(/{{c_type}}/, c_type)
|
98
|
+
.gsub(/{{ruby_type}}/, ruby_type)
|
99
|
+
.gsub(/{{ruby_klass}}/, ruby_klass)
|
100
|
+
.gsub(/{{alt_ruby_klass}}/, alt_ruby_klass)
|
101
|
+
.gsub(/{{ruby_klass_human}}/, ruby_klass_human)
|
102
|
+
else
|
103
|
+
''
|
104
|
+
end
|
105
|
+
end.map(&:strip).join("\n\n")
|
106
|
+
end
|
107
|
+
|
108
|
+
HEADERS = %{
|
109
|
+
// Include the Ruby headers and goodies
|
110
|
+
#include "ruby.h"
|
111
|
+
|
112
|
+
#include "assert.h"
|
113
|
+
|
114
|
+
#include "yepCore.h"
|
115
|
+
#include "yepLibrary.h"
|
116
|
+
#include "yepMath.h"
|
117
|
+
}.strip.freeze
|
118
|
+
|
119
|
+
PRIMARY = %{
|
120
|
+
// Defining a space for information and references about the module to be stored
|
121
|
+
// internally
|
122
|
+
VALUE cRyeppp;
|
123
|
+
}.strip.freeze
|
124
|
+
|
125
|
+
# verb_name is in [Add, Subtract, Multiply]
|
126
|
+
FUNCS = Proc.new do |verb_name|
|
127
|
+
%{#{if verb_name == 'Multiply'
|
128
|
+
typed_variants(%{
|
129
|
+
static VALUE multiply_v64{{type}}s64{{type}}_v64{{type}}(VALUE self, VALUE x, VALUE multiply_by) {
|
130
|
+
enum YepStatus status;
|
131
|
+
long i;
|
132
|
+
VALUE new_ary;
|
133
|
+
VALUE *x_a = RARRAY_PTR(x);
|
134
|
+
long l = RARRAY_LEN(x);
|
135
|
+
Yep64{{type}} mult_by = (Yep64{{type}})NUM2DBL(multiply_by);
|
136
|
+
|
137
|
+
/* Allocate arrays of inputs and outputs */
|
138
|
+
#{allocate_yep64_typed_array_and_assert(%w{x y}, 'l')}
|
139
|
+
|
140
|
+
#{initialize_yeppp}
|
141
|
+
|
142
|
+
#{load_ruby_array_into_yeppp_array_parameterized('x', 'i', 'l')}
|
143
|
+
|
144
|
+
/* Perform the operation */
|
145
|
+
status = yepCore_Multiply_V64{{type}}S64{{type}}_V64{{type}}(yep_x, mult_by, yep_y, (YepSize)l);
|
146
|
+
assert(status == YepStatusOk);
|
147
|
+
|
148
|
+
#{load_ruby_array_from_yeppp_array_parameterized('y', 'i', 'l')}
|
149
|
+
|
150
|
+
#{deinitialize_yeppp}
|
151
|
+
|
152
|
+
#{release_array_memory(%w{x y})}
|
153
|
+
|
154
|
+
return new_ary;
|
155
|
+
}
|
156
|
+
}).strip.freeze
|
157
|
+
end
|
158
|
+
}
|
159
|
+
|
160
|
+
#{typed_variants(%{
|
161
|
+
// #{verb_name} Arrays of Fixnums.
|
162
|
+
static VALUE #{verb_name.downcase}_v64{{type}}v64{{type}}_v64{{type}}(VALUE self, VALUE x, VALUE y) {
|
163
|
+
enum YepStatus status;
|
164
|
+
VALUE new_ary;
|
165
|
+
long i;
|
166
|
+
VALUE *x_a = RARRAY_PTR(x);
|
167
|
+
VALUE *y_a = RARRAY_PTR(y);
|
168
|
+
long l = RARRAY_LEN(x);
|
169
|
+
|
170
|
+
/* Allocate arrays of inputs and outputs */
|
171
|
+
#{allocate_yep64_typed_array_and_assert(%w{x y z}, 'l')}
|
172
|
+
|
173
|
+
#{initialize_yeppp}
|
174
|
+
|
175
|
+
#{load_ruby_array_into_yeppp_array_parameterized('x', 'i', 'l')}
|
176
|
+
#{load_ruby_array_into_yeppp_array_parameterized('y', 'i', 'l')}
|
177
|
+
|
178
|
+
/* Perform the addition */
|
179
|
+
status = yepCore_#{verb_name}_V64{{type}}V64{{type}}_V64{{type}}(yep_x, yep_y, yep_z, (YepSize)l);
|
180
|
+
assert(status == YepStatusOk);
|
181
|
+
|
182
|
+
#{load_ruby_array_from_yeppp_array_parameterized('z', 'i', 'l')}
|
183
|
+
|
184
|
+
#{deinitialize_yeppp}
|
185
|
+
|
186
|
+
#{release_array_memory(%w{x y z})}
|
187
|
+
|
188
|
+
return new_ary;
|
189
|
+
}
|
190
|
+
})}
|
191
|
+
}
|
192
|
+
end
|
193
|
+
|
194
|
+
DOT_PRODUCT = %{
|
195
|
+
// Get the dot product of 2 Arrays.
|
196
|
+
static VALUE dotproduct_v64fv64f_s64f(VALUE self, VALUE x, VALUE y) {
|
197
|
+
enum YepStatus status;
|
198
|
+
long i;
|
199
|
+
Yep64f dp;
|
200
|
+
VALUE *x_a = RARRAY_PTR(x);
|
201
|
+
VALUE *y_a = RARRAY_PTR(y);
|
202
|
+
long l = RARRAY_LEN(x);
|
203
|
+
|
204
|
+
/* Allocate arrays of inputs and outputs */
|
205
|
+
#{allocate_yep64_typed_array_and_assert(%w{x y}, 'l', :type => 'f')}
|
206
|
+
|
207
|
+
#{initialize_yeppp}
|
208
|
+
|
209
|
+
#{load_ruby_array_into_yeppp_array('x', 'i', 'l', 'f', [:integer, :float])}
|
210
|
+
#{load_ruby_array_into_yeppp_array('y', 'i', 'l', 'f', [:integer, :float])}
|
211
|
+
|
212
|
+
/* Perform the operation */
|
213
|
+
status = yepCore_DotProduct_V64fV64f_S64f(yep_x, yep_y, &dp, (YepSize)l);
|
214
|
+
assert(status == YepStatusOk);
|
215
|
+
|
216
|
+
#{deinitialize_yeppp}
|
217
|
+
|
218
|
+
#{release_array_memory(%w{x y})}
|
219
|
+
|
220
|
+
return DBL2NUM((double)dp);
|
221
|
+
}
|
222
|
+
}.strip
|
223
|
+
|
224
|
+
MIN_MAX = typed_variants(%w{Min Max}.map do |kind|
|
225
|
+
%{
|
226
|
+
// Get the #{kind.downcase} value from an Array.
