blob: 4945d60f4ddaf1b7caedaa8056eafcf6b17e2a31 [file] [log] [blame]
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
<html>
<head>
<title>Bytecode for the Dalvik VM</title>
<link rel=stylesheet href="dalvik-bytecode.css">
</head>
<body>
<h1>Bytecode for the Dalvik VM</h1>
<p>Copyright &copy; 2007 The Android Open Source Project
<h2>General Design</h2>
<ul>
<li>The machine model and calling conventions are meant to approximately
imitate common real architectures and C-style calling conventions:
<ul>
<li>The VM is register-based, and frames are fixed in size upon creation.
Each frame consists of a particular number of registers (specified by
the method) as well as any adjunct data needed to execute the method,
such as (but not limited to) the program counter and a reference to the
<code>.dex</code> file that contains the method.
</li>
<li>Registers are 32 bits wide. Adjacent register pairs are used for 64-bit
values.
</li>
<li>In terms of bitwise representation, <code>(Object) null == (int)
0</code>.
</li>
<li>The <i>N</i> arguments to a method land in the last <i>N</i> registers
of the method's invocation frame, in order. Wide arguments consume
two registers. Instance methods are passed a <code>this</code> reference
as their first argument.
</li>
</ul>
<li>The storage unit in the instruction stream is a 16-bit unsigned quantity.
Some bits in some instructions are ignored / must-be-zero.
</li>
<li>Instructions aren't gratuitously limited to a particular type. For
example, instructions that move 32-bit register values without interpretation
don't have to specify whether they are moving ints or floats.
</li>
<li>There are separately enumerated and indexed constant pools for
references to strings, types, fields, and methods.
</li>
<li>Bitwise literal data is represented in-line in the instruction stream.</li>
<li>Because, in practice, it is uncommon for a method to need more than
16 registers, and because needing more than eight registers <i>is</i>
reasonably common, many instructions may only address the first 16
registers. When reasonably possible, instructions allow references to
up to the first 256 registers. In cases where an instruction variant isn't
available to address a desired register, it is expected that the register
contents get moved from the original register to a low register (before the
operation) and/or moved from a low result register to a high register
(after the operation).
</li>
<li>There are several "pseudo-instructions" that are used to hold
variable-length data referred to by regular instructions (for example,
<code>fill-array-data</code>). Such instructions must never be
encountered during the normal flow of execution. In addition, the
instructions must be located on even-numbered bytecode offsets (that is,
4-byte aligned). In order to meet this requirement, dex generation tools
should emit an extra <code>nop</code> instruction as a spacer if such an
instruction would otherwise be unaligned. Finally, though not required,
it is expected that most tools will choose to emit these instructions at
the ends of methods, since otherwise it would likely be the case that
additional instructions would be needed to branch around them.
</li>
<li>When installed on a running system, some instructions may be altered,
changing their format, as an install-time static linking optimization.
This is to allow for faster execution once linkage is known.
See the associated
<a href="instruction-formats.html">instruction formats document</a>
for the suggested variants. The word "suggested" is used advisedly;
it is not mandatory to implement these.
</li>
<li>Human-syntax and mnemonics:
<ul>
<li>Dest-then-source ordering for arguments.</li>
<li>Some opcodes have a disambiguating suffix with respect to the type(s)
they operate on: Type-general 64-bit opcodes
are suffixed with <code>-wide</code>.
Type-specific opcodes are suffixed with their type (or a
straightforward abbreviation), one of: <code>-boolean</code>
<code>-byte</code> <code>-char</code> <code>-short</code>
<code>-int</code> <code>-long</code> <code>-float</code>
<code>-double</code> <code>-object</code> <code>-string</code>
<code>-class</code> <code>-void</code>. Type-general 32-bit opcodes
are unmarked.
</li>
<li>Some opcodes have a disambiguating suffix to distinguish
otherwise-identical operations that have different instruction layouts
or options. These suffixes are separated from the main names with a slash
("<code>/</code>") and mainly exist at all to make there be a one-to-one
mapping with static constants in the code that generates and interprets
executables (that is, to reduce ambiguity for humans).
</li>
</ul>
</li>
<li>See the <a href="instruction-formats.html">instruction formats
document</a> for more details about the various instruction formats
(listed under "Op &amp; Format") as well as details about the opcode
syntax.
</li>
</ul>
<h2>Summary of Instruction Set</h2>
<table class="instruc">
<thead>
<tr>
<th>Op &amp; Format</th>
<th>Mnemonic / Syntax</th>
<th>Arguments</th>
<th>Description</th>
</tr>
</thead>
<tbody>
<tr>
<td>00 10x</td>
<td>nop</td>
<td>&nbsp;</td>
<td>Waste cycles.</td>
</tr>
<tr>
<td>01 12x</td>
<td>move vA, vB</td>
<td><code>A:</code> destination register (4 bits)<br/>
<code>B:</code> source register (4 bits)</td>
<td>Move the contents of one non-object register to another.</td>
</tr>
<tr>
<td>02 22x</td>
<td>move/from16 vAA, vBBBB</td>
<td><code>A:</code> destination register (8 bits)<br/>
<code>B:</code> source register (16 bits)</td>
<td>Move the contents of one non-object register to another.</td>
</tr>
<tr>
<td>03 32x</td>
<td>move/16 vAAAA, vBBBB</td>
<td><code>A:</code> destination register (16 bits)<br/>
<code>B:</code> source register (16 bits)</td>
<td>Move the contents of one non-object register to another.</td>
</tr>
<tr>
<td>04 12x</td>
<td>move-wide vA, vB</td>
<td><code>A:</code> destination register pair (4 bits)<br/>
<code>B:</code> source register pair (4 bits)</td>
<td>Move the contents of one register-pair to another.
<p><b>Note:</b>
It is legal to move from <code>v<i>N</i></code> to either
<code>v<i>N-1</i></code> or <code>v<i>N+1</i></code>, so implementations
must arrange for both halves of a register pair to be read before
anything is written.</p>
</td>
</tr>
<tr>
<td>05 22x</td>
<td>move-wide/from16 vAA, vBBBB</td>
<td><code>A:</code> destination register pair (8 bits)<br/>
<code>B:</code> source register pair (16 bits)</td>
<td>Move the contents of one register-pair to another.
