cranelift_codegen/isa/x64/encoding/rex.rs
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594
//! Encodes instructions in the standard x86 encoding mode. This is called
//! IA-32E mode in the Intel manuals but corresponds to the addition of the
//! REX-prefix format (hence the name of this module) that allowed encoding
//! instructions in both compatibility mode (32-bit instructions running on a
//! 64-bit OS) and in 64-bit mode (using the full 64-bit address space).
//!
//! For all of the routines that take both a memory-or-reg operand (sometimes
//! called "E" in the Intel documentation, see the Intel Developer's manual,
//! vol. 2, section A.2) and a reg-only operand ("G" in Intel-ese), the order is
//! always G first, then E. The term "enc" in the following means "hardware
//! register encoding number".
use super::ByteSink;
use crate::isa::x64::inst::args::{Amode, OperandSize};
use crate::isa::x64::inst::{regs, Inst, LabelUse};
use crate::machinst::{MachBuffer, Reg, RegClass};
pub(crate) fn low8_will_sign_extend_to_64(x: u32) -> bool {
let xs = (x as i32) as i64;
xs == ((xs << 56) >> 56)
}
pub(crate) fn low8_will_sign_extend_to_32(x: u32) -> bool {
let xs = x as i32;
xs == ((xs << 24) >> 24)
}
/// Encode the ModR/M byte.
#[inline(always)]
pub fn encode_modrm(m0d: u8, enc_reg_g: u8, rm_e: u8) -> u8 {
debug_assert!(m0d < 4);
debug_assert!(enc_reg_g < 8);
debug_assert!(rm_e < 8);
((m0d & 3) << 6) | ((enc_reg_g & 7) << 3) | (rm_e & 7)
}
#[inline(always)]
pub(crate) fn encode_sib(shift: u8, enc_index: u8, enc_base: u8) -> u8 {
debug_assert!(shift < 4);
debug_assert!(enc_index < 8);
debug_assert!(enc_base < 8);
((shift & 3) << 6) | ((enc_index & 7) << 3) | (enc_base & 7)
}
/// Get the encoding number of a GPR.
#[inline(always)]
pub(crate) fn int_reg_enc(reg: impl Into<Reg>) -> u8 {
let reg = reg.into();
debug_assert!(reg.is_real(), "reg = {reg:?}");
debug_assert_eq!(reg.class(), RegClass::Int);
reg.to_real_reg().unwrap().hw_enc()
}
/// Get the encoding number of any register.
#[inline(always)]
pub(crate) fn reg_enc(reg: impl Into<Reg>) -> u8 {
let reg = reg.into();
debug_assert!(reg.is_real());
reg.to_real_reg().unwrap().hw_enc()
}
/// A small bit field to record a REX prefix specification:
/// - bit 0 set to 1 indicates REX.W must be 0 (cleared).
/// - bit 1 set to 1 indicates the REX prefix must always be emitted.
#[repr(transparent)]
#[derive(Clone, Copy)]
pub struct RexFlags(u8);
impl RexFlags {
/// By default, set the W field, and don't always emit.
#[inline(always)]
pub fn set_w() -> Self {
Self(0)
}
/// Creates a new RexPrefix for which the REX.W bit will be cleared.
#[inline(always)]
pub fn clear_w() -> Self {
Self(1)
}
/// True if 64-bit operands are used.
#[inline(always)]
pub fn must_clear_w(&self) -> bool {
(self.0 & 1) != 0
}
/// Require that the REX prefix is emitted.
#[inline(always)]
pub fn always_emit(&mut self) -> &mut Self {
self.0 = self.0 | 2;
self
}
/// True if the REX prefix must always be emitted.
#[inline(always)]
pub fn must_always_emit(&self) -> bool {
(self.0 & 2) != 0
}
/// Emit the rex prefix if the referenced register would require it for 8-bit operations.
#[inline(always)]
pub fn always_emit_if_8bit_needed(&mut self, reg: Reg) -> &mut Self {
let enc_reg = int_reg_enc(reg);
if enc_reg >= 4 && enc_reg <= 7 {
self.always_emit();
}
self
}
/// Emit a unary instruction.