|
227
|
+
static VALUE #{kind.downcase}_v64{{type}}_s64{{type}}(VALUE self, VALUE x) {
|
228
|
+
enum YepStatus status;
|
229
|
+
long i;
|
230
|
+
Yep64{{type}} #{kind.downcase};
|
231
|
+
VALUE *x_a = RARRAY_PTR(x);
|
232
|
+
long l = RARRAY_LEN(x);
|
233
|
+
|
234
|
+
/* Allocate arrays of inputs and outputs */
|
235
|
+
#{allocate_yep64_typed_array_and_assert('x', 'l')}
|
236
|
+
|
237
|
+
#{initialize_yeppp}
|
238
|
+
|
239
|
+
#{load_ruby_array_into_yeppp_array_parameterized('x', 'i', 'l')}
|
240
|
+
|
241
|
+
/* Perform the operation */
|
242
|
+
status = yepCore_#{kind}_V64{{type}}_S64{{type}}(yep_x, &#{kind.downcase}, (YepSize)l);
|
243
|
+
assert(status == YepStatusOk);
|
244
|
+
|
245
|
+
#{deinitialize_yeppp}
|
246
|
+
|
247
|
+
#{release_array_memory(%w{x})}
|
248
|
+
|
249
|
+
return {{ruby_type}}2NUM(({{c_type}})#{kind.downcase});
|
250
|
+
}
|
251
|
+
}.strip
|
252
|
+
end.join("\n\n"))
|
253
|
+
|
254
|
+
PAIRWISE_MIN_MAX = typed_variants(%w{Min Max}.map do |kind|
|
255
|
+
%{
|
256
|
+
// Get the pairwise #{kind.downcase}ima from Arrays.
|
257
|
+
static VALUE #{kind.downcase}_v64{{type}}v64{{type}}_v64{{type}}(VALUE self, VALUE x, VALUE y) {
|
258
|
+
enum YepStatus status;
|
259
|
+
long i;
|
260
|
+
VALUE new_ary;
|
261
|
+
VALUE *x_a = RARRAY_PTR(x);
|
262
|
+
VALUE *y_a = RARRAY_PTR(y);
|
263
|
+
long l = RARRAY_LEN(x);
|
264
|
+
|
265
|
+
/* Allocate arrays of inputs and outputs */
|
266
|
+
#{allocate_yep64_typed_array_and_assert(%w{x y z}, 'l')}
|
267
|
+
|
268
|
+
#{initialize_yeppp}
|
269
|
+
|
270
|
+
#{load_ruby_array_into_yeppp_array_parameterized('x', 'i', 'l')}
|
271
|
+
#{load_ruby_array_into_yeppp_array_parameterized('y', 'i', 'l')}
|
272
|
+
|
273
|
+
/* Perform the operation */
|
274
|
+
status = yepCore_#{kind}_V64{{type}}V64{{type}}_V64{{type}}(yep_x, yep_y, yep_z, (YepSize)l);
|
275
|
+
assert(status == YepStatusOk);
|
276
|
+
|
277
|
+
#{load_ruby_array_from_yeppp_array_parameterized('z', 'i', 'l')}
|
278
|
+
|
279
|
+
#{deinitialize_yeppp}
|
280
|
+
|
281
|
+
#{release_array_memory(%w{x y z})}
|
282
|
+
|
283
|
+
return new_ary;
|
284
|
+
}
|
285
|
+
}.strip
|
286
|
+
end.join("\n\n"), :only_type => 'f')
|
287
|
+
|
288
|
+
CONSTANT_MIN_MAX = typed_variants(%w{Min Max}.map do |kind|
|
289
|
+
%{
|
290
|
+
// Get the #{kind.downcase}ima from an Array and a constant.
|
291
|
+
static VALUE #{kind.downcase}_v64{{type}}s64{{type}}_v64{{type}}(VALUE self, VALUE x, VALUE c) {
|
292
|
+
enum YepStatus status;
|
293
|
+
long i;
|
294
|
+
VALUE new_ary;
|
295
|
+
VALUE *x_a = RARRAY_PTR(x);
|
296
|
+
long l = RARRAY_LEN(x);
|
297
|
+
Yep64f konst = (Yep64f)NUM2{{ruby_type}}(c);
|
298
|
+
|
299
|
+
/* Allocate arrays of inputs and outputs */
|
300
|
+
#{allocate_yep64_typed_array_and_assert(%w{x y}, 'l')}
|
301
|
+
|
302
|
+
#{initialize_yeppp}
|
303
|
+
|
304
|
+
#{load_ruby_array_into_yeppp_array_parameterized('x', 'i', 'l')}
|
305
|
+
|
306
|
+
/* Perform the operation */
|
307
|
+
status = yepCore_#{kind}_V64{{type}}S64{{type}}_V64{{type}}(yep_x, konst, yep_y, (YepSize)l);
|
308
|
+
assert(status == YepStatusOk);
|
309
|
+
|
310
|
+
#{load_ruby_array_from_yeppp_array_parameterized('y', 'i', 'l')}
|
311
|
+
|
312
|
+
#{deinitialize_yeppp}
|
313
|
+
|
314
|
+
#{release_array_memory(%w{x y})}
|
315
|
+
|
316
|
+
return new_ary;
|
317
|
+
}
|
318
|
+
}.strip
|
319
|
+
end.join("\n\n"), :only_type => 'f')
|
320
|
+
|
321
|
+
NEGATE = typed_variants(%{
|
322
|
+
// Negate an Array.
