1 // Copyright 2015 The Go Authors. All rights reserved.
2 // Use of this source code is governed by a BSD-style
3 // license that can be found in the LICENSE file.
7 // Generic opcodes typically specify a width. The inputs and outputs
8 // of that op are the given number of bits wide. There is no notion of
9 // "sign", so Add32 can be used both for signed and unsigned 32-bit
12 // Signed/unsigned is explicit with the extension ops
13 // (SignExt*/ZeroExt*) and implicit as the arg to some opcodes
14 // (e.g. the second argument to shifts is unsigned). If not mentioned,
15 // all args take signed inputs, or don't care whether their inputs
16 // are signed or unsigned.
18 var genericOps = []opData{
20 // Types must be consistent with Go typing. Add, for example, must take two values
21 // of the same type and produces that same type.
22 {name: "Add8", argLength: 2, commutative: true}, // arg0 + arg1
23 {name: "Add16", argLength: 2, commutative: true},
24 {name: "Add32", argLength: 2, commutative: true},
25 {name: "Add64", argLength: 2, commutative: true},
26 {name: "AddPtr", argLength: 2}, // For address calculations. arg0 is a pointer and arg1 is an int.
27 {name: "Add32F", argLength: 2, commutative: true},
28 {name: "Add64F", argLength: 2, commutative: true},
30 {name: "Sub8", argLength: 2}, // arg0 - arg1
31 {name: "Sub16", argLength: 2},
32 {name: "Sub32", argLength: 2},
33 {name: "Sub64", argLength: 2},
34 {name: "SubPtr", argLength: 2},
35 {name: "Sub32F", argLength: 2},
36 {name: "Sub64F", argLength: 2},
38 {name: "Mul8", argLength: 2, commutative: true}, // arg0 * arg1
39 {name: "Mul16", argLength: 2, commutative: true},
40 {name: "Mul32", argLength: 2, commutative: true},
41 {name: "Mul64", argLength: 2, commutative: true},
42 {name: "Mul32F", argLength: 2, commutative: true},
43 {name: "Mul64F", argLength: 2, commutative: true},
45 {name: "Div32F", argLength: 2}, // arg0 / arg1
46 {name: "Div64F", argLength: 2},
48 {name: "Hmul32", argLength: 2, commutative: true},
49 {name: "Hmul32u", argLength: 2, commutative: true},
50 {name: "Hmul64", argLength: 2, commutative: true},
51 {name: "Hmul64u", argLength: 2, commutative: true},
53 {name: "Mul32uhilo", argLength: 2, typ: "(UInt32,UInt32)", commutative: true}, // arg0 * arg1, returns (hi, lo)
54 {name: "Mul64uhilo", argLength: 2, typ: "(UInt64,UInt64)", commutative: true}, // arg0 * arg1, returns (hi, lo)
56 {name: "Mul32uover", argLength: 2, typ: "(UInt32,Bool)", commutative: true}, // Let x = arg0*arg1 (full 32x32-> 64 unsigned multiply), returns (uint32(x), (uint32(x) != x))
57 {name: "Mul64uover", argLength: 2, typ: "(UInt64,Bool)", commutative: true}, // Let x = arg0*arg1 (full 64x64->128 unsigned multiply), returns (uint64(x), (uint64(x) != x))
59 // Weird special instructions for use in the strength reduction of divides.
60 // These ops compute unsigned (arg0 + arg1) / 2, correct to all
61 // 32/64 bits, even when the intermediate result of the add has 33/65 bits.
62 // These ops can assume arg0 >= arg1.
63 // Note: these ops aren't commutative!
64 {name: "Avg32u", argLength: 2, typ: "UInt32"}, // 32-bit platforms only
65 {name: "Avg64u", argLength: 2, typ: "UInt64"}, // 64-bit platforms only
67 // For Div16, Div32 and Div64, AuxInt non-zero means that the divisor has been proved to be not -1
68 // or that the dividend is not the most negative value.
69 {name: "Div8", argLength: 2}, // arg0 / arg1, signed
70 {name: "Div8u", argLength: 2}, // arg0 / arg1, unsigned
71 {name: "Div16", argLength: 2, aux: "Bool"},
72 {name: "Div16u", argLength: 2},
73 {name: "Div32", argLength: 2, aux: "Bool"},
74 {name: "Div32u", argLength: 2},
75 {name: "Div64", argLength: 2, aux: "Bool"},
76 {name: "Div64u", argLength: 2},
77 {name: "Div128u", argLength: 3}, // arg0:arg1 / arg2 (128-bit divided by 64-bit), returns (q, r)
79 // For Mod16, Mod32 and Mod64, AuxInt non-zero means that the divisor has been proved to be not -1.
80 {name: "Mod8", argLength: 2}, // arg0 % arg1, signed
81 {name: "Mod8u", argLength: 2}, // arg0 % arg1, unsigned
82 {name: "Mod16", argLength: 2, aux: "Bool"},
83 {name: "Mod16u", argLength: 2},
84 {name: "Mod32", argLength: 2, aux: "Bool"},
85 {name: "Mod32u", argLength: 2},
86 {name: "Mod64", argLength: 2, aux: "Bool"},
87 {name: "Mod64u", argLength: 2},
89 {name: "And8", argLength: 2, commutative: true}, // arg0 & arg1
90 {name: "And16", argLength: 2, commutative: true},
91 {name: "And32", argLength: 2, commutative: true},
92 {name: "And64", argLength: 2, commutative: true},
94 {name: "Or8", argLength: 2, commutative: true}, // arg0 | arg1
95 {name: "Or16", argLength: 2, commutative: true},
96 {name: "Or32", argLength: 2, commutative: true},
97 {name: "Or64", argLength: 2, commutative: true},
99 {name: "Xor8", argLength: 2, commutative: true}, // arg0 ^ arg1
100 {name: "Xor16", argLength: 2, commutative: true},
101 {name: "Xor32", argLength: 2, commutative: true},
102 {name: "Xor64", argLength: 2, commutative: true},
104 // For shifts, AxB means the shifted value has A bits and the shift amount has B bits.
105 // Shift amounts are considered unsigned.
106 // If arg1 is known to be nonnegative and less than the number of bits in arg0,
107 // then auxInt may be set to 1.
108 // This enables better code generation on some platforms.