<p><b>Note:</b>
Implementation considerations are the same as <code>move-wide</code>,
above.</p>
</td>
</tr>
<tr>
<td>06 32x</td>
<td>move-wide/16 vAAAA, vBBBB</td>
<td><code>A:</code> destination register pair (16 bits)<br/>
<code>B:</code> source register pair (16 bits)</td>
<td>Move the contents of one register-pair to another.
<p><b>Note:</b>
Implementation considerations are the same as <code>move-wide</code>,
above.</p>
</td>
</tr>
<tr>
<td>07 12x</td>
<td>move-object vA, vB</td>
<td><code>A:</code> destination register (4 bits)<br/>
<code>B:</code> source register (4 bits)</td>
<td>Move the contents of one object-bearing register to another.</td>
</tr>
<tr>
<td>08 22x</td>
<td>move-object/from16 vAA, vBBBB</td>
<td><code>A:</code> destination register (8 bits)<br/>
<code>B:</code> source register (16 bits)</td>
<td>Move the contents of one object-bearing register to another.</td>
</tr>
<tr>
<td>09 32x</td>
<td>move-object/16 vAAAA, vBBBB</td>
<td><code>A:</code> destination register (16 bits)<br/>
<code>B:</code> source register (16 bits)</td>
<td>Move the contents of one object-bearing register to another.</td>
</tr>
<tr>
<td>0a 11x</td>
<td>move-result vAA</td>
<td><code>A:</code> destination register (8 bits)</td>
<td>Move the single-word non-object result of the most recent
<code>invoke-<i>kind</i></code> into the indicated register.
This must be done as the instruction immediately after an
<code>invoke-<i>kind</i></code> whose (single-word, non-object) result
is not to be ignored; anywhere else is invalid.</td>
</tr>
<tr>
<td>0b 11x</td>
<td>move-result-wide vAA</td>
<td><code>A:</code> destination register pair (8 bits)</td>
<td>Move the double-word result of the most recent
<code>invoke-<i>kind</i></code> into the indicated register pair.
This must be done as the instruction immediately after an
<code>invoke-<i>kind</i></code> whose (double-word) result
is not to be ignored; anywhere else is invalid.</td>
</tr>
<tr>
<td>0c 11x</td>
<td>move-result-object vAA</td>
<td><code>A:</code> destination register (8 bits)</td>
<td>Move the object result of the most recent <code>invoke-<i>kind</i></code>
into the indicated register. This must be done as the instruction
immediately after an <code>invoke-<i>kind</i></code> or
<code>filled-new-array</code>
whose (object) result is not to be ignored; anywhere else is invalid.</td>
</tr>
<tr>
<td>0d 11x</td>
<td>move-exception vAA</td>
<td><code>A:</code> destination register (8 bits)</td>
<td>Save a just-caught exception into the given register. This should
be the first instruction of any exception handler whose caught
exception is not to be ignored, and this instruction may <i>only</i>
ever occur as the first instruction of an exception handler; anywhere
else is invalid.</td>
</tr>
<tr>
<td>0e 10x</td>
<td>return-void</td>
<td>&nbsp;</td>
<td>Return from a <code>void</code> method.</td>
</tr>
<tr>
<td>0f 11x</td>
<td>return vAA</td>
<td><code>A:</code> return value register (8 bits)</td>
<td>Return from a single-width (32-bit) non-object value-returning
method.
</td>
</tr>
<tr>
<td>10 11x</td>
<td>return-wide vAA</td>
<td><code>A:</code> return value register-pair (8 bits)</td>
<td>Return from a double-width (64-bit) value-returning method.</td>
</tr>
<tr>
<td>11 11x</td>
<td>return-object vAA</td>
<td><code>A:</code> return value register (8 bits)</td>
<td>Return from an object-returning method.</td>
</tr>
<tr>
<td>12 11n</td>
<td>const/4 vA, #+B</td>
<td><code>A:</code> destination register (4 bits)<br/>
<code>B:</code> signed int (4 bits)</td>
<td>Move the given literal value (sign-extended to 32 bits) into
the specified register.</td>
</tr>
<tr>
<td>13 21s</td>
<td>const/16 vAA, #+BBBB</td>
<td><code>A:</code> destination register (8 bits)<br/>
<code>B:</code> signed int (16 bits)</td>
<td>Move the given literal value (sign-extended to 32 bits) into
the specified register.</td>
</tr>
<tr>
<td>14 31i</td>
<td>const vAA, #+BBBBBBBB</td>
<td><code>A:</code> destination register (8 bits)<br/>
<code>B:</code> arbitrary 32-bit constant</td>
<td>Move the given literal value into the specified register.</td>
</tr>
<tr>
<td>15 21h</td>
<td>const/high16 vAA, #+BBBB0000</td>
<td><code>A:</code> destination register (8 bits)<br/>
<code>B:</code> signed int (16 bits)</td>
<td>Move the given literal value (right-zero-extended to 32 bits) into
the specified register.</td>
</tr>
<tr>
<td>16 21s</td>
<td>const-wide/16 vAA, #+BBBB</td>
<td><code>A:</code> destination register (8 bits)<br/>
<code>B:</code> signed int (16 bits)</td>
<td>Move the given literal value (sign-extended to 64 bits) into
the specified register-pair.</td>
</tr>
<tr>
<td>17 31i</td>
<td>const-wide/32 vAA, #+BBBBBBBB</td>
<td><code>A:</code> destination register (8 bits)<br/>
<code>B:</code> signed int (32 bits)</td>
<td>Move the given literal value (sign-extended to 64 bits) into
the specified register-pair.</td>
</tr>
<tr>
<td>18 51l</td>
<td>const-wide vAA, #+BBBBBBBBBBBBBBBB</td>
<td><code>A:</code> destination register (8 bits)<br/>
<code>B:</code> arbitrary double-width (64-bit) constant</td>
<td>Move the given literal value into
the specified register-pair.</td>
</tr>
<tr>
<td>19 21h</td>
<td>const-wide/high16 vAA, #+BBBB000000000000</td>
<td><code>A:</code> destination register (8 bits)<br/>
<code>B:</code> signed int (16 bits)</td>
<td>Move the given literal value (right-zero-extended to 64 bits) into
the specified register-pair.</td>
</tr>
<tr>
<td>1a 21c</td>
<td>const-string vAA, string@BBBB</td>
<td><code>A:</code> destination register (8 bits)<br/>
<code>B:</code> string index</td>
<td>Move a reference to the string specified by the given index into the
specified register.</td>
</tr>
<tr>
<td>1b 31c</td>
<td>const-string/jumbo vAA, string@BBBBBBBB</td>
<td><code>A:</code> destination register (8 bits)<br/>
<code>B:</code> string index</td>
<td>Move a reference to the string specified by the given index into the
specified register.</td>
</tr>
<tr>
<td>1c 21c</td>
<td>const-class vAA, type@BBBB</td>
<td><code>A:</code> destination register (8 bits)<br/>
<code>B:</code> type index</td>
<td>Move a reference to the class specified by the given index into the
specified register. In the case where the indicated type is primitive,
this will store a reference to the primitive type's degenerate
class.</td>
</tr>
<tr>
<td>1d 11x</td>
<td>monitor-enter vAA</td>
<td><code>A:</code> reference-bearing register (8 bits)</td>
<td>Acquire the monitor for the indicated object.</td>
</tr>
<tr>
<td>1e 11x</td>
<td>monitor-exit vAA</td>
<td><code>A:</code> reference-bearing register (8 bits)</td>
<td>Release the monitor for the indicated object.