|
323
|
+
static VALUE negate_v64{{type}}_s64{{type}}(VALUE self, VALUE x) {
|
324
|
+
enum YepStatus status;
|
325
|
+
long i;
|
326
|
+
VALUE new_ary;
|
327
|
+
VALUE *x_a = RARRAY_PTR(x);
|
328
|
+
long l = RARRAY_LEN(x);
|
329
|
+
|
330
|
+
/* Allocate arrays of inputs and outputs */
|
331
|
+
#{allocate_yep64_typed_array_and_assert(%w{x y}, 'l')}
|
332
|
+
|
333
|
+
#{initialize_yeppp}
|
334
|
+
|
335
|
+
#{load_ruby_array_into_yeppp_array_parameterized('x', 'i', 'l')}
|
336
|
+
|
337
|
+
/* Perform the negation */
|
338
|
+
status = yepCore_Negate_V64{{type}}_V64{{type}}(yep_x, yep_y, (YepSize)l);
|
339
|
+
assert(status == YepStatusOk);
|
340
|
+
|
341
|
+
#{load_ruby_array_from_yeppp_array_parameterized('y', 'i', 'l')}
|
342
|
+
|
343
|
+
#{deinitialize_yeppp}
|
344
|
+
|
345
|
+
#{release_array_memory(%w{x y})}
|
346
|
+
|
347
|
+
return new_ary;
|
348
|
+
}
|
349
|
+
}).freeze
|
350
|
+
|
351
|
+
SUMS = %w{Sum SumAbs SumSquares}.map do |kind|
|
352
|
+
%{
|
353
|
+
static VALUE #{kind.downcase}_v64f_s64f(VALUE self, VALUE x) {
|
354
|
+
enum YepStatus status;
|
355
|
+
long i;
|
356
|
+
Yep64f sum;
|
357
|
+
long l = RARRAY_LEN(x);
|
358
|
+
VALUE *x_a = RARRAY_PTR(x);
|
359
|
+
|
360
|
+
/* Allocate arrays of inputs and outputs */
|
361
|
+
#{allocate_yep64_typed_array_and_assert('x', 'l', :type => 'f')}
|
362
|
+
|
363
|
+
#{initialize_yeppp}
|
364
|
+
|
365
|
+
#{load_ruby_array_into_yeppp_array('x', 'i', 'l', 'f', [:integer, :float])}
|
366
|
+
|
367
|
+
/* Perform the operation */
|
368
|
+
status = yepCore_#{kind}_V64f_S64f(yep_x, &sum, (YepSize)l);
|
369
|
+
assert(status == YepStatusOk);
|
370
|
+
|
371
|
+
#{deinitialize_yeppp}
|
372
|
+
|
373
|
+
#{release_array_memory(%w{x})}
|
374
|
+
|
375
|
+
return DBL2NUM((double)sum);
|
376
|
+
}
|
377
|
+
}.strip
|
378
|
+
end.join("\n\n").freeze
|
379
|
+
|
380
|
+
MATHS_KINDS = %w{Log Exp Sin Cos Tan}.freeze
|
381
|
+
MATHS = MATHS_KINDS.map do |kind|
|
382
|
+
%{
|
383
|
+
static VALUE #{kind.downcase}_v64f_v64f(VALUE self, VALUE x) {
|
384
|
+
enum YepStatus status;
|
385
|
+
long i;
|
386
|
+
VALUE new_ary;
|
387
|
+
long l = RARRAY_LEN(x);
|
388
|
+
VALUE *x_a = RARRAY_PTR(x);
|
389
|
+
|
390
|
+
/* Allocate arrays of inputs and outputs */
|
391
|
+
#{allocate_yep64_typed_array_and_assert(%w{x y}, 'l', :type => 'f')}
|
392
|
+
|
393
|
+
#{initialize_yeppp}
|
394
|
+
|
395
|
+
#{load_ruby_array_into_yeppp_array('x', 'i', 'l', 'f', [:integer, :float])}
|
396
|
+
|
397
|
+
/* Perform the operation */
|
398
|
+
status = yepMath_#{kind}_V64f_V64f(yep_x, yep_y, (YepSize)l);
|
399
|
+
assert(status == YepStatusOk);
|
400
|
+
|
401
|
+
#{load_ruby_array_from_yeppp_array('y', 'i', 'l', 'f')}
|
402
|
+
|
403
|
+
#{deinitialize_yeppp}
|
404
|
+
|
405
|
+
#{release_array_memory(%w{x y})}
|
406
|
+
|
407
|
+
return new_ary;
|
408
|
+
}
|
409
|
+
}.strip
|
410
|
+
end.join("\n\n").freeze
|
411
|
+
|
412
|
+
POLYNOMIAL = %{
|
413
|
+
// x is the coefficients in standard form
|
414
|
+
// where is the set of points at which to evaluate x
|
415
|
+
static VALUE evaluatepolynomial_v64fv64f_v64f(VALUE self, VALUE x, VALUE where) {
|
416
|
+
enum YepStatus status;
|
417
|
+
long i;
|
418
|
+
VALUE new_ary;
|
419
|
+
long x_l = RARRAY_LEN(x);
|
420
|
+
long y_l = RARRAY_LEN(where);
|
421
|
+
VALUE *x_a = RARRAY_PTR(x);
|
422
|
+
VALUE *y_a = RARRAY_PTR(where);
|
423
|
+
|
424
|
+
/* Allocate arrays of inputs and outputs */
|
425
|
+
#{allocate_yep64_typed_array_and_assert(%w{x}, 'x_l', :assert => false, :type => 'f')}
|
426
|
+
#{allocate_yep64_typed_array_and_assert(%w{y z}, 'y_l', :assert => false, :type => 'f')}
|
427
|
+
assert(yep_x != NULL);
|
428
|
+
assert(yep_y != NULL);
|
429
|
+
assert(yep_z != NULL);
|
430
|
+
|
431
|
+
#{initialize_yeppp}
|
432
|
+
|
433
|
+
// Yeppp! polynomial evaluation works in reverse standard form, so we have
|
434
|
+
// to load yep_x in reverse.
|
435
|
+
#{load_ruby_array_into_yeppp_array('x', 'i', 'x_l', 'f', [:integer, :float], :reverse => true)}
|
436
|
+
#{load_ruby_array_into_yeppp_array('y', 'i', 'y_l', 'f', [:integer, :float])}
|
437
|
+
|
438
|
+
/* Perform the operation */
|
439
|
+
status = yepMath_EvaluatePolynomial_V64fV64f_V64f(yep_x, yep_y, yep_z, (YepSize)x_l, (YepSize)y_l);
|
440
|
+
assert(status == YepStatusOk);
|
441
|
+
|
442
|
+
#{load_ruby_array_from_yeppp_array('z', 'i', 'y_l', 'f')}
|
443
|
+
|
444
|
+
#{deinitialize_yeppp}
|
445
|
+
|
446
|
+
#{release_array_memory(%w{x y z})}
|
447
|
+
|
448
|
+
return new_ary;
|
449
|
+
}
|
450
|
+
}.strip.freeze
|
451
|
+
|
452
|
+
INITIALIZER = %{
|
453
|
+
// The initialization method for this module
|
454
|
+
void Init_ryeppp() {
|
455
|
+
cRyeppp = rb_define_class("Ryeppp", rb_cObject);
|
456
|
+
|
457
|
+
/* Addition */
|
458
|
+
rb_define_singleton_method(cRyeppp, "add_v64fv64f_v64f", add_v64fv64f_v64f, 2);
|
459
|
+
rb_define_singleton_method(cRyeppp, "add_v64sv64s_v64s", add_v64sv64s_v64s, 2);
|
460
|
+
|
461
|
+
/* Subtraction */
|
462
|
+
rb_define_singleton_method(cRyeppp, "subtract_v64fv64f_v64f", subtract_v64fv64f_v64f, 2);
|
463
|
+
rb_define_singleton_method(cRyeppp, "subtract_v64sv64s_v64s", subtract_v64sv64s_v64s, 2);
|
464
|
+
|
465
|
+
/* Multiplication */
|
466
|
+
rb_define_singleton_method(cRyeppp, "multiply_v64fs64f_v64f", multiply_v64fs64f_v64f, 2);
|
467
|
+
rb_define_singleton_method(cRyeppp, "multiply_v64sv64s_v64s", multiply_v64sv64s_v64s, 2);
|
468
|
+
rb_define_singleton_method(cRyeppp, "multiply_v64fv64f_v64f", multiply_v64fv64f_v64f, 2);
|
469
|
+
rb_define_singleton_method(cRyeppp, "multiply_v64ss64s_v64s", multiply_v64ss64s_v64s, 2);
|
470
|
+
|
471
|
+
/* Dot Product */
|
472
|
+
rb_define_singleton_method(cRyeppp, "dotproduct_v64fv64f_s64f", dotproduct_v64fv64f_s64f, 2);
|
473
|
+
|
474
|
+
/* Minimum */
|
475
|
+
rb_define_singleton_method(cRyeppp, "min_v64f_s64f", min_v64f_s64f, 1);
|
476
|
+
rb_define_singleton_method(cRyeppp, "min_v64s_s64s", min_v64s_s64s, 1);
|
477
|
+
|
478
|
+
/* Maximum */
|
479
|
+
rb_define_singleton_method(cRyeppp, "max_v64f_s64f", max_v64f_s64f, 1);
|
480
|
+
rb_define_singleton_method(cRyeppp, "max_v64s_s64s", max_v64s_s64s, 1);
|
481
|
+
|
482
|
+
/* Pairwise Minima */
|
483
|
+
rb_define_singleton_method(cRyeppp, "min_v64fv64f_v64f", min_v64fv64f_v64f, 2);
|
484
|
+
// Pairwise signed min is not available.
