src/devices/tech/dalvik/dalvik-bytecode.jd

page.title=Bytecode for the Dalvik VM
@jd:body

<!--
    Copyright 2013 The Android Open Source Project

    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.
-->
<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>When used for bit values (such as integers and floating point
    numbers), registers are considered 32 bits wide. Adjacent register
    pairs are used for 64-bit values. There is no alignment requirement
    for register pairs.
  </li>
  <li>When used for object references, registers are considered wide enough
    to hold exactly one such reference.
  </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 are limited to only addressing
  the first 16
  registers. When reasonably possible, instructions allow references to
  up to the first 256 registers. In addition, some instructions have variants
  that allow for much larger register counts, including a pair of catch-all
  <code>move</code> instructions that can address registers in the range
  <code>v0</code> &ndash; <code>v65535</code>.
  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 payloads, which are 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
  must 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 name suffix to indicate the type(s)
    they operate on:
    <ul>
    <li>Type-general 32-bit opcodes are unmarked.</li>
    <li>Type-general 64-bit opcodes are suffixed with <code>-wide</code>.</li>
    <li>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>.</li>
    </ul>
  </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>
  <li>In the descriptions here, the width of a value (indicating, e.g., the
    range of a constant or the number of registers possibly addressed) is
    emphasized by the use of a character per four bits of width.
  </li>
  <li>For example, in the instruction
    "<code>move-wide/from16 vAA, vBBBB</code>":
    <ul>
    <li>"<code>move</code>" is the base opcode, indicating the base operation
    (move a register's value).</li>
    <li>"<code>wide</code>" is the name suffix, indicating that it operates
    on wide (64 bit) data.</li>
    <li>"<code>from16</code>" is the opcode suffix, indicating a variant
    that has a 16-bit register reference as a source.</li>
    <li>"<code>vAA</code>" is the destination register (implied by the
    operation; again, the rule is that destination arguments always come
    first), which must be in the range <code>v0</code> &ndash;
    <code>v255</code>.</li>
    <li>"<code>vBBBB</code>" is the source register, which must be in the
    range <code>v0</code> &ndash; <code>v65535</code>.</li>
    </ul>
  </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>
<li>See the <a href="dex-format.html"><code>.dex</code> file format
  document</a> for more details about where the bytecode fits into
  the bigger picture.
</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.
    <p><b>Note:</b>
    Data-bearing pseudo-instructions are tagged with this opcode, in which
    case the high-order byte of the opcode unit indicates the nature of
    the data. See "<code>packed-switch-payload</code> Format",
    "<code>sparse-switch-payload</code> Format", and
    "<code>fill-array-data-payload</code> Format" below.</p>
  </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 must
    be the first instruction of any exception handler whose caught
    exception is not to be ignored, and this instruction must <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 {vC, vD, vE, vF, vG}, type@BBBB</td>
  <td>
    <code>A:</code> array size and argument word count (4 bits)<br/>
    <code>B:</code> type index (16 bits)<br/>
    <code>C..G:</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>, but reference
    types are acceptable). 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 an immediately 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-payload</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 must 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 must 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-payload</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-payload</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,
    setting <code>a</code> to <code>0</code> if <code>b == c</code>,
    <code>1</code> if <code>b &gt; c</code>,
    or <code>-1</code> if <code>b &lt; c</code>.
    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>x &lt; y</code> 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 of the 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 must 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 must 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> {vC, vD, vE, vF, vG}, meth@BBBB<br/>
    6e: invoke-virtual<br/>
    6f: invoke-super<br/>
    70: invoke-direct<br/>
    71: invoke-static<br/>
    72: invoke-interface
  </td>
  <td>
    <code>A:</code> argument word count (4 bits)<br/>
    <code>B:</code> method reference index (16 bits)<br/>
    <code>C..G:</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>private</code>, <code>static</code>,
    or <code>final</code>, and is also 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). The same method restrictions hold as for
    <code>invoke-virtual</code>.</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 reference 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>packed-switch-payload 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>sparse-switch-payload 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>fill-array-data-payload 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>