<p><b>Note:</b>
If this instruction needs to throw an exception, it must do
so as if the pc has already advanced past the instruction.
It may be useful to think of this as the instruction successfully
executing (in a sense), and the exception getting thrown <i>after</i>
the instruction but <i>before</i> the next one gets a chance to
run. This definition makes it possible for a method to use
a monitor cleanup catch-all (e.g., <code>finally</code>) block as
the monitor cleanup for that block itself, as a way to handle the
arbitrary exceptions that might get thrown due to the historical
implementation of <code>Thread.stop()</code>, while still managing
to have proper monitor hygiene.</p>
</td>
</tr>
<tr>
<td>1f 21c</td>
<td>check-cast vAA, type@BBBB</td>
<td><code>A:</code> reference-bearing register (8 bits)<br/>
<code>B:</code> type index (16 bits)</td>
<td>Throw a <code>ClassCastException</code> if the reference in the
given register cannot be cast to the indicated type.
<p><b>Note:</b> Since <code>A</code> must always be a reference
(and not a primitive value), this will necessarily fail at runtime
(that is, it will throw an exception) if <code>B</code> refers to a
primitive type.</p>
</td>
</tr>
<tr>
<td>20 22c</td>
<td>instance-of vA, vB, type@CCCC</td>
<td><code>A:</code> destination register (4 bits)<br/>
<code>B:</code> reference-bearing register (4 bits)<br/>
<code>C:</code> type index (16 bits)</td>
<td>Store in the given destination register <code>1</code>
if the indicated reference is an instance of the given type,
or <code>0</code> if not.
<p><b>Note:</b> Since <code>B</code> must always be a reference
(and not a primitive value), this will always result
in <code>0</code> being stored if <code>C</code> refers to a primitive
type.</td>
</tr>
<tr>
<td>21 12x</td>
<td>array-length vA, vB</td>
<td><code>A:</code> destination register (4 bits)<br/>
<code>B:</code> array reference-bearing register (4 bits)</td>
<td>Store in the given destination register the length of the indicated
array, in entries</td>
</tr>
<tr>
<td>22 21c</td>
<td>new-instance vAA, type@BBBB</td>
<td><code>A:</code> destination register (8 bits)<br/>
<code>B:</code> type index</td>
<td>Construct a new instance of the indicated type, storing a
reference to it in the destination. The type must refer to a
non-array class.</td>
</tr>
<tr>
<td>23 22c</td>
<td>new-array vA, vB, type@CCCC</td>
<td><code>A:</code> destination register (8 bits)<br/>
<code>B:</code> size register<br/>
<code>C:</code> type index</td>
<td>Construct a new array of the indicated type and size. The type
must be an array type.</td>
</tr>
<tr>
<td>24 35c</td>
<td>filled-new-array {vD, vE, vF, vG, vA}, type@CCCC</td>
<td><code>B:</code> array size and argument word count (4 bits)<br/>
<code>C:</code> type index (16 bits)<br/>
<code>D..G, A:</code> argument registers (4 bits each)</td>
<td>Construct an array of the given type and size, filling it with the
supplied contents. The type must be an array type. The array's
contents must be single-word (that is,
no arrays of <code>long</code> or <code>double</code>). The constructed
instance is stored as a "result" in the same way that the method invocation
instructions store their results, so the constructed instance must
be moved to a register with a subsequent
<code>move-result-object</code> instruction (if it is to be used).</td>
</tr>
<tr>
<td>25 3rc</td>
<td>filled-new-array/range {vCCCC .. vNNNN}, type@BBBB</td>
<td><code>A:</code> array size and argument word count (8 bits)<br/>
<code>B:</code> type index (16 bits)<br/>
<code>C:</code> first argument register (16 bits)<br/>
<code>N = A + C - 1</code></td>
<td>Construct an array of the given type and size, filling it with
the supplied contents. Clarifications and restrictions are the same
as <code>filled-new-array</code>, described above.</td>
</tr>
<tr>
<td>26 31t</td>
<td>fill-array-data vAA, +BBBBBBBB <i>(with supplemental data as specified
below in "<code>fill-array-data</code> Format")</i></td>
<td><code>A:</code> array reference (8 bits)<br/>
<code>B:</code> signed "branch" offset to table data pseudo-instruction
(32 bits)
</td>
<td>Fill the given array with the indicated data. The reference must be
to an array of primitives, and the data table must match it in type and
must contain no more elements than will fit in the array. That is,
the array may be larger than the table, and if so, only the initial
elements of the array are set, leaving the remainder alone.