|
485
|
+
|
486
|
+
/* Pairwise Maxima */
|
487
|
+
rb_define_singleton_method(cRyeppp, "max_v64fv64f_v64f", max_v64fv64f_v64f, 2);
|
488
|
+
// Pairwise signed max is not available.
|
489
|
+
|
490
|
+
/* Constant Minima */
|
491
|
+
rb_define_singleton_method(cRyeppp, "min_v64fs64f_v64f", min_v64fs64f_v64f, 2);
|
492
|
+
// Constant signed min is not available.
|
493
|
+
|
494
|
+
/* Constant Maxima */
|
495
|
+
rb_define_singleton_method(cRyeppp, "max_v64fs64f_v64f", max_v64fs64f_v64f, 2);
|
496
|
+
// Constant signed max is not available.
|
497
|
+
|
498
|
+
/* Negation */
|
499
|
+
rb_define_singleton_method(cRyeppp, "negate_v64f_s64f", negate_v64f_s64f, 1);
|
500
|
+
rb_define_singleton_method(cRyeppp, "negate_v64s_s64s", negate_v64s_s64s, 1);
|
501
|
+
|
502
|
+
/* Sums */
|
503
|
+
rb_define_singleton_method(cRyeppp, "sum_v64f_s64f", sum_v64f_s64f, 1);
|
504
|
+
// Signed sum is not available.
|
505
|
+
rb_define_singleton_method(cRyeppp, "sumabs_v64f_s64f", sumabs_v64f_s64f, 1);
|
506
|
+
// Signed abs sum is not available.
|
507
|
+
rb_define_singleton_method(cRyeppp, "sumsquares_v64f_s64f", sumsquares_v64f_s64f, 1);
|
508
|
+
// Signed squares sum is not available.
|
509
|
+
|
510
|
+
/* Maths */
|
511
|
+
#{MATHS_KINDS.map do |kind|
|
512
|
+
%{rb_define_singleton_method(cRyeppp, "#{kind.downcase}_v64f_v64f", #{kind.downcase}_v64f_v64f, 1);}
|
513
|
+
end.join("\n")}
|
514
|
+
|
515
|
+
/* Polynomial */
|
516
|
+
rb_define_singleton_method(cRyeppp, "evaluatepolynomial_v64fv64f_v64f", evaluatepolynomial_v64fv64f_v64f, 2);
|
517
|
+
}
|
518
|
+
}.strip.freeze
|
data/lib/ryeppp.rb
ADDED
data/lib/ryeppp/bench.rb
ADDED
@@ -0,0 +1,233 @@
|
|
1
|
+
require 'benchmark'
|
2
|
+
require 'inline'
|
3
|
+
|
4
|
+
$:<< './lib'
|
5
|
+
require 'ryeppp'
|
6
|
+
|
7
|
+
VERBOSE = %w{1 t true y yes}.include?(ENV['VERBOSE'])
|
8
|
+
|
9
|
+
def puts_with_pounds(s)
|
10
|
+
puts "\n\n#{s} #{'#' * (79 - s.size)}"
|
11
|
+
end
|
12
|
+
|
13
|
+
class Array
|
14
|
+
def sum
|
15
|
+
self.inject(0){|sum, o| sum + o}
|
16
|
+
end
|
17
|
+
def sumabs
|
18
|
+
self.inject(0){|sum, o| sum + o.abs}
|
19
|
+
end
|
20
|
+
def sumsquares
|
21
|
+
self.inject(0){|sum, o| sum + o**2}
|
22
|
+
end
|
23
|
+
inline do |builder|
|
24
|
+
builder.c %{
|
25
|
+
static VALUE c_sum() {
|
26
|
+
long n = RARRAY_LEN(self);
|
27
|
+
double sum = 0.0;
|
28
|
+
|
29
|
+
VALUE *x_a = RARRAY_PTR(self);
|
30
|
+
|
31
|
+
long i;
|
32
|
+
for (i=0; i<n; i++) {
|
33
|
+
if (TYPE(x_a[i]) != T_FIXNUM && TYPE(x_a[i]) != T_BIGNUM && TYPE(x_a[i]) != T_FLOAT) {
|
34
|
+
rb_raise(rb_eTypeError, "input was not all integers and floats");
|
35
|
+
}
|
36
|
+
sum += NUM2DBL(x_a[i]);
|
37
|
+
}
|
38
|
+
return DBL2NUM(sum);
|
39
|
+
}
|
40
|
+
}
|
41
|
+
end
|
42
|
+
inline do |builder|
|
43
|
+
builder.c %{
|
44
|
+
static VALUE c_sumabs() {
|
45
|
+
long n = RARRAY_LEN(self);
|
46
|
+
double sum = 0.0;
|
47
|
+
|
48
|
+
VALUE *x_a = RARRAY_PTR(self);
|
49
|
+
|
50
|
+
long i;
|
51
|
+
for (i=0; i<n; i++) {
|
52
|
+
if (TYPE(x_a[i]) != T_FIXNUM && TYPE(x_a[i]) != T_BIGNUM && TYPE(x_a[i]) != T_FLOAT) {
|
53
|
+
rb_raise(rb_eTypeError, "input was not all integers and floats");
|
54
|
+
}
|
55
|
+
sum += (NUM2DBL(x_a[i]) < 0) ? (-1.0 * NUM2DBL(x_a[i])) : NUM2DBL(x_a[i]);
|
56
|
+
}
|
57
|
+
return DBL2NUM(sum);
|
58
|
+
}
|
59
|
+
}
|
60
|
+
end
|
61
|
+
inline do |builder|
|
62
|
+
builder.c %{
|
63
|
+
static VALUE c_sumsquares() {
|
64
|
+
long n = RARRAY_LEN(self);
|
65
|
+
double sum = 0.0;
|
66
|
+
|
67
|
+
VALUE *x_a = RARRAY_PTR(self);
|
68
|
+
|
69
|
+
long i;
|
70
|
+
for (i=0; i<n; i++) {
|
71
|
+
if (TYPE(x_a[i]) != T_FIXNUM && TYPE(x_a[i]) != T_BIGNUM && TYPE(x_a[i]) != T_FLOAT) {
|
72
|
+
rb_raise(rb_eTypeError, "input was not all integers and floats");
|
73
|
+
}
|
74
|
+
sum += NUM2DBL(x_a[i]) * NUM2DBL(x_a[i]);
|
75
|
+
}
|
76
|
+
return DBL2NUM(sum);
|
77
|
+
}
|
78
|
+
}
|
79
|
+
end
|
80
|
+
|
81
|
+
def dot_product_f(v2)
|
82
|
+
self.each.with_index.inject(0) do |sum, (o, idx)|
|
83
|
+
sum + o.to_f * v2[idx]
|
84
|
+
end
|
85
|
+
end
|
86
|
+
inline do |builder|
|
87
|
+
builder.c %{
|
88
|
+
static VALUE c_dot_product_f(VALUE v2) {
|
89
|
+
long n = RARRAY_LEN(self);
|
90
|
+
double sum = 0.0;
|
91
|
+
|
92
|
+
VALUE *x_a = RARRAY_PTR(self);
|
93
|
+
VALUE *y_a = RARRAY_PTR(v2);