109 {name: "Lsh8x8", argLength: 2, aux: "Bool"}, // arg0 << arg1
110 {name: "Lsh8x16", argLength: 2, aux: "Bool"},
111 {name: "Lsh8x32", argLength: 2, aux: "Bool"},
112 {name: "Lsh8x64", argLength: 2, aux: "Bool"},
113 {name: "Lsh16x8", argLength: 2, aux: "Bool"},
114 {name: "Lsh16x16", argLength: 2, aux: "Bool"},
115 {name: "Lsh16x32", argLength: 2, aux: "Bool"},
116 {name: "Lsh16x64", argLength: 2, aux: "Bool"},
117 {name: "Lsh32x8", argLength: 2, aux: "Bool"},
118 {name: "Lsh32x16", argLength: 2, aux: "Bool"},
119 {name: "Lsh32x32", argLength: 2, aux: "Bool"},
120 {name: "Lsh32x64", argLength: 2, aux: "Bool"},
121 {name: "Lsh64x8", argLength: 2, aux: "Bool"},
122 {name: "Lsh64x16", argLength: 2, aux: "Bool"},
123 {name: "Lsh64x32", argLength: 2, aux: "Bool"},
124 {name: "Lsh64x64", argLength: 2, aux: "Bool"},
126 {name: "Rsh8x8", argLength: 2, aux: "Bool"}, // arg0 >> arg1, signed
127 {name: "Rsh8x16", argLength: 2, aux: "Bool"},
128 {name: "Rsh8x32", argLength: 2, aux: "Bool"},
129 {name: "Rsh8x64", argLength: 2, aux: "Bool"},
130 {name: "Rsh16x8", argLength: 2, aux: "Bool"},
131 {name: "Rsh16x16", argLength: 2, aux: "Bool"},
132 {name: "Rsh16x32", argLength: 2, aux: "Bool"},
133 {name: "Rsh16x64", argLength: 2, aux: "Bool"},
134 {name: "Rsh32x8", argLength: 2, aux: "Bool"},
135 {name: "Rsh32x16", argLength: 2, aux: "Bool"},
136 {name: "Rsh32x32", argLength: 2, aux: "Bool"},
137 {name: "Rsh32x64", argLength: 2, aux: "Bool"},
138 {name: "Rsh64x8", argLength: 2, aux: "Bool"},
139 {name: "Rsh64x16", argLength: 2, aux: "Bool"},
140 {name: "Rsh64x32", argLength: 2, aux: "Bool"},
141 {name: "Rsh64x64", argLength: 2, aux: "Bool"},
143 {name: "Rsh8Ux8", argLength: 2, aux: "Bool"}, // arg0 >> arg1, unsigned
144 {name: "Rsh8Ux16", argLength: 2, aux: "Bool"},
145 {name: "Rsh8Ux32", argLength: 2, aux: "Bool"},
146 {name: "Rsh8Ux64", argLength: 2, aux: "Bool"},
147 {name: "Rsh16Ux8", argLength: 2, aux: "Bool"},
148 {name: "Rsh16Ux16", argLength: 2, aux: "Bool"},
149 {name: "Rsh16Ux32", argLength: 2, aux: "Bool"},
150 {name: "Rsh16Ux64", argLength: 2, aux: "Bool"},
151 {name: "Rsh32Ux8", argLength: 2, aux: "Bool"},
152 {name: "Rsh32Ux16", argLength: 2, aux: "Bool"},
153 {name: "Rsh32Ux32", argLength: 2, aux: "Bool"},
154 {name: "Rsh32Ux64", argLength: 2, aux: "Bool"},
155 {name: "Rsh64Ux8", argLength: 2, aux: "Bool"},
156 {name: "Rsh64Ux16", argLength: 2, aux: "Bool"},
157 {name: "Rsh64Ux32", argLength: 2, aux: "Bool"},
158 {name: "Rsh64Ux64", argLength: 2, aux: "Bool"},
160 // 2-input comparisons
161 {name: "Eq8", argLength: 2, commutative: true, typ: "Bool"}, // arg0 == arg1
162 {name: "Eq16", argLength: 2, commutative: true, typ: "Bool"},
163 {name: "Eq32", argLength: 2, commutative: true, typ: "Bool"},
164 {name: "Eq64", argLength: 2, commutative: true, typ: "Bool"},
165 {name: "EqPtr", argLength: 2, commutative: true, typ: "Bool"},
166 {name: "EqInter", argLength: 2, typ: "Bool"}, // arg0 or arg1 is nil; other cases handled by frontend
167 {name: "EqSlice", argLength: 2, typ: "Bool"}, // arg0 or arg1 is nil; other cases handled by frontend
168 {name: "Eq32F", argLength: 2, commutative: true, typ: "Bool"},
169 {name: "Eq64F", argLength: 2, commutative: true, typ: "Bool"},
171 {name: "Neq8", argLength: 2, commutative: true, typ: "Bool"}, // arg0 != arg1
172 {name: "Neq16", argLength: 2, commutative: true, typ: "Bool"},
173 {name: "Neq32", argLength: 2, commutative: true, typ: "Bool"},
174 {name: "Neq64", argLength: 2, commutative: true, typ: "Bool"},
175 {name: "NeqPtr", argLength: 2, commutative: true, typ: "Bool"},
176 {name: "NeqInter", argLength: 2, typ: "Bool"}, // arg0 or arg1 is nil; other cases handled by frontend
177 {name: "NeqSlice", argLength: 2, typ: "Bool"}, // arg0 or arg1 is nil; other cases handled by frontend
178 {name: "Neq32F", argLength: 2, commutative: true, typ: "Bool"},
179 {name: "Neq64F", argLength: 2, commutative: true, typ: "Bool"},
181 {name: "Less8", argLength: 2, typ: "Bool"}, // arg0 < arg1, signed
182 {name: "Less8U", argLength: 2, typ: "Bool"}, // arg0 < arg1, unsigned
183 {name: "Less16", argLength: 2, typ: "Bool"},
184 {name: "Less16U", argLength: 2, typ: "Bool"},
185 {name: "Less32", argLength: 2, typ: "Bool"},
186 {name: "Less32U", argLength: 2, typ: "Bool"},
187 {name: "Less64", argLength: 2, typ: "Bool"},
188 {name: "Less64U", argLength: 2, typ: "Bool"},
189 {name: "Less32F", argLength: 2, typ: "Bool"},
190 {name: "Less64F", argLength: 2, typ: "Bool"},
192 {name: "Leq8", argLength: 2, typ: "Bool"}, // arg0 <= arg1, signed
193 {name: "Leq8U", argLength: 2, typ: "Bool"}, // arg0 <= arg1, unsigned
194 {name: "Leq16", argLength: 2, typ: "Bool"},
195 {name: "Leq16U", argLength: 2, typ: "Bool"},