#[inline(always)]
pub fn emit_one_op<BS: ByteSink + ?Sized>(&self, sink: &mut BS, enc_e: u8) {
// Register Operand coded in Opcode Byte
// REX.R and REX.X unused
// REX.B == 1 accesses r8-r15
let w = if self.must_clear_w() { 0 } else { 1 };
let r = 0;
let x = 0;
let b = (enc_e >> 3) & 1;
let rex = 0x40 | (w << 3) | (r << 2) | (x << 1) | b;
if rex != 0x40 || self.must_always_emit() {
sink.put1(rex);
}
}
/// Emit a binary instruction.
#[inline(always)]
pub fn emit_two_op<BS: ByteSink + ?Sized>(&self, sink: &mut BS, enc_g: u8, enc_e: u8) {
let w = if self.must_clear_w() { 0 } else { 1 };
let r = (enc_g >> 3) & 1;
let x = 0;
let b = (enc_e >> 3) & 1;
let rex = 0x40 | (w << 3) | (r << 2) | (x << 1) | b;
if rex != 0x40 || self.must_always_emit() {
sink.put1(rex);
}
}
/// Emit a ternary instruction.
#[inline(always)]
pub fn emit_three_op<BS: ByteSink + ?Sized>(
&self,
sink: &mut BS,
enc_g: u8,
enc_index: u8,
enc_base: u8,
) {
let w = if self.must_clear_w() { 0 } else { 1 };
let r = (enc_g >> 3) & 1;
let x = (enc_index >> 3) & 1;
let b = (enc_base >> 3) & 1;
let rex = 0x40 | (w << 3) | (r << 2) | (x << 1) | b;
if rex != 0x40 || self.must_always_emit() {
sink.put1(rex);
}
}
}
/// Generate the proper Rex flags for the given operand size.
impl From<OperandSize> for RexFlags {
fn from(size: OperandSize) -> Self {
match size {
OperandSize::Size64 => RexFlags::set_w(),
_ => RexFlags::clear_w(),
}
}
}
/// Generate Rex flags for an OperandSize/register tuple.
impl From<(OperandSize, Reg)> for RexFlags {
fn from((size, reg): (OperandSize, Reg)) -> Self {
let mut rex = RexFlags::from(size);
if size == OperandSize::Size8 {
rex.always_emit_if_8bit_needed(reg);
}
rex
}
}
/// Allows using the same opcode byte in different "opcode maps" to allow for more instruction
/// encodings. See appendix A in the Intel Software Developer's Manual, volume 2A, for more details.
#[allow(missing_docs)]
#[derive(PartialEq)]
pub enum OpcodeMap {
None,
_0F,
_0F38,
_0F3A,
}
impl OpcodeMap {
/// Normally the opcode map is specified as bytes in the instruction, but some x64 encoding
/// formats pack this information as bits in a prefix (e.g. VEX / EVEX).
pub(crate) fn bits(&self) -> u8 {
match self {
OpcodeMap::None => 0b00,
OpcodeMap::_0F => 0b01,
OpcodeMap::_0F38 => 0b10,
OpcodeMap::_0F3A => 0b11,
}
}
}
impl Default for OpcodeMap {
fn default() -> Self {
Self::None
}
}
/// We may need to include one or more legacy prefix bytes before the REX prefix. This enum
/// covers only the small set of possibilities that we actually need.
#[derive(PartialEq)]
pub enum LegacyPrefixes {
/// No prefix bytes.
None,
/// Operand Size Override -- here, denoting "16-bit operation".
_66,
/// The Lock prefix.
_F0,
/// Operand size override and Lock.
_66F0,
/// REPNE, but no specific meaning here -- is just an opcode extension.
_F2,
/// REP/REPE, but no specific meaning here -- is just an opcode extension.
_F3,
/// Operand size override and same effect as F3.
_66F3,
}
impl LegacyPrefixes {
/// Emit the legacy prefix as bytes (e.g. in REX instructions).