|
94
|
+
|
95
|
+
long i;
|
96
|
+
for (i=0; i<n; i++) {
|
97
|
+
if (TYPE(x_a[i]) != T_FIXNUM && TYPE(x_a[i]) != T_BIGNUM && TYPE(x_a[i]) != T_FLOAT) {
|
98
|
+
rb_raise(rb_eTypeError, "input was not all integers and floats");
|
99
|
+
}
|
100
|
+
if (TYPE(y_a[i]) != T_FIXNUM && TYPE(y_a[i]) != T_BIGNUM && TYPE(y_a[i]) != T_FLOAT) {
|
101
|
+
rb_raise(rb_eTypeError, "input was not all integers and floats");
|
102
|
+
}
|
103
|
+
sum += NUM2DBL(x_a[i]) * NUM2DBL(y_a[i]);
|
104
|
+
}
|
105
|
+
return DBL2NUM(sum);
|
106
|
+
}
|
107
|
+
}
|
108
|
+
end
|
109
|
+
|
110
|
+
def evaluate_polynomial(x)
|
111
|
+
if self.size == 1
|
112
|
+
self.first
|
113
|
+
else
|
114
|
+
self.last + x * self[0..-2].evaluate_polynomial(x)
|
115
|
+
end
|
116
|
+
end
|
117
|
+
def evaluate_polynomial_iter(x)
|
118
|
+
self.map.with_index do |o, i|
|
119
|
+
self[i] * x ** (self.size - 1 - i)
|
120
|
+
end.sum
|
121
|
+
end
|
122
|
+
inline do |builder|
|
123
|
+
builder.prefix %{
|
124
|
+
static inline double heron_evaluate(VALUE *x_a, long i, double x) {
|
125
|
+
if (i == 0) {
|
126
|
+
return NUM2DBL(x_a[i]);
|
127
|
+
} else {
|
128
|
+
return NUM2DBL(x_a[i]) + x * heron_evaluate(x_a, i - 1, x);
|
129
|
+
}
|
130
|
+
}
|
131
|
+
}
|
132
|
+
builder.c %{
|
133
|
+
static VALUE c_evaluate_polynomial(VALUE x_v) {
|
134
|
+
long n = RARRAY_LEN(self);
|
135
|
+
|
136
|
+
VALUE *x_a = RARRAY_PTR(self);
|
137
|
+
double x = NUM2DBL(x_v);
|
138
|
+
|
139
|
+
long i;
|
140
|
+
for (i=0; i<n; i++) {
|
141
|
+
if (TYPE(x_a[i]) != T_FIXNUM && TYPE(x_a[i]) != T_BIGNUM && TYPE(x_a[i]) != T_FLOAT) {
|
142
|
+
rb_raise(rb_eTypeError, "input was not all integers and floats");
|
143
|
+
}
|
144
|
+
}
|
145
|
+
return DBL2NUM(heron_evaluate(x_a, n - 1, x));
|
146
|
+
}
|
147
|
+
}
|
148
|
+
end
|
149
|
+
inline do |builder|
|
150
|
+
builder.include('<math.h>')
|
151
|
+
builder.c %{
|
152
|
+
static VALUE c_evaluate_polynomial_iter(VALUE x_v) {
|
153
|
+
long n = RARRAY_LEN(self);
|
154
|
+
|
155
|
+
VALUE *x_a = RARRAY_PTR(self);
|
156
|
+
double x = NUM2DBL(x_v);
|
157
|
+
double res = 0;
|
158
|
+
|
159
|
+
long i;
|
160
|
+
for (i=0; i<n; i++) {
|
161
|
+
if (TYPE(x_a[i]) != T_FIXNUM && TYPE(x_a[i]) != T_BIGNUM && TYPE(x_a[i]) != T_FLOAT) {
|
162
|
+
rb_raise(rb_eTypeError, "input was not all integers and floats");
|
163
|
+
}
|
164
|
+
res += NUM2DBL(x_a[i]) * pow(x, (n - 1 - i));
|
165
|
+
}
|
166
|
+
return DBL2NUM(res);
|
167
|
+
}
|
168
|
+
}
|
169
|
+
end
|
170
|
+
end
|
171
|
+
|
172
|
+
WIDTH = 40
|
173
|
+
V_f = (0..1_024*1_024).to_a.map{Random.rand}
|
174
|
+
|
175
|
+
# Dot Product
|
176
|
+
puts_with_pounds "Dot Product"
|
177
|
+
n = 1
|
178
|
+
if VERBOSE
|
179
|
+
puts "dot_product_f: #{V_f.dot_product_f(V_f)}"
|
180
|
+
puts "c_dot_product_f: #{V_f.c_dot_product_f(V_f)}"
|
181
|
+
puts "Ryeppp.dotproduct_v64fv64f_s64f: #{Ryeppp.dotproduct_v64fv64f_s64f(V_f, V_f)}"
|
182
|
+
end
|
183
|
+
Benchmark.bm(WIDTH) do |x|
|
184
|
+
x.report("dot_product_f:") { n.times { V_f.dot_product_f(V_f) } }
|
185
|
+
x.report("c_dot_product_f:") { n.times { V_f.c_dot_product_f(V_f) } }
|
186
|
+
x.report("Ryeppp.dotproduct_v64fv64f_s64f:") { n.times { Ryeppp.dotproduct_v64fv64f_s64f(V_f, V_f) } }
|
187
|
+
end
|
188
|
+
|
189
|
+
# Sums
|
190
|
+
puts_with_pounds "Sums"
|
191
|
+
n = 1
|
192
|
+
Benchmark.bm(WIDTH) do |x|
|
193
|
+
['', 'abs', 'squares'].each do |suffix|
|
194
|
+
x.report("sum#{suffix}:") { n.times { V_f.send("sum#{suffix}") } }
|
195
|
+
x.report("c_sum#{suffix}:") { n.times { V_f.send("c_sum#{suffix}") } }
|
196
|
+
x.report("Ryeppp.sum#{suffix}_v64f_s64f:") { n.times { Ryeppp.send("sum#{suffix}_v64f_s64f", V_f) } }
|
197
|
+
end
|
198
|
+
end
|
199
|
+
|
200
|
+
# Math Functions
|
201
|
+
puts_with_pounds "Math Functions"
|
202
|
+
n = 1
|
203
|
+
Benchmark.bm(WIDTH) do |x|
|
204
|
+
%w{log exp sin cos tan}.each do |f|
|
205
|
+
x.report("#{f}:") { n.times { V_f.each{|o| Math.send(f, o)} } }
|
206
|
+
x.report("Ryeppp.#{f}_v64f_v64f:") { n.times { Ryeppp.send("#{f}_v64f_v64f", V_f) } }
|
207
|
+
end
|
208
|
+
end
|
209
|
+
|
210
|
+
# Polynomial
|
211
|
+
puts_with_pounds "Polynomial"
|
212
|
+
n = 1
|
213
|
+
P_SMALL = (0..1_024).to_a.map{Random.rand}
|
214
|
+
P_LARGE = (0..1_024*1_024).to_a.map{Random.rand}
|
215
|
+
X = Random.rand
|
