pulley_interpreter/
lib.rs

1//! The pulley bytecode for fast interpreters.
2
3#![cfg_attr(docsrs, feature(doc_auto_cfg))]
4#![cfg_attr(pulley_tail_calls, feature(explicit_tail_calls))]
5#![cfg_attr(pulley_tail_calls, allow(incomplete_features, unstable_features))]
6#![deny(missing_docs)]
7#![no_std]
8
9#[cfg(feature = "std")]
10#[macro_use]
11extern crate std;
12
13#[cfg(feature = "decode")]
14extern crate alloc;
15
16/// Calls the given macro with each opcode.
17///
18/// # Instruction Guidelines
19///
20/// We're inventing an instruction set here which naturally brings a whole set
21/// of design questions. Note that this is explicitly intended to be only ever
22/// used for Pulley where there are a different set of design constraints than
23/// other instruction sets (e.g. general-purpose CPU ISAs). Some examples of
24/// constraints for Pulley are:
25///
26/// * Instructions must be portable to many architectures.
27/// * The Pulley ISA is mostly target-independent as the compilation target is
28///   currently only parameterized on pointer width and endianness.
29/// * Pulley instructions should be balance of time-to-decode and code size. For
30///   example super fancy bit-packing tricks might be tough to decode in
31///   software but might be worthwhile if it's quite common and greatly reduces
32///   the size of bytecode. There's not a hard-and-fast answer here, but a
33///   balance to be made.
34/// * Many "macro ops" are present to reduce the size of compiled bytecode so
35///   there is a wide set of duplicate functionality between opcodes (and this
36///   is expected).
37///
38/// Given all this it's also useful to have a set of guidelines used to name and
39/// develop Pulley instructions. As of the time of this writing it's still
40/// pretty early days for Pulley so some of these guidelines may change over
41/// time. Additionally instructions don't necessarily all follow these
42/// conventions and that may also change over time. With that in mind, here's a
43/// rough set of guidelines:
44///
45/// * Most instructions are prefixed with `x`, `f`, or `v`, indicating which
46///   type of register they're operating on. (e.g. `xadd32` operates on the `x`
47///   integer registers and `fadd32` operates on the `f` float registers).
48///
49/// * Most instructions are suffixed or otherwise contain the bit width they're
50///   operating on. For example `xadd32` is a 32-bit addition.
51///
52/// * If an instruction operates on signed or unsigned data (such as division
53///   and remainder), then the instruction is suffixed with `_s` or `_u`.
54///
55/// * Instructions operate on either 32 or 64-bit parts of a register.
56///   Instructions modifying only 32-bits of a register always modify the "low"
57///   part of a register and leave the upper part unmodified. This is intended
58///   to help 32-bit platforms where if most operations are 32-bit there's no
59///   need for extra instructions to sign or zero extend and modify the upper
60///   half of the register.
61///
62/// * Binops use `BinaryOperands<T>` for the destination and argument registers.
63///
64/// * Instructions operating on memory contain a few pieces of information:
65///
66///   ```text
67///   xload16le_u32_offset32
68///   │└─┬┘└┤└┤ └┬┘ └──┬───┘
69///   │  │  │ │  │     ▼
70///   │  │  │ │  │     addressing mode
71///   │  │  │ │  ▼
72///   │  │  │ │  width of register modified + sign-extension (optional)
73///   │  │  │ ▼
74///   │  │  │ endianness of the operation (le/be)
75///   │  │  ▼
76///   │  │  bit-width of the operation
77///   │  ▼
78///   │  what's happening (load/store)
79///   ▼
80///   register being operated on (x/f/z)
81///   ```
82///
83/// More guidelines might get added here over time, and if you have any
84/// questions feel free to raise them and we can try to add them here as well!
85#[macro_export]
86macro_rules! for_each_op {
87    ( $macro:ident ) => {
88        $macro! {
89            /// Transfer control the address in the `lr` register.
90            ret = Ret;
91
92            /// Transfer control to the PC at the given offset and set the `lr`
93            /// register to the PC just after this instruction.
94            ///
95            /// This instruction generally assumes that the Pulley ABI is being
96            /// respected where arguments are in argument registers (starting at
97            /// x0 for integer arguments) and results are in result registers.
98            /// This instruction itself assume that all arguments are already in
99            /// their registers. Subsequent instructions below enable moving
100            /// arguments into the correct registers as part of the same call
101            /// instruction.
102            call = Call { offset: PcRelOffset };
103            /// Like `call`, but also `x0 = arg1`
104            call1 = Call1 { arg1: XReg, offset: PcRelOffset };
105            /// Like `call`, but also `x0, x1 = arg1, arg2`
106            call2 = Call2 { arg1: XReg, arg2: XReg, offset: PcRelOffset };
107            /// Like `call`, but also `x0, x1, x2 = arg1, arg2, arg3`
108            call3 = Call3 { arg1: XReg, arg2: XReg, arg3: XReg, offset: PcRelOffset };
109            /// Like `call`, but also `x0, x1, x2, x3 = arg1, arg2, arg3, arg4`
110            call4 = Call4 { arg1: XReg, arg2: XReg, arg3: XReg, arg4: XReg, offset: PcRelOffset };
111
112            /// Transfer control to the PC in `reg` and set `lr` to the PC just
113            /// after this instruction.
114            call_indirect = CallIndirect { reg: XReg };
115
116            /// Unconditionally transfer control to the PC at the given offset.
117            jump = Jump { offset: PcRelOffset };
118
119            /// Unconditionally transfer control to the PC at specified
120            /// register.
121            xjump = XJump { reg: XReg };
122
123            /// Conditionally transfer control to the given PC offset if
124            /// `low32(cond)` contains a non-zero value.
125            br_if32 = BrIf { cond: XReg, offset: PcRelOffset };
126
127            /// Conditionally transfer control to the given PC offset if
128            /// `low32(cond)` contains a zero value.
129            br_if_not32 = BrIfNot { cond: XReg, offset: PcRelOffset };
130
131            /// Branch if `a == b`.
132            br_if_xeq32 = BrIfXeq32 { a: XReg, b: XReg, offset: PcRelOffset };
133            /// Branch if `a != `b.
134            br_if_xneq32 = BrIfXneq32 { a: XReg, b: XReg, offset: PcRelOffset };
135            /// Branch if signed `a < b`.
136            br_if_xslt32 = BrIfXslt32 { a: XReg, b: XReg, offset: PcRelOffset };
137            /// Branch if signed `a <= b`.
138            br_if_xslteq32 = BrIfXslteq32 { a: XReg, b: XReg, offset: PcRelOffset };
139            /// Branch if unsigned `a < b`.
140            br_if_xult32 = BrIfXult32 { a: XReg, b: XReg, offset: PcRelOffset };
141            /// Branch if unsigned `a <= b`.
142            br_if_xulteq32 = BrIfXulteq32 { a: XReg, b: XReg, offset: PcRelOffset };
143            /// Branch if `a == b`.
144            br_if_xeq64 = BrIfXeq64 { a: XReg, b: XReg, offset: PcRelOffset };
145            /// Branch if `a != `b.
146            br_if_xneq64 = BrIfXneq64 { a: XReg, b: XReg, offset: PcRelOffset };
147            /// Branch if signed `a < b`.
148            br_if_xslt64 = BrIfXslt64 { a: XReg, b: XReg, offset: PcRelOffset };
149            /// Branch if signed `a <= b`.
150            br_if_xslteq64 = BrIfXslteq64 { a: XReg, b: XReg, offset: PcRelOffset };
151            /// Branch if unsigned `a < b`.
152            br_if_xult64 = BrIfXult64 { a: XReg, b: XReg, offset: PcRelOffset };
153            /// Branch if unsigned `a <= b`.
154            br_if_xulteq64 = BrIfXulteq64 { a: XReg, b: XReg, offset: PcRelOffset };
155
156            /// Branch if `a == b`.
157            br_if_xeq32_i8 = BrIfXeq32I8 { a: XReg, b: i8, offset: PcRelOffset };
158            /// Branch if `a == b`.
159            br_if_xeq32_i32 = BrIfXeq32I32 { a: XReg, b: i32, offset: PcRelOffset };
160            /// Branch if `a != `b.
161            br_if_xneq32_i8 = BrIfXneq32I8 { a: XReg, b: i8, offset: PcRelOffset };
162            /// Branch if `a != `b.
163            br_if_xneq32_i32 = BrIfXneq32I32 { a: XReg, b: i32, offset: PcRelOffset };
164            /// Branch if signed `a < b`.
165            br_if_xslt32_i8 = BrIfXslt32I8 { a: XReg, b: i8, offset: PcRelOffset };
166            /// Branch if signed `a < b`.
167            br_if_xslt32_i32 = BrIfXslt32I32 { a: XReg, b: i32, offset: PcRelOffset };
168            /// Branch if signed `a > b`.
169            br_if_xsgt32_i8 = BrIfXsgt32I8 { a: XReg, b: i8, offset: PcRelOffset };
170            /// Branch if signed `a > b`.
171            br_if_xsgt32_i32 = BrIfXsgt32I32 { a: XReg, b: i32, offset: PcRelOffset };
172            /// Branch if signed `a <= b`.
173            br_if_xslteq32_i8 = BrIfXslteq32I8 { a: XReg, b: i8, offset: PcRelOffset };
174            /// Branch if signed `a <= b`.
175            br_if_xslteq32_i32 = BrIfXslteq32I32 { a: XReg, b: i32, offset: PcRelOffset };
176            /// Branch if signed `a >= b`.
177            br_if_xsgteq32_i8 = BrIfXsgteq32I8 { a: XReg, b: i8, offset: PcRelOffset };
178            /// Branch if signed `a >= b`.
179            br_if_xsgteq32_i32 = BrIfXsgteq32I32 { a: XReg, b: i32, offset: PcRelOffset };
180            /// Branch if unsigned `a < b`.
181            br_if_xult32_u8 = BrIfXult32U8 { a: XReg, b: u8, offset: PcRelOffset };
182            /// Branch if unsigned `a < b`.
183            br_if_xult32_u32 = BrIfXult32U32 { a: XReg, b: u32, offset: PcRelOffset };
184            /// Branch if unsigned `a <= b`.
185            br_if_xulteq32_u8 = BrIfXulteq32U8 { a: XReg, b: u8, offset: PcRelOffset };
186            /// Branch if unsigned `a <= b`.
187            br_if_xulteq32_u32 = BrIfXulteq32U32 { a: XReg, b: u32, offset: PcRelOffset };
188            /// Branch if unsigned `a > b`.
189            br_if_xugt32_u8 = BrIfXugt32U8 { a: XReg, b: u8, offset: PcRelOffset };
190            /// Branch if unsigned `a > b`.
191            br_if_xugt32_u32 = BrIfXugt32U32 { a: XReg, b: u32, offset: PcRelOffset };
192            /// Branch if unsigned `a >= b`.
193            br_if_xugteq32_u8 = BrIfXugteq32U8 { a: XReg, b: u8, offset: PcRelOffset };
194            /// Branch if unsigned `a >= b`.
195            br_if_xugteq32_u32 = BrIfXugteq32U32 { a: XReg, b: u32, offset: PcRelOffset };
196
197            /// Branch if `a == b`.
198            br_if_xeq64_i8 = BrIfXeq64I8 { a: XReg, b: i8, offset: PcRelOffset };
199            /// Branch if `a == b`.