196 {name: "Leq32", argLength: 2, typ: "Bool"},
197 {name: "Leq32U", argLength: 2, typ: "Bool"},
198 {name: "Leq64", argLength: 2, typ: "Bool"},
199 {name: "Leq64U", argLength: 2, typ: "Bool"},
200 {name: "Leq32F", argLength: 2, typ: "Bool"},
201 {name: "Leq64F", argLength: 2, typ: "Bool"},
203 // the type of a CondSelect is the same as the type of its first
204 // two arguments, which should be register-width scalars; the third
205 // argument should be a boolean
206 {name: "CondSelect", argLength: 3}, // arg2 ? arg0 : arg1
209 {name: "AndB", argLength: 2, commutative: true, typ: "Bool"}, // arg0 && arg1 (not shortcircuited)
210 {name: "OrB", argLength: 2, commutative: true, typ: "Bool"}, // arg0 || arg1 (not shortcircuited)
211 {name: "EqB", argLength: 2, commutative: true, typ: "Bool"}, // arg0 == arg1
212 {name: "NeqB", argLength: 2, commutative: true, typ: "Bool"}, // arg0 != arg1
213 {name: "Not", argLength: 1, typ: "Bool"}, // !arg0, boolean
216 {name: "Neg8", argLength: 1}, // -arg0
217 {name: "Neg16", argLength: 1},
218 {name: "Neg32", argLength: 1},
219 {name: "Neg64", argLength: 1},
220 {name: "Neg32F", argLength: 1},
221 {name: "Neg64F", argLength: 1},
223 {name: "Com8", argLength: 1}, // ^arg0
224 {name: "Com16", argLength: 1},
225 {name: "Com32", argLength: 1},
226 {name: "Com64", argLength: 1},
228 {name: "Ctz8", argLength: 1}, // Count trailing (low order) zeroes (returns 0-8)
229 {name: "Ctz16", argLength: 1}, // Count trailing (low order) zeroes (returns 0-16)
230 {name: "Ctz32", argLength: 1}, // Count trailing (low order) zeroes (returns 0-32)
231 {name: "Ctz64", argLength: 1}, // Count trailing (low order) zeroes (returns 0-64)
232 {name: "Ctz8NonZero", argLength: 1}, // same as above, but arg[0] known to be non-zero, returns 0-7
233 {name: "Ctz16NonZero", argLength: 1}, // same as above, but arg[0] known to be non-zero, returns 0-15
234 {name: "Ctz32NonZero", argLength: 1}, // same as above, but arg[0] known to be non-zero, returns 0-31
235 {name: "Ctz64NonZero", argLength: 1}, // same as above, but arg[0] known to be non-zero, returns 0-63
236 {name: "BitLen8", argLength: 1}, // Number of bits in arg[0] (returns 0-8)
237 {name: "BitLen16", argLength: 1}, // Number of bits in arg[0] (returns 0-16)
238 {name: "BitLen32", argLength: 1}, // Number of bits in arg[0] (returns 0-32)
239 {name: "BitLen64", argLength: 1}, // Number of bits in arg[0] (returns 0-64)
241 {name: "Bswap16", argLength: 1}, // Swap bytes
242 {name: "Bswap32", argLength: 1}, // Swap bytes
243 {name: "Bswap64", argLength: 1}, // Swap bytes
245 {name: "BitRev8", argLength: 1}, // Reverse the bits in arg[0]
246 {name: "BitRev16", argLength: 1}, // Reverse the bits in arg[0]
247 {name: "BitRev32", argLength: 1}, // Reverse the bits in arg[0]
248 {name: "BitRev64", argLength: 1}, // Reverse the bits in arg[0]
250 {name: "PopCount8", argLength: 1}, // Count bits in arg[0]
251 {name: "PopCount16", argLength: 1}, // Count bits in arg[0]
252 {name: "PopCount32", argLength: 1}, // Count bits in arg[0]
253 {name: "PopCount64", argLength: 1}, // Count bits in arg[0]
255 // RotateLeftX instructions rotate the X bits of arg[0] to the left
256 // by the low lg_2(X) bits of arg[1], interpreted as an unsigned value.
257 // Note that this works out regardless of the bit width or signedness of
258 // arg[1]. In particular, RotateLeft by x is the same as RotateRight by -x.
259 {name: "RotateLeft64", argLength: 2},
260 {name: "RotateLeft32", argLength: 2},
261 {name: "RotateLeft16", argLength: 2},
262 {name: "RotateLeft8", argLength: 2},
267 // ±0 → ±0 (sign preserved)
270 {name: "Sqrt", argLength: 1}, // √arg0 (floating point, double precision)
271 {name: "Sqrt32", argLength: 1}, // √arg0 (floating point, single precision)
273 // Round to integer, float64 only.
275 // ±∞ → ±∞ (sign preserved)
276 // ±0 → ±0 (sign preserved)
278 {name: "Floor", argLength: 1}, // round arg0 toward -∞
279 {name: "Ceil", argLength: 1}, // round arg0 toward +∞
280 {name: "Trunc", argLength: 1}, // round arg0 toward 0
281 {name: "Round", argLength: 1}, // round arg0 to nearest, ties away from 0
282 {name: "RoundToEven", argLength: 1}, // round arg0 to nearest, ties to even
284 // Modify the sign bit
285 {name: "Abs", argLength: 1}, // absolute value arg0
286 {name: "Copysign", argLength: 2}, // copy sign from arg0 to arg1
288 // Float min/max implementation, if hardware is available.
289 {name: "Min64F", argLength: 2}, // min(arg0,arg1)
290 {name: "Min32F", argLength: 2}, // min(arg0,arg1)
291 {name: "Max64F", argLength: 2}, // max(arg0,arg1)
292 {name: "Max32F", argLength: 2}, // max(arg0,arg1)
295 // Fused-multiply-add, float64 only.