</td>
</tr>
<tr>
<td>27 11x</td>
<td>throw vAA</td>
<td><code>A:</code> exception-bearing register (8 bits)<br/></td>
<td>Throw the indicated exception.</td>
</tr>
<tr>
<td>28 10t</td>
<td>goto +AA</td>
<td><code>A:</code> signed branch offset (8 bits)</td>
<td>Unconditionally jump to the indicated instruction.
<p><b>Note:</b>
The branch offset may not be <code>0</code>. (A spin
loop may be legally constructed either with <code>goto/32</code> or
by including a <code>nop</code> as a target before the branch.)</p>
</td>
</tr>
<tr>
<td>29 20t</td>
<td>goto/16 +AAAA</td>
<td><code>A:</code> signed branch offset (16 bits)<br/></td>
<td>Unconditionally jump to the indicated instruction.
<p><b>Note:</b>
The branch offset may not be <code>0</code>. (A spin
loop may be legally constructed either with <code>goto/32</code> or
by including a <code>nop</code> as a target before the branch.)</p>
</td>
</tr>
<tr>
<td>2a 30t</td>
<td>goto/32 +AAAAAAAA</td>
<td><code>A:</code> signed branch offset (32 bits)<br/></td>
<td>Unconditionally jump to the indicated instruction.</td>
</tr>
<tr>
<td>2b 31t</td>
<td>packed-switch vAA, +BBBBBBBB <i>(with supplemental data as
specified below in "<code>packed-switch</code> Format")</i></td>
<td><code>A:</code> register to test<br/>
<code>B:</code> signed "branch" offset to table data pseudo-instruction
(32 bits)
</td>
<td>Jump to a new instruction based on the value in the
given register, using a table of offsets corresponding to each value
in a particular integral range, or fall through to the next
instruction if there is no match.
</td>
</tr>
<tr>
<td>2c 31t</td>
<td>sparse-switch vAA, +BBBBBBBB <i>(with supplemental data as
specified below in "<code>sparse-switch</code> Format")</i></td>
<td><code>A:</code> register to test<br/>
<code>B:</code> signed "branch" offset to table data pseudo-instruction
(32 bits)
</td>
<td>Jump to a new instruction based on the value in the given
register, using an ordered table of value-offset pairs, or fall
through to the next instruction if there is no match.
</td>
</tr>
<tr>
<td>2d..31 23x</td>
<td>cmp<i>kind</i> vAA, vBB, vCC<br/>
2d: cmpl-float <i>(lt bias)</i><br/>
2e: cmpg-float <i>(gt bias)</i><br/>
2f: cmpl-double <i>(lt bias)</i><br/>
30: cmpg-double <i>(gt bias)</i><br/>
31: cmp-long
</td>
<td><code>A:</code> destination register (8 bits)<br/>
<code>B:</code> first source register or pair<br/>
<code>C:</code> second source register or pair</td>
<td>Perform the indicated floating point or <code>long</code> comparison,
storing <code>0</code> if the two arguments are equal, <code>1</code>
if the second argument is larger, or <code>-1</code> if the first
argument is larger. The "bias" listed for the floating point operations
indicates how <code>NaN</code> comparisons are treated: "Gt bias"
instructions return <code>1</code> for <code>NaN</code> comparisons,
and "lt bias" instructions return
<code>-1</code>.
<p>For example, to check to see if floating point
<code>a &lt; b</code>, then it is advisable to use
<code>cmpg-float</code>; a result of <code>-1</code> indicates that
the test was true, and the other values indicate it was false either
due to a valid comparison or because one or the other values was
<code>NaN</code>.</p>
</td>
</tr>
<tr>
<td>32..37 22t</td>
<td>if-<i>test</i> vA, vB, +CCCC<br/>
32: if-eq<br/>
33: if-ne<br/>
34: if-lt<br/>
35: if-ge<br/>
36: if-gt<br/>
37: if-le<br/>
</td>
<td><code>A:</code> first register to test (4 bits)<br/>
<code>B:</code> second register to test (4 bits)<br/>
<code>C:</code> signed branch offset (16 bits)</td>
<td>Branch to the given destination if the given two registers' values
compare as specified.
<p><b>Note:</b>
The branch offset may not be <code>0</code>. (A spin
loop may be legally constructed either by branching around a
backward <code>goto</code> or by including a <code>nop</code> as
a target before the branch.)</p>
</td>
</tr>
<tr>
<td>38..3d 21t</td>
<td>if-<i>test</i>z vAA, +BBBB<br/>
38: if-eqz<br/>
39: if-nez<br/>
3a: if-ltz<br/>
3b: if-gez<br/>
3c: if-gtz<br/>
3d: if-lez<br/>
</td>
<td><code>A:</code> register to test (8 bits)<br/>
<code>B:</code> signed branch offset (16 bits)</td>
<td>Branch to the given destination if the given register's value compares
with 0 as specified.
<p><b>Note:</b>
The branch offset may not be <code>0</code>. (A spin
loop may be legally constructed either by branching around a
backward <code>goto</code> or by including a <code>nop</code> as
a target before the branch.)</p>
</td>
</tr>
<tr>
<td>3e..43 10x</td>
<td><i>(unused)</i></td>
<td>&nbsp;</td>
<td><i>(unused)</i></td>
</tr>
<tr>
<td>44..51 23x</td>
<td><i>arrayop</i> vAA, vBB, vCC<br/>
44: aget<br/>
45: aget-wide<br/>
46: aget-object<br/>
47: aget-boolean<br/>
48: aget-byte<br/>
49: aget-char<br/>
4a: aget-short<br/>
4b: aput<br/>
4c: aput-wide<br/>
4d: aput-object<br/>
4e: aput-boolean<br/>
4f: aput-byte<br/>
50: aput-char<br/>
51: aput-short
</td>
<td><code>A:</code> value register or pair; may be source or dest
(8 bits)<br/>
<code>B:</code> array register (8 bits)<br/>
<code>C:</code> index register (8 bits)</td>
<td>Perform the identified array operation at the identified index of
the given array, loading or storing into the value register.</td>
</tr>
<tr>
<td>52..5f 22c</td>
<td>i<i>instanceop</i> vA, vB, field@CCCC<br/>
52: iget<br/>
53: iget-wide<br/>
54: iget-object<br/>
55: iget-boolean<br/>
56: iget-byte<br/>
57: iget-char<br/>
58: iget-short<br/>
59: iput<br/>
5a: iput-wide<br/>
5b: iput-object<br/>
5c: iput-boolean<br/>
5d: iput-byte<br/>
5e: iput-char<br/>
5f: iput-short
</td>
<td><code>A:</code> value register or pair; may be source or dest
(4 bits)<br/>
<code>B:</code> object register (4 bits)<br/>
<code>C:</code> instance field reference index (16 bits)</td>
<td>Perform the identified object instance field operation with
the identified field, loading or storing into the value register.
