gsubr.go 28.7 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55
// Derived from Inferno utils/6c/txt.c
// http://code.google.com/p/inferno-os/source/browse/utils/6c/txt.c
//
//	Copyright © 1994-1999 Lucent Technologies Inc.  All rights reserved.
//	Portions Copyright © 1995-1997 C H Forsyth (forsyth@terzarima.net)
//	Portions Copyright © 1997-1999 Vita Nuova Limited
//	Portions Copyright © 2000-2007 Vita Nuova Holdings Limited (www.vitanuova.com)
//	Portions Copyright © 2004,2006 Bruce Ellis
//	Portions Copyright © 2005-2007 C H Forsyth (forsyth@terzarima.net)
//	Revisions Copyright © 2000-2007 Lucent Technologies Inc. and others
//	Portions Copyright © 2009 The Go Authors.  All rights reserved.
//
// Permission is hereby granted, free of charge, to any person obtaining a copy
// of this software and associated documentation files (the "Software"), to deal
// in the Software without restriction, including without limitation the rights
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
// copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:
//
// The above copyright notice and this permission notice shall be included in
// all copies or substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
// THE SOFTWARE.

package main

import (
	"cmd/internal/obj"
	"cmd/internal/obj/x86"
	"fmt"
)
import "cmd/internal/gc"

// TODO(rsc): Can make this bigger if we move
// the text segment up higher in 6l for all GOOS.
// At the same time, can raise StackBig in ../../runtime/stack.h.
var unmappedzero int64 = 4096

var resvd = []int{
	x86.REG_DI, // for movstring
	x86.REG_SI, // for movstring

	x86.REG_AX, // for divide
	x86.REG_CX, // for shift
	x86.REG_DX, // for divide
	x86.REG_SP, // for stack
}

func ginit() {
56
	for i := 0; i < len(reg); i++ {
57 58
		reg[i] = 1
	}
59
	for i := x86.REG_AX; i <= x86.REG_R15; i++ {
60 61
		reg[i] = 0
	}
62
	for i := x86.REG_X0; i <= x86.REG_X15; i++ {
63 64 65
		reg[i] = 0
	}

66
	for i := 0; i < len(resvd); i++ {
67 68 69 70 71 72 73 74 75 76 77 78 79
		reg[resvd[i]]++
	}

	if gc.Nacl {
		reg[x86.REG_BP]++
		reg[x86.REG_R15]++
	} else if obj.Framepointer_enabled != 0 {
		// BP is part of the calling convention of framepointer_enabled.
		reg[x86.REG_BP]++
	}
}

func gclean() {
80
	for i := 0; i < len(resvd); i++ {
81 82 83 84 85 86 87 88 89
		reg[resvd[i]]--
	}
	if gc.Nacl {
		reg[x86.REG_BP]--
		reg[x86.REG_R15]--
	} else if obj.Framepointer_enabled != 0 {
		reg[x86.REG_BP]--
	}

90
	for i := x86.REG_AX; i <= x86.REG_R15; i++ {
91 92 93 94
		if reg[i] != 0 {
			gc.Yyerror("reg %v left allocated\n", gc.Ctxt.Rconv(i))
		}
	}
95
	for i := x86.REG_X0; i <= x86.REG_X15; i++ {
96 97 98 99 100 101
		if reg[i] != 0 {
			gc.Yyerror("reg %v left allocated\n", gc.Ctxt.Rconv(i))
		}
	}
}

102
func anyregalloc() bool {
103 104
	var j int

105
	for i := x86.REG_AX; i <= x86.REG_R15; i++ {
106 107 108 109 110 111 112 113
		if reg[i] == 0 {
			goto ok
		}
		for j = 0; j < len(resvd); j++ {
			if resvd[j] == i {
				goto ok
			}
		}
114
		return true
115 116 117
	ok:
	}

118
	return false
119 120 121 122 123 124 125 126 127 128 129 130 131
}

var regpc [x86.REG_R15 + 1 - x86.REG_AX]uint32

/*
 * allocate register of type t, leave in n.
 * if o != N, o is desired fixed register.
 * caller must regfree(n).
 */
func regalloc(n *gc.Node, t *gc.Type, o *gc.Node) {
	if t == nil {
		gc.Fatal("regalloc: t nil")
	}
132
	et := int(gc.Simtype[t.Etype])
133

