1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
|
<HTML>
<HEAD>
<TITLE>timesoftfloat</TITLE>
</HEAD>
<BODY>
<H1>Berkeley TestFloat Release 3b: <CODE>timesoftfloat</CODE></H1>
<P>
John R. Hauser<BR>
2016 July 22<BR>
</P>
<H2>Overview</H2>
<P>
The <CODE>timesoftfloat</CODE> program provides a simple way to evaluate the
speed of the floating-point operations of the Berkeley SoftFloat library.
Program <CODE>timesoftfloat</CODE> is included with the Berkeley TestFloat
package, a small collection of programs for testing that an implementation of
floating-point conforms to the IEEE Standard for Binary Floating-Point
Arithmetic.
Although <CODE>timesoftfloat</CODE> does not test floating-point correctness
like the other TestFloat programs, nevertheless <CODE>timesoftfloat</CODE> is a
partner to TestFloat’s <CODE>testsoftfloat</CODE> program.
For more about TestFloat generally and <CODE>testsoftfloat</CODE> specifically,
see file
<A HREF="TestFloat-general.html"><NOBR><CODE>TestFloat-general.html</CODE></NOBR></A>.
</P>
<P>
Ordinarily, <CODE>timesoftfloat</CODE> will measure a function’s speed
separately for each rounding mode defined by the IEEE Floating-Point Standard,
one after the other.
If an operation is not supposed to require rounding, it will by default be
timed only with the rounding mode set to <CODE>near_even</CODE> (nearest/even).
In the same way, if an operation is affected by the way in which underflow
tininess is detected, <CODE>timesoftfloat</CODE> times the function with
tininess detected both before rounding and after rounding.
For <NOBR>80-bit</NOBR> double-extended-precision operations affected by
rounding precision control, <CODE>timesoftfloat</CODE> also times the function
for each of the three rounding precision modes, one after the other.
Evaluation of a function can be limited to a single rounding mode, a single
tininess mode, and/or a single rounding precision with appropriate command-line
options.
</P>
<P>
For each function and mode evaluated, <CODE>timesoftfloat</CODE> reports the
measured speed of the function in Mop/s, or “millions of operations per
second”.
The speeds reported by <CODE>timesoftfloat</CODE> may be affected somewhat by
other software executing at the same time as <CODE>timesoftfloat</CODE>.
Be aware also that the exact execution time of any SoftFloat function depends
partly on the values of arguments and the state of the processor’s caches
at the time the function is called.
Your actual experience with SoftFloat may differ from the speeds reported by
<CODE>timesoftfloat</CODE> for all these reasons.
</P>
<P>
Note that the remainder operations for larger formats (<CODE>f64_rem</CODE>,
<CODE>extF80_rem</CODE>, and <CODE>f128_rem</CODE>) can be markedly slower than
other operations, particularly for double-extended-precision
(<CODE>extF80_rem</CODE>) and quadruple precision (<CODE>f128_rem</CODE>).
This is inherent to the remainder operation itself and is not a failing of the
SoftFloat implementation.
</P>
<H2>Command Syntax</H2>
<P>
The <CODE>timesoftfloat</CODE> program is executed as a command with this
syntax:
<BLOCKQUOTE>
<PRE>
timesoftfloat [<option>...] <function>
</PRE>
</BLOCKQUOTE>
Square brackets (<CODE>[ ]</CODE>) denote optional arguments,
<CODE><option></CODE> is a supported option, and
<CODE><function></CODE> is the name of either a testable function or a
function set.
The available options and function sets are documented below.
If <CODE>timesoftfloat</CODE> is executed without any arguments, a summary of
usage is written.
</P>
<H2>Options</H2>
<P>
The <CODE>timesoftfloat</CODE> program accepts several command options.
If mutually contradictory options are given, the last one has priority.