216
|
+
if VERBOSE
|
217
|
+
puts "evaluate_polynomial: #{P_SMALL.evaluate_polynomial(X)}"
|
218
|
+
puts "evaluate_polynomial_iter: #{P_SMALL.evaluate_polynomial_iter(X)}"
|
219
|
+
puts "c_evaluate_polynomial: #{P_SMALL.c_evaluate_polynomial(X)}"
|
220
|
+
puts "c_evaluate_polynomial_iter: #{P_SMALL.c_evaluate_polynomial_iter(X)}"
|
221
|
+
puts "Ryeppp.evaluatepolynomial_v64fv64f_v64f: #{Ryeppp.evaluatepolynomial_v64fv64f_v64f(P_SMALL, [X])}"
|
222
|
+
puts "evaluate_polynomial_iter: #{P_LARGE.evaluate_polynomial_iter(X)}"
|
223
|
+
puts "c_evaluate_polynomial_iter: #{P_LARGE.c_evaluate_polynomial_iter(X)}"
|
224
|
+
puts "Ryeppp.evaluatepolynomial_v64fv64f_v64f: #{Ryeppp.evaluatepolynomial_v64fv64f_v64f(P_LARGE, [X])}"
|
225
|
+
end
|
226
|
+
Benchmark.bm(WIDTH) do |x|
|
227
|
+
x.report("evaluate_polynomial:") { n.times { P_SMALL.evaluate_polynomial(X) } }
|
228
|
+
x.report("c_evaluate_polynomial:") { n.times { P_SMALL.c_evaluate_polynomial(X) } }
|
229
|
+
x.report("Ryeppp.evaluatepolynomial_v64fv64f_v64f:") { n.times { Ryeppp.evaluatepolynomial_v64fv64f_v64f(P_SMALL, [X]) } }
|
230
|
+
x.report("evaluate_polynomial_iter:") { n.times { P_LARGE.evaluate_polynomial_iter(X) } }
|
231
|
+
x.report("c_evaluate_polynomial_iter:") { n.times { P_LARGE.c_evaluate_polynomial_iter(X) } }
|
232
|
+
x.report("Ryeppp.evaluatepolynomial_v64fv64f_v64f:") { n.times { Ryeppp.evaluatepolynomial_v64fv64f_v64f(P_LARGE, [X]) } }
|
233
|
+
end
|
data/ryeppp.gemspec
ADDED
@@ -0,0 +1,29 @@
|
|
1
|
+
# coding: utf-8
|
2
|
+
lib = File.expand_path('../lib', __FILE__)
|
3
|
+
$LOAD_PATH.unshift(lib) unless $LOAD_PATH.include?(lib)
|
4
|
+
require 'ryeppp/version'
|
5
|
+
|
6
|
+
Gem::Specification.new do |spec|
|
7
|
+
spec.name = "ryeppp"
|
8
|
+
spec.version = Ryeppp::VERSION
|
9
|
+
spec.authors = ["Brad Cater"]
|
10
|
+
spec.email = ["bradcater@gmail.com"]
|
11
|
+
spec.description = %q{This gem provides bindings to the Yeppp! library. According to the documentation, "Yeppp! is a high-performance SIMD-optimized mathematical library for x86, ARM, and MIPS processors on Windows, Android, Mac OS X, and GNU/Linux systems."}
|
12
|
+
spec.summary = %q{This gem provides bindings to the Yeppp! library.}
|
13
|
+
spec.homepage = "https://github.com/bradcater/ruby-yeppp"
|
14
|
+
spec.license = "MIT"
|
15
|
+
|
16
|
+
spec.files = `git ls-files`.split($/)
|
17
|
+
spec.executables = spec.files.grep(%r{^bin/}) { |f| File.basename(f) }
|
18
|
+
spec.test_files = spec.files.grep(%r{^(test|spec|features)/})
|
19
|
+
spec.require_paths = ["lib"]
|
20
|
+
|
21
|
+
spec.extensions << "ext/ryeppp/extconf.rb"
|
22
|
+
|
23
|
+
spec.add_development_dependency "bundler", "~> 1.3"
|
24
|
+
spec.add_development_dependency "rake-compiler"
|
25
|
+
spec.add_development_dependency "rake", ">= 1.9.1"
|
26
|
+
spec.add_development_dependency "rspec", ">= 2.13.0"
|
27
|
+
spec.add_development_dependency "rspec", ">= 2.13.0"
|
28
|
+
spec.add_development_dependency "RubyInline", "~> 3.12.2"
|
29
|
+
end
|
data/spec/ryeppp_spec.rb
ADDED
@@ -0,0 +1,166 @@
|
|
1
|
+
require 'spec_helper'
|
2
|
+
|
3
|
+
describe Ryeppp do
|
4
|
+
# Addition
|
5
|
+
it 'should add vectors of Fixnums' do
|
6
|
+
Ryeppp.add_v64sv64s_v64s([1], [2]).should eq([3])
|
7
|
+
expect{Ryeppp.add_v64sv64s_v64s([1, 'a'], [2, 'b'])}.to raise_error(TypeError)
|
8
|
+
expect{Ryeppp.add_v64sv64s_v64s([2**63], [1])}.to raise_error(RangeError)
|
9
|
+
expect{Ryeppp.add_v64sv64s_v64s([-2**63], [1])}.to raise_error(RangeError)
|
10
|
+
end
|
11
|
+
it 'should add vectors of Floats' do
|
12
|
+
Ryeppp.add_v64fv64f_v64f([1.1], [1.1]).should eq([2.2])
|
13
|
+
expect{Ryeppp.add_v64fv64f_v64f([1.1, 'a'], [2.2, 'b'])}.to raise_error(TypeError)
|
14
|
+
end
|
15
|
+
|
16
|
+
# Subtraction
|
17
|
+
it 'should subtract vectors of Fixnums' do
|
18
|
+
Ryeppp.subtract_v64sv64s_v64s([1], [2]).should eq([-1])
|
19
|
+
expect{Ryeppp.subtract_v64sv64s_v64s([1, 'a'], [2, 'b'])}.to raise_error(TypeError)
|
20
|
+
expect{Ryeppp.subtract_v64sv64s_v64s([2**63], [2])}.to raise_error(RangeError)
|
21
|
+
expect{Ryeppp.subtract_v64sv64s_v64s([-2**63], [2])}.to raise_error(RangeError)
|
22
|
+
end
|
23
|
+
it 'should subtract vectors of Floats' do
|
24
|
+
Ryeppp.subtract_v64fv64f_v64f([1.1], [1.1]).should eq([0])
|
25
|
+