296 // When a*b+c is exactly zero (before rounding), then the result is +0 or -0.
297 // The 0's sign is determined according to the standard rules for the
298 // addition (-0 if both a*b and c are -0, +0 otherwise).
300 // Otherwise, when a*b+c rounds to zero, then the resulting 0's sign is
301 // determined by the sign of the exact result a*b+c.
302 // See section 6.3 in ieee754.
304 // When the multiply is an infinity times a zero, the result is NaN.
305 // See section 7.2 in ieee754.
306 {name: "FMA", argLength: 3}, // compute (a*b)+c without intermediate rounding
308 // Data movement. Max argument length for Phi is indefinite.
309 {name: "Phi", argLength: -1, zeroWidth: true}, // select an argument based on which predecessor block we came from
310 {name: "Copy", argLength: 1}, // output = arg0
311 // Convert converts between pointers and integers.
312 // We have a special op for this so as to not confuse GC
313 // (particularly stack maps). It takes a memory arg so it
314 // gets correctly ordered with respect to GC safepoints.
315 // It gets compiled to nothing, so its result must in the same
316 // register as its argument. regalloc knows it can use any
317 // allocatable integer register for OpConvert.
318 // arg0=ptr/int arg1=mem, output=int/ptr
319 {name: "Convert", argLength: 2, zeroWidth: true, resultInArg0: true},
321 // constants. Constant values are stored in the aux or
323 {name: "ConstBool", aux: "Bool"}, // auxint is 0 for false and 1 for true
324 {name: "ConstString", aux: "String"}, // value is aux.(string)
325 {name: "ConstNil", typ: "BytePtr"}, // nil pointer
326 {name: "Const8", aux: "Int8"}, // auxint is sign-extended 8 bits
327 {name: "Const16", aux: "Int16"}, // auxint is sign-extended 16 bits
328 {name: "Const32", aux: "Int32"}, // auxint is sign-extended 32 bits
329 // Note: ConstX are sign-extended even when the type of the value is unsigned.
330 // For instance, uint8(0xaa) is stored as auxint=0xffffffffffffffaa.
331 {name: "Const64", aux: "Int64"}, // value is auxint
332 // Note: for both Const32F and Const64F, we disallow encoding NaNs.
333 // Signaling NaNs are tricky because if you do anything with them, they become quiet.
334 // Particularly, converting a 32 bit sNaN to 64 bit and back converts it to a qNaN.
335 // See issue 36399 and 36400.
336 // Encodings of +inf, -inf, and -0 are fine.
337 {name: "Const32F", aux: "Float32"}, // value is math.Float64frombits(uint64(auxint)) and is exactly representable as float 32
338 {name: "Const64F", aux: "Float64"}, // value is math.Float64frombits(uint64(auxint))
339 {name: "ConstInterface"}, // nil interface
340 {name: "ConstSlice"}, // nil slice
342 // Constant-like things
343 {name: "InitMem", zeroWidth: true}, // memory input to the function.
344 {name: "Arg", aux: "SymOff", symEffect: "Read", zeroWidth: true}, // argument to the function. aux=GCNode of arg, off = offset in that arg.
346 // Like Arg, these are generic ops that survive lowering. AuxInt is a register index, and the actual output register for each index is defined by the architecture.
347 // AuxInt = integer argument index (not a register number). ABI-specified spill loc obtained from function
348 {name: "ArgIntReg", aux: "NameOffsetInt8", zeroWidth: true}, // argument to the function in an int reg.
349 {name: "ArgFloatReg", aux: "NameOffsetInt8", zeroWidth: true}, // argument to the function in a float reg.
351 // The address of a variable. arg0 is the base pointer.
352 // If the variable is a global, the base pointer will be SB and
353 // the Aux field will be a *obj.LSym.
354 // If the variable is a local, the base pointer will be SP and
355 // the Aux field will be a *gc.Node.
356 {name: "Addr", argLength: 1, aux: "Sym", symEffect: "Addr"}, // Address of a variable. Arg0=SB. Aux identifies the variable.
357 {name: "LocalAddr", argLength: 2, aux: "Sym", symEffect: "Addr"}, // Address of a variable. Arg0=SP. Arg1=mem. Aux identifies the variable.
359 {name: "SP", zeroWidth: true}, // stack pointer
360 {name: "SB", typ: "Uintptr", zeroWidth: true}, // static base pointer (a.k.a. globals pointer)
361 {name: "Invalid"}, // unused value
362 {name: "SPanchored", typ: "Uintptr", argLength: 2, zeroWidth: true}, // arg0 = SP, arg1 = mem. Result is identical to arg0, but cannot be scheduled before memory state arg1.
365 {name: "Load", argLength: 2}, // Load from arg0. arg1=memory
366 {name: "Dereference", argLength: 2}, // Load from arg0. arg1=memory. Helper op for arg/result passing, result is an otherwise not-SSA-able "value".
367 {name: "Store", argLength: 3, typ: "Mem", aux: "Typ"}, // Store arg1 to arg0. arg2=memory, aux=type. Returns memory.
368 // Normally we require that the source and destination of Move do not overlap.
369 // There is an exception when we know all the loads will happen before all
370 // the stores. In that case, overlap is ok. See
371 // memmove inlining in generic.rules. When inlineablememmovesize (in ../rewrite.go)
372 // returns true, we must do all loads before all stores, when lowering Move.
373 // The type of Move is used for the write barrier pass to insert write barriers
374 // and for alignment on some architectures.
375 // For pointerless types, it is possible for the type to be inaccurate.
376 // For type alignment and pointer information, use the type in Aux;
377 // for type size, use the size in AuxInt.
378 // The "inline runtime.memmove" rewrite rule generates Moves with inaccurate types,
379 // such as type byte instead of the more accurate type [8]byte.
380 {name: "Move", argLength: 3, typ: "Mem", aux: "TypSize"}, // arg0=destptr, arg1=srcptr, arg2=mem, auxint=size, aux=type. Returns memory.
381 {name: "Zero", argLength: 2, typ: "Mem", aux: "TypSize"}, // arg0=destptr, arg1=mem, auxint=size, aux=type. Returns memory.