200            br_if_xeq64_i32 = BrIfXeq64I32 { a: XReg, b: i32, offset: PcRelOffset };
201            /// Branch if `a != `b.
202            br_if_xneq64_i8 = BrIfXneq64I8 { a: XReg, b: i8, offset: PcRelOffset };
203            /// Branch if `a != `b.
204            br_if_xneq64_i32 = BrIfXneq64I32 { a: XReg, b: i32, offset: PcRelOffset };
205            /// Branch if signed `a < b`.
206            br_if_xslt64_i8 = BrIfXslt64I8 { a: XReg, b: i8, offset: PcRelOffset };
207            /// Branch if signed `a < b`.
208            br_if_xslt64_i32 = BrIfXslt64I32 { a: XReg, b: i32, offset: PcRelOffset };
209            /// Branch if signed `a > b`.
210            br_if_xsgt64_i8 = BrIfXsgt64I8 { a: XReg, b: i8, offset: PcRelOffset };
211            /// Branch if signed `a > b`.
212            br_if_xsgt64_i32 = BrIfXsgt64I32 { a: XReg, b: i32, offset: PcRelOffset };
213            /// Branch if signed `a <= b`.
214            br_if_xslteq64_i8 = BrIfXslteq64I8 { a: XReg, b: i8, offset: PcRelOffset };
215            /// Branch if signed `a <= b`.
216            br_if_xslteq64_i32 = BrIfXslteq64I32 { a: XReg, b: i32, offset: PcRelOffset };
217            /// Branch if signed `a >= b`.
218            br_if_xsgteq64_i8 = BrIfXsgteq64I8 { a: XReg, b: i8, offset: PcRelOffset };
219            /// Branch if signed `a >= b`.
220            br_if_xsgteq64_i32 = BrIfXsgteq64I32 { a: XReg, b: i32, offset: PcRelOffset };
221            /// Branch if unsigned `a < b`.
222            br_if_xult64_u8 = BrIfXult64U8 { a: XReg, b: u8, offset: PcRelOffset };
223            /// Branch if unsigned `a < b`.
224            br_if_xult64_u32 = BrIfXult64U32 { a: XReg, b: u32, offset: PcRelOffset };
225            /// Branch if unsigned `a <= b`.
226            br_if_xulteq64_u8 = BrIfXulteq64U8 { a: XReg, b: u8, offset: PcRelOffset };
227            /// Branch if unsigned `a <= b`.
228            br_if_xulteq64_u32 = BrIfXulteq64U32 { a: XReg, b: u32, offset: PcRelOffset };
229            /// Branch if unsigned `a > b`.
230            br_if_xugt64_u8 = BrIfXugt64U8 { a: XReg, b: u8, offset: PcRelOffset };
231            /// Branch if unsigned `a > b`.
232            br_if_xugt64_u32 = BrIfXugt64U32 { a: XReg, b: u32, offset: PcRelOffset };
233            /// Branch if unsigned `a >= b`.
234            br_if_xugteq64_u8 = BrIfXugteq64U8 { a: XReg, b: u8, offset: PcRelOffset };
235            /// Branch if unsigned `a >= b`.
236            br_if_xugteq64_u32 = BrIfXugteq64U32 { a: XReg, b: u32, offset: PcRelOffset };
237
238            /// Branch to the label indicated by `low32(idx)`.
239            ///
240            /// After this instruction are `amt` instances of `PcRelOffset`
241            /// and the `idx` selects which one will be branched to. The value
242            /// of `idx` is clamped to `amt - 1` (e.g. the last offset is the
243            /// "default" one.
244            br_table32 = BrTable32 { idx: XReg, amt: u32 };
245
246            /// Move between `x` registers.
247            xmov = Xmov { dst: XReg, src: XReg };
248
249            /// Set `dst = 0`
250            xzero = Xzero { dst: XReg };
251            /// Set `dst = 1`
252            xone = Xone { dst: XReg };
253            /// Set `dst = sign_extend(imm8)`.
254            xconst8 = Xconst8 { dst: XReg, imm: i8 };
255            /// Set `dst = sign_extend(imm16)`.
256            xconst16 = Xconst16 { dst: XReg, imm: i16 };
257            /// Set `dst = sign_extend(imm32)`.
258            xconst32 = Xconst32 { dst: XReg, imm: i32 };
259            /// Set `dst = imm64`.
260            xconst64 = Xconst64 { dst: XReg, imm: i64 };
261
262            /// 32-bit wrapping addition: `low32(dst) = low32(src1) + low32(src2)`.
263            ///
264            /// The upper 32-bits of `dst` are unmodified.
265            xadd32 = Xadd32 { operands: BinaryOperands<XReg> };
266            /// Same as `xadd32` but `src2` is a zero-extended 8-bit immediate.
267            xadd32_u8 = Xadd32U8 { dst: XReg, src1: XReg, src2: u8 };
268            /// Same as `xadd32` but `src2` is a 32-bit immediate.
269            xadd32_u32 = Xadd32U32 { dst: XReg, src1: XReg, src2: u32 };
270
271            /// 64-bit wrapping addition: `dst = src1 + src2`.
272            xadd64 = Xadd64 { operands: BinaryOperands<XReg> };
273            /// Same as `xadd64` but `src2` is a zero-extended 8-bit immediate.
274            xadd64_u8 = Xadd64U8 { dst: XReg, src1: XReg, src2: u8 };
275            /// Same as `xadd64` but `src2` is a zero-extended 32-bit immediate.
276            xadd64_u32 = Xadd64U32 { dst: XReg, src1: XReg, src2: u32 };
277
278            /// `low32(dst) = low32(src1) * low32(src2) + low32(src3)`
279            xmadd32 = Xmadd32 { dst: XReg, src1: XReg, src2: XReg, src3: XReg };
280            /// `dst = src1 * src2 + src3`
281            xmadd64 = Xmadd64 { dst: XReg, src1: XReg, src2: XReg, src3: XReg };
282
283            /// 32-bit wrapping subtraction: `low32(dst) = low32(src1) - low32(src2)`.
284            ///
285            /// The upper 32-bits of `dst` are unmodified.
286            xsub32 = Xsub32 { operands: BinaryOperands<XReg> };
287            /// Same as `xsub32` but `src2` is a zero-extended 8-bit immediate.
288            xsub32_u8 = Xsub32U8 { dst: XReg, src1: XReg, src2: u8 };
289            /// Same as `xsub32` but `src2` is a 32-bit immediate.
290            xsub32_u32 = Xsub32U32 { dst: XReg, src1: XReg, src2: u32 };
291
292            /// 64-bit wrapping subtraction: `dst = src1 - src2`.
293            xsub64 = Xsub64 { operands: BinaryOperands<XReg> };
294            /// Same as `xsub64` but `src2` is a zero-extended 8-bit immediate.
295            xsub64_u8 = Xsub64U8 { dst: XReg, src1: XReg, src2: u8 };
296            /// Same as `xsub64` but `src2` is a zero-extended 32-bit immediate.
297            xsub64_u32 = Xsub64U32 { dst: XReg, src1: XReg, src2: u32 };
298
299            /// `low32(dst) = low32(src1) * low32(src2)`
300            xmul32 = XMul32 { operands: BinaryOperands<XReg> };
301            /// Same as `xmul64` but `src2` is a sign-extended 8-bit immediate.
302            xmul32_s8 = Xmul32S8 { dst: XReg, src1: XReg, src2: i8 };
303            /// Same as `xmul32` but `src2` is a sign-extended 32-bit immediate.
304            xmul32_s32 = Xmul32S32 { dst: XReg, src1: XReg, src2: i32 };
305
306            /// `dst = src1 * src2`
307            xmul64 = XMul64 { operands: BinaryOperands<XReg> };
308            /// Same as `xmul64` but `src2` is a sign-extended 8-bit immediate.
309            xmul64_s8 = Xmul64S8 { dst: XReg, src1: XReg, src2: i8 };
310            /// Same as `xmul64` but `src2` is a sign-extended 64-bit immediate.
311            xmul64_s32 = Xmul64S32 { dst: XReg, src1: XReg, src2: i32 };
312
313            /// `low32(dst) = trailing_zeros(low32(src))`
314            xctz32 = Xctz32 { dst: XReg, src: XReg };
315            /// `dst = trailing_zeros(src)`
316            xctz64 = Xctz64 { dst: XReg, src: XReg };
317
318            /// `low32(dst) = leading_zeros(low32(src))`
319            xclz32 = Xclz32 { dst: XReg, src: XReg };
320            /// `dst = leading_zeros(src)`
321            xclz64 = Xclz64 { dst: XReg, src: XReg };
322
323            /// `low32(dst) = count_ones(low32(src))`
324            xpopcnt32 = Xpopcnt32 { dst: XReg, src: XReg };
325            /// `dst = count_ones(src)`
326            xpopcnt64 = Xpopcnt64 { dst: XReg, src: XReg };
327
328            /// `low32(dst) = rotate_left(low32(src1), low32(src2))`
329            xrotl32 = Xrotl32 { operands: BinaryOperands<XReg> };
330            /// `dst = rotate_left(src1, src2)`
331            xrotl64 = Xrotl64 { operands: BinaryOperands<XReg> };
332
333            /// `low32(dst) = rotate_right(low32(src1), low32(src2))`
334            xrotr32 = Xrotr32 { operands: BinaryOperands<XReg> };
335            /// `dst = rotate_right(src1, src2)`
336            xrotr64 = Xrotr64 { operands: BinaryOperands<XReg> };
337
338            /// `low32(dst) = low32(src1) << low5(src2)`
339            xshl32 = Xshl32 { operands: BinaryOperands<XReg> };
340            /// `low32(dst) = low32(src1) >> low5(src2)`
341            xshr32_s = Xshr32S { operands: BinaryOperands<XReg> };
342            /// `low32(dst) = low32(src1) >> low5(src2)`
343            xshr32_u = Xshr32U { operands: BinaryOperands<XReg> };
344            /// `dst = src1 << low5(src2)`
345            xshl64 = Xshl64 { operands: BinaryOperands<XReg> };
346            /// `dst = src1 >> low6(src2)`
347            xshr64_s = Xshr64S { operands: BinaryOperands<XReg> };
348            /// `dst = src1 >> low6(src2)`
349            xshr64_u = Xshr64U { operands: BinaryOperands<XReg> };
350
351            /// `low32(dst) = low32(src1) << low5(src2)`
352            xshl32_u6 = Xshl32U6 { operands: BinaryOperands<XReg, XReg, U6> };
353            /// `low32(dst) = low32(src1) >> low5(src2)`
354            xshr32_s_u6 = Xshr32SU6 { operands: BinaryOperands<XReg, XReg, U6> };
355            /// `low32(dst) = low32(src1) >> low5(src2)`
356            xshr32_u_u6 = Xshr32UU6 { operands: BinaryOperands<XReg, XReg, U6> };
357            /// `dst = src1 << low5(src2)`
358            xshl64_u6 = Xshl64U6 { operands: BinaryOperands<XReg, XReg, U6> };