<p><b>Note:</b> These opcodes are reasonable candidates for static linking,
altering the field argument to be a more direct offset.</p>
</td>
</tr>
<tr>
<td>60..6d 21c</td>
<td>s<i>staticop</i> vAA, field@BBBB<br/>
60: sget<br/>
61: sget-wide<br/>
62: sget-object<br/>
63: sget-boolean<br/>
64: sget-byte<br/>
65: sget-char<br/>
66: sget-short<br/>
67: sput<br/>
68: sput-wide<br/>
69: sput-object<br/>
6a: sput-boolean<br/>
6b: sput-byte<br/>
6c: sput-char<br/>
6d: sput-short
</td>
<td><code>A:</code> value register or pair; may be source or dest
(8 bits)<br/>
<code>B:</code> static field reference index (16 bits)</td>
<td>Perform the identified object static field operation with the identified
static field, loading or storing into the value register.
<p><b>Note:</b> These opcodes are reasonable candidates for static linking,
altering the field argument to be a more direct offset.</p>
</td>
</tr>
<tr>
<td>6e..72 35c</td>
<td>invoke-<i>kind</i> {vD, vE, vF, vG, vA}, meth@CCCC<br/>
6e: invoke-virtual<br/>
6f: invoke-super<br/>
70: invoke-direct<br/>
71: invoke-static<br/>
72: invoke-interface
</td>
<td><code>B:</code> argument word count (4 bits)<br/>
<code>C:</code> method index (16 bits)<br/>
<code>D..G, A:</code> argument registers (4 bits each)</td>
<td>Call the indicated method. The result (if any) may be stored
with an appropriate <code>move-result*</code> variant as the immediately
subsequent instruction.
<p><code>invoke-virtual</code> is used to invoke a normal virtual
method (a method that is not <code>static</code> or <code>final</code>,
and is not a constructor).</p>
<p><code>invoke-super</code> is used to invoke the closest superclass's
virtual method (as opposed to the one with the same <code>method_id</code>
in the calling class).</p>
<p><code>invoke-direct</code> is used to invoke a non-<code>static</code>
direct method (that is, an instance method that is by its nature
non-overridable, namely either a <code>private</code> instance method
or a constructor).</p>
<p><code>invoke-static</code> is used to invoke a <code>static</code>
method (which is always considered a direct method).</p>
<p><code>invoke-interface</code> is used to invoke an
<code>interface</code> method, that is, on an object whose concrete
class isn't known, using a <code>method_id</code> that refers to
an <code>interface</code>.</p>
<p><b>Note:</b> These opcodes are reasonable candidates for static linking,
altering the method argument to be a more direct offset
(or pair thereof).</p>
</td>
</tr>
<tr>
<td>73 10x</td>
<td><i>(unused)</i></td>
<td>&nbsp;</td>
<td><i>(unused)</i></td>
</tr>
<tr>
<td>74..78 3rc</td>
<td>invoke-<i>kind</i>/range {vCCCC .. vNNNN}, meth@BBBB<br/>
74: invoke-virtual/range<br/>
75: invoke-super/range<br/>
76: invoke-direct/range<br/>
77: invoke-static/range<br/>
78: invoke-interface/range
</td>
<td><code>A:</code> argument word count (8 bits)<br/>
<code>B:</code> method index (16 bits)<br/>
<code>C:</code> first argument register (16 bits)<br/>
<code>N = A + C - 1</code></td>
<td>Call the indicated method. See first <code>invoke-<i>kind</i></code>
description above for details, caveats, and suggestions.
</td>
</tr>
<tr>
<td>79..7a 10x</td>
<td><i>(unused)</i></td>
<td>&nbsp;</td>
<td><i>(unused)</i></td>
</tr>
<tr>
<td>7b..8f 12x</td>
<td><i>unop</i> vA, vB<br/>
7b: neg-int<br/>
7c: not-int<br/>
7d: neg-long<br/>
7e: not-long<br/>
7f: neg-float<br/>
80: neg-double<br/>
81: int-to-long<br/>
82: int-to-float<br/>
83: int-to-double<br/>
84: long-to-int<br/>
85: long-to-float<br/>
86: long-to-double<br/>
87: float-to-int<br/>
88: float-to-long<br/>
89: float-to-double<br/>
8a: double-to-int<br/>
8b: double-to-long<br/>
8c: double-to-float<br/>
8d: int-to-byte<br/>
8e: int-to-char<br/>
8f: int-to-short
</td>
<td><code>A:</code> destination register or pair (4 bits)<br/>
<code>B:</code> source register or pair (4 bits)</td>
<td>Perform the identified unary operation on the source register,
storing the result in the destination register.</td>
</tr>
<tr>
<td>90..af 23x</td>
<td><i>binop</i> vAA, vBB, vCC<br/>
90: add-int<br/>
91: sub-int<br/>
92: mul-int<br/>
93: div-int<br/>
94: rem-int<br/>
95: and-int<br/>
96: or-int<br/>
97: xor-int<br/>
98: shl-int<br/>
99: shr-int<br/>
9a: ushr-int<br/>
9b: add-long<br/>
9c: sub-long<br/>
9d: mul-long<br/>
9e: div-long<br/>
9f: rem-long<br/>
a0: and-long<br/>
a1: or-long<br/>
a2: xor-long<br/>
a3: shl-long<br/>
a4: shr-long<br/>
a5: ushr-long<br/>
a6: add-float<br/>
a7: sub-float<br/>
a8: mul-float<br/>
a9: div-float<br/>