134
	var i int
135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161
	switch et {
	case gc.TINT8,
		gc.TUINT8,
		gc.TINT16,
		gc.TUINT16,
		gc.TINT32,
		gc.TUINT32,
		gc.TINT64,
		gc.TUINT64,
		gc.TPTR32,
		gc.TPTR64,
		gc.TBOOL:
		if o != nil && o.Op == gc.OREGISTER {
			i = int(o.Val.U.Reg)
			if i >= x86.REG_AX && i <= x86.REG_R15 {
				goto out
			}
		}

		for i = x86.REG_AX; i <= x86.REG_R15; i++ {
			if reg[i] == 0 {
				regpc[i-x86.REG_AX] = uint32(obj.Getcallerpc(&n))
				goto out
			}
		}

		gc.Flusherrors()
162
		for i := 0; i+x86.REG_AX <= x86.REG_R15; i++ {
163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203
			fmt.Printf("%d %p\n", i, regpc[i])
		}
		gc.Fatal("out of fixed registers")

	case gc.TFLOAT32,
		gc.TFLOAT64:
		if o != nil && o.Op == gc.OREGISTER {
			i = int(o.Val.U.Reg)
			if i >= x86.REG_X0 && i <= x86.REG_X15 {
				goto out
			}
		}

		for i = x86.REG_X0; i <= x86.REG_X15; i++ {
			if reg[i] == 0 {
				goto out
			}
		}
		gc.Fatal("out of floating registers")

	case gc.TCOMPLEX64,
		gc.TCOMPLEX128:
		gc.Tempname(n, t)
		return
	}

	gc.Fatal("regalloc: unknown type %v", gc.Tconv(t, 0))
	return

out:
	reg[i]++
	gc.Nodreg(n, t, i)
}

func regfree(n *gc.Node) {
	if n.Op == gc.ONAME {
		return
	}
	if n.Op != gc.OREGISTER && n.Op != gc.OINDREG {
		gc.Fatal("regfree: not a register")
	}
204
	i := int(n.Val.U.Reg)
205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259
	if i == x86.REG_SP {
		return
	}
	if i < 0 || i >= len(reg) {
		gc.Fatal("regfree: reg out of range")
	}
	if reg[i] <= 0 {
		gc.Fatal("regfree: reg not allocated")
	}
	reg[i]--
	if reg[i] == 0 && x86.REG_AX <= i && i <= x86.REG_R15 {
		regpc[i-x86.REG_AX] = 0
	}
}

/*
 * generate
 *	as $c, reg
 */
func gconreg(as int, c int64, reg int) {
	var nr gc.Node

	switch as {
	case x86.AADDL,
		x86.AMOVL,
		x86.ALEAL:
		gc.Nodreg(&nr, gc.Types[gc.TINT32], reg)

	default:
		gc.Nodreg(&nr, gc.Types[gc.TINT64], reg)
	}

	ginscon(as, c, &nr)
}

/*
 * generate
 *	as $c, n
 */
func ginscon(as int, c int64, n2 *gc.Node) {
	var n1 gc.Node

	switch as {
	case x86.AADDL,
		x86.AMOVL,
		x86.ALEAL:
		gc.Nodconst(&n1, gc.Types[gc.TINT32], c)

	default:
		gc.Nodconst(&n1, gc.Types[gc.TINT64], c)
	}

	if as != x86.AMOVQ && (c < -(1<<31) || c >= 1<<31) {
		// cannot have 64-bit immediate in ADD, etc.
		// instead, MOV into register first.
260
		var ntmp gc.Node
261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306
		regalloc(&ntmp, gc.Types[gc.TINT64], nil)

		gins(x86.AMOVQ, &n1, &ntmp)
		gins(as, &ntmp, n2)
		regfree(&ntmp)
		return
	}

	gins(as, &n1, n2)
}

/*
 * set up nodes representing 2^63
 */
var bigi gc.Node

var bigf gc.Node

var bignodes_did int

func bignodes() {
	if bignodes_did != 0 {
		return
	}
	bignodes_did = 1

	gc.Nodconst(&bigi, gc.Types[gc.TUINT64], 1)
	gc.Mpshiftfix(bigi.Val.U.Xval, 63)

	bigf = bigi
	bigf.Type = gc.Types[gc.TFLOAT64]
	bigf.Val.Ctype = gc.CTFLT
	bigf.Val.U.Fval = new(gc.Mpflt)
	gc.Mpmovefixflt(bigf.Val.U.Fval, bigi.Val.U.Xval)
}