#[inline(always)]
pub(crate) fn emit<BS: ByteSink + ?Sized>(&self, sink: &mut BS) {
match self {
Self::_66 => sink.put1(0x66),
Self::_F0 => sink.put1(0xF0),
Self::_66F0 => {
// I don't think the order matters, but in any case, this is the same order that
// the GNU assembler uses.
sink.put1(0x66);
sink.put1(0xF0);
}
Self::_F2 => sink.put1(0xF2),
Self::_F3 => sink.put1(0xF3),
Self::_66F3 => {
sink.put1(0x66);
sink.put1(0xF3);
}
Self::None => (),
}
}
/// Emit the legacy prefix as bits (e.g. for EVEX instructions).
#[inline(always)]
pub(crate) fn bits(&self) -> u8 {
match self {
Self::None => 0b00,
Self::_66 => 0b01,
Self::_F3 => 0b10,
Self::_F2 => 0b11,
_ => panic!(
"VEX and EVEX bits can only be extracted from single prefixes: None, 66, F3, F2"
),
}
}
}
impl Default for LegacyPrefixes {
fn default() -> Self {
Self::None
}
}
/// This is the core 'emit' function for instructions that reference memory.
///
/// For an instruction that has as operands a reg encoding `enc_g` and a memory address `mem_e`,
/// create and emit:
/// - first the legacy prefixes, if any
/// - then the REX prefix, if needed
/// - then caller-supplied opcode byte(s) (`opcodes` and `num_opcodes`),
/// - then the MOD/RM byte,
/// - then optionally, a SIB byte,
/// - and finally optionally an immediate that will be derived from the `mem_e` operand.
///
/// For most instructions up to and including SSE4.2, that will be the whole instruction: this is
/// what we call "standard" instructions, and abbreviate "std" in the name here. VEX-prefixed
/// instructions will require their own emitter functions.
///
/// This will also work for 32-bits x86 instructions, assuming no REX prefix is provided.
///
/// The opcodes are written bigendianly for the convenience of callers. For example, if the opcode
/// bytes to be emitted are, in this order, F3 0F 27, then the caller should pass `opcodes` ==
/// 0xF3_0F_27 and `num_opcodes` == 3.
///
/// The register operand is represented here not as a `Reg` but as its hardware encoding, `enc_g`.
/// `rex` can specify special handling for the REX prefix. By default, the REX prefix will
/// indicate a 64-bit operation and will be deleted if it is redundant (0x40). Note that for a
/// 64-bit operation, the REX prefix will normally never be redundant, since REX.W must be 1 to
/// indicate a 64-bit operation.
pub(crate) fn emit_std_enc_mem(
sink: &mut MachBuffer<Inst>,
prefixes: LegacyPrefixes,
opcodes: u32,
mut num_opcodes: usize,
enc_g: u8,
mem_e: &Amode,
rex: RexFlags,
bytes_at_end: u8,
) {
// General comment for this function: the registers in `mem_e` must be
// 64-bit integer registers, because they are part of an address
// expression. But `enc_g` can be derived from a register of any class.
if let Some(trap_code) = mem_e.get_flags().trap_code() {
sink.add_trap(trap_code);
}
prefixes.emit(sink);
// After prefixes, first emit the REX byte depending on the kind of
// addressing mode that's being used.
match *mem_e {
Amode::ImmReg { base, .. } => {
let enc_e = int_reg_enc(base);
rex.emit_two_op(sink, enc_g, enc_e);
}
Amode::ImmRegRegShift {
base: reg_base,
index: reg_index,
..
} => {
let enc_base = int_reg_enc(*reg_base);
let enc_index = int_reg_enc(*reg_index);
rex.emit_three_op(sink, enc_g, enc_index, enc_base);
}
Amode::RipRelative { .. } => {
// note REX.B = 0.
rex.emit_two_op(sink, enc_g, 0);
}
}
// Now the opcode(s). These include any other prefixes the caller
// hands to us.
while num_opcodes > 0 {
num_opcodes -= 1;
sink.put1(((opcodes >> (num_opcodes << 3)) & 0xFF) as u8);
}
// And finally encode the mod/rm bytes and all further information.