359            /// `dst = src1 >> low6(src2)`
360            xshr64_s_u6 = Xshr64SU6 { operands: BinaryOperands<XReg, XReg, U6> };
361            /// `dst = src1 >> low6(src2)`
362            xshr64_u_u6 = Xshr64UU6 { operands: BinaryOperands<XReg, XReg, U6> };
363
364            /// `low32(dst) = -low32(src)`
365            xneg32 = Xneg32 { dst: XReg, src: XReg };
366            /// `dst = -src`
367            xneg64 = Xneg64 { dst: XReg, src: XReg };
368
369            /// `low32(dst) = src1 == src2`
370            xeq64 = Xeq64 { operands: BinaryOperands<XReg> };
371            /// `low32(dst) = src1 != src2`
372            xneq64 = Xneq64 { operands: BinaryOperands<XReg> };
373            /// `low32(dst) = src1 < src2` (signed)
374            xslt64 = Xslt64 { operands: BinaryOperands<XReg> };
375            /// `low32(dst) = src1 <= src2` (signed)
376            xslteq64 = Xslteq64 { operands: BinaryOperands<XReg> };
377            /// `low32(dst) = src1 < src2` (unsigned)
378            xult64 = Xult64 { operands: BinaryOperands<XReg> };
379            /// `low32(dst) = src1 <= src2` (unsigned)
380            xulteq64 = Xulteq64 { operands: BinaryOperands<XReg> };
381            /// `low32(dst) = low32(src1) == low32(src2)`
382            xeq32 = Xeq32 { operands: BinaryOperands<XReg> };
383            /// `low32(dst) = low32(src1) != low32(src2)`
384            xneq32 = Xneq32 { operands: BinaryOperands<XReg> };
385            /// `low32(dst) = low32(src1) < low32(src2)` (signed)
386            xslt32 = Xslt32 { operands: BinaryOperands<XReg> };
387            /// `low32(dst) = low32(src1) <= low32(src2)` (signed)
388            xslteq32 = Xslteq32 { operands: BinaryOperands<XReg> };
389            /// `low32(dst) = low32(src1) < low32(src2)` (unsigned)
390            xult32 = Xult32 { operands: BinaryOperands<XReg> };
391            /// `low32(dst) = low32(src1) <= low32(src2)` (unsigned)
392            xulteq32 = Xulteq32 { operands: BinaryOperands<XReg> };
393
394            /// `low32(dst) = zext(*(ptr + offset))`
395            xload8_u32_offset32 = XLoad8U32Offset32 { dst: XReg, ptr: XReg, offset: i32 };
396            /// `low32(dst) = sext(*(ptr + offset))`
397            xload8_s32_offset32 = XLoad8S32Offset32 { dst: XReg, ptr: XReg, offset: i32 };
398            /// `low32(dst) = zext(*(ptr + offset))`
399            xload16le_u32_offset32 = XLoad16LeU32Offset32 { dst: XReg, ptr: XReg, offset: i32 };
400            /// `low32(dst) = sext(*(ptr + offset))`
401            xload16le_s32_offset32 = XLoad16LeS32Offset32 { dst: XReg, ptr: XReg, offset: i32 };
402            /// `low32(dst) = *(ptr + offset)`
403            xload32le_offset32 = XLoad32LeOffset32 { dst: XReg, ptr: XReg, offset: i32 };
404
405            /// `dst = zext(*(ptr + offset))`
406            xload8_u64_offset32 = XLoad8U64Offset32 { dst: XReg, ptr: XReg, offset: i32 };
407            /// `dst = sext(*(ptr + offset))`
408            xload8_s64_offset32 = XLoad8S64Offset32 { dst: XReg, ptr: XReg, offset: i32 };
409            /// `dst = zext(*(ptr + offset))`
410            xload16le_u64_offset32 = XLoad16LeU64Offset32 { dst: XReg, ptr: XReg, offset: i32 };
411            /// `dst = sext(*(ptr + offset))`
412            xload16le_s64_offset32 = XLoad16LeS64Offset32 { dst: XReg, ptr: XReg, offset: i32 };
413            /// `dst = zext(*(ptr + offset))`
414            xload32le_u64_offset32 = XLoad32LeU64Offset32 { dst: XReg, ptr: XReg, offset: i32 };
415            /// `dst = sext(*(ptr + offset))`
416            xload32le_s64_offset32 = XLoad32LeS64Offset32 { dst: XReg, ptr: XReg, offset: i32 };
417            /// `dst = *(ptr + offset)`
418            xload64le_offset32 = XLoad64LeOffset32 { dst: XReg, ptr: XReg, offset: i32 };
419
420            /// `*(ptr + offset) = low8(src)`
421            xstore8_offset32 = XStore8Offset32 { ptr: XReg, offset: i32, src: XReg };
422            /// `*(ptr + offset) = low16(src)`
423            xstore16le_offset32 = XStore16LeOffset32 { ptr: XReg, offset: i32, src: XReg };
424            /// `*(ptr + offset) = low32(src)`
425            xstore32le_offset32 = XStore32LeOffset32 { ptr: XReg, offset: i32, src: XReg };
426            /// `*(ptr + offset) = low64(src)`
427            xstore64le_offset32 = XStore64LeOffset32 { ptr: XReg, offset: i32, src: XReg };
428
429            /// `low32(dst) = zext(*(ptr + offset))`
430            xload8_u32_offset8 = XLoad8U32Offset8 { dst: XReg, ptr: XReg, offset: u8 };
431            /// `low32(dst) = sext(*(ptr + offset))`
432            xload8_s32_offset8 = XLoad8S32Offset8 { dst: XReg, ptr: XReg, offset: u8 };
433            /// `low32(dst) = zext(*(ptr + offset))`
434            xload16le_u32_offset8 = XLoad16LeU32Offset8 { dst: XReg, ptr: XReg, offset: u8 };
435            /// `low32(dst) = sext(*(ptr + offset))`
436            xload16le_s32_offset8 = XLoad16LeS32Offset8 { dst: XReg, ptr: XReg, offset: u8 };
437            /// `low32(dst) = *(ptr + offset)`
438            xload32le_offset8 = XLoad32LeOffset8 { dst: XReg, ptr: XReg, offset: u8 };
439
440            /// `dst = zext(*(ptr + offset))`
441            xload8_u64_offset8 = XLoad8U64Offset8 { dst: XReg, ptr: XReg, offset: u8 };
442            /// `dst = sext(*(ptr + offset))`
443            xload8_s64_offset8 = XLoad8S64Offset8 { dst: XReg, ptr: XReg, offset: u8 };
444            /// `dst = zext(*(ptr + offset))`
445            xload16le_u64_offset8 = XLoad16LeU64Offset8 { dst: XReg, ptr: XReg, offset: u8 };
446            /// `dst = sext(*(ptr + offset))`
447            xload16le_s64_offset8 = XLoad16LeS64Offset8 { dst: XReg, ptr: XReg, offset: u8 };
448            /// `dst = zext(*(ptr + offset))`
449            xload32le_u64_offset8 = XLoad32LeU64Offset8 { dst: XReg, ptr: XReg, offset: u8 };
450            /// `dst = sext(*(ptr + offset))`
451            xload32le_s64_offset8 = XLoad32LeS64Offset8 { dst: XReg, ptr: XReg, offset: u8 };
452            /// `dst = *(ptr + offset)`
453            xload64le_offset8 = XLoad64LeOffset8 { dst: XReg, ptr: XReg, offset: u8 };
454
455            /// `*(ptr + offset) = low8(src)`
456            xstore8_offset8 = XStore8Offset8 { ptr: XReg, offset: u8, src: XReg };
457            /// `*(ptr + offset) = low16(src)`
458            xstore16le_offset8 = XStore16LeOffset8 { ptr: XReg, offset: u8, src: XReg };
459            /// `*(ptr + offset) = low32(src)`
460            xstore32le_offset8 = XStore32LeOffset8 { ptr: XReg, offset: u8, src: XReg };
461            /// `*(ptr + offset) = low64(src)`
462            xstore64le_offset8 = XStore64LeOffset8 { ptr: XReg, offset: u8, src: XReg };
463
464            // wasm addressing modes
465            //
466            // g32 = 32-bit guest, arithmetic is zero-extending the `addr`
467            //       to the host pointer width.
468
469            /// `low32(dst) = zext_8_32(*(base + zext(addr) + offset))`
470            xload8_u32_g32 = XLoad8U32G32 { dst: XReg, base: XReg, addr: XReg, offset: u8 };
471            /// `low32(dst) = sext_8_32(*(base + zext(addr) + offset))`
472            xload8_s32_g32 = XLoad8S32G32 { dst: XReg, base: XReg, addr: XReg, offset: u8 };
473            /// `low32(dst) = zext_16_32(*(base + zext(addr) + offset))`
474            xload16le_u32_g32 = XLoad16LeU32G32 { dst: XReg, base: XReg, addr: XReg, offset: u8 };
475            /// `low32(dst) = sext_16_32(*(base + zext(addr) + offset))`
476            xload16le_s32_g32 = XLoad16LeS32G32 { dst: XReg, base: XReg, addr: XReg, offset: u8 };
477            /// `low32(dst) = *(base + zext(addr) + offset)`
478            xload32le_g32 = XLoad32LeG32 { dst: XReg, base: XReg, addr: XReg, offset: u8 };
479            /// `dst = *(base + zext(addr) + offset)`
480            xload64le_g32 = XLoad64LeG32 { dst: XReg, base: XReg, addr: XReg, offset: u8 };
481
482            /// `*(base + zext(addr) + offset) = low8(src)`
483            xstore8_g32 = XStore8G32 { base: XReg, addr: XReg, offset: u8, src: XReg };
484            /// `*(base + zext(addr) + offset) = low16(src)`
485            xstore16le_g32 = XStore16LeG32 { base: XReg, addr: XReg, offset: u8, src: XReg };
486            /// `*(base + zext(addr) + offset) = low32(src)`
487            xstore32le_g32 = XStore32LeG32 { base: XReg, addr: XReg, offset: u8, src: XReg };
488            /// `*(base + zext(addr) + offset) = src`
489            xstore64le_g32 = XStore64LeG32 { base: XReg, addr: XReg, offset: u8, src: XReg };
490
491            /// `push lr; push fp; fp = sp`
492            push_frame = PushFrame ;
493            /// `sp = fp; pop fp; pop lr`
494            pop_frame = PopFrame ;
495
496            /// Macro-instruction to enter a function, allocate some stack, and
497            /// then save some registers.
498            ///
499            /// This is equivalent to `push_frame`, `stack_alloc32 amt`, then
500            /// saving all of `regs` to the top of the stack just allocated.
501            push_frame_save = PushFrameSave { amt: u16, regs: UpperRegSet<XReg> };
502            /// Inverse of `push_frame_save`. Restores `regs` from the top of
503            /// the stack, then runs `stack_free32 amt`, then runs `pop_frame`.