/*
 * generate move:
 *	t = f
 * hard part is conversions.
 */
func gmove(f *gc.Node, t *gc.Node) {
	if gc.Debug['M'] != 0 {
		fmt.Printf("gmove %v -> %v\n", gc.Nconv(f, obj.FmtLong), gc.Nconv(t, obj.FmtLong))
	}

307 308 309
	ft := gc.Simsimtype(f.Type)
	tt := gc.Simsimtype(t.Type)
	cvt := t.Type
310 311 312 313 314 315 316

	if gc.Iscomplex[ft] != 0 || gc.Iscomplex[tt] != 0 {
		gc.Complexmove(f, t)
		return
	}

	// cannot have two memory operands
317 318
	var r1 gc.Node
	var a int
319
	if gc.Ismem(f) && gc.Ismem(t) {
320 321 322 323 324
		goto hard
	}

	// convert constant to desired type
	if f.Op == gc.OLITERAL {
325
		var con gc.Node
326 327 328 329 330
		gc.Convconst(&con, t.Type, &f.Val)
		f = &con
		ft = tt // so big switch will choose a simple mov

		// some constants can't move directly to memory.
331
		if gc.Ismem(t) {
332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532
			// float constants come from memory.
			if gc.Isfloat[tt] != 0 {
				goto hard
			}

			// 64-bit immediates are really 32-bit sign-extended
			// unless moving into a register.
			if gc.Isint[tt] != 0 {
				if gc.Mpcmpfixfix(con.Val.U.Xval, gc.Minintval[gc.TINT32]) < 0 {
					goto hard
				}
				if gc.Mpcmpfixfix(con.Val.U.Xval, gc.Maxintval[gc.TINT32]) > 0 {
					goto hard
				}
			}
		}
	}

	// value -> value copy, only one memory operand.
	// figure out the instruction to use.
	// break out of switch for one-instruction gins.
	// goto rdst for "destination must be register".
	// goto hard for "convert to cvt type first".
	// otherwise handle and return.

	switch uint32(ft)<<16 | uint32(tt) {
	default:
		gc.Fatal("gmove %v -> %v", gc.Tconv(f.Type, obj.FmtLong), gc.Tconv(t.Type, obj.FmtLong))

		/*
		 * integer copy and truncate
		 */
	case gc.TINT8<<16 | gc.TINT8, // same size
		gc.TINT8<<16 | gc.TUINT8,
		gc.TUINT8<<16 | gc.TINT8,
		gc.TUINT8<<16 | gc.TUINT8,
		gc.TINT16<<16 | gc.TINT8,
		// truncate
		gc.TUINT16<<16 | gc.TINT8,
		gc.TINT32<<16 | gc.TINT8,
		gc.TUINT32<<16 | gc.TINT8,
		gc.TINT64<<16 | gc.TINT8,
		gc.TUINT64<<16 | gc.TINT8,
		gc.TINT16<<16 | gc.TUINT8,
		gc.TUINT16<<16 | gc.TUINT8,
		gc.TINT32<<16 | gc.TUINT8,
		gc.TUINT32<<16 | gc.TUINT8,
		gc.TINT64<<16 | gc.TUINT8,
		gc.TUINT64<<16 | gc.TUINT8:
		a = x86.AMOVB

	case gc.TINT16<<16 | gc.TINT16, // same size
		gc.TINT16<<16 | gc.TUINT16,
		gc.TUINT16<<16 | gc.TINT16,
		gc.TUINT16<<16 | gc.TUINT16,
		gc.TINT32<<16 | gc.TINT16,
		// truncate
		gc.TUINT32<<16 | gc.TINT16,
		gc.TINT64<<16 | gc.TINT16,
		gc.TUINT64<<16 | gc.TINT16,
		gc.TINT32<<16 | gc.TUINT16,
		gc.TUINT32<<16 | gc.TUINT16,
		gc.TINT64<<16 | gc.TUINT16,
		gc.TUINT64<<16 | gc.TUINT16:
		a = x86.AMOVW

	case gc.TINT32<<16 | gc.TINT32, // same size
		gc.TINT32<<16 | gc.TUINT32,
		gc.TUINT32<<16 | gc.TINT32,
		gc.TUINT32<<16 | gc.TUINT32:
		a = x86.AMOVL

	case gc.TINT64<<16 | gc.TINT32, // truncate
		gc.TUINT64<<16 | gc.TINT32,
		gc.TINT64<<16 | gc.TUINT32,
		gc.TUINT64<<16 | gc.TUINT32:
		a = x86.AMOVQL

	case gc.TINT64<<16 | gc.TINT64, // same size
		gc.TINT64<<16 | gc.TUINT64,
		gc.TUINT64<<16 | gc.TINT64,
		gc.TUINT64<<16 | gc.TUINT64:
		a = x86.AMOVQ