383 // Memory operations with write barriers.
384 // Expand to runtime calls. Write barrier will be removed if write on stack.
385 {name: "StoreWB", argLength: 3, typ: "Mem", aux: "Typ"}, // Store arg1 to arg0. arg2=memory, aux=type. Returns memory.
386 {name: "MoveWB", argLength: 3, typ: "Mem", aux: "TypSize"}, // arg0=destptr, arg1=srcptr, arg2=mem, auxint=size, aux=type. Returns memory.
387 {name: "ZeroWB", argLength: 2, typ: "Mem", aux: "TypSize"}, // arg0=destptr, arg1=mem, auxint=size, aux=type. Returns memory.
388 {name: "WBend", argLength: 1, typ: "Mem"}, // Write barrier code is done, interrupting is now allowed.
390 // WB invokes runtime.gcWriteBarrier. This is not a normal
391 // call: it takes arguments in registers, doesn't clobber
392 // general-purpose registers (the exact clobber set is
393 // arch-dependent), and is not a safe-point.
394 {name: "WB", argLength: 1, typ: "(BytePtr,Mem)", aux: "Int64"}, // arg0=mem, auxint=# of buffer entries needed. Returns buffer pointer and memory.
396 {name: "HasCPUFeature", argLength: 0, typ: "bool", aux: "Sym", symEffect: "None"}, // aux=place that this feature flag can be loaded from
398 // PanicBounds and PanicExtend generate a runtime panic.
399 // Their arguments provide index values to use in panic messages.
400 // Both PanicBounds and PanicExtend have an AuxInt value from the BoundsKind type (in ../op.go).
401 // PanicBounds' index is int sized.
402 // PanicExtend's index is int64 sized. (PanicExtend is only used on 32-bit archs.)
403 {name: "PanicBounds", argLength: 3, aux: "Int64", typ: "Mem", call: true}, // arg0=idx, arg1=len, arg2=mem, returns memory.
404 {name: "PanicExtend", argLength: 4, aux: "Int64", typ: "Mem", call: true}, // arg0=idxHi, arg1=idxLo, arg2=len, arg3=mem, returns memory.
406 // Function calls. Arguments to the call have already been written to the stack.
407 // Return values appear on the stack. The method receiver, if any, is treated
408 // as a phantom first argument.
409 // TODO(josharian): ClosureCall and InterCall should have Int32 aux
410 // to match StaticCall's 32 bit arg size limit.
411 // TODO(drchase,josharian): could the arg size limit be bundled into the rules for CallOff?
413 // Before lowering, LECalls receive their fixed inputs (first), memory (last),
414 // and a variable number of input values in the middle.
415 // They produce a variable number of result values.
416 // These values are not necessarily "SSA-able"; they can be too large,
417 // but in that case inputs are loaded immediately before with OpDereference,
418 // and outputs are stored immediately with OpStore.
420 // After call expansion, Calls have the same fixed-middle-memory arrangement of inputs,
421 // with the difference that the "middle" is only the register-resident inputs,
422 // and the non-register inputs are instead stored at ABI-defined offsets from SP
423 // (and the stores thread through the memory that is ultimately an input to the call).
424 // Outputs follow a similar pattern; register-resident outputs are the leading elements
425 // of a Result-typed output, with memory last, and any memory-resident outputs have been
426 // stored to ABI-defined locations. Each non-memory input or output fits in a register.
428 // Subsequent architecture-specific lowering only changes the opcode.
430 {name: "ClosureCall", argLength: -1, aux: "CallOff", call: true}, // arg0=code pointer, arg1=context ptr, arg2..argN-1 are register inputs, argN=memory. auxint=arg size. Returns Result of register results, plus memory.
431 {name: "StaticCall", argLength: -1, aux: "CallOff", call: true}, // call function aux.(*obj.LSym), arg0..argN-1 are register inputs, argN=memory. auxint=arg size. Returns Result of register results, plus memory.
432 {name: "InterCall", argLength: -1, aux: "CallOff", call: true}, // interface call. arg0=code pointer, arg1..argN-1 are register inputs, argN=memory, auxint=arg size. Returns Result of register results, plus memory.
433 {name: "TailCall", argLength: -1, aux: "CallOff", call: true}, // tail call function aux.(*obj.LSym), arg0..argN-1 are register inputs, argN=memory. auxint=arg size. Returns Result of register results, plus memory.
435 {name: "ClosureLECall", argLength: -1, aux: "CallOff", call: true}, // late-expanded closure call. arg0=code pointer, arg1=context ptr, arg2..argN-1 are inputs, argN is mem. auxint = arg size. Result is tuple of result(s), plus mem.
436 {name: "StaticLECall", argLength: -1, aux: "CallOff", call: true}, // late-expanded static call function aux.(*ssa.AuxCall.Fn). arg0..argN-1 are inputs, argN is mem. auxint = arg size. Result is tuple of result(s), plus mem.
437 {name: "InterLECall", argLength: -1, aux: "CallOff", call: true}, // late-expanded interface call. arg0=code pointer, arg1..argN-1 are inputs, argN is mem. auxint = arg size. Result is tuple of result(s), plus mem.
438 {name: "TailLECall", argLength: -1, aux: "CallOff", call: true}, // late-expanded static tail call function aux.(*ssa.AuxCall.Fn). arg0..argN-1 are inputs, argN is mem. auxint = arg size. Result is tuple of result(s), plus mem.