aa: rem-float<br/>
ab: add-double<br/>
ac: sub-double<br/>
ad: mul-double<br/>
ae: div-double<br/>
af: rem-double
</td>
<td><code>A:</code> destination register or pair (8 bits)<br/>
<code>B:</code> first source register or pair (8 bits)<br/>
<code>C:</code> second source register or pair (8 bits)</td>
<td>Perform the identified binary operation on the two source registers,
storing the result in the first source register.</td>
</tr>
<tr>
<td>b0..cf 12x</td>
<td><i>binop</i>/2addr vA, vB<br/>
b0: add-int/2addr<br/>
b1: sub-int/2addr<br/>
b2: mul-int/2addr<br/>
b3: div-int/2addr<br/>
b4: rem-int/2addr<br/>
b5: and-int/2addr<br/>
b6: or-int/2addr<br/>
b7: xor-int/2addr<br/>
b8: shl-int/2addr<br/>
b9: shr-int/2addr<br/>
ba: ushr-int/2addr<br/>
bb: add-long/2addr<br/>
bc: sub-long/2addr<br/>
bd: mul-long/2addr<br/>
be: div-long/2addr<br/>
bf: rem-long/2addr<br/>
c0: and-long/2addr<br/>
c1: or-long/2addr<br/>
c2: xor-long/2addr<br/>
c3: shl-long/2addr<br/>
c4: shr-long/2addr<br/>
c5: ushr-long/2addr<br/>
c6: add-float/2addr<br/>
c7: sub-float/2addr<br/>
c8: mul-float/2addr<br/>
c9: div-float/2addr<br/>
ca: rem-float/2addr<br/>
cb: add-double/2addr<br/>
cc: sub-double/2addr<br/>
cd: mul-double/2addr<br/>
ce: div-double/2addr<br/>
cf: rem-double/2addr
</td>
<td><code>A:</code> destination and first source register or pair
(4 bits)<br/>
<code>B:</code> second source register or pair (4 bits)</td>
<td>Perform the identified binary operation on the two source registers,
storing the result in the first source register.</td>
</tr>
<tr>
<td>d0..d7 22s</td>
<td><i>binop</i>/lit16 vA, vB, #+CCCC<br/>
d0: add-int/lit16<br/>
d1: rsub-int (reverse subtract)<br/>
d2: mul-int/lit16<br/>
d3: div-int/lit16<br/>
d4: rem-int/lit16<br/>
d5: and-int/lit16<br/>
d6: or-int/lit16<br/>
d7: xor-int/lit16
</td>
<td><code>A:</code> destination register (4 bits)<br/>
<code>B:</code> source register (4 bits)<br/>
<code>C:</code> signed int constant (16 bits)</td>
<td>Perform the indicated binary op on the indicated register (first
argument) and literal value (second argument), storing the result in
the destination register.
<p><b>Note:</b>
<code>rsub-int</code> does not have a suffix since this version is the
main opcode of its family. Also, see below for details on its semantics.
</p>
</td>
</tr>
<tr>
<td>d8..e2 22b</td>
<td><i>binop</i>/lit8 vAA, vBB, #+CC<br/>
d8: add-int/lit8<br/>
d9: rsub-int/lit8<br/>
da: mul-int/lit8<br/>
db: div-int/lit8<br/>
dc: rem-int/lit8<br/>
dd: and-int/lit8<br/>
de: or-int/lit8<br/>
df: xor-int/lit8<br/>
e0: shl-int/lit8<br/>
e1: shr-int/lit8<br/>
e2: ushr-int/lit8
</td>
<td><code>A:</code> destination register (8 bits)<br/>
<code>B:</code> source register (8 bits)<br/>
<code>C:</code> signed int constant (8 bits)</td>
<td>Perform the indicated binary op on the indicated register (first
argument) and literal value (second argument), storing the result
in the destination register.
<p><b>Note:</b> See below for details on the semantics of
<code>rsub-int</code>.</p>
</td>
</tr>
<tr>
<td>e3..ff 10x</td>
<td><i>(unused)</i></td>
<td>&nbsp;</td>
<td><i>(unused)</i></td>
</tr>
</tbody>
</table>
<h2><code>packed-switch</code> Format</h2>
<table class="supplement">
<thead>
<tr>
<th>Name</th>
<th>Format</th>
<th>Description</th>
</tr>
</thead>
<tbody>
<tr>
<td>ident</td>
<td>ushort = 0x0100</td>
<td>identifying pseudo-opcode</td>
</tr>
<tr>
<td>size</td>
<td>ushort</td>
<td>number of entries in the table</td>
</tr>
<tr>
<td>first_key</td>
<td>int</td>
<td>first (and lowest) switch case value</td>
</tr>
<tr>
<td>targets</td>
<td>int[]</td>
<td>list of <code>size</code> relative branch targets. The targets are
relative to the address of the switch opcode, not of this table.
</td>
</tr>
</tbody>
</table>
<p><b>Note:</b> The total number of code units for an instance of this
table is <code>(size * 2) + 4</code>.</p>
<h2><code>sparse-switch</code> Format</h2>
<table class="supplement">
<thead>
<tr>
<th>Name</th>
<th>Format</th>
<th>Description</th>
</tr>
</thead>
<tbody>
<tr>
<td>ident</td>
<td>ushort = 0x0200</td>
<td>identifying pseudo-opcode</td>
</tr>
<tr>
<td>size</td>
<td>ushort</td>
<td>number of entries in the table</td>
</tr>
<tr>
<td>keys</td>
<td>int[]</td>
<td>list of <code>size</code> key values, sorted low-to-high</td>
</tr>
<tr>
<td>targets</td>
<td>int[]</td>
<td>list of <code>size</code> relative branch targets, each corresponding
to the key value at the same index. The targets are
relative to the address of the switch opcode, not of this table.
