1page.title=Dalvik VM Instruction Formats 2@jd:body 3 4<!-- 5 Copyright 2013 The Android Open Source Project 6 7 Licensed under the Apache License, Version 2.0 (the "License"); 8 you may not use this file except in compliance with the License. 9 You may obtain a copy of the License at 10 11 http://www.apache.org/licenses/LICENSE-2.0 12 13 Unless required by applicable law or agreed to in writing, software 14 distributed under the License is distributed on an "AS IS" BASIS, 15 WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 16 See the License for the specific language governing permissions and 17 limitations under the License. 18--> 19<h2>Introduction and Overview</h2> 20 21<p>This document lists the instruction formats used by Dalvik bytecode 22and is meant to be used in conjunction with the 23<a href="dalvik-bytecode.html">bytecode reference document</a>.</p> 24 25<h3>Bitwise descriptions</h3> 26 27<p>The first column in the format table lists the bitwise layout of 28the format. It consists of one or more space-separated "words" each of 29which describes a 16-bit code unit. Each character in a word 30represents four bits, read from high bits to low, with vertical bars 31("<code>|</code>") interspersed to aid in reading. Uppercase letters 32in sequence from "<code>A</code>" are used to indicate fields within 33the format (which then get defined further by the syntax column). The term 34"<code>op</code>" is used to indicate the position of an eight-bit 35opcode within the format. A slashed zero 36("<code>Ø</code>") is used to indicate that all bits must be 37zero in the indicated position.</p> 38 39<p>For the most part, lettering proceeds from earlier code units to 40later code units, and low-order to high-order within a code unit. 41However, there are a few exceptions to this general rule, which are 42done in order to make the naming of similar-meaning parts be the same 43across different instruction formats. These cases are noted explicitly 44in the format descriptions.</p> 45 46<p>For example, the format "<code>B|A|<i>op</i> CCCC</code>" indicates 47that the format consists of two 16-bit code units. The first word 48consists of the opcode in the low eight bits and a pair of four-bit 49values in the high eight bits; and the second word consists of a single 5016-bit value.</p> 51 52<h3>Format IDs</h3> 53 54<p>The second column in the format table indicates the short identifier 55for the format, which is used in other documents and in code to identify 56the format.</p> 57 58<p>Most format IDs consist of three characters, two digits followed by a 59letter. The first digit indicates the number of 16-bit code units in the 60format. The second digit indicates the maximum number of registers that the 61format contains (maximum, since some formats can accomodate a variable 62number of registers), with the special designation "<code>r</code>" indicating 63that a range of registers is encoded. The final letter semi-mnemonically 64indicates the type of any extra data encoded by the format. For example, 65format "<code>21t</code>" is of length two, contains one register reference, 66and additionally contains a branch target.</p> 67 68<p>Suggested static linking formats have an additional 69"<code>s</code>" suffix, making them four characters total. Similarly, 70suggested "inline" linking formats have an additional "<code>i</code>" 71suffix. (In this context, inline linking is like static linking, 72except with more direct ties into a virtual machine's implementation.) 73Finally, a couple oddball suggested formats (e.g., 74"<code>20bc</code>") include two pieces of data which are both 75represented in its format ID.