expect{Ryeppp.subtract_v64fv64f_v64f([1.1, 'a'], [2.2, 'b'])}.to raise_error(TypeError)
|
26
|
+
end
|
27
|
+
|
28
|
+
# Multiplication
|
29
|
+
it 'should multiply a vector of Fixnums by a constant' do
|
30
|
+
Ryeppp.multiply_v64ss64s_v64s([1, 2, 3], 2).should eq([2, 4, 6])
|
31
|
+
expect{Ryeppp.multiply_v64ss64s_v64s([1, 'a'], 2)}.to raise_error(TypeError)
|
32
|
+
expect{Ryeppp.multiply_v64ss64s_v64s([2**63], 2)}.to raise_error(RangeError)
|
33
|
+
expect{Ryeppp.multiply_v64ss64s_v64s([-2**63], 2)}.to raise_error(RangeError)
|
34
|
+
end
|
35
|
+
it 'should multiply a vector of Floats by a constant' do
|
36
|
+
Ryeppp.multiply_v64fs64f_v64f([1.0, 2.0, 3.0], 2.0).should eq([2.0, 4.0, 6.0])
|
37
|
+
expect{Ryeppp.multiply_v64fs64f_v64f([1, 'a'], 2)}.to raise_error(TypeError)
|
38
|
+
end
|
39
|
+
it 'should multiply vectors of Fixnums' do
|
40
|
+
Ryeppp.multiply_v64sv64s_v64s([2], [3]).should eq([6])
|
41
|
+
expect{Ryeppp.multiply_v64sv64s_v64s([1, 'a'], [2, 'b'])}.to raise_error(TypeError)
|
42
|
+
expect{Ryeppp.multiply_v64sv64s_v64s([2**63], [2])}.to raise_error(RangeError)
|
43
|
+
expect{Ryeppp.multiply_v64sv64s_v64s([-2**63], [2])}.to raise_error(RangeError)
|
44
|
+
end
|
45
|
+
it 'should multiply vectors of Floats' do
|
46
|
+
Ryeppp.multiply_v64fv64f_v64f([2.5], [3.5]).should eq([8.75])
|
47
|
+
expect{Ryeppp.multiply_v64fv64f_v64f([1.1, 'a'], [2.2, 'b'])}.to raise_error(TypeError)
|
48
|
+
end
|
49
|
+
|
50
|
+
# Dot Product
|
51
|
+
it 'should do the dot product of vectors of Floats' do
|
52
|
+
Ryeppp.dotproduct_v64fv64f_s64f([1, 2, 3], [4, 5, 6]).should eq(32.0)
|
53
|
+
expect{Ryeppp.dotproduct_v64fv64f_s64f([1, 2, 'a'], [4, 5, 'b'])}.to raise_error(TypeError)
|
54
|
+
end
|
55
|
+
|
56
|
+
# Minimum
|
57
|
+
it 'should find the minimum in a vector of Fixnums' do
|
58
|
+
Ryeppp.min_v64s_s64s([3, 2, 1]).should eq(1)
|
59
|
+
expect{Ryeppp.min_v64s_s64s([1, 'a'])}.to raise_error(TypeError)
|
60
|
+
expect{Ryeppp.min_v64s_s64s([2**63])}.to raise_error(RangeError)
|
61
|
+
expect{Ryeppp.min_v64s_s64s([-2**63])}.to raise_error(RangeError)
|
62
|
+
end
|
63
|
+
it 'should find the minimum in a vector of Floats' do
|
64
|
+
Ryeppp.min_v64f_s64f([1.0, 2.0, 3.0]).should eq(1.0)
|
65
|
+
expect{Ryeppp.min_v64f_s64f([1.0, 'a'])}.to raise_error(TypeError)
|
66
|
+
end
|
67
|
+
|
68
|
+
# Maximum
|
69
|
+
it 'should find the maximum in a vector of Fixnums' do
|
70
|
+
Ryeppp.max_v64s_s64s([3, 2, 1]).should eq(3)
|
71
|
+
expect{Ryeppp.max_v64s_s64s([1, 'a'])}.to raise_error(TypeError)
|
72
|
+
expect{Ryeppp.max_v64s_s64s([2**63])}.to raise_error(RangeError)
|
73
|
+
expect{Ryeppp.max_v64s_s64s([-2**63])}.to raise_error(RangeError)
|
74
|
+
end
|
75
|
+
it 'should find the maximum in a vector of Floats' do
|
76
|
+
Ryeppp.max_v64f_s64f([1.0, 2.0, 3.0]).should eq(3.0)
|
77
|
+
expect{Ryeppp.max_v64f_s64f([1.0, 'a'])}.to raise_error(TypeError)
|
78
|
+
end
|
79
|
+
|
80
|
+
# Pairwise Minima
|
81
|
+
it 'should find the pairwise minima in vectors of Floats' do
|
82
|
+
Ryeppp.min_v64fv64f_v64f([1.0, 2.0, 3.0], [3.0, 2.0, 1.0]).should eq([1.0, 2.0, 1.0])
|
83
|
+
expect{Ryeppp.min_v64fv64f_v64f([1.0, 'a'], [2.0, 'b'])}.to raise_error(TypeError)
|
84
|
+
end
|
85
|
+
# Pairwise Maxima
|
86
|
+
it 'should find the pairwise maxima in vectors of Floats' do
|
87
|
+
Ryeppp.max_v64fv64f_v64f([1.0, 2.0, 3.0], [3.0, 2.0, 1.0]).should eq([3.0, 2.0, 3.0])
|
88
|
+
expect{Ryeppp.max_v64fv64f_v64f([1.0, 'a'], [2.0, 'b'])}.to raise_error(TypeError)
|
89
|
+
end
|
90
|
+
|
91
|
+
# Constant Minima
|
92
|
+
it 'should find the minima in a vector of Floats and a constant' do
|
93
|
+
Ryeppp.min_v64fs64f_v64f([1.0, 2.0, 3.0], 2.0).should eq([1.0, 2.0, 2.0])
|
94
|
+
expect{Ryeppp.min_v64fs64f_v64f([1.0, 'a'], 2.0)}.to raise_error(TypeError)
|
95
|
+
end
|
96
|
+
# Constant Maxima
|
97
|
+
it 'should find the maxima in a vector of Floats and a constant' do
|
98
|
+
Ryeppp.max_v64fs64f_v64f([1.0, 2.0, 3.0], 2.5).should eq([2.5, 2.5, 3.0])
|
99
|
+
expect{Ryeppp.max_v64fs64f_v64f([1.0, 'a'], 2.5)}.to raise_error(TypeError)
|
100
|
+
end
|
101
|
+
|
102
|
+
# Negation
|
103
|
+
it 'should negate vectors of Fixnums' do
|
104
|
+
Ryeppp.negate_v64s_s64s([1]).should eq([-1])
|
105
|
+
expect{Ryeppp.negate_v64s_s64s([1, 'a'])}.to raise_error(TypeError)
|
106
|
+