504            pop_frame_restore = PopFrameRestore { amt: u16, regs: UpperRegSet<XReg> };
505
506            /// `sp = sp.checked_sub(amt)`
507            stack_alloc32 = StackAlloc32 { amt: u32 };
508
509            /// `sp = sp + amt`
510            stack_free32 = StackFree32 { amt: u32 };
511
512            /// `dst = zext(low8(src))`
513            zext8 = Zext8 { dst: XReg, src: XReg };
514            /// `dst = zext(low16(src))`
515            zext16 = Zext16 { dst: XReg, src: XReg };
516            /// `dst = zext(low32(src))`
517            zext32 = Zext32 { dst: XReg, src: XReg };
518            /// `dst = sext(low8(src))`
519            sext8 = Sext8 { dst: XReg, src: XReg };
520            /// `dst = sext(low16(src))`
521            sext16 = Sext16 { dst: XReg, src: XReg };
522            /// `dst = sext(low32(src))`
523            sext32 = Sext32 { dst: XReg, src: XReg };
524
525            /// `low32(dst) = |low32(src)|`
526            xabs32 = XAbs32 { dst: XReg, src: XReg };
527            /// `dst = |src|`
528            xabs64 = XAbs64 { dst: XReg, src: XReg };
529
530            /// `low32(dst) = low32(src1) / low32(src2)` (signed)
531            xdiv32_s = XDiv32S { operands: BinaryOperands<XReg> };
532
533            /// `dst = src1 / src2` (signed)
534            xdiv64_s = XDiv64S { operands: BinaryOperands<XReg> };
535
536            /// `low32(dst) = low32(src1) / low32(src2)` (unsigned)
537            xdiv32_u = XDiv32U { operands: BinaryOperands<XReg> };
538
539            /// `dst = src1 / src2` (unsigned)
540            xdiv64_u = XDiv64U { operands: BinaryOperands<XReg> };
541
542            /// `low32(dst) = low32(src1) % low32(src2)` (signed)
543            xrem32_s = XRem32S { operands: BinaryOperands<XReg> };
544
545            /// `dst = src1 / src2` (signed)
546            xrem64_s = XRem64S { operands: BinaryOperands<XReg> };
547
548            /// `low32(dst) = low32(src1) % low32(src2)` (unsigned)
549            xrem32_u = XRem32U { operands: BinaryOperands<XReg> };
550
551            /// `dst = src1 / src2` (unsigned)
552            xrem64_u = XRem64U { operands: BinaryOperands<XReg> };
553
554            /// `low32(dst) = low32(src1) & low32(src2)`
555            xband32 = XBand32 { operands: BinaryOperands<XReg> };
556            /// Same as `xband64` but `src2` is a sign-extended 8-bit immediate.
557            xband32_s8 = Xband32S8 { dst: XReg, src1: XReg, src2: i8 };
558            /// Same as `xband32` but `src2` is a sign-extended 32-bit immediate.
559            xband32_s32 = Xband32S32 { dst: XReg, src1: XReg, src2: i32 };
560            /// `dst = src1 & src2`
561            xband64 = XBand64 { operands: BinaryOperands<XReg> };
562            /// Same as `xband64` but `src2` is a sign-extended 8-bit immediate.
563            xband64_s8 = Xband64S8 { dst: XReg, src1: XReg, src2: i8 };
564            /// Same as `xband64` but `src2` is a sign-extended 32-bit immediate.
565            xband64_s32 = Xband64S32 { dst: XReg, src1: XReg, src2: i32 };
566            /// `low32(dst) = low32(src1) | low32(src2)`
567            xbor32 = XBor32 { operands: BinaryOperands<XReg> };
568            /// Same as `xbor64` but `src2` is a sign-extended 8-bit immediate.
569            xbor32_s8 = Xbor32S8 { dst: XReg, src1: XReg, src2: i8 };
570            /// Same as `xbor32` but `src2` is a sign-extended 32-bit immediate.
571            xbor32_s32 = Xbor32S32 { dst: XReg, src1: XReg, src2: i32 };
572            /// `dst = src1 | src2`
573            xbor64 = XBor64 { operands: BinaryOperands<XReg> };
574            /// Same as `xbor64` but `src2` is a sign-extended 8-bit immediate.
575            xbor64_s8 = Xbor64S8 { dst: XReg, src1: XReg, src2: i8 };
576            /// Same as `xbor64` but `src2` is a sign-extended 32-bit immediate.
577            xbor64_s32 = Xbor64S32 { dst: XReg, src1: XReg, src2: i32 };
578
579            /// `low32(dst) = low32(src1) ^ low32(src2)`
580            xbxor32 = XBxor32 { operands: BinaryOperands<XReg> };
581            /// Same as `xbxor64` but `src2` is a sign-extended 8-bit immediate.
582            xbxor32_s8 = Xbxor32S8 { dst: XReg, src1: XReg, src2: i8 };
583            /// Same as `xbxor32` but `src2` is a sign-extended 32-bit immediate.
584            xbxor32_s32 = Xbxor32S32 { dst: XReg, src1: XReg, src2: i32 };
585            /// `dst = src1 ^ src2`
586            xbxor64 = XBxor64 { operands: BinaryOperands<XReg> };
587            /// Same as `xbxor64` but `src2` is a sign-extended 8-bit immediate.
588            xbxor64_s8 = Xbxor64S8 { dst: XReg, src1: XReg, src2: i8 };
589            /// Same as `xbxor64` but `src2` is a sign-extended 32-bit immediate.
590            xbxor64_s32 = Xbxor64S32 { dst: XReg, src1: XReg, src2: i32 };
591
592            /// `low32(dst) = !low32(src1)`
593            xbnot32 = XBnot32 { dst: XReg, src: XReg };
594            /// `dst = !src1`
595            xbnot64 = XBnot64 { dst: XReg, src: XReg };
596
597            /// `low32(dst) = min(low32(src1), low32(src2))` (unsigned)
598            xmin32_u = Xmin32U { operands: BinaryOperands<XReg> };
599            /// `low32(dst) = min(low32(src1), low32(src2))` (signed)
600            xmin32_s = Xmin32S { operands: BinaryOperands<XReg> };
601            /// `low32(dst) = max(low32(src1), low32(src2))` (unsigned)
602            xmax32_u = Xmax32U { operands: BinaryOperands<XReg> };
603            /// `low32(dst) = max(low32(src1), low32(src2))` (signed)
604            xmax32_s = Xmax32S { operands: BinaryOperands<XReg> };
605            /// `dst = min(src1, src2)` (unsigned)
606            xmin64_u = Xmin64U { operands: BinaryOperands<XReg> };
607            /// `dst = min(src1, src2)` (signed)
608            xmin64_s = Xmin64S { operands: BinaryOperands<XReg> };
609            /// `dst = max(src1, src2)` (unsigned)
610            xmax64_u = Xmax64U { operands: BinaryOperands<XReg> };
611            /// `dst = max(src1, src2)` (signed)
612            xmax64_s = Xmax64S { operands: BinaryOperands<XReg> };
613
614            /// `low32(dst) = low32(cond) ? low32(if_nonzero) : low32(if_zero)`
615            xselect32 = XSelect32 { dst: XReg, cond: XReg, if_nonzero: XReg, if_zero: XReg };
616            /// `dst = low32(cond) ? if_nonzero : if_zero`
617            xselect64 = XSelect64 { dst: XReg, cond: XReg, if_nonzero: XReg, if_zero: XReg };
618
619            /// `trapif(addr > bound_ptr - size)` (unsigned)
620            xbc32_bound_trap = XBc32BoundTrap {
621                addr: XReg,
622                bound: XReg,
623                size: u8
624            };
625            /// `trapif(addr > *(bound_ptr + bound_off) - size)` (unsigned)
626            ///
627            /// Note that the `bound_ptr + bound_off` load loads a
628            /// host-native-endian pointer-sized value.
629            xbc32_boundne_trap = XBc32BoundNeTrap {
630                addr: XReg,
631                bound_ptr: XReg,
632                bound_off: u8,
633                size: u8
634            };
635            /// `trapif(addr >= bound_ptr)` (unsigned)
636            xbc32_strict_bound_trap = XBc32StrictBoundTrap {
637                addr: XReg,
638                bound: XReg
639            };
640            /// `trapif(addr >= *(bound_ptr + bound_off))` (unsigned)
641            xbc32_strict_boundne_trap = XBc32StrictBoundNeTrap {
642                addr: XReg,
643                bound_ptr: XReg,
644                bound_off: u8
645            };
646        }
647    };
648}
649
650/// Calls the given macro with each extended opcode.
651#[macro_export]
652macro_rules! for_each_extended_op {
653    ( $macro:ident ) => {
654        $macro! {
655            /// Raise a trap.
656            trap = Trap;
657
658            /// Do nothing.
659            nop = Nop;
660
661            /// A special opcode to halt interpreter execution and yield control
662            /// back to the host.
663            ///
664            /// This opcode results in `DoneReason::CallIndirectHost` where the
665            /// `id` here is shepherded along to the embedder. It's up to the
666            /// embedder to determine what to do with the `id` and the current
667            /// state of registers and the stack.
668            ///
669            /// In Wasmtime this is used to implement interpreter-to-host calls.
670            /// This is modeled as a `call` instruction where the first
671            /// parameter is the native function pointer to invoke and all
672            /// remaining parameters for the native function are in following
673            /// parameter positions (e.g. `x1`, `x2`, ...). The results of the
674            /// host call are then store in `x0`.
675            ///
676            /// Handling this in Wasmtime is done through a "relocation" which
677            /// is resolved at link-time when raw bytecode from Cranelift is
678            /// assembled into the final object that Wasmtime will interpret.
679            call_indirect_host = CallIndirectHost { id: u8 };
680
681            /// Gets the special "fp" register and moves it into `dst`.
682            xmov_fp = XmovFp { dst: XReg };
683
684            /// Gets the special "lr" register and moves it into `dst`.
685            xmov_lr = XmovLr { dst: XReg };
686
687            /// `dst = byteswap(low32(src))`
688            bswap32 = Bswap32 { dst: XReg, src: XReg };
689            /// `dst = byteswap(src)`
690            bswap64 = Bswap64 { dst: XReg, src: XReg };
691
692            /// 32-bit checked unsigned addition: `low32(dst) = low32(src1) +
693            /// low32(src2)`.
694            ///
695            /// The upper 32-bits of `dst` are unmodified. Traps if the addition
696            /// overflows.
697            xadd32_uoverflow_trap = Xadd32UoverflowTrap { operands: BinaryOperands<XReg> };
698
699            /// 64-bit checked unsigned addition: `dst = src1 + src2`.