</P>
<H3><CODE>-help</CODE></H3>
<P>
The <CODE>-help</CODE> option causes a summary of program usage to be written,
after which the program exits.
</P>
<H3><CODE>-precision32, -precision64, -precision80</CODE></H3>
<P>
For <NOBR>80-bit</NOBR> double-extended-precision funcions affected by
rounding precision control, the <CODE>-precision32</CODE> option restricts
timing of an operation to only the cases in which the rounding precision is
<NOBR>32 bits</NOBR>, equivalent to <NOBR>32-bit</NOBR> single-precision.
Other rounding precision choices are not timed.
Likewise, <CODE>-precision64</CODE> fixes the rounding precision to
<NOBR>64 bits</NOBR>, equivalent to <NOBR>64-bit</NOBR> double-precision;
and <CODE>-precision80</CODE> fixes the rounding precision to the full
<NOBR>80 bits</NOBR> of the double-extended-precision format.
All these options are ignored for operations not affected by rounding precision
control.
</P>
<H3><CODE>-rnear_even, -rnear_maxMag, -rminMag, -rmin, -rmax</CODE></H3>
<P>
The <CODE>-rnear_even</CODE> option restricts timing of an operation to only
the cases in which the rounding mode is nearest/even.
Other rounding mode choices are not timed.
Likewise, <CODE>-rnear_maxMag</CODE> forces rounding to nearest/maximum
magnitude (nearest-away), <CODE>-rminMag</CODE> forces rounding to minimum
magnitude (toward zero), <CODE>-rmin</CODE> forces rounding to minimum (down,
toward negative infinity), and <CODE>-rmax</CODE> forces rounding to maximum
(up, toward positive infinity).
These options are ignored for operations that are exact and thus do not round.
</P>
<H3><CODE>-tininessbefore, -tininessafter</CODE></H3>
<P>
The <CODE>-tininessbefore</CODE> option restricts timing of an operation to
only the cases in which tininess on underflow is detected before rounding.
Likewise, <CODE>-tininessafter</CODE> restricts measurement to only the cases
in which tininess on underflow is detected after rounding.
</P>
<H3><CODE>-notexact, -exact</CODE></H3>
<P>
For functions that round to an integer (conversions to integer types and the
<CODE>roundToInt</CODE> functions), the <CODE>-notexact</CODE> option restricts
timing of an operation to only the cases for which the
<CODE><I>exact</I></CODE> operand (specifying whether the <I>inexact</I>
exception flag may be raised) is <CODE>false</CODE>.
Likewise, the <CODE>-exact</CODE> option restricts measurement to only the
cases for which the <CODE><I>exact</I></CODE> operand is <CODE>true</CODE>.
</P>
<H2>Function Sets</H2>
<P>
Just as <CODE>timesoftfloat</CODE> can time a function for all five rounding
modes in sequence, multiple functions can be timed with a single execution of
<CODE>timesoftfloat</CODE>.
Three sets are recognized:
<CODE>-all1</CODE>, <CODE>-all2</CODE>, and <CODE>-all</CODE>.
The set <CODE>-all1</CODE> is all one-operand operations, <CODE>-all2</CODE> is
all two-operand operations, and <CODE>-all</CODE> is obviously all operations.
A function set is used in place of a function name in the
<CODE>timesoftfloat</CODE> command line, such as
<BLOCKQUOTE>
<PRE>
timesoftfloat [<option>...] -all1
</PRE>
</BLOCKQUOTE>
</P>
<P>
For the purpose of deciding the number of operands of an operation, any
<CODE><I>roundingMode</I></CODE> and <CODE><I>exact</I></CODE> arguments are
ignored.
(Such arguments specify the rounding mode and whether the <I>inexact</I>
exception flag may be raised, respectively.)
Thus, functions that convert to integer type and the <CODE>roundToInt</CODE>
functions are included in the set of one-operand operations timed by
<CODE>-all1</CODE>.
</P>
</BODY>
|