emit_modrm_sib_disp(sink, enc_g, mem_e, bytes_at_end, None)
}
pub(crate) fn emit_modrm_sib_disp(
sink: &mut MachBuffer<Inst>,
enc_g: u8,
mem_e: &Amode,
bytes_at_end: u8,
evex_scaling: Option<i8>,
) {
match *mem_e {
Amode::ImmReg { simm32, base, .. } => {
let enc_e = int_reg_enc(base);
let mut imm = Imm::new(simm32, evex_scaling);
// Most base registers allow for a single ModRM byte plus an
// optional immediate. If rsp is the base register, however, then a
// SIB byte must be used.
let enc_e_low3 = enc_e & 7;
if enc_e_low3 != regs::ENC_RSP {
// If the base register is rbp and there's no offset then force
// a 1-byte zero offset since otherwise the encoding would be
// invalid.
if enc_e_low3 == regs::ENC_RBP {
imm.force_immediate();
}
sink.put1(encode_modrm(imm.m0d(), enc_g & 7, enc_e & 7));
imm.emit(sink);
} else {
// Displacement from RSP is encoded with a SIB byte where
// the index and base are both encoded as RSP's encoding of
// 0b100. This special encoding means that the index register
// isn't used and the base is 0b100 with or without a
// REX-encoded 4th bit (e.g. rsp or r12)
sink.put1(encode_modrm(imm.m0d(), enc_g & 7, 0b100));
sink.put1(0b00_100_100);
imm.emit(sink);
}
}
Amode::ImmRegRegShift {
simm32,
base: reg_base,
index: reg_index,
shift,
..
} => {
let enc_base = int_reg_enc(*reg_base);
let enc_index = int_reg_enc(*reg_index);
// Encoding of ModRM/SIB bytes don't allow the index register to
// ever be rsp. Note, though, that the encoding of r12, whose three
// lower bits match the encoding of rsp, is explicitly allowed with
// REX bytes so only rsp is disallowed.
assert!(enc_index != regs::ENC_RSP);
// If the offset is zero then there is no immediate. Note, though,
// that if the base register's lower three bits are `101` then an
// offset must be present. This is a special case in the encoding of
// the SIB byte and requires an explicit displacement with rbp/r13.
let mut imm = Imm::new(simm32, evex_scaling);
if enc_base & 7 == regs::ENC_RBP {
imm.force_immediate();
}
// With the above determined encode the ModRM byte, then the SIB
// byte, then any immediate as necessary.
sink.put1(encode_modrm(imm.m0d(), enc_g & 7, 0b100));
sink.put1(encode_sib(shift, enc_index & 7, enc_base & 7));
imm.emit(sink);
}
Amode::RipRelative { ref target } => {
// RIP-relative is mod=00, rm=101.
sink.put1(encode_modrm(0b00, enc_g & 7, 0b101));
let offset = sink.cur_offset();
sink.use_label_at_offset(offset, *target, LabelUse::JmpRel32);
// N.B.: some instructions (XmmRmRImm format for example)
// have bytes *after* the RIP-relative offset. The
// addressed location is relative to the end of the
// instruction, but the relocation is nominally relative
// to the end of the u32 field. So, to compensate for
// this, we emit a negative extra offset in the u32 field
// initially, and the relocation will add to it.
sink.put4(-(i32::from(bytes_at_end)) as u32);
}
}
}
#[derive(Copy, Clone)]
enum Imm {
None,
Imm8(i8),
Imm32(i32),
}
impl Imm {
/// Classifies the 32-bit immediate `val` as how this can be encoded
/// with ModRM/SIB bytes.