700            xadd64_uoverflow_trap = Xadd64UoverflowTrap { operands: BinaryOperands<XReg> };
701
702            /// `dst = high64(src1 * src2)` (signed)
703            xmulhi64_s = XMulHi64S { operands: BinaryOperands<XReg> };
704            /// `dst = high64(src1 * src2)` (unsigned)
705            xmulhi64_u = XMulHi64U { operands: BinaryOperands<XReg> };
706
707            /// low32(dst) = if low32(src) == 0 { 0 } else { -1 }
708            xbmask32 = Xbmask32 { dst: XReg, src: XReg };
709            /// dst = if src == 0 { 0 } else { -1 }
710            xbmask64 = Xbmask64 { dst: XReg, src: XReg };
711
712            /// `dst = zext(*(ptr + offset))`
713            xload16be_u64_offset32 = XLoad16BeU64Offset32 { dst: XReg, ptr: XReg, offset: i32 };
714            /// `dst = sext(*(ptr + offset))`
715            xload16be_s64_offset32 = XLoad16BeS64Offset32 { dst: XReg, ptr: XReg, offset: i32 };
716            /// `dst = zext(*(ptr + offset))`
717            xload32be_u64_offset32 = XLoad32BeU64Offset32 { dst: XReg, ptr: XReg, offset: i32 };
718            /// `dst = sext(*(ptr + offset))`
719            xload32be_s64_offset32 = XLoad32BeS64Offset32 { dst: XReg, ptr: XReg, offset: i32 };
720            /// `dst = *(ptr + offset)`
721            xload64be_offset32 = XLoad64BeOffset32 { dst: XReg, ptr: XReg, offset: i32 };
722
723            /// `*(ptr + offset) = low16(src)`
724            xstore16be_offset32 = XStore16BeOffset32 { ptr: XReg, offset: i32, src: XReg };
725            /// `*(ptr + offset) = low32(src)`
726            xstore32be_offset32 = XStore32BeOffset32 { ptr: XReg, offset: i32, src: XReg };
727            /// `*(ptr + offset) = low64(src)`
728            xstore64be_offset32 = XStore64BeOffset32 { ptr: XReg, offset: i32, src: XReg };
729
730            /// `low32(dst) = zext(*(ptr + offset))`
731            fload32be_offset32 = Fload32BeOffset32 { dst: FReg, ptr: XReg, offset: i32 };
732            /// `dst = *(ptr + offset)`
733            fload64be_offset32 = Fload64BeOffset32 { dst: FReg, ptr: XReg, offset: i32 };
734            /// `*(ptr + offset) = low32(src)`
735            fstore32be_offset32 = Fstore32BeOffset32 { ptr: XReg, offset: i32, src: FReg };
736            /// `*(ptr + offset) = src`
737            fstore64be_offset32 = Fstore64BeOffset32 { ptr: XReg, offset: i32, src: FReg };
738
739            /// `low32(dst) = zext(*(ptr + offset))`
740            fload32le_offset32 = Fload32LeOffset32 { dst: FReg, ptr: XReg, offset: i32 };
741            /// `dst = *(ptr + offset)`
742            fload64le_offset32 = Fload64LeOffset32 { dst: FReg, ptr: XReg, offset: i32 };
743            /// `*(ptr + offset) = low32(src)`
744            fstore32le_offset32 = Fstore32LeOffset32 { ptr: XReg, offset: i32, src: FReg };
745            /// `*(ptr + offset) = src`
746            fstore64le_offset32 = Fstore64LeOffset32 { ptr: XReg, offset: i32, src: FReg };
747
748            /// `dst = *(ptr + offset)`
749            vload128le_offset32 = VLoad128Offset32 { dst: VReg, ptr: XReg, offset: i32 };
750            /// `*(ptr + offset) = src`
751            vstore128le_offset32 = Vstore128LeOffset32 { ptr: XReg, offset: i32, src: VReg };
752
753            /// Move between `f` registers.
754            fmov = Fmov { dst: FReg, src: FReg };
755            /// Move between `v` registers.
756            vmov = Vmov { dst: VReg, src: VReg };
757
758            /// `low32(dst) = bitcast low32(src) as i32`
759            bitcast_int_from_float_32 = BitcastIntFromFloat32 { dst: XReg, src: FReg };
760            /// `dst = bitcast src as i64`
761            bitcast_int_from_float_64 = BitcastIntFromFloat64 { dst: XReg, src: FReg };
762            /// `low32(dst) = bitcast low32(src) as f32`
763            bitcast_float_from_int_32 = BitcastFloatFromInt32 { dst: FReg, src: XReg };
764            /// `dst = bitcast src as f64`
765            bitcast_float_from_int_64 = BitcastFloatFromInt64 { dst: FReg, src: XReg };
766
767            /// `low32(dst) = bits`
768            fconst32 = FConst32 { dst: FReg, bits: u32 };
769            /// `dst = bits`
770            fconst64 = FConst64 { dst: FReg, bits: u64 };
771
772            /// `low32(dst) = zext(src1 == src2)`
773            feq32 = Feq32 { dst: XReg, src1: FReg, src2: FReg };
774            /// `low32(dst) = zext(src1 != src2)`
775            fneq32 = Fneq32 { dst: XReg, src1: FReg, src2: FReg };
776            /// `low32(dst) = zext(src1 < src2)`
777            flt32 = Flt32 { dst: XReg, src1: FReg, src2: FReg };
778            /// `low32(dst) = zext(src1 <= src2)`
779            flteq32 = Flteq32 { dst: XReg, src1: FReg, src2: FReg };
780            /// `low32(dst) = zext(src1 == src2)`
781            feq64 = Feq64 { dst: XReg, src1: FReg, src2: FReg };
782            /// `low32(dst) = zext(src1 != src2)`
783            fneq64 = Fneq64 { dst: XReg, src1: FReg, src2: FReg };
784            /// `low32(dst) = zext(src1 < src2)`
785            flt64 = Flt64 { dst: XReg, src1: FReg, src2: FReg };
786            /// `low32(dst) = zext(src1 <= src2)`
787            flteq64 = Flteq64 { dst: XReg, src1: FReg, src2: FReg };
788
789            /// `low32(dst) = low32(cond) ? low32(if_nonzero) : low32(if_zero)`
790            fselect32 = FSelect32 { dst: FReg, cond: XReg, if_nonzero: FReg, if_zero: FReg };
791            /// `dst = low32(cond) ? if_nonzero : if_zero`
792            fselect64 = FSelect64 { dst: FReg, cond: XReg, if_nonzero: FReg, if_zero: FReg };
793
794            /// `low32(dst) = demote(src)`
795            f32_from_f64 = F32FromF64 { dst: FReg, src: FReg };
796            /// `(st) = promote(low32(src))`
797            f64_from_f32 = F64FromF32 { dst: FReg, src: FReg };
798
799            /// `low32(dst) = checked_f32_from_signed(low32(src))`
800            f32_from_x32_s = F32FromX32S { dst: FReg, src: XReg };
801            /// `low32(dst) = checked_f32_from_unsigned(low32(src))`
802            f32_from_x32_u = F32FromX32U { dst: FReg, src: XReg };
803            /// `low32(dst) = checked_f32_from_signed(src)`
804            f32_from_x64_s = F32FromX64S { dst: FReg, src: XReg };
805            /// `low32(dst) = checked_f32_from_unsigned(src)`
806            f32_from_x64_u = F32FromX64U { dst: FReg, src: XReg };
807            /// `dst = checked_f64_from_signed(low32(src))`
808            f64_from_x32_s = F64FromX32S { dst: FReg, src: XReg };
809            /// `dst = checked_f64_from_unsigned(low32(src))`
810            f64_from_x32_u = F64FromX32U { dst: FReg, src: XReg };
811            /// `dst = checked_f64_from_signed(src)`
812            f64_from_x64_s = F64FromX64S { dst: FReg, src: XReg };
813            /// `dst = checked_f64_from_unsigned(src)`
814            f64_from_x64_u = F64FromX64U { dst: FReg, src: XReg };
815
816            /// `low32(dst) = checked_signed_from_f32(low32(src))`
817            x32_from_f32_s = X32FromF32S { dst: XReg, src: FReg };
818            /// `low32(dst) = checked_unsigned_from_f32(low32(src))`
819            x32_from_f32_u = X32FromF32U { dst: XReg, src: FReg };
820            /// `low32(dst) = checked_signed_from_f64(src)`
821            x32_from_f64_s = X32FromF64S { dst: XReg, src: FReg };
822            /// `low32(dst) = checked_unsigned_from_f64(src)`
823            x32_from_f64_u = X32FromF64U { dst: XReg, src: FReg };
824            /// `dst = checked_signed_from_f32(low32(src))`
825            x64_from_f32_s = X64FromF32S { dst: XReg, src: FReg };
826            /// `dst = checked_unsigned_from_f32(low32(src))`
827            x64_from_f32_u = X64FromF32U { dst: XReg, src: FReg };
828            /// `dst = checked_signed_from_f64(src)`
829            x64_from_f64_s = X64FromF64S { dst: XReg, src: FReg };
830            /// `dst = checked_unsigned_from_f64(src)`
831            x64_from_f64_u = X64FromF64U { dst: XReg, src: FReg };
832
833            /// `low32(dst) = saturating_signed_from_f32(low32(src))`
834            x32_from_f32_s_sat = X32FromF32SSat { dst: XReg, src: FReg };
835            /// `low32(dst) = saturating_unsigned_from_f32(low32(src))`
836            x32_from_f32_u_sat = X32FromF32USat { dst: XReg, src: FReg };
837            /// `low32(dst) = saturating_signed_from_f64(src)`
838            x32_from_f64_s_sat = X32FromF64SSat { dst: XReg, src: FReg };
839            /// `low32(dst) = saturating_unsigned_from_f64(src)`
840            x32_from_f64_u_sat = X32FromF64USat { dst: XReg, src: FReg };
841            /// `dst = saturating_signed_from_f32(low32(src))`
842            x64_from_f32_s_sat = X64FromF32SSat { dst: XReg, src: FReg };
843            /// `dst = saturating_unsigned_from_f32(low32(src))`
844            x64_from_f32_u_sat = X64FromF32USat { dst: XReg, src: FReg };
845            /// `dst = saturating_signed_from_f64(src)`
846            x64_from_f64_s_sat = X64FromF64SSat { dst: XReg, src: FReg };
847            /// `dst = saturating_unsigned_from_f64(src)`
848            x64_from_f64_u_sat = X64FromF64USat { dst: XReg, src: FReg };
849
850            /// `low32(dst) = copysign(low32(src1), low32(src2))`
851            fcopysign32 = FCopySign32 { operands: BinaryOperands<FReg> };
852            /// `dst = copysign(src1, src2)`
853            fcopysign64 = FCopySign64 { operands: BinaryOperands<FReg> };
854
855            /// `low32(dst) = low32(src1) + low32(src2)`
856            fadd32 = Fadd32 { operands: BinaryOperands<FReg> };
857            /// `low32(dst) = low32(src1) - low32(src2)`
858            fsub32 = Fsub32 { operands: BinaryOperands<FReg> };
859            /// `low128(dst) = low128(src1) - low128(src2)`
860            vsubf32x4 = Vsubf32x4 { operands: BinaryOperands<VReg> };
861            /// `low32(dst) = low32(src1) * low32(src2)`
862            fmul32 = Fmul32 { operands: BinaryOperands<FReg> };
863            /// `low128(dst) = low128(src1) * low128(src2)`
864            vmulf32x4 = Vmulf32x4 { operands: BinaryOperands<VReg> };
865            /// `low32(dst) = low32(src1) / low32(src2)`
866            fdiv32 = Fdiv32 { operands: BinaryOperands<FReg> };
867            /// `low128(dst) = low128(src1) / low128(src2)`
868            vdivf32x4 = Vdivf32x4 { operands: BinaryOperands<VReg> };
869            /// `low32(dst) = ieee_maximum(low32(src1), low32(src2))`
870            fmaximum32 = Fmaximum32 { operands: BinaryOperands<FReg> };
871            /// `low32(dst) = ieee_minimum(low32(src1), low32(src2))`
872            fminimum32 = Fminimum32 { operands: BinaryOperands<FReg> };