440 // Conversions: signed extensions, zero (unsigned) extensions, truncations
441 {name: "SignExt8to16", argLength: 1, typ: "Int16"},
442 {name: "SignExt8to32", argLength: 1, typ: "Int32"},
443 {name: "SignExt8to64", argLength: 1, typ: "Int64"},
444 {name: "SignExt16to32", argLength: 1, typ: "Int32"},
445 {name: "SignExt16to64", argLength: 1, typ: "Int64"},
446 {name: "SignExt32to64", argLength: 1, typ: "Int64"},
447 {name: "ZeroExt8to16", argLength: 1, typ: "UInt16"},
448 {name: "ZeroExt8to32", argLength: 1, typ: "UInt32"},
449 {name: "ZeroExt8to64", argLength: 1, typ: "UInt64"},
450 {name: "ZeroExt16to32", argLength: 1, typ: "UInt32"},
451 {name: "ZeroExt16to64", argLength: 1, typ: "UInt64"},
452 {name: "ZeroExt32to64", argLength: 1, typ: "UInt64"},
453 {name: "Trunc16to8", argLength: 1},
454 {name: "Trunc32to8", argLength: 1},
455 {name: "Trunc32to16", argLength: 1},
456 {name: "Trunc64to8", argLength: 1},
457 {name: "Trunc64to16", argLength: 1},
458 {name: "Trunc64to32", argLength: 1},
460 {name: "Cvt32to32F", argLength: 1},
461 {name: "Cvt32to64F", argLength: 1},
462 {name: "Cvt64to32F", argLength: 1},
463 {name: "Cvt64to64F", argLength: 1},
464 {name: "Cvt32Fto32", argLength: 1},
465 {name: "Cvt32Fto64", argLength: 1},
466 {name: "Cvt64Fto32", argLength: 1},
467 {name: "Cvt64Fto64", argLength: 1},
468 {name: "Cvt32Fto64F", argLength: 1},
469 {name: "Cvt64Fto32F", argLength: 1},
470 {name: "CvtBoolToUint8", argLength: 1},
472 // Force rounding to precision of type.
473 {name: "Round32F", argLength: 1},
474 {name: "Round64F", argLength: 1},
476 // Automatically inserted safety checks
477 {name: "IsNonNil", argLength: 1, typ: "Bool"}, // arg0 != nil
478 {name: "IsInBounds", argLength: 2, typ: "Bool"}, // 0 <= arg0 < arg1. arg1 is guaranteed >= 0.
479 {name: "IsSliceInBounds", argLength: 2, typ: "Bool"}, // 0 <= arg0 <= arg1. arg1 is guaranteed >= 0.
480 {name: "NilCheck", argLength: 2, typ: "Void"}, // arg0=ptr, arg1=mem. Panics if arg0 is nil. Returns void.
483 {name: "GetG", argLength: 1, zeroWidth: true}, // runtime.getg() (read g pointer). arg0=mem
484 {name: "GetClosurePtr"}, // get closure pointer from dedicated register
485 {name: "GetCallerPC"}, // for getcallerpc intrinsic
486 {name: "GetCallerSP", argLength: 1}, // for getcallersp intrinsic. arg0=mem.
488 // Indexing operations
489 {name: "PtrIndex", argLength: 2}, // arg0=ptr, arg1=index. Computes ptr+sizeof(*v.type)*index, where index is extended to ptrwidth type
490 {name: "OffPtr", argLength: 1, aux: "Int64"}, // arg0 + auxint (arg0 and result are pointers)
493 {name: "SliceMake", argLength: 3}, // arg0=ptr, arg1=len, arg2=cap
494 {name: "SlicePtr", argLength: 1, typ: "BytePtr"}, // ptr(arg0)
495 {name: "SliceLen", argLength: 1}, // len(arg0)
496 {name: "SliceCap", argLength: 1}, // cap(arg0)
497 // SlicePtrUnchecked, like SlicePtr, extracts the pointer from a slice.
498 // SlicePtr values are assumed non-nil, because they are guarded by bounds checks.
499 // SlicePtrUnchecked values can be nil.
500 {name: "SlicePtrUnchecked", argLength: 1},
502 // Complex (part/whole)
503 {name: "ComplexMake", argLength: 2}, // arg0=real, arg1=imag
504 {name: "ComplexReal", argLength: 1}, // real(arg0)
505 {name: "ComplexImag", argLength: 1}, // imag(arg0)
508 {name: "StringMake", argLength: 2}, // arg0=ptr, arg1=len
509 {name: "StringPtr", argLength: 1, typ: "BytePtr"}, // ptr(arg0)
510 {name: "StringLen", argLength: 1, typ: "Int"}, // len(arg0)
513 {name: "IMake", argLength: 2}, // arg0=itab, arg1=data
514 {name: "ITab", argLength: 1, typ: "Uintptr"}, // arg0=interface, returns itable field
515 {name: "IData", argLength: 1}, // arg0=interface, returns data field
518 {name: "StructMake0"}, // Returns struct with 0 fields.
519 {name: "StructMake1", argLength: 1}, // arg0=field0. Returns struct.
520 {name: "StructMake2", argLength: 2}, // arg0,arg1=field0,field1. Returns struct.
521 {name: "StructMake3", argLength: 3}, // arg0..2=field0..2. Returns struct.
522 {name: "StructMake4", argLength: 4}, // arg0..3=field0..3. Returns struct.
523 {name: "StructSelect", argLength: 1, aux: "Int64"}, // arg0=struct, auxint=field index. Returns the auxint'th field.
526 {name: "ArrayMake0"}, // Returns array with 0 elements
527 {name: "ArrayMake1", argLength: 1}, // Returns array with 1 element
528 {name: "ArraySelect", argLength: 1, aux: "Int64"}, // arg0=array, auxint=index. Returns a[i].
530 // Spill&restore ops for the register allocator. These are
531 // semantically identical to OpCopy; they do not take/return
532 // stores like regular memory ops do. We can get away without memory
533 // args because we know there is no aliasing of spill slots on the stack.
534 {name: "StoreReg", argLength: 1},
535 {name: "LoadReg", argLength: 1},
537 // Used during ssa construction. Like Copy, but the arg has not been specified yet.
538 {name: "FwdRef", aux: "Sym", symEffect: "None"},
540 // Unknown value. Used for Values whose values don't matter because they are dead code.
543 {name: "VarDef", argLength: 1, aux: "Sym", typ: "Mem", symEffect: "None", zeroWidth: true}, // aux is a *gc.Node of a variable that is about to be initialized. arg0=mem, returns mem
544 // TODO: what's the difference between VarLive and KeepAlive?
545 {name: "VarLive", argLength: 1, aux: "Sym", symEffect: "Read", zeroWidth: true}, // aux is a *gc.Node of a variable that must be kept live. arg0=mem, returns mem
546 {name: "KeepAlive", argLength: 2, typ: "Mem", zeroWidth: true}, // arg[0] is a value that must be kept alive until this mark. arg[1]=mem, returns mem
548 // InlMark marks the start of an inlined function body. Its AuxInt field
549 // distinguishes which entry in the local inline tree it is marking.
