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
|
<?xml version="1.0" encoding="utf-8" ?>
<!DOCTYPE appref SYSTEM "appref.dtd">
<appref>
<header>
<copyright>
<year>1997</year><year>2018</year>
<holder>Ericsson AB. All Rights Reserved.</holder>
</copyright>
<legalnotice>
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
</legalnotice>
<title>HiPE</title>
<prepared></prepared>
<responsible></responsible>
<docno></docno>
<approved></approved>
<checked></checked>
<date></date>
<rev></rev>
<file>hipe.xml</file>
</header>
<app>HiPE</app>
<appsummary>The HiPE Application</appsummary>
<description>
<note>
<p>
HiPE and execution of HiPE compiled code only have limited support by
the OTP team at Ericsson. The OTP team only does limited maintenance
of HiPE and does not actively develop HiPE. HiPE is mainly supported
by the HiPE team at Uppsala University.
</p>
</note>
<p>
The normal way to native-compile an Erlang module using HiPE is to include the atom native
in the Erlang compiler options, as in:</p>
<pre>
1> <input>c(my_module, [native]).</input></pre>
<p>Options to the HiPE compiler are then passed as follows:</p>
<pre>
1> <input>c(my_module, [native,{hipe,Options}]).</input></pre>
<p>For on-line help in the Erlang shell, call <c>hipe:help()</c>.
Details on HiPE compiler options are given by <c>hipe:help_options()</c>.</p>
</description>
<section>
<title>Feature Limitations</title>
<p>
The HiPE compiler is in general compliant with the normal BEAM compiler,
with respect to semantic behavior. There are however features in the BEAM compiler
and the runtime system that have limited or no support for HiPE compiled modules.
</p>
<taglist>
<tag>Stack traces</tag>
<item><p>Stack traces returned from <seealso marker="erts:erlang#get_stacktrace/0">
<c>erlang:get_stacktrace/0</c></seealso> or as part of <c>'EXIT'</c> terms
can look incomplete if HiPE compiled functions are involved. Typically a stack trace
will contain only BEAM compiled functions or only HiPE compiled functions, depending
on where the exception was raised.</p>
<p>Source code line numbers in stack traces are also not supported by HiPE compiled functions.</p>
</item>
<tag>Tracing</tag>
<item><p>Erlang call trace is not supported by HiPE. Calling
<seealso marker="erts:erlang#trace_pattern/3"><c>erlang:trace_pattern({M,F,A}, ...)</c></seealso>
does not have any effect on HiPE compiled modules.</p>
</item>
<tag>NIFs</tag>
<item><p>Modules compiled with HiPE can not call <seealso marker="erts:erlang#load_nif-2">
<c>erlang:load_nif/2</c></seealso> to load NIFs.</p>
</item>
<tag>-on_load</tag>
<item><p>Modules compiled with HiPE can not use
<seealso marker="doc/reference_manual:code_loading#on_load"><c>-on_load()</c></seealso>
directives.</p>
</item>
</taglist>
</section>
<section>
<title>Performance Limitations</title>
<p>
The HiPE compiler does in general produce faster code than the
BEAM compiler. There are however some situation when HiPE
compiled code will perform worse than BEAM code.
</p>
<taglist>
<tag>Mode switches</tag>
<item><p>Every time a process changes from executing code in a
HiPE compiled module to a BEAM compiled module (or vice versa),
it will do a mode switch. This involves a certain amount of
CPU overhead which can have a negative net impact if the
process is switching back and forth without getting enough done in
each mode.</p>
</item>
<tag>Optimization for <c>receive</c> with unique references</tag>
<item>
<p>
The BEAM compiler can do an optimization when a receive
statement is only waiting for messages containing a reference
created before the receive. All messages that existed in the
queue when the reference was created will be bypassed, as they
cannot possibly contain the reference. HiPE currently has an
optimization similar this, but it is not guaranteed to
bypass all messages. In the worst case scenario, it cannot
bypass any messages at all.
</p>
<p>
An example of this is when <c>gen_server:call()</c> waits for
the reply message.
</p>
</item>
<tag>Garbage collection after BIFs</tag>
<item>
<p>
The condition for determining whether a garbage collection
is needed or not has changed in later releases. HiPE has not
been updated regarding this which may cause premature garbage
collections after BIF calls.
</p>
</item>
</taglist>
</section>
<section>
<title>Stability Issues</title>
<taglist>
<tag>Not checking reduction count on function returns</tag>
<item>
<p>
BEAM checks the reduction count and schedules out the executing
process if needed both when calling a function and when returning
from a function call that was not called using a tail call.
HiPE only checks the reduction count when calling a function.
</p>
<p>
The runtime system might need to schedule out a process
in order to reclaim memory. If the process isn't scheduled
out soon after the process has entered this state, memory
consumption will quickly grow. Maintaining this state is also
quite expensive performance wise.
</p>
<p>
Processes executing code that performs large recursions and
produce data after returning from recursive calls may have to
be scheduled out when returning from a function call. Since
HiPE does not check reductions on returns, processes executing
such HiPE compiled code may cause huge peeks in memory
consumption as well as severe performance degradation.
</p>
</item>
<tag>Not bumping appropriate amount of reductions in <c>receive</c> statements</tag>
<item>
<p>
The process signaling improvements made in ERTS version
10.0 moved potentially significant amounts of work into the
receive statement from other places. In order to account for
this work, the reduction count should be bumped on the
executing process. Reductions are not bumped when entering
the <c>receive</c> statement from HiPE compiled code.
</p>
</item>
</taglist>
</section>
<section>
<title>SEE ALSO</title>
<p>
<seealso marker="stdlib:c">c(3)</seealso>,
<seealso marker="compiler:compile">compile(3)</seealso>
</p>
</section>
</appref>
|