873            /// `low32(dst) = ieee_trunc(low32(src))`
874            ftrunc32 = Ftrunc32 { dst: FReg, src: FReg };
875            /// `low128(dst) = ieee_trunc(low128(src))`
876            vtrunc32x4 = Vtrunc32x4 { dst: VReg, src: VReg };
877            /// `low128(dst) = ieee_trunc(low128(src))`
878            vtrunc64x2 = Vtrunc64x2 { dst: VReg, src: VReg };
879            /// `low32(dst) = ieee_floor(low32(src))`
880            ffloor32 = Ffloor32 { dst: FReg, src: FReg };
881            /// `low128(dst) = ieee_floor(low128(src))`
882            vfloor32x4 = Vfloor32x4 { dst: VReg, src: VReg };
883            /// `low128(dst) = ieee_floor(low128(src))`
884            vfloor64x2 = Vfloor64x2 { dst: VReg, src: VReg };
885            /// `low32(dst) = ieee_ceil(low32(src))`
886            fceil32 = Fceil32 { dst: FReg, src: FReg };
887            /// `low128(dst) = ieee_ceil(low128(src))`
888            vceil32x4 = Vceil32x4 { dst: VReg, src: VReg };
889            /// `low128(dst) = ieee_ceil(low128(src))`
890            vceil64x2 = Vceil64x2 { dst: VReg, src: VReg };
891            /// `low32(dst) = ieee_nearest(low32(src))`
892            fnearest32 = Fnearest32 { dst: FReg, src: FReg };
893            /// `low32(dst) = ieee_sqrt(low32(src))`
894            fsqrt32 = Fsqrt32 { dst: FReg, src: FReg };
895            /// `low32(dst) = ieee_sqrt(low32(src))`
896            vsqrt32x4 = Vsqrt32x4 { dst: VReg, src: VReg };
897            /// `low32(dst) = ieee_sqrt(low32(src))`
898            vsqrt64x2 = Vsqrt64x2 { dst: VReg, src: VReg };
899            /// `low32(dst) = -low32(src)`
900            fneg32 = Fneg32 { dst: FReg, src: FReg };
901            /// `low128(dst) = -low128(src)`
902            vnegf32x4 = Vnegf32x4 { dst: VReg, src: VReg };
903            /// `low32(dst) = |low32(src)|`
904            fabs32 = Fabs32 { dst: FReg, src: FReg };
905
906            /// `dst = src1 + src2`
907            fadd64 = Fadd64 { operands: BinaryOperands<FReg> };
908            /// `dst = src1 - src2`
909            fsub64 = Fsub64 { operands: BinaryOperands<FReg> };
910            /// `dst = src1 * src2`
911            fmul64 = Fmul64 { operands: BinaryOperands<FReg> };
912            /// `dst = src1 / src2`
913            fdiv64 = Fdiv64 { operands: BinaryOperands<FReg> };
914            /// `dst = src1 / src2`
915            vdivf64x2 = VDivF64x2 { operands: BinaryOperands<VReg> };
916            /// `dst = ieee_maximum(src1, src2)`
917            fmaximum64 = Fmaximum64 { operands: BinaryOperands<FReg> };
918            /// `dst = ieee_minimum(src1, src2)`
919            fminimum64 = Fminimum64 { operands: BinaryOperands<FReg> };
920            /// `dst = ieee_trunc(src)`
921            ftrunc64 = Ftrunc64 { dst: FReg, src: FReg };
922            /// `dst = ieee_floor(src)`
923            ffloor64 = Ffloor64 { dst: FReg, src: FReg };
924            /// `dst = ieee_ceil(src)`
925            fceil64 = Fceil64 { dst: FReg, src: FReg };
926            /// `dst = ieee_nearest(src)`
927            fnearest64 = Fnearest64 { dst: FReg, src: FReg };
928            /// `low128(dst) = ieee_nearest(low128(src))`
929            vnearest32x4 = Vnearest32x4 { dst: VReg, src: VReg };
930            /// `low128(dst) = ieee_nearest(low128(src))`
931            vnearest64x2 = Vnearest64x2 { dst: VReg, src: VReg };
932            /// `dst = ieee_sqrt(src)`
933            fsqrt64 = Fsqrt64 { dst: FReg, src: FReg };
934            /// `dst = -src`
935            fneg64 = Fneg64 { dst: FReg, src: FReg };
936            /// `dst = |src|`
937            fabs64 = Fabs64 { dst: FReg, src: FReg };
938
939            /// `dst = imm`
940            vconst128 = Vconst128 { dst: VReg, imm: u128 };
941
942            /// `dst = src1 + src2`
943            vaddi8x16 = VAddI8x16 { operands: BinaryOperands<VReg> };
944            /// `dst = src1 + src2`
945            vaddi16x8 = VAddI16x8 { operands: BinaryOperands<VReg> };
946            /// `dst = src1 + src2`
947            vaddi32x4 = VAddI32x4 { operands: BinaryOperands<VReg> };
948            /// `dst = src1 + src2`
949            vaddi64x2 = VAddI64x2 { operands: BinaryOperands<VReg> };
950            /// `dst = src1 + src2`
951            vaddf32x4 = VAddF32x4 { operands: BinaryOperands<VReg> };
952            /// `dst = src1 + src2`
953            vaddf64x2 = VAddF64x2 { operands: BinaryOperands<VReg> };
954
955            /// `dst = satruating_add(src1, src2)`
956            vaddi8x16_sat = VAddI8x16Sat { operands: BinaryOperands<VReg> };
957            /// `dst = satruating_add(src1, src2)`
958            vaddu8x16_sat = VAddU8x16Sat { operands: BinaryOperands<VReg> };
959            /// `dst = satruating_add(src1, src2)`
960            vaddi16x8_sat = VAddI16x8Sat { operands: BinaryOperands<VReg> };
961            /// `dst = satruating_add(src1, src2)`
962            vaddu16x8_sat = VAddU16x8Sat { operands: BinaryOperands<VReg> };
963
964            /// `dst = [src1[0] + src1[1], ..., src2[6] + src2[7]]`
965            vaddpairwisei16x8_s = VAddpairwiseI16x8S { operands: BinaryOperands<VReg> };
966            /// `dst = [src1[0] + src1[1], ..., src2[2] + src2[3]]`
967            vaddpairwisei32x4_s = VAddpairwiseI32x4S { operands: BinaryOperands<VReg> };
968
969            /// `dst = src1 << src2`
970            vshli8x16 = VShlI8x16 { operands: BinaryOperands<VReg, VReg, XReg> };
971            /// `dst = src1 << src2`
972            vshli16x8 = VShlI16x8 { operands: BinaryOperands<VReg, VReg, XReg> };
973            /// `dst = src1 << src2`
974            vshli32x4 = VShlI32x4 { operands: BinaryOperands<VReg, VReg, XReg> };
975            /// `dst = src1 << src2`
976            vshli64x2 = VShlI64x2 { operands: BinaryOperands<VReg, VReg, XReg> };
977            /// `dst = src1 >> src2` (signed)
978            vshri8x16_s = VShrI8x16S { operands: BinaryOperands<VReg, VReg, XReg> };
979            /// `dst = src1 >> src2` (signed)
980            vshri16x8_s = VShrI16x8S { operands: BinaryOperands<VReg, VReg, XReg> };
981            /// `dst = src1 >> src2` (signed)
982            vshri32x4_s = VShrI32x4S { operands: BinaryOperands<VReg, VReg, XReg> };
983            /// `dst = src1 >> src2` (signed)
984            vshri64x2_s = VShrI64x2S { operands: BinaryOperands<VReg, VReg, XReg> };
985            /// `dst = src1 >> src2` (unsigned)
986            vshri8x16_u = VShrI8x16U { operands: BinaryOperands<VReg, VReg, XReg> };
987            /// `dst = src1 >> src2` (unsigned)
988            vshri16x8_u = VShrI16x8U { operands: BinaryOperands<VReg, VReg, XReg> };
989            /// `dst = src1 >> src2` (unsigned)
990            vshri32x4_u = VShrI32x4U { operands: BinaryOperands<VReg, VReg, XReg> };
991            /// `dst = src1 >> src2` (unsigned)
992            vshri64x2_u = VShrI64x2U { operands: BinaryOperands<VReg, VReg, XReg> };
993
994            /// `dst = splat(low8(src))`
995            vsplatx8 = VSplatX8 { dst: VReg, src: XReg };
996            /// `dst = splat(low16(src))`
997            vsplatx16 = VSplatX16 { dst: VReg, src: XReg };
998            /// `dst = splat(low32(src))`
999            vsplatx32 = VSplatX32 { dst: VReg, src: XReg };
1000            /// `dst = splat(src)`
1001            vsplatx64 = VSplatX64 { dst: VReg, src: XReg };
1002            /// `dst = splat(low32(src))`
1003            vsplatf32 = VSplatF32 { dst: VReg, src: FReg };
1004            /// `dst = splat(src)`
1005            vsplatf64 = VSplatF64 { dst: VReg, src: FReg };
1006
1007            /// Load the 64-bit source as i8x8 and sign-extend to i16x8.
1008            vload8x8_s_offset32 = VLoad8x8SOffset32 { dst: VReg, ptr: XReg, offset: i32 };
1009            /// Load the 64-bit source as u8x8 and zero-extend to i16x8.
1010            vload8x8_u_offset32 = VLoad8x8UOffset32 { dst: VReg, ptr: XReg, offset: i32 };
1011            /// Load the 64-bit source as i16x4 and sign-extend to i32x4.
1012            vload16x4le_s_offset32 = VLoad16x4LeSOffset32 { dst: VReg, ptr: XReg, offset: i32 };
1013            /// Load the 64-bit source as u16x4 and zero-extend to i32x4.
1014            vload16x4le_u_offset32 = VLoad16x4LeUOffset32 { dst: VReg, ptr: XReg, offset: i32 };
1015            /// Load the 64-bit source as i32x2 and sign-extend to i64x2.
1016            vload32x2le_s_offset32 = VLoad32x2LeSOffset32 { dst: VReg, ptr: XReg, offset: i32 };
1017            /// Load the 64-bit source as u32x2 and zero-extend to i64x2.
1018            vload32x2le_u_offset32 = VLoad32x2LeUOffset32 { dst: VReg, ptr: XReg, offset: i32 };
1019
1020            /// `dst = src1 & src2`
1021            vband128 = VBand128 { operands: BinaryOperands<VReg> };
1022            /// `dst = src1 | src2`
1023            vbor128 = VBor128 { operands: BinaryOperands<VReg> };
1024            /// `dst = src1 ^ src2`
1025            vbxor128 = VBxor128 { operands: BinaryOperands<VReg> };
1026            /// `dst = !src1`
1027            vbnot128 = VBnot128 { dst: VReg, src: VReg };
1028            /// `dst = (c & x) | (!c & y)`
1029            vbitselect128 = VBitselect128 { dst: VReg, c: VReg, x: VReg, y: VReg };
1030            /// Collect high bits of each lane into the low 32-bits of the
1031            /// destination.
1032            vbitmask8x16 = Vbitmask8x16 { dst: XReg, src: VReg };
1033            /// Collect high bits of each lane into the low 32-bits of the
1034            /// destination.
1035            vbitmask16x8 = Vbitmask16x8 { dst: XReg, src: VReg };
1036            /// Collect high bits of each lane into the low 32-bits of the
1037            /// destination.
1038            vbitmask32x4 = Vbitmask32x4 { dst: XReg, src: VReg };
1039            /// Collect high bits of each lane into the low 32-bits of the
1040            /// destination.
1041            vbitmask64x2 = Vbitmask64x2 { dst: XReg, src: VReg };
1042            /// Store whether all lanes are nonzero in `dst`.
1043            valltrue8x16 = Valltrue8x16 { dst: XReg, src: VReg };
1044            /// Store whether all lanes are nonzero in `dst`.
1045            valltrue16x8 = Valltrue16x8 { dst: XReg, src: VReg };
1046            /// Store whether all lanes are nonzero in `dst`.
1047            valltrue32x4 = Valltrue32x4 { dst: XReg, src: VReg };
1048            /// Store whether any lanes are nonzero in `dst`.
1049            valltrue64x2 = Valltrue64x2 { dst: XReg, src: VReg };
1050            /// Store whether any lanes are nonzero in `dst`.