</p> 76 77<p>The full list of typecode letters are as follows. Note that some 78forms have different sizes, depending on the format:</p> 79 80<table class="letters"> 81<thead> 82<tr> 83 <th>Mnemonic</th> 84 <th>Bit Sizes</th> 85 <th>Meaning</th> 86</tr> 87</thead> 88<tbody> 89<tr> 90 <td>b</td> 91 <td>8</td> 92 <td>immediate signed <b>b</b>yte</td> 93</tr> 94<tr> 95 <td>c</td> 96 <td>16, 32</td> 97 <td><b>c</b>onstant pool index</td> 98</tr> 99<tr> 100 <td>f</td> 101 <td>16</td> 102 <td>inter<b>f</b>ace constants (only used in statically linked formats) 103 </td> 104</tr> 105<tr> 106 <td>h</td> 107 <td>16</td> 108 <td>immediate signed <b>h</b>at (high-order bits of a 32- or 64-bit 109 value; low-order bits are all <code>0</code>) 110 </td> 111</tr> 112<tr> 113 <td>i</td> 114 <td>32</td> 115 <td>immediate signed <b>i</b>nt, or 32-bit float</td> 116</tr> 117<tr> 118 <td>l</td> 119 <td>64</td> 120 <td>immediate signed <b>l</b>ong, or 64-bit double</td> 121</tr> 122<tr> 123 <td>m</td> 124 <td>16</td> 125 <td><b>m</b>ethod constants (only used in statically linked formats)</td> 126</tr> 127<tr> 128 <td>n</td> 129 <td>4</td> 130 <td>immediate signed <b>n</b>ibble</td> 131</tr> 132<tr> 133 <td>s</td> 134 <td>16</td> 135 <td>immediate signed <b>s</b>hort</td> 136</tr> 137<tr> 138 <td>t</td> 139 <td>8, 16, 32</td> 140 <td>branch <b>t</b>arget</td> 141</tr> 142<tr> 143 <td>x</td> 144 <td>0</td> 145 <td>no additional data</td> 146</tr> 147</tbody> 148</table> 149 150<h3>Syntax</h3> 151 152<p>The third column of the format table indicates the human-oriented 153syntax for instructions which use the indicated format. Each instruction 154starts with the named opcode and is optionally followed by one or 155more arguments, themselves separated with commas.</p> 156 157<p>Wherever an argument refers to a field from the first column, the 158letter for that field is indicated in the syntax, repeated once for 159each four bits of the field. For example, an eight-bit field labeled 160"<code>BB</code>" in the first column would also be labeled 161"<code>BB</code>" in the syntax column.</p> 162 163<p>Arguments which name a register have the form "<code>v<i>X</i></code>". 164The prefix "<code>v</code>" was chosen instead of the more common 165"<code>r</code>" exactly to avoid conflicting with (non-virtual) architectures 166on which a Dalvik virtual machine might be implemented which themselves 167use the prefix "<code>r</code>" for their registers. (That is, this 168decision makes it possible to talk about both virtual and real registers 169together without the need for circumlocution.)</p> 170 171<p>Arguments which indicate a literal value have the form 172"<code>#+<i>X</i></code>". Some formats indicate literals that only 173have non-zero bits in their high-order bits; for these, the zeroes 174are represented explicitly in the syntax, even though they do not 175appear in the bitwise representation.</p> 176 177<p>Arguments which indicate a relative instruction address offset have the 178form "<code>+<i>X</i></code>".</p> 179 180<p>Arguments which indicate a literal constant pool index have the form 181"<code><i>kind</i>@<i>X</i></code>", where "<code><i>kind</i></code>" 182indicates which constant pool is being referred to. Each opcode that 183uses such a format explicitly allows only one kind of constant; see 184the opcode reference to figure out the correspondence. The four 185kinds of constant pool are "<code>string</code>" (string pool index), 186"<code>type</code>" (type pool index), "<code>field</code>" (field 187pool index), and "<code>meth</code>" (method pool index).</p> 188 189<p>Similar to the representation of constant pool indices, there are 190also suggested (optional) forms that indicate prelinked offsets or 191indices. There are two types of suggested prelinked value: vtable offsets 192(indicated as "<code>vtaboff</code>") and field offsets (indicated as 193"<code>fieldoff</code>").</p> 194 195<p>In the cases where a format value isn't explictly part of the syntax 196but instead picks a variant, each variant is listed with the prefix 197"<code>[<i>X</i>=<i>N</i>]</code>" (e.g., "<code>[A=2]</code>") to indicate 198the correspondence.