expect{Ryeppp.negate_v64s_s64s([2**63])}.to raise_error(RangeError)
|
107
|
+
expect{Ryeppp.negate_v64s_s64s([-2**63])}.to raise_error(RangeError)
|
108
|
+
end
|
109
|
+
it 'should negate vectors of Floats' do
|
110
|
+
Ryeppp.negate_v64f_s64f([1.1]).should eq([-1.1])
|
111
|
+
expect{Ryeppp.negate_v64f_s64f([1.1, 'a'])}.to raise_error(TypeError)
|
112
|
+
end
|
113
|
+
|
114
|
+
# Sum
|
115
|
+
it 'should sum a vector of Floats' do
|
116
|
+
Ryeppp.sum_v64f_s64f([1, 2, 3]).should eq(6.0)
|
117
|
+
expect{Ryeppp.sum_v64f_s64f([1, 'a'])}.to raise_error(TypeError)
|
118
|
+
end
|
119
|
+
# Sum Absolute Values
|
120
|
+
it 'should sum absolute values of a vector of Floats' do
|
121
|
+
Ryeppp.sumabs_v64f_s64f([1.0, -1.0, 2.0]).should eq(4.0)
|
122
|
+
expect{Ryeppp.sumabs_v64f_s64f([1.0, -1.0, 'a'])}.to raise_error(TypeError)
|
123
|
+
end
|
124
|
+
# Sum Square Values
|
125
|
+
it 'should sum square values of a vector of Floats' do
|
126
|
+
Ryeppp.sumsquares_v64f_s64f([1.0, -1.0, 2.0]).should eq(6.0)
|
127
|
+
expect{Ryeppp.sumsquares_v64f_s64f([1.0, -1.0, 'a'])}.to raise_error(TypeError)
|
128
|
+
end
|
129
|
+
|
130
|
+
# Log
|
131
|
+
it 'should find the natural logarithm of elements of a vector of Floats' do
|
132
|
+
Ryeppp.log_v64f_v64f([1.0, 2.0, 3.0]).map{|o| o.round(5)}.should eq([0.0, 0.69315, 1.09861])
|
133
|
+
expect{Ryeppp.log_v64f_v64f([1.0, 'a'])}.to raise_error(TypeError)
|
134
|
+
end
|
135
|
+
# Exp
|
136
|
+
it 'should find the base e exponent of elements of a vector of Floats' do
|
137
|
+
Ryeppp.exp_v64f_v64f([1.0, 2.0, 3.0]).map{|o| o.round(5)}.should eq([2.71828, 7.38906, 20.08554])
|
138
|
+
expect{Ryeppp.exp_v64f_v64f([1.0, 'a'])}.to raise_error(TypeError)
|
139
|
+
end
|
140
|
+
# Sin
|
141
|
+
it 'should find the sine of elements of a vector of Floats' do
|
142
|
+
Ryeppp.sin_v64f_v64f([1.0, 2.0, 3.0]).map{|o| o.round(5)}.should eq([0.84147, 0.9093, 0.14112])
|
143
|
+
expect{Ryeppp.sin_v64f_v64f([1.0, 'a'])}.to raise_error(TypeError)
|
144
|
+
end
|
145
|
+
# Cos
|
146
|
+
it 'should find the cosine of elements of a vector of Floats' do
|
147
|
+
Ryeppp.cos_v64f_v64f([1.0, 2.0, 3.0]).map{|o| o.round(5)}.should eq([0.5403, -0.41615, -0.98999])
|
148
|
+
expect{Ryeppp.cos_v64f_v64f([1.0, 'a'])}.to raise_error(TypeError)
|
149
|
+
end
|
150
|
+
# Tan
|
151
|
+
it 'should find the tangent of elements of a vector of Floats' do
|
152
|
+
Ryeppp.tan_v64f_v64f([1.0, 2.0, 3.0]).map{|o| o.round(5)}.should eq([1.55741, -2.18504, -0.14255])
|
153
|
+
expect{Ryeppp.tan_v64f_v64f([1.0, 'a'])}.to raise_error(TypeError)
|
154
|
+
end
|
155
|
+
|
156
|
+
# Polynomial
|
157
|
+
it 'should evaluate a polynomial for a vector of Floats' do
|
158
|
+
# x^2 + x + 1
|
159
|
+
# evaluated at x=1.
|
160
|
+
Ryeppp.evaluatepolynomial_v64fv64f_v64f([1.0, 1.0, 1.0], [1.0]).should eq([3.0])
|
161
|
+
# -5x^3 - 4x^2 + 2x + 1
|
162
|
+
# evaluated at x=0, x=1, and x=2.
|
163
|
+
Ryeppp.evaluatepolynomial_v64fv64f_v64f([-5, -4, 2, 1], [0, 1, 2]).should eq([1.0, -6.0, -51.0])
|
164
|
+
expect{Ryeppp.evaluatepolynomial_v64fv64f_v64f([1, 'a'], [0, 1])}.to raise_error(TypeError)
|
165
|
+
end
|
166
|
+
end
|
data/spec/spec_helper.rb
ADDED
metadata
CHANGED
@@ -1,7 +1,7 @@
|
|
1
1
|
--- !ruby/object:Gem::Specification
|
2
2
|
name: ryeppp
|
3
3
|
version: !ruby/object:Gem::Version
|
4
|
-
version: 0.0.
|
4
|
+
version: 0.0.3
|
5
5
|
prerelease:
|
6
6
|
platform: ruby
|
7
7
|
authors:
|
@@ -117,7 +117,20 @@ extensions:
|
|
117
117
|
- ext/ryeppp/extconf.rb
|
118
118
|
extra_rdoc_files: []
|
119
119
|
files:
|
120
|
+
- .gitignore
|
121
|
+
- .rspec
|
122
|
+
- Gemfile
|
123
|
+
- LICENSE.txt
|
124
|
+
- README.md
|
125
|
+
- Rakefile
|
120
126
|
- ext/ryeppp/extconf.rb
|
127
|
+
- ext/templates/ryeppp.c.rb
|
128
|
+
- lib/ryeppp.rb
|
129
|
+
- lib/ryeppp/bench.rb
|
130
|
+
- lib/ryeppp/version.rb
|
131
|
+
- ryeppp.gemspec
|
132
|
+
- spec/ryeppp_spec.rb
|
133
|
+
- spec/spec_helper.rb
|
121
134
|
homepage: https://github.com/bradcater/ruby-yeppp
|
122
135
|
licenses:
|
123
136
|
- MIT
|
@@ -143,4 +156,6 @@ rubygems_version: 1.8.25
|
|
143
156
|
signing_key:
|
144
157
|
specification_version: 3
|
145
158
|
summary: This gem provides bindings to the Yeppp! library.
|
146
|
-
test_files:
|
159
|
+
test_files:
|
160
|
+
- spec/ryeppp_spec.rb
|
161
|
+
- spec/spec_helper.rb
|