1051            vanytrue8x16 = Vanytrue8x16 { dst: XReg, src: VReg };
1052            /// Store whether any lanes are nonzero in `dst`.
1053            vanytrue16x8 = Vanytrue16x8 { dst: XReg, src: VReg };
1054            /// Store whether any lanes are nonzero in `dst`.
1055            vanytrue32x4 = Vanytrue32x4 { dst: XReg, src: VReg };
1056            /// Store whether any lanes are nonzero in `dst`.
1057            vanytrue64x2 = Vanytrue64x2 { dst: XReg, src: VReg };
1058
1059            /// Int-to-float conversion (same as `f32_from_x32_s`)
1060            vf32x4_from_i32x4_s = VF32x4FromI32x4S { dst: VReg, src: VReg };
1061            /// Int-to-float conversion (same as `f32_from_x32_u`)
1062            vf32x4_from_i32x4_u = VF32x4FromI32x4U { dst: VReg, src: VReg };
1063            /// Int-to-float conversion (same as `f64_from_x64_s`)
1064            vf64x2_from_i64x2_s = VF64x2FromI64x2S { dst: VReg, src: VReg };
1065            /// Int-to-float conversion (same as `f64_from_x64_u`)
1066            vf64x2_from_i64x2_u = VF64x2FromI64x2U { dst: VReg, src: VReg };
1067            /// Float-to-int conversion (same as `x32_from_f32_s`
1068            vi32x4_from_f32x4_s = VI32x4FromF32x4S { dst: VReg, src: VReg };
1069            /// Float-to-int conversion (same as `x32_from_f32_u`
1070            vi32x4_from_f32x4_u = VI32x4FromF32x4U { dst: VReg, src: VReg };
1071            /// Float-to-int conversion (same as `x64_from_f64_s`
1072            vi64x2_from_f64x2_s = VI64x2FromF64x2S { dst: VReg, src: VReg };
1073            /// Float-to-int conversion (same as `x64_from_f64_u`
1074            vi64x2_from_f64x2_u = VI64x2FromF64x2U { dst: VReg, src: VReg };
1075
1076            /// Widens the low lanes of the input vector, as signed, to twice
1077            /// the width.
1078            vwidenlow8x16_s = VWidenLow8x16S { dst: VReg, src: VReg };
1079            /// Widens the low lanes of the input vector, as unsigned, to twice
1080            /// the width.
1081            vwidenlow8x16_u = VWidenLow8x16U { dst: VReg, src: VReg };
1082            /// Widens the low lanes of the input vector, as signed, to twice
1083            /// the width.
1084            vwidenlow16x8_s = VWidenLow16x8S { dst: VReg, src: VReg };
1085            /// Widens the low lanes of the input vector, as unsigned, to twice
1086            /// the width.
1087            vwidenlow16x8_u = VWidenLow16x8U { dst: VReg, src: VReg };
1088            /// Widens the low lanes of the input vector, as signed, to twice
1089            /// the width.
1090            vwidenlow32x4_s = VWidenLow32x4S { dst: VReg, src: VReg };
1091            /// Widens the low lanes of the input vector, as unsigned, to twice
1092            /// the width.
1093            vwidenlow32x4_u = VWidenLow32x4U { dst: VReg, src: VReg };
1094            /// Widens the high lanes of the input vector, as signed, to twice
1095            /// the width.
1096            vwidenhigh8x16_s = VWidenHigh8x16S { dst: VReg, src: VReg };
1097            /// Widens the high lanes of the input vector, as unsigned, to twice
1098            /// the width.
1099            vwidenhigh8x16_u = VWidenHigh8x16U { dst: VReg, src: VReg };
1100            /// Widens the high lanes of the input vector, as signed, to twice
1101            /// the width.
1102            vwidenhigh16x8_s = VWidenHigh16x8S { dst: VReg, src: VReg };
1103            /// Widens the high lanes of the input vector, as unsigned, to twice
1104            /// the width.
1105            vwidenhigh16x8_u = VWidenHigh16x8U { dst: VReg, src: VReg };
1106            /// Widens the high lanes of the input vector, as signed, to twice
1107            /// the width.
1108            vwidenhigh32x4_s = VWidenHigh32x4S { dst: VReg, src: VReg };
1109            /// Widens the high lanes of the input vector, as unsigned, to twice
1110            /// the width.
1111            vwidenhigh32x4_u = VWidenHigh32x4U { dst: VReg, src: VReg };
1112
1113            /// Narrows the two 16x8 vectors, assuming all input lanes are
1114            /// signed, to half the width. Narrowing is signed and saturating.
1115            vnarrow16x8_s = Vnarrow16x8S { operands: BinaryOperands<VReg> };
1116            /// Narrows the two 16x8 vectors, assuming all input lanes are
1117            /// signed, to half the width. Narrowing is unsigned and saturating.
1118            vnarrow16x8_u = Vnarrow16x8U { operands: BinaryOperands<VReg> };
1119            /// Narrows the two 32x4 vectors, assuming all input lanes are
1120            /// signed, to half the width. Narrowing is signed and saturating.
1121            vnarrow32x4_s = Vnarrow32x4S { operands: BinaryOperands<VReg> };
1122            /// Narrows the two 32x4 vectors, assuming all input lanes are
1123            /// signed, to half the width. Narrowing is unsigned and saturating.
1124            vnarrow32x4_u = Vnarrow32x4U { operands: BinaryOperands<VReg> };
1125            /// Narrows the two 64x2 vectors, assuming all input lanes are
1126            /// signed, to half the width. Narrowing is signed and saturating.
1127            vnarrow64x2_s = Vnarrow64x2S { operands: BinaryOperands<VReg> };
1128            /// Narrows the two 64x2 vectors, assuming all input lanes are
1129            /// signed, to half the width. Narrowing is unsigned and saturating.
1130            vnarrow64x2_u = Vnarrow64x2U { operands: BinaryOperands<VReg> };
1131            /// Narrows the two 64x2 vectors, assuming all input lanes are
1132            /// unsigned, to half the width. Narrowing is unsigned and saturating.
1133            vunarrow64x2_u = Vunarrow64x2U { operands: BinaryOperands<VReg> };
1134            /// Promotes the low two lanes of the f32x4 input to f64x2.
1135            vfpromotelow = VFpromoteLow { dst: VReg, src: VReg };
1136            /// Demotes the two f64x2 lanes to f32x2 and then extends with two
1137            /// more zero lanes.
1138            vfdemote = VFdemote { dst: VReg, src: VReg };
1139
1140            /// `dst = src1 - src2`
1141            vsubi8x16 = VSubI8x16 { operands: BinaryOperands<VReg> };
1142            /// `dst = src1 - src2`
1143            vsubi16x8 = VSubI16x8 { operands: BinaryOperands<VReg> };
1144            /// `dst = src1 - src2`
1145            vsubi32x4 = VSubI32x4 { operands: BinaryOperands<VReg> };
1146            /// `dst = src1 - src2`
1147            vsubi64x2 = VSubI64x2 { operands: BinaryOperands<VReg> };
1148            /// `dst = src1 - src2`
1149            vsubf64x2 = VSubF64x2 { operands: BinaryOperands<VReg> };
1150
1151            /// `dst = saturating_sub(src1, src2)`
1152            vsubi8x16_sat = VSubI8x16Sat { operands: BinaryOperands<VReg> };
1153            /// `dst = saturating_sub(src1, src2)`
1154            vsubu8x16_sat = VSubU8x16Sat { operands: BinaryOperands<VReg> };
1155            /// `dst = saturating_sub(src1, src2)`
1156            vsubi16x8_sat = VSubI16x8Sat { operands: BinaryOperands<VReg> };
1157            /// `dst = saturating_sub(src1, src2)`
1158            vsubu16x8_sat = VSubU16x8Sat { operands: BinaryOperands<VReg> };
1159
1160            /// `dst = src1 * src2`
1161            vmuli8x16 = VMulI8x16 { operands: BinaryOperands<VReg> };
1162            /// `dst = src1 * src2`
1163            vmuli16x8 = VMulI16x8 { operands: BinaryOperands<VReg> };
1164            /// `dst = src1 * src2`
1165            vmuli32x4 = VMulI32x4 { operands: BinaryOperands<VReg> };
1166            /// `dst = src1 * src2`
1167            vmuli64x2 = VMulI64x2 { operands: BinaryOperands<VReg> };
1168            /// `dst = src1 * src2`
1169            vmulf64x2 = VMulF64x2 { operands: BinaryOperands<VReg> };
1170
1171            /// `dst = signed_saturate(src1 * src2 + (1 << (Q - 1)) >> Q)`
1172            vqmulrsi16x8 = VQmulrsI16x8 { operands: BinaryOperands<VReg> };
1173
1174            /// `dst = count_ones(src)`
1175            vpopcnt8x16 = VPopcnt8x16 { dst: VReg, src: VReg };
1176
1177            /// `low32(dst) = zext(src[lane])`
1178            xextractv8x16 = XExtractV8x16 { dst: XReg, src: VReg, lane: u8 };
1179            /// `low32(dst) = zext(src[lane])`
1180            xextractv16x8 = XExtractV16x8 { dst: XReg, src: VReg, lane: u8 };
1181            /// `low32(dst) = src[lane]`
1182            xextractv32x4 = XExtractV32x4 { dst: XReg, src: VReg, lane: u8 };
1183            /// `dst = src[lane]`
1184            xextractv64x2 = XExtractV64x2 { dst: XReg, src: VReg, lane: u8 };
1185            /// `low32(dst) = src[lane]`
1186            fextractv32x4 = FExtractV32x4 { dst: FReg, src: VReg, lane: u8 };
1187            /// `dst = src[lane]`
1188            fextractv64x2 = FExtractV64x2 { dst: FReg, src: VReg, lane: u8 };
1189
1190            /// `dst = src1; dst[lane] = src2`
1191            vinsertx8 = VInsertX8 { operands: BinaryOperands<VReg, VReg, XReg>, lane: u8 };
1192            /// `dst = src1; dst[lane] = src2`
1193            vinsertx16 = VInsertX16 { operands: BinaryOperands<VReg, VReg, XReg>, lane: u8 };
1194            /// `dst = src1; dst[lane] = src2`
1195            vinsertx32 = VInsertX32 { operands: BinaryOperands<VReg, VReg, XReg>, lane: u8 };
1196            /// `dst = src1; dst[lane] = src2`
1197            vinsertx64 = VInsertX64 { operands: BinaryOperands<VReg, VReg, XReg>, lane: u8 };
1198            /// `dst = src1; dst[lane] = src2`
1199            vinsertf32 = VInsertF32 { operands: BinaryOperands<VReg, VReg, FReg>, lane: u8 };
1200            /// `dst = src1; dst[lane] = src2`
1201            vinsertf64 = VInsertF64 { operands: BinaryOperands<VReg, VReg, FReg>, lane: u8 };
1202
1203            /// `dst = src == dst`
1204            veq8x16 = Veq8x16 { operands: BinaryOperands<VReg> };
1205            /// `dst = src != dst`