</p> 199 200<h2>The Formats</h2> 201 202<table class="format"> 203<thead> 204<tr> 205 <th>Format</th> 206 <th>ID</th> 207 <th>Syntax</th> 208 <th>Notable Opcodes Covered</th> 209</tr> 210</thead> 211<tbody> 212<tr> 213 <td><i>N/A</i></td> 214 <td>00x</td> 215 <td><i><code>N/A</code></i></td> 216 <td><i>pseudo-format used for unused opcodes; suggested for use as the 217 nominal format for a breakpoint opcode</i></td> 218</tr> 219<tr> 220 <td>ØØ|<i>op</i></td> 221 <td>10x</td> 222 <td><i><code>op</code></i></td> 223 <td> </td> 224</tr> 225<tr> 226 <td rowspan="2">B|A|<i>op</i></td> 227 <td>12x</td> 228 <td><i><code>op</code></i> vA, vB</td> 229 <td> </td> 230</tr> 231<tr> 232 <td>11n</td> 233 <td><i><code>op</code></i> vA, #+B</td> 234 <td> </td> 235</tr> 236<tr> 237 <td rowspan="2">AA|<i>op</i></td> 238 <td>11x</td> 239 <td><i><code>op</code></i> vAA</td> 240 <td> </td> 241</tr> 242<tr> 243 <td>10t</td> 244 <td><i><code>op</code></i> +AA</td> 245 <td>goto</td> 246</tr> 247<tr> 248 <td>ØØ|<i>op</i> AAAA</td></td> 249 <td>20t</td> 250 <td><i><code>op</code></i> +AAAA</td> 251 <td>goto/16</td> 252</tr> 253<tr> 254 <td>AA|<i>op</i> BBBB</td></td> 255 <td>20bc</td> 256 <td><i><code>op</code></i> AA, kind@BBBB</td> 257 <td><i>suggested format for statically determined verification errors; 258 A is the type of error and B is an index into a type-appropriate 259 table (e.g. method references for a no-such-method error)</i></td> 260</tr> 261<tr> 262 <td rowspan="5">AA|<i>op</i> BBBB</td> 263 <td>22x</td> 264 <td><i><code>op</code></i> vAA, vBBBB</td> 265 <td> </td> 266</tr> 267<tr> 268 <td>21t</td> 269 <td><i><code>op</code></i> vAA, +BBBB</td> 270 <td> </td> 271</tr> 272<tr> 273 <td>21s</td> 274 <td><i><code>op</code></i> vAA, #+BBBB</td> 275 <td> </td> 276</tr> 277<tr> 278 <td>21h</td> 279 <td><i><code>op</code></i> vAA, #+BBBB0000<br/> 280 <i><code>op</code></i> vAA, #+BBBB000000000000 281 </td> 282 <td> </td> 283</tr> 284<tr> 285 <td>21c</td> 286 <td><i><code>op</code></i> vAA, type@BBBB<br/> 287 <i><code>op</code></i> vAA, field@BBBB<br/> 288 <i><code>op</code></i> vAA, string@BBBB 289 </td> 290 <td>check-cast<br/> 291 const-class<br/> 292 const-string 293 </td> 294</tr> 295<tr> 296 <td rowspan="2">AA|<i>op</i> CC|BB</td> 297 <td>23x</td> 298 <td><i><code>op</code></i> vAA, vBB, vCC</td> 299 <td> </td> 300</tr> 301<tr> 302 <td>22b</td> 303 <td><i><code>op</code></i> vAA, vBB, #+CC</td> 304 <td> </td> 305</tr> 306<tr> 307 <td rowspan="4">B|A|<i>op</i> CCCC</td> 308 <td>22t</td> 309 <td><i><code>op</code></i> vA, vB, +CCCC</td> 310 <td> </td> 311</tr> 312<tr> 313 <td>22s</td> 314 <td><i><code>op</code></i> vA, vB, #+CCCC</td> 315 <td> </td> 316</tr> 317<tr> 318 <td>22c</td> 319 <td><i><code>op</code></i> vA, vB, type@CCCC<br/> 320 <i><code>op</code></i> vA, vB, field@CCCC 321 </td> 322 <td>instance-of</td> 323</tr> 324<tr> 325 <td>22cs</td> 326 <td><i><code>op</code></i> vA, vB, fieldoff@CCCC</td> 327 <td><i>suggested format for statically linked field access instructions of 328 format 22c</i> 329 </td> 330</tr> 331<tr> 332 <td>ØØ|<i>op</i> AAAA<sub>lo</sub> AAAA<sub>hi</sub></td></td> 333 <td>30t</td> 334 <td><i><code>op</code></i> +AAAAAAAA</td> 335 <td>goto/32</td> 336</tr> 337<tr> 338 <td>ØØ|<i>op</i> AAAA BBBB</td> 339 <td>32x</td> 340 <td><i><code>op</code></i> vAAAA, vBBBB</td> 341 <td> </td> 342</tr> 343<tr> 344 <td rowspan="3">AA|<i>op</i> BBBB<sub>lo</sub> BBBB<sub>hi</sub></td> 345 <td>31i</td> 346 <td><i><code>op</code></i> vAA, #+BBBBBBBB</td> 347 <td> </td> 348</tr> 349<tr> 350 <td>31t</td> 351 <td><i><code>op</code></i> vAA, +BBBBBBBB</td> 352 <td> </td> 353</tr> 354<tr> 355 <td>31c</td> 356 <td><i><code>op</code></i> vAA, string@BBBBBBBB</td> 357 <td>const-string/jumbo</td> 358</tr> 359<tr> 360 <td rowspan="3">A|G|<i>op</i> BBBB F|E|D|C</td> 361 <td>35c</td> 