550 {name: "InlMark", argLength: 1, aux: "Int32", typ: "Void"}, // arg[0]=mem, returns void.
552 // Ops for breaking 64-bit operations on 32-bit architectures
553 {name: "Int64Make", argLength: 2, typ: "UInt64"}, // arg0=hi, arg1=lo
554 {name: "Int64Hi", argLength: 1, typ: "UInt32"}, // high 32-bit of arg0
555 {name: "Int64Lo", argLength: 1, typ: "UInt32"}, // low 32-bit of arg0
557 {name: "Add32carry", argLength: 2, commutative: true, typ: "(UInt32,Flags)"}, // arg0 + arg1, returns (value, carry)
558 {name: "Add32withcarry", argLength: 3, commutative: true}, // arg0 + arg1 + arg2, arg2=carry (0 or 1)
560 {name: "Sub32carry", argLength: 2, typ: "(UInt32,Flags)"}, // arg0 - arg1, returns (value, carry)
561 {name: "Sub32withcarry", argLength: 3}, // arg0 - arg1 - arg2, arg2=carry (0 or 1)
563 {name: "Add64carry", argLength: 3, commutative: true, typ: "(UInt64,UInt64)"}, // arg0 + arg1 + arg2, arg2 must be 0 or 1. returns (value, value>>64)
564 {name: "Sub64borrow", argLength: 3, typ: "(UInt64,UInt64)"}, // arg0 - (arg1 + arg2), arg2 must be 0 or 1. returns (value, value>>64&1)
566 {name: "Signmask", argLength: 1, typ: "Int32"}, // 0 if arg0 >= 0, -1 if arg0 < 0
567 {name: "Zeromask", argLength: 1, typ: "UInt32"}, // 0 if arg0 == 0, 0xffffffff if arg0 != 0
568 {name: "Slicemask", argLength: 1}, // 0 if arg0 == 0, -1 if arg0 > 0, undef if arg0<0. Type is native int size.
570 {name: "SpectreIndex", argLength: 2}, // arg0 if 0 <= arg0 < arg1, 0 otherwise. Type is native int size.
571 {name: "SpectreSliceIndex", argLength: 2}, // arg0 if 0 <= arg0 <= arg1, 0 otherwise. Type is native int size.
573 {name: "Cvt32Uto32F", argLength: 1}, // uint32 -> float32, only used on 32-bit arch
574 {name: "Cvt32Uto64F", argLength: 1}, // uint32 -> float64, only used on 32-bit arch
575 {name: "Cvt32Fto32U", argLength: 1}, // float32 -> uint32, only used on 32-bit arch
576 {name: "Cvt64Fto32U", argLength: 1}, // float64 -> uint32, only used on 32-bit arch
577 {name: "Cvt64Uto32F", argLength: 1}, // uint64 -> float32, only used on archs that has the instruction
578 {name: "Cvt64Uto64F", argLength: 1}, // uint64 -> float64, only used on archs that has the instruction
579 {name: "Cvt32Fto64U", argLength: 1}, // float32 -> uint64, only used on archs that has the instruction
580 {name: "Cvt64Fto64U", argLength: 1}, // float64 -> uint64, only used on archs that has the instruction
582 // pseudo-ops for breaking Tuple
583 {name: "Select0", argLength: 1, zeroWidth: true}, // the first component of a tuple
584 {name: "Select1", argLength: 1, zeroWidth: true}, // the second component of a tuple
585 {name: "SelectN", argLength: 1, aux: "Int64"}, // arg0=result, auxint=field index. Returns the auxint'th member.
586 {name: "SelectNAddr", argLength: 1, aux: "Int64"}, // arg0=result, auxint=field index. Returns the address of auxint'th member. Used for un-SSA-able result types.
587 {name: "MakeResult", argLength: -1}, // arg0 .. are components of a "Result" (like the result from a Call). The last arg should be memory (like the result from a call).
589 // Atomic operations used for semantically inlining sync/atomic and
590 // runtime/internal/atomic. Atomic loads return a new memory so that
591 // the loads are properly ordered with respect to other loads and
593 {name: "AtomicLoad8", argLength: 2, typ: "(UInt8,Mem)"}, // Load from arg0. arg1=memory. Returns loaded value and new memory.
594 {name: "AtomicLoad32", argLength: 2, typ: "(UInt32,Mem)"}, // Load from arg0. arg1=memory. Returns loaded value and new memory.
595 {name: "AtomicLoad64", argLength: 2, typ: "(UInt64,Mem)"}, // Load from arg0. arg1=memory. Returns loaded value and new memory.
596 {name: "AtomicLoadPtr", argLength: 2, typ: "(BytePtr,Mem)"}, // Load from arg0. arg1=memory. Returns loaded value and new memory.
597 {name: "AtomicLoadAcq32", argLength: 2, typ: "(UInt32,Mem)"}, // Load from arg0. arg1=memory. Lock acquisition, returns loaded value and new memory.
598 {name: "AtomicLoadAcq64", argLength: 2, typ: "(UInt64,Mem)"}, // Load from arg0. arg1=memory. Lock acquisition, returns loaded value and new memory.
599 {name: "AtomicStore8", argLength: 3, typ: "Mem", hasSideEffects: true}, // Store arg1 to *arg0. arg2=memory. Returns memory.
600 {name: "AtomicStore32", argLength: 3, typ: "Mem", hasSideEffects: true}, // Store arg1 to *arg0. arg2=memory. Returns memory.
601 {name: "AtomicStore64", argLength: 3, typ: "Mem", hasSideEffects: true}, // Store arg1 to *arg0. arg2=memory. Returns memory.
602 {name: "AtomicStorePtrNoWB", argLength: 3, typ: "Mem", hasSideEffects: true}, // Store arg1 to *arg0. arg2=memory. Returns memory.
603 {name: "AtomicStoreRel32", argLength: 3, typ: "Mem", hasSideEffects: true}, // Store arg1 to *arg0. arg2=memory. Lock release, returns memory.
604 {name: "AtomicStoreRel64", argLength: 3, typ: "Mem", hasSideEffects: true}, // Store arg1 to *arg0. arg2=memory. Lock release, returns memory.