1206            vneq8x16 = Vneq8x16 { operands: BinaryOperands<VReg> };
1207            /// `dst = src < dst` (signed)
1208            vslt8x16 = Vslt8x16 { operands: BinaryOperands<VReg> };
1209            /// `dst = src <= dst` (signed)
1210            vslteq8x16 = Vslteq8x16 { operands: BinaryOperands<VReg> };
1211            /// `dst = src < dst` (unsigned)
1212            vult8x16 = Vult8x16 { operands: BinaryOperands<VReg> };
1213            /// `dst = src <= dst` (unsigned)
1214            vulteq8x16 = Vulteq8x16 { operands: BinaryOperands<VReg> };
1215            /// `dst = src == dst`
1216            veq16x8 = Veq16x8 { operands: BinaryOperands<VReg> };
1217            /// `dst = src != dst`
1218            vneq16x8 = Vneq16x8 { operands: BinaryOperands<VReg> };
1219            /// `dst = src < dst` (signed)
1220            vslt16x8 = Vslt16x8 { operands: BinaryOperands<VReg> };
1221            /// `dst = src <= dst` (signed)
1222            vslteq16x8 = Vslteq16x8 { operands: BinaryOperands<VReg> };
1223            /// `dst = src < dst` (unsigned)
1224            vult16x8 = Vult16x8 { operands: BinaryOperands<VReg> };
1225            /// `dst = src <= dst` (unsigned)
1226            vulteq16x8 = Vulteq16x8 { operands: BinaryOperands<VReg> };
1227            /// `dst = src == dst`
1228            veq32x4 = Veq32x4 { operands: BinaryOperands<VReg> };
1229            /// `dst = src != dst`
1230            vneq32x4 = Vneq32x4 { operands: BinaryOperands<VReg> };
1231            /// `dst = src < dst` (signed)
1232            vslt32x4 = Vslt32x4 { operands: BinaryOperands<VReg> };
1233            /// `dst = src <= dst` (signed)
1234            vslteq32x4 = Vslteq32x4 { operands: BinaryOperands<VReg> };
1235            /// `dst = src < dst` (unsigned)
1236            vult32x4 = Vult32x4 { operands: BinaryOperands<VReg> };
1237            /// `dst = src <= dst` (unsigned)
1238            vulteq32x4 = Vulteq32x4 { operands: BinaryOperands<VReg> };
1239            /// `dst = src == dst`
1240            veq64x2 = Veq64x2 { operands: BinaryOperands<VReg> };
1241            /// `dst = src != dst`
1242            vneq64x2 = Vneq64x2 { operands: BinaryOperands<VReg> };
1243            /// `dst = src < dst` (signed)
1244            vslt64x2 = Vslt64x2 { operands: BinaryOperands<VReg> };
1245            /// `dst = src <= dst` (signed)
1246            vslteq64x2 = Vslteq64x2 { operands: BinaryOperands<VReg> };
1247            /// `dst = src < dst` (unsigned)
1248            vult64x2 = Vult64x2 { operands: BinaryOperands<VReg> };
1249            /// `dst = src <= dst` (unsigned)
1250            vulteq64x2 = Vulteq64x2 { operands: BinaryOperands<VReg> };
1251
1252            /// `dst = -src`
1253            vneg8x16 = Vneg8x16 { dst: VReg, src: VReg };
1254            /// `dst = -src`
1255            vneg16x8 = Vneg16x8 { dst: VReg, src: VReg };
1256            /// `dst = -src`
1257            vneg32x4 = Vneg32x4 { dst: VReg, src: VReg };
1258            /// `dst = -src`
1259            vneg64x2 = Vneg64x2 { dst: VReg, src: VReg };
1260            /// `dst = -src`
1261            vnegf64x2 = VnegF64x2 { dst: VReg, src: VReg };
1262
1263            /// `dst = min(src1, src2)` (signed)
1264            vmin8x16_s = Vmin8x16S { operands: BinaryOperands<VReg> };
1265            /// `dst = min(src1, src2)` (unsigned)
1266            vmin8x16_u = Vmin8x16U { operands: BinaryOperands<VReg> };
1267            /// `dst = min(src1, src2)` (signed)
1268            vmin16x8_s = Vmin16x8S { operands: BinaryOperands<VReg> };
1269            /// `dst = min(src1, src2)` (unsigned)
1270            vmin16x8_u = Vmin16x8U { operands: BinaryOperands<VReg> };
1271            /// `dst = max(src1, src2)` (signed)
1272            vmax8x16_s = Vmax8x16S { operands: BinaryOperands<VReg> };
1273            /// `dst = max(src1, src2)` (unsigned)
1274            vmax8x16_u = Vmax8x16U { operands: BinaryOperands<VReg> };
1275            /// `dst = max(src1, src2)` (signed)
1276            vmax16x8_s = Vmax16x8S { operands: BinaryOperands<VReg> };
1277            /// `dst = max(src1, src2)` (unsigned)
1278            vmax16x8_u = Vmax16x8U { operands: BinaryOperands<VReg> };
1279
1280            /// `dst = min(src1, src2)` (signed)
1281            vmin32x4_s = Vmin32x4S { operands: BinaryOperands<VReg> };
1282            /// `dst = min(src1, src2)` (unsigned)
1283            vmin32x4_u = Vmin32x4U { operands: BinaryOperands<VReg> };
1284            /// `dst = max(src1, src2)` (signed)
1285            vmax32x4_s = Vmax32x4S { operands: BinaryOperands<VReg> };
1286            /// `dst = max(src1, src2)` (unsigned)
1287            vmax32x4_u = Vmax32x4U { operands: BinaryOperands<VReg> };
1288
1289            /// `dst = |src|`
1290            vabs8x16 = Vabs8x16 { dst: VReg, src: VReg };
1291            /// `dst = |src|`
1292            vabs16x8 = Vabs16x8 { dst: VReg, src: VReg };
1293            /// `dst = |src|`
1294            vabs32x4 = Vabs32x4 { dst: VReg, src: VReg };
1295            /// `dst = |src|`
1296            vabs64x2 = Vabs64x2 { dst: VReg, src: VReg };
1297
1298            /// `dst = |src|`
1299            vabsf32x4 = Vabsf32x4 { dst: VReg, src: VReg };
1300            /// `dst = |src|`
1301            vabsf64x2 = Vabsf64x2 { dst: VReg, src: VReg };
1302            /// `dst = ieee_maximum(src1, src2)`
1303            vmaximumf32x4 = Vmaximumf32x4 { operands: BinaryOperands<VReg> };
1304            /// `dst = ieee_maximum(src1, src2)`
1305            vmaximumf64x2 = Vmaximumf64x2 { operands: BinaryOperands<VReg> };
1306            /// `dst = ieee_minimum(src1, src2)`
1307            vminimumf32x4 = Vminimumf32x4 { operands: BinaryOperands<VReg> };
1308            /// `dst = ieee_minimum(src1, src2)`
1309            vminimumf64x2 = Vminimumf64x2 { operands: BinaryOperands<VReg> };
1310
1311            /// `dst = shuffle(src1, src2, mask)`
1312            vshuffle = VShuffle { dst: VReg, src1: VReg, src2: VReg, mask: u128 };
1313
1314            /// `dst = swizzle(src1, src2)`
1315            vswizzlei8x16 = Vswizzlei8x16 { operands: BinaryOperands<VReg> };
1316
1317            /// `dst = (src1 + src2 + 1) // 2`
1318            vavground8x16 = Vavground8x16 { operands: BinaryOperands<VReg> };
1319            /// `dst = (src1 + src2 + 1) // 2`
1320            vavground16x8 = Vavground16x8 { operands: BinaryOperands<VReg> };
1321
1322            /// `dst = src == dst`
1323            veqf32x4 = VeqF32x4 { operands: BinaryOperands<VReg> };
1324            /// `dst = src != dst`
1325            vneqf32x4 = VneqF32x4 { operands: BinaryOperands<VReg> };
1326            /// `dst = src < dst`
1327            vltf32x4 = VltF32x4 { operands: BinaryOperands<VReg> };
1328            /// `dst = src <= dst`
1329            vlteqf32x4 = VlteqF32x4 { operands: BinaryOperands<VReg> };
1330            /// `dst = src == dst`
1331            veqf64x2 = VeqF64x2 { operands: BinaryOperands<VReg> };
1332            /// `dst = src != dst`
1333            vneqf64x2 = VneqF64x2 { operands: BinaryOperands<VReg> };
1334            /// `dst = src < dst`
1335            vltf64x2 = VltF64x2 { operands: BinaryOperands<VReg> };
1336            /// `dst = src <= dst`
1337            vlteqf64x2 = VlteqF64x2 { operands: BinaryOperands<VReg> };
1338
1339            /// `dst = ieee_fma(a, b, c)`
1340            vfma32x4 = Vfma32x4 { dst: VReg, a: VReg, b: VReg, c: VReg };
1341            /// `dst = ieee_fma(a, b, c)`
1342            vfma64x2 = Vfma64x2 { dst: VReg, a: VReg, b: VReg, c: VReg };
1343
1344            /// `dst = low32(cond) ? if_nonzero : if_zero`
1345            vselect = Vselect { dst: VReg, cond: XReg, if_nonzero: VReg, if_zero: VReg };
1346
1347            /// `dst_hi:dst_lo = lhs_hi:lhs_lo + rhs_hi:rhs_lo`
1348            xadd128 = Xadd128 {
1349                dst_lo: XReg,
1350                dst_hi: XReg,
1351                lhs_lo: XReg,
1352                lhs_hi: XReg,
1353                rhs_lo: XReg,
1354                rhs_hi: XReg
1355            };
1356            /// `dst_hi:dst_lo = lhs_hi:lhs_lo - rhs_hi:rhs_lo`
1357            xsub128 = Xsub128 {
1358                dst_lo: XReg,
1359                dst_hi: XReg,
1360                lhs_lo: XReg,
1361                lhs_hi: XReg,
1362                rhs_lo: XReg,
1363                rhs_hi: XReg
1364            };
1365            /// `dst_hi:dst_lo = sext(lhs) * sext(rhs)`
1366            xwidemul64_s = Xwidemul64S {
1367                dst_lo: XReg,
1368                dst_hi: XReg,
1369                lhs: XReg,
1370                rhs: XReg
1371            };
1372            /// `dst_hi:dst_lo = zext(lhs) * zext(rhs)`
1373            xwidemul64_u = Xwidemul64U {
1374                dst_lo: XReg,
1375                dst_hi: XReg,
1376                lhs: XReg,
1377                rhs: XReg
1378            };
1379        }
1380    };
1381}
1382
1383#[cfg(feature = "decode")]
1384pub mod decode;
1385#[cfg(feature = "disas")]
1386pub mod disas;
1387#[cfg(feature = "encode")]
1388pub mod encode;
1389#[cfg(feature = "interp")]
1390pub mod interp;
1391#[cfg(feature = "profile")]
1392pub mod profile;
1393#[cfg(all(not(feature = "profile"), feature = "interp"))]
1394mod profile_disabled;
1395#[cfg(all(not(feature = "profile"), feature = "interp"))]
1396use profile_disabled as profile;
1397
1398pub mod regs;
1399pub use regs::*;
1400
1401pub mod imms;
1402pub use imms::*;
1403
1404pub mod op;
1405pub use op::*;
1406
1407pub mod opcode;
1408pub use opcode::*;
1409
1410#[cfg(any(feature = "encode", feature = "decode"))]
1411pub(crate) unsafe fn unreachable_unchecked<T>() -> T {
1412    #[cfg(debug_assertions)]
1413    unreachable!();
1414
1415    #[cfg(not(debug_assertions))]
1416    unsafe {
1417        core::hint::unreachable_unchecked()
1418    }
1419}