362 <td><i>[<code>A=5</code>] <code>op</code></i> {vC, vD, vE, vF, vG}, 363 meth@BBBB<br/> 364 <i>[<code>A=5</code>] <code>op</code></i> {vC, vD, vE, vF, vG}, 365 type@BBBB<br/> 366 <i>[<code>A=4</code>] <code>op</code></i> {vC, vD, vE, vF}, 367 <i><code>kind</code></i>@BBBB<br/> 368 <i>[<code>A=3</code>] <code>op</code></i> {vC, vD, vE}, 369 <i><code>kind</code></i>@BBBB<br/> 370 <i>[<code>A=2</code>] <code>op</code></i> {vC, vD}, 371 <i><code>kind</code></i>@BBBB<br/> 372 <i>[<code>A=1</code>] <code>op</code></i> {vC}, 373 <i><code>kind</code></i>@BBBB<br/> 374 <i>[<code>A=0</code>] <code>op</code></i> {}, 375 <i><code>kind</code></i>@BBBB<br/> 376 <p><i>The unusual choice in lettering here reflects a desire to make 377 the count and the reference index have the same label as in format 378 3rc.</i></p> 379 </td> 380 <td> </td> 381</tr> 382<tr> 383 <td>35ms</td> 384 <td><i>[<code>A=5</code>] <code>op</code></i> {vC, vD, vE, vF, vG}, 385 vtaboff@BBBB<br/> 386 <i>[<code>A=4</code>] <code>op</code></i> {vC, vD, vE, vF}, 387 vtaboff@BBBB<br/> 388 <i>[<code>A=3</code>] <code>op</code></i> {vC, vD, vE}, 389 vtaboff@BBBB<br/> 390 <i>[<code>A=2</code>] <code>op</code></i> {vC, vD}, 391 vtaboff@BBBB<br/> 392 <i>[<code>A=1</code>] <code>op</code></i> {vC}, 393 vtaboff@BBBB<br/> 394 <p><i>The unusual choice in lettering here reflects a desire to make 395 the count and the reference index have the same label as in format 396 3rms.</i></p> 397 </td> 398 <td><i>suggested format for statically linked <code>invoke-virtual</code> 399 and <code>invoke-super</code> instructions of format 35c</i> 400 </td> 401</tr> 402<tr> 403 <td>35mi</td> 404 <td><i>[<code>A=5</code>] <code>op</code></i> {vC, vD, vE, vF, vG}, 405 inline@BBBB<br/> 406 <i>[<code>A=4</code>] <code>op</code></i> {vC, vD, vE, vF}, 407 inline@BBBB<br/> 408 <i>[<code>A=3</code>] <code>op</code></i> {vC, vD, vE}, 409 inline@BBBB<br/> 410 <i>[<code>A=2</code>] <code>op</code></i> {vC, vD}, 411 inline@BBBB<br/> 412 <i>[<code>A=1</code>] <code>op</code></i> {vC}, 413 inline@BBBB<br/> 414 <p><i>The unusual choice in lettering here reflects a desire to make 415 the count and the reference index have the same label as in format 416 3rmi.</i></p> 417 </td> 418 <td><i>suggested format for inline linked <code>invoke-static</code> 419 and <code>invoke-virtual</code> instructions of format 35c</i> 420 </td> 421</tr> 422<tr> 423 <td rowspan="3">AA|<i>op</i> BBBB CCCC</td> 424 <td>3rc</td> 425 <td><i><code>op</code></i> {vCCCC .. vNNNN}, meth@BBBB<br/> 426 <i><code>op</code></i> {vCCCC .. vNNNN}, type@BBBB<br/> 427 <p><i>where <code>NNNN = CCCC+AA-1</code>, that is <code>A</code> 428 determines the count <code>0..255</code>, and <code>C</code> 429 determines the first register</i></p> 430 </td> 431 <td> </td> 432</tr> 433<tr> 434 <td>3rms</td> 435 <td><i><code>op</code></i> {vCCCC .. vNNNN}, vtaboff@BBBB<br/> 436 <p><i>where <code>NNNN = CCCC+AA-1</code>, that is <code>A</code> 437 determines the count <code>0..255</code>, and <code>C</code> 438 determines the first register</i></p> 439 </td> 440 <td><i>suggested format for statically linked <code>invoke-virtual</code> 441 and <code>invoke-super</code> instructions of format <code>3rc</code></i> 442 </td> 443</tr> 444<tr> 445 <td>3rmi</td> 446 <td><i><code>op</code></i> {vCCCC .. vNNNN}, inline@BBBB<br/> 447 <p><i>where <code>NNNN = CCCC+AA-1</code>, that is <code>A</code> 448 determines the count <code>0..255</code>, and <code>C</code> 449 determines the first register</i></p> 450 </td> 451 <td><i>suggested format for inline linked <code>invoke-static</code> 452 and <code>invoke-virtual</code> instructions of format 3rc</i> 453 </td> 454</tr> 455<tr> 456 <td>AA|<i>op</i> BBBB<sub>lo</sub> BBBB BBBB BBBB<sub>hi</sub></td> 457 <td>51l</td> 458 <td><i><code>op</code></i> vAA, #+BBBBBBBBBBBBBBBB</td> 459 <td>const-wide</td> 460</tr> 461</tbody> 462</table> 463