///
/// For `evex_scaling` according to Section 2.7.5 of Intel's manual:
///
/// > EVEX-encoded instructions always use a compressed displacement scheme
/// > by multiplying disp8 in conjunction with a scaling factor N that is
/// > determined based on the vector length, the value of EVEX.b bit
/// > (embedded broadcast) and the input element size of the instruction
///
/// The `evex_scaling` factor provided here is `Some(N)` for EVEX
/// instructions. This is taken into account where the `Imm` value
/// contained is the raw byte offset.
fn new(val: i32, evex_scaling: Option<i8>) -> Imm {
if val == 0 {
return Imm::None;
}
match evex_scaling {
Some(scaling) => {
if val % i32::from(scaling) == 0 {
let scaled = val / i32::from(scaling);
if low8_will_sign_extend_to_32(scaled as u32) {
return Imm::Imm8(scaled as i8);
}
}
Imm::Imm32(val)
}
None => match i8::try_from(val) {
Ok(val) => Imm::Imm8(val),
Err(_) => Imm::Imm32(val),
},
}
}
/// Forces `Imm::None` to become `Imm::Imm8(0)`, used for special cases
/// where some base registers require an immediate.
fn force_immediate(&mut self) {
if let Imm::None = self {
*self = Imm::Imm8(0);
}
}
/// Returns the two "mod" bits present at the upper bits of the mod/rm
/// byte.
fn m0d(&self) -> u8 {
match self {
Imm::None => 0b00,
Imm::Imm8(_) => 0b01,
Imm::Imm32(_) => 0b10,
}
}
fn emit<BS: ByteSink + ?Sized>(&self, sink: &mut BS) {
match self {
Imm::None => {}
Imm::Imm8(n) => sink.put1(*n as u8),
Imm::Imm32(n) => sink.put4(*n as u32),
}
}
}
/// This is the core 'emit' function for instructions that do not reference memory.
///
/// This is conceptually the same as emit_modrm_sib_enc_ge, except it is for the case where the E
/// operand is a register rather than memory. Hence it is much simpler.
pub(crate) fn emit_std_enc_enc<BS: ByteSink + ?Sized>(
sink: &mut BS,
prefixes: LegacyPrefixes,
opcodes: u32,
mut num_opcodes: usize,
enc_g: u8,
enc_e: u8,
rex: RexFlags,
) {
// EncG and EncE can be derived from registers of any class, and they
// don't even have to be from the same class. For example, for an
// integer-to-FP conversion insn, one might be RegClass::I64 and the other
// RegClass::V128.
// The legacy prefixes.
prefixes.emit(sink);
// The rex byte.
rex.emit_two_op(sink, enc_g, enc_e);
// All other prefixes and opcodes.
while num_opcodes > 0 {
num_opcodes -= 1;
sink.put1(((opcodes >> (num_opcodes << 3)) & 0xFF) as u8);
}
// Now the mod/rm byte. The instruction we're generating doesn't access
// memory, so there is no SIB byte or immediate -- we're done.
sink.put1(encode_modrm(0b11, enc_g & 7, enc_e & 7));
}
// These are merely wrappers for the above two functions that facilitate passing
// actual `Reg`s rather than their encodings.
pub(crate) fn emit_std_reg_mem(
sink: &mut MachBuffer<Inst>,
prefixes: LegacyPrefixes,
opcodes: u32,
num_opcodes: usize,
reg_g: Reg,
mem_e: &Amode,
rex: RexFlags,
bytes_at_end: u8,
) {
let enc_g = reg_enc(reg_g);
emit_std_enc_mem(
sink,
prefixes,
opcodes,
num_opcodes,
enc_g,
mem_e,
rex,
bytes_at_end,
);
}
pub(crate) fn emit_std_reg_reg<BS: ByteSink + ?Sized>(
sink: &mut BS,
prefixes: LegacyPrefixes,
opcodes: u32,
num_opcodes: usize,
reg_g: Reg,
reg_e: Reg,
rex: RexFlags,
) {
let enc_g = reg_enc(reg_g);
let enc_e = reg_enc(reg_e);
emit_std_enc_enc(sink, prefixes, opcodes, num_opcodes, enc_g, enc_e, rex);
}
/// Write a suitable number of bits from an imm64 to the sink.
pub(crate) fn emit_simm<BS: ByteSink + ?Sized>(sink: &mut BS, size: u8, simm32: u32) {
match size {
8 | 4 => sink.put4(simm32),
2 => sink.put2(simm32 as u16),
1 => sink.put1(simm32 as u8),
_ => unreachable!(),
}
}