		/*
		 * integer up-conversions
		 */
	case gc.TINT8<<16 | gc.TINT16, // sign extend int8
		gc.TINT8<<16 | gc.TUINT16:
		a = x86.AMOVBWSX

		goto rdst

	case gc.TINT8<<16 | gc.TINT32,
		gc.TINT8<<16 | gc.TUINT32:
		a = x86.AMOVBLSX
		goto rdst

	case gc.TINT8<<16 | gc.TINT64,
		gc.TINT8<<16 | gc.TUINT64:
		a = x86.AMOVBQSX
		goto rdst

	case gc.TUINT8<<16 | gc.TINT16, // zero extend uint8
		gc.TUINT8<<16 | gc.TUINT16:
		a = x86.AMOVBWZX

		goto rdst

	case gc.TUINT8<<16 | gc.TINT32,
		gc.TUINT8<<16 | gc.TUINT32:
		a = x86.AMOVBLZX
		goto rdst

	case gc.TUINT8<<16 | gc.TINT64,
		gc.TUINT8<<16 | gc.TUINT64:
		a = x86.AMOVBQZX
		goto rdst

	case gc.TINT16<<16 | gc.TINT32, // sign extend int16
		gc.TINT16<<16 | gc.TUINT32:
		a = x86.AMOVWLSX

		goto rdst

	case gc.TINT16<<16 | gc.TINT64,
		gc.TINT16<<16 | gc.TUINT64:
		a = x86.AMOVWQSX
		goto rdst

	case gc.TUINT16<<16 | gc.TINT32, // zero extend uint16
		gc.TUINT16<<16 | gc.TUINT32:
		a = x86.AMOVWLZX

		goto rdst

	case gc.TUINT16<<16 | gc.TINT64,
		gc.TUINT16<<16 | gc.TUINT64:
		a = x86.AMOVWQZX
		goto rdst

	case gc.TINT32<<16 | gc.TINT64, // sign extend int32
		gc.TINT32<<16 | gc.TUINT64:
		a = x86.AMOVLQSX

		goto rdst

		// AMOVL into a register zeros the top of the register,
	// so this is not always necessary, but if we rely on AMOVL
	// the optimizer is almost certain to screw with us.
	case gc.TUINT32<<16 | gc.TINT64, // zero extend uint32
		gc.TUINT32<<16 | gc.TUINT64:
		a = x86.AMOVLQZX

		goto rdst

		/*
		* float to integer
		 */
	case gc.TFLOAT32<<16 | gc.TINT32:
		a = x86.ACVTTSS2SL

		goto rdst

	case gc.TFLOAT64<<16 | gc.TINT32:
		a = x86.ACVTTSD2SL
		goto rdst

	case gc.TFLOAT32<<16 | gc.TINT64:
		a = x86.ACVTTSS2SQ
		goto rdst

	case gc.TFLOAT64<<16 | gc.TINT64:
		a = x86.ACVTTSD2SQ
		goto rdst

		// convert via int32.
	case gc.TFLOAT32<<16 | gc.TINT16,
		gc.TFLOAT32<<16 | gc.TINT8,
		gc.TFLOAT32<<16 | gc.TUINT16,
		gc.TFLOAT32<<16 | gc.TUINT8,
		gc.TFLOAT64<<16 | gc.TINT16,
		gc.TFLOAT64<<16 | gc.TINT8,
		gc.TFLOAT64<<16 | gc.TUINT16,
		gc.TFLOAT64<<16 | gc.TUINT8:
		cvt = gc.Types[gc.TINT32]

		goto hard

		// convert via int64.
	case gc.TFLOAT32<<16 | gc.TUINT32,
		gc.TFLOAT64<<16 | gc.TUINT32:
		cvt = gc.Types[gc.TINT64]