605 {name: "AtomicExchange32", argLength: 3, typ: "(UInt32,Mem)", hasSideEffects: true}, // Store arg1 to *arg0. arg2=memory. Returns old contents of *arg0 and new memory.
606 {name: "AtomicExchange64", argLength: 3, typ: "(UInt64,Mem)", hasSideEffects: true}, // Store arg1 to *arg0. arg2=memory. Returns old contents of *arg0 and new memory.
607 {name: "AtomicAdd32", argLength: 3, typ: "(UInt32,Mem)", hasSideEffects: true}, // Do *arg0 += arg1. arg2=memory. Returns sum and new memory.
608 {name: "AtomicAdd64", argLength: 3, typ: "(UInt64,Mem)", hasSideEffects: true}, // Do *arg0 += arg1. arg2=memory. Returns sum and new memory.
609 {name: "AtomicCompareAndSwap32", argLength: 4, typ: "(Bool,Mem)", hasSideEffects: true}, // if *arg0==arg1, then set *arg0=arg2. Returns true if store happens and new memory.
610 {name: "AtomicCompareAndSwap64", argLength: 4, typ: "(Bool,Mem)", hasSideEffects: true}, // if *arg0==arg1, then set *arg0=arg2. Returns true if store happens and new memory.
611 {name: "AtomicCompareAndSwapRel32", argLength: 4, typ: "(Bool,Mem)", hasSideEffects: true}, // if *arg0==arg1, then set *arg0=arg2. Lock release, reports whether store happens and new memory.
612 {name: "AtomicAnd8", argLength: 3, typ: "Mem", hasSideEffects: true}, // *arg0 &= arg1. arg2=memory. Returns memory.
613 {name: "AtomicAnd32", argLength: 3, typ: "Mem", hasSideEffects: true}, // *arg0 &= arg1. arg2=memory. Returns memory.
614 {name: "AtomicOr8", argLength: 3, typ: "Mem", hasSideEffects: true}, // *arg0 |= arg1. arg2=memory. Returns memory.
615 {name: "AtomicOr32", argLength: 3, typ: "Mem", hasSideEffects: true}, // *arg0 |= arg1. arg2=memory. Returns memory.
617 // Atomic operation variants
618 // These variants have the same semantics as above atomic operations.
619 // But they are used for generating more efficient code on certain modern machines, with run-time CPU feature detection.
620 // Currently, they are used on ARM64 only.
621 {name: "AtomicAdd32Variant", argLength: 3, typ: "(UInt32,Mem)", hasSideEffects: true}, // Do *arg0 += arg1. arg2=memory. Returns sum and new memory.
622 {name: "AtomicAdd64Variant", argLength: 3, typ: "(UInt64,Mem)", hasSideEffects: true}, // Do *arg0 += arg1. arg2=memory. Returns sum and new memory.
623 {name: "AtomicExchange32Variant", argLength: 3, typ: "(UInt32,Mem)", hasSideEffects: true}, // Store arg1 to *arg0. arg2=memory. Returns old contents of *arg0 and new memory.
624 {name: "AtomicExchange64Variant", argLength: 3, typ: "(UInt64,Mem)", hasSideEffects: true}, // Store arg1 to *arg0. arg2=memory. Returns old contents of *arg0 and new memory.
625 {name: "AtomicCompareAndSwap32Variant", argLength: 4, typ: "(Bool,Mem)", hasSideEffects: true}, // if *arg0==arg1, then set *arg0=arg2. Returns true if store happens and new memory.
626 {name: "AtomicCompareAndSwap64Variant", argLength: 4, typ: "(Bool,Mem)", hasSideEffects: true}, // if *arg0==arg1, then set *arg0=arg2. Returns true if store happens and new memory.
627 {name: "AtomicAnd8Variant", argLength: 3, typ: "Mem", hasSideEffects: true}, // *arg0 &= arg1. arg2=memory. Returns memory.
628 {name: "AtomicAnd32Variant", argLength: 3, typ: "Mem", hasSideEffects: true}, // *arg0 &= arg1. arg2=memory. Returns memory.
629 {name: "AtomicOr8Variant", argLength: 3, typ: "Mem", hasSideEffects: true}, // *arg0 |= arg1. arg2=memory. Returns memory.
630 {name: "AtomicOr32Variant", argLength: 3, typ: "Mem", hasSideEffects: true}, // *arg0 |= arg1. arg2=memory. Returns memory.
632 // Publication barrier
633 {name: "PubBarrier", argLength: 1, hasSideEffects: true}, // Do data barrier. arg0=memory.
635 // Clobber experiment op
636 {name: "Clobber", argLength: 0, typ: "Void", aux: "SymOff", symEffect: "None"}, // write an invalid pointer value to the given pointer slot of a stack variable
637 {name: "ClobberReg", argLength: 0, typ: "Void"}, // clobber a register
639 // Prefetch instruction
640 {name: "PrefetchCache", argLength: 2, hasSideEffects: true}, // Do prefetch arg0 to cache. arg0=addr, arg1=memory.
641 {name: "PrefetchCacheStreamed", argLength: 2, hasSideEffects: true}, // Do non-temporal or streamed prefetch arg0 to cache. arg0=addr, arg1=memory.
644 // kind controls successors implicit exit
645 // ----------------------------------------------------------
646 // Exit [return mem] [] yes
647 // Ret [return mem] [] yes
648 // RetJmp [return mem] [] yes
650 // If [boolean Value] [then, else]
651 // First [] [always, never]
653 var genericBlocks = []blockData{
654 {name: "Plain"}, // a single successor
655 {name: "If", controls: 1}, // if Controls[0] goto Succs[0] else goto Succs[1]
656 {name: "Defer", controls: 1}, // Succs[0]=defer queued, Succs[1]=defer recovered. Controls[0] is call op (of memory type)
657 {name: "Ret", controls: 1}, // no successors, Controls[0] value is memory result
658 {name: "RetJmp", controls: 1}, // no successors, Controls[0] value is a tail call
659 {name: "Exit", controls: 1}, // no successors, Controls[0] value generates a panic
660 {name: "JumpTable", controls: 1}, // multiple successors, the integer Controls[0] selects which one
662 // transient block state used for dead code removal
663 {name: "First"}, // 2 successors, always takes the first one (second is dead)
667 archs = append(archs, arch{
670 blocks: genericBlocks,