</td>
</tr>
</tbody>
</table>
<p><b>Note:</b> The total number of code units for an instance of this
table is <code>(size * 4) + 2</code>.</p>
<h2><code>fill-array-data</code> Format</h2>
<table class="supplement">
<thead>
<tr>
<th>Name</th>
<th>Format</th>
<th>Description</th>
</tr>
</thead>
<tbody>
<tr>
<td>ident</td>
<td>ushort = 0x0300</td>
<td>identifying pseudo-opcode</td>
</tr>
<tr>
<td>element_width</td>
<td>ushort</td>
<td>number of bytes in each element</td>
</tr>
<tr>
<td>size</td>
<td>uint</td>
<td>number of elements in the table</td>
</tr>
<tr>
<td>data</td>
<td>ubyte[]</td>
<td>data values</td>
</tr>
</tbody>
</table>
<p><b>Note:</b> The total number of code units for an instance of this
table is <code>(size * element_width + 1) / 2 + 4</code>.</p>
<h2>Mathematical Operation Details</h2>
<p><b>Note:</b> Floating point operations must follow IEEE 754 rules, using
round-to-nearest and gradual underflow, except where stated otherwise.</p>
<table class="math">
<thead>
<tr>
<th>Opcode</th>
<th>C Semantics</th>
<th>Notes</th>
</tr>
</thead>
<tbody>
<tr>
<td>neg-int</td>
<td>int32 a;<br/>
int32 result = -a;
</td>
<td>Unary twos-complement.</td>
</tr>
<tr>
<td>not-int</td>
<td>int32 a;<br/>
int32 result = ~a;
</td>
<td>Unary ones-complement.</td>
</tr>
<tr>
<td>neg-long</td>
<td>int64 a;<br/>
int64 result = -a;
</td>
<td>Unary twos-complement.</td>
</tr>
<tr>
<td>not-long</td>
<td>int64 a;<br/>
int64 result = ~a;
</td>
<td>Unary ones-complement.</td>
</tr>
<tr>
<td>neg-float</td>
<td>float a;<br/>
float result = -a;
</td>
<td>Floating point negation.</td>
</tr>
<tr>
<td>neg-double</td>
<td>double a;<br/>
double result = -a;
</td>
<td>Floating point negation.</td>
</tr>
<tr>
<td>int-to-long</td>
<td>int32 a;<br/>
int64 result = (int64) a;
</td>
<td>Sign extension of <code>int32</code> into <code>int64</code>.</td>
</tr>
<tr>
<td>int-to-float</td>
<td>int32 a;<br/>
float result = (float) a;
</td>
<td>Conversion of <code>int32</code> to <code>float</code>, using
round-to-nearest. This loses precision for some values.
</td>
</tr>
<tr>
<td>int-to-double</td>
<td>int32 a;<br/>
double result = (double) a;
</td>
<td>Conversion of <code>int32</code> to <code>double</code>.</td>
</tr>
<tr>
<td>long-to-int</td>
<td>int64 a;<br/>
int32 result = (int32) a;
</td>
<td>Truncation of <code>int64</code> into <code>int32</code>.</td>
</tr>
<tr>
<td>long-to-float</td>
<td>int64 a;<br/>
float result = (float) a;
</td>
<td>Conversion of <code>int64</code> to <code>float</code>, using
round-to-nearest. This loses precision for some values.
</td>
</tr>
<tr>
<td>long-to-double</td>
<td>int64 a;<br/>
double result = (double) a;
</td>
<td>Conversion of <code>int64</code> to <code>double</code>, using
round-to-nearest. This loses precision for some values.
</td>
</tr>
<tr>
<td>float-to-int</td>
<td>float a;<br/>
int32 result = (int32) a;
</td>
<td>Conversion of <code>float</code> to <code>int32</code>, using
round-toward-zero. <code>NaN</code> and <code>-0.0</code> (negative zero)
convert to the integer <code>0</code>. Infinities and values with
too large a magnitude to be represented get converted to either
<code>0x7fffffff</code> or <code>-0x80000000</code> depending on sign.
</td>
</tr>
<tr>
<td>float-to-long</td>
<td>float a;<br/>
int64 result = (int64) a;
</td>
<td>Conversion of <code>float</code> to <code>int64</code>, using
round-toward-zero. The same special case rules as for
<code>float-to-int</code> apply here, except that out-of-range values
get converted to either <code>0x7fffffffffffffff</code> or
<code>-0x8000000000000000</code> depending on sign.
</td>
</tr>
<tr>
<td>float-to-double</td>
<td>float a;<br/>
double result = (double) a;
</td>
<td>Conversion of <code>float</code> to <code>double</code>, preserving
the value exactly.
</td>
</tr>
<tr>
<td>double-to-int</td>
<td>double a;<br/>
int32 result = (int32) a;
</td>
<td>Conversion of <code>double</code> to <code>int32</code>, using
round-toward-zero. The same special case rules as for
<code>float-to-int</code> apply here.
</td>
</tr>
<tr>
<td>double-to-long</td>
<td>double a;<br/>
int64 result = (int64) a;
</td>
<td>Conversion of <code>double</code> to <code>int64</code>, using
round-toward-zero. The same special case rules as for
<code>float-to-long</code> apply here.
</td>
</tr>
<tr>
<td>double-to-float</td>
<td>double a;<br/>
float result = (float) a;
</td>
<td>Conversion of <code>double</code> to <code>float</code>, using
round-to-nearest. This loses precision for some values.
</td>
</tr>
<tr>
<td>int-to-byte</td>
<td>int32 a;<br/>
int32 result = (a &lt;&lt; 24) &gt;&gt; 24;
</td>
<td>Truncation of <code>int32</code> to <code>int8</code>, sign
extending the result.
</td>
</tr>
<tr>
<td>int-to-char</td>
<td>int32 a;<br/>
int32 result = a &amp; 0xffff;
</td>
<td>Truncation of <code>int32</code> to <code>uint16</code>, without
sign extension.
</td>
</tr>
<tr>
<td>int-to-short</td>
<td>int32 a;<br/>
int32 result = (a &lt;&lt; 16) &gt;&gt; 16;
</td>
<td>Truncation of <code>int32</code> to <code>int16</code>, sign
extending the result.
</td>
</tr>
<tr>
<td>add-int</td>
<td>int32 a, b;<br/>
int32 result = a + b;
</td>
<td>Twos-complement addition.</td>
</tr>
<tr>
<td>sub-int</td>
<td>int32 a, b;<br/>
int32 result = a - b;
</td>
<td>Twos-complement subtraction.</td>
</tr>
<tr>
<td>rsub-int</td>
<td>int32 a, b;<br/>
int32 result = b - a;
</td>
<td>Twos-complement reverse subtraction.</td>
</tr>
<tr>
<td>mul-int</td>
<td>int32 a, b;<br/>
int32 result = a * b;
</td>
<td>Twos-complement multiplication.</td>
</tr>
<tr>
<td>div-int</td>
<td>int32 a, b;<br/>
int32 result = a / b;
</td>
<td>Twos-complement division, rounded towards zero (that is, truncated to
integer). This throws <code>ArithmeticException</code> if
<code>b == 0</code>.
</td>
</tr>
<tr>
<td>rem-int</td>
<td>int32 a, b;<br/>
int32 result = a % b;
</td>
<td>Twos-complement remainder after division. The sign of the result
is the same as that of <code>a</code>, and it is more precisely
defined as <code>result == a - (a / b) * b</code>. This throws
<code>ArithmeticException</code> if <code>b == 0</code>.
</td>
</tr>
<tr>
<td>and-int</td>
<td>int32 a, b;<br/>
int32 result = a &amp; b;
</td>
<td>Bitwise AND.</td>
</tr>
<tr>
<td>or-int</td>
<td>int32 a, b;<br/>
int32 result = a | b;
</td>
<td>Bitwise OR.</td>
</tr>
<tr>
<td>xor-int</td>
<td>int32 a, b;<br/>
int32 result = a ^ b;
</td>
<td>Bitwise XOR.</td>
</tr>
<tr>
<td>shl-int</td>
<td>int32 a, b;<br/>
int32 result = a &lt;&lt; (b &amp; 0x1f);
</td>
<td>Bitwise shift left (with masked argument).</td>
</tr>
<tr>
<td>shr-int</td>
<td>int32 a, b;<br/>
int32 result = a &gt;&gt; (b &amp; 0x1f);
</td>
<td>Bitwise signed shift right (with masked argument).</td>
</tr>
<tr>
<td>ushr-int</td>
<td>uint32 a, b;<br/>
int32 result = a &gt;&gt; (b &amp; 0x1f);
</td>
<td>Bitwise unsigned shift right (with masked argument).</td>
</tr>
<tr>
<td>add-long</td>
<td>int64 a, b;<br/>
int64 result = a + b;
</td>
<td>Twos-complement addition.</td>
</tr>
<tr>
<td>sub-long</td>
<td>int64 a, b;<br/>
int64 result = a - b;
</td>
<td>Twos-complement subtraction.</td>
</tr>
<tr>
<td>mul-long</td>
<td>int64 a, b;<br/>
int64 result = a * b;
</td>
<td>Twos-complement multiplication.</td>
</tr>
<tr>
<td>div-long</td>
<td>int64 a, b;<br/>
int64 result = a / b;
</td>
<td>Twos-complement division, rounded towards zero (that is, truncated to
integer). This throws <code>ArithmeticException</code> if
<code>b == 0</code>.
</td>
</tr>
<tr>
<td>rem-long</td>
<td>int64 a, b;<br/>
int64 result = a % b;
</td>
<td>Twos-complement remainder after division. The sign of the result
is the same as that of <code>a</code>, and it is more precisely
defined as <code>result == a - (a / b) * b</code>. This throws
<code>ArithmeticException</code> if <code>b == 0</code>.
</td>
</tr>
<tr>
<td>and-long</td>
<td>int64 a, b;<br/>
int64 result = a &amp; b;
</td>
<td>Bitwise AND.</td>
</tr>
<tr>
<td>or-long</td>
<td>int64 a, b;<br/>
int64 result = a | b;
</td>
<td>Bitwise OR.</td>
</tr>
<tr>
<td>xor-long</td>
<td>int64 a, b;<br/>
int64 result = a ^ b;
</td>
<td>Bitwise XOR.</td>
</tr>
<tr>
<td>shl-long</td>
<td>int64 a, b;<br/>
int64 result = a &lt;&lt; (b &amp; 0x3f);
</td>
<td>Bitwise shift left (with masked argument).</td>
</tr>
<tr>
<td>shr-long</td>
<td>int64 a, b;<br/>
int64 result = a &gt;&gt; (b &amp; 0x3f);
</td>
<td>Bitwise signed shift right (with masked argument).</td>
</tr>
<tr>
<td>ushr-long</td>
<td>uint64 a, b;<br/>
int64 result = a &gt;&gt; (b &amp; 0x3f);
</td>
<td>Bitwise unsigned shift right (with masked argument).</td>
</tr>
<tr>
<td>add-float</td>
<td>float a, b;<br/>
float result = a + b;
</td>
<td>Floating point addition.</td>
</tr>
<tr>
<td>sub-float</td>
<td>float a, b;<br/>
float result = a - b;
</td>
<td>Floating point subtraction.</td>
</tr>
<tr>
<td>mul-float</td>
<td>float a, b;<br/>
float result = a * b;
</td>
<td>Floating point multiplication.</td>
</tr>
<tr>
<td>div-float</td>
<td>float a, b;<br/>
float result = a / b;
</td>
<td>Floating point division.</td>
</tr>
<tr>
<td>rem-float</td>
<td>float a, b;<br/>
float result = a % b;
</td>
<td>Floating point remainder after division. This function is different
than IEEE 754 remainder and is defined as
<code>result == a - roundTowardZero(a / b) * b</code>.
</td>
</tr>
<tr>
<td>add-double</td>
<td>double a, b;<br/>
double result = a + b;
</td>
<td>Floating point addition.</td>
</tr>
<tr>
<td>sub-double</td>
<td>double a, b;<br/>
double result = a - b;
</td>
<td>Floating point subtraction.</td>
</tr>
<tr>
<td>mul-double</td>
<td>double a, b;<br/>
double result = a * b;
</td>
<td>Floating point multiplication.</td>
</tr>
<tr>
<td>div-double</td>
<td>double a, b;<br/>
double result = a / b;
</td>
<td>Floating point division.</td>
</tr>
<tr>
<td>rem-double</td>
<td>double a, b;<br/>
double result = a % b;
</td>
<td>Floating point remainder after division. This function is different
than IEEE 754 remainder and is defined as
<code>result == a - roundTowardZero(a / b) * b</code>.
</td>
</tr>
</tbody>
</table>
</body>
</html>