		goto hard

		// algorithm is:
	//	if small enough, use native float64 -> int64 conversion.
	//	otherwise, subtract 2^63, convert, and add it back.
	case gc.TFLOAT32<<16 | gc.TUINT64,
		gc.TFLOAT64<<16 | gc.TUINT64:
533
		a := x86.ACVTTSS2SQ
534 535 536 537 538

		if ft == gc.TFLOAT64 {
			a = x86.ACVTTSD2SQ
		}
		bignodes()
539
		var r1 gc.Node
540
		regalloc(&r1, gc.Types[ft], nil)
541
		var r2 gc.Node
542
		regalloc(&r2, gc.Types[tt], t)
543
		var r3 gc.Node
544
		regalloc(&r3, gc.Types[ft], nil)
545
		var r4 gc.Node
546 547 548
		regalloc(&r4, gc.Types[tt], nil)
		gins(optoas(gc.OAS, f.Type), f, &r1)
		gins(optoas(gc.OCMP, f.Type), &bigf, &r1)
549
		p1 := gc.Gbranch(optoas(gc.OLE, f.Type), nil, +1)
550
		gins(a, &r1, &r2)
551
		p2 := gc.Gbranch(obj.AJMP, nil, 0)
552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610
		gc.Patch(p1, gc.Pc)
		gins(optoas(gc.OAS, f.Type), &bigf, &r3)
		gins(optoas(gc.OSUB, f.Type), &r3, &r1)
		gins(a, &r1, &r2)
		gins(x86.AMOVQ, &bigi, &r4)
		gins(x86.AXORQ, &r4, &r2)
		gc.Patch(p2, gc.Pc)
		gmove(&r2, t)
		regfree(&r4)
		regfree(&r3)
		regfree(&r2)
		regfree(&r1)
		return

		/*
		 * integer to float
		 */
	case gc.TINT32<<16 | gc.TFLOAT32:
		a = x86.ACVTSL2SS

		goto rdst

	case gc.TINT32<<16 | gc.TFLOAT64:
		a = x86.ACVTSL2SD
		goto rdst

	case gc.TINT64<<16 | gc.TFLOAT32:
		a = x86.ACVTSQ2SS
		goto rdst

	case gc.TINT64<<16 | gc.TFLOAT64:
		a = x86.ACVTSQ2SD
		goto rdst

		// convert via int32
	case gc.TINT16<<16 | gc.TFLOAT32,
		gc.TINT16<<16 | gc.TFLOAT64,
		gc.TINT8<<16 | gc.TFLOAT32,
		gc.TINT8<<16 | gc.TFLOAT64,
		gc.TUINT16<<16 | gc.TFLOAT32,
		gc.TUINT16<<16 | gc.TFLOAT64,
		gc.TUINT8<<16 | gc.TFLOAT32,
		gc.TUINT8<<16 | gc.TFLOAT64:
		cvt = gc.Types[gc.TINT32]

		goto hard

		// convert via int64.
	case gc.TUINT32<<16 | gc.TFLOAT32,
		gc.TUINT32<<16 | gc.TFLOAT64:
		cvt = gc.Types[gc.TINT64]

		goto hard

		// algorithm is:
	//	if small enough, use native int64 -> uint64 conversion.
	//	otherwise, halve (rounding to odd?), convert, and double.
	case gc.TUINT64<<16 | gc.TFLOAT32,
		gc.TUINT64<<16 | gc.TFLOAT64:
611
		a := x86.ACVTSQ2SS
612 613 614 615

		if tt == gc.TFLOAT64 {
			a = x86.ACVTSQ2SD
		}
616
		var zero gc.Node
617
		gc.Nodconst(&zero, gc.Types[gc.TUINT64], 0)
618
		var one gc.Node
619
		gc.Nodconst(&one, gc.Types[gc.TUINT64], 1)
620
		var r1 gc.Node
621
		regalloc(&r1, f.Type, f)
622
		var r2 gc.Node
623
		regalloc(&r2, t.Type, t)
624
		var r3 gc.Node
625
		regalloc(&r3, f.Type, nil)
626
		var r4 gc.Node
627 628 629
		regalloc(&r4, f.Type, nil)
		gmove(f, &r1)
		gins(x86.ACMPQ, &r1, &zero)
630
		p1 := gc.Gbranch(x86.AJLT, nil, +1)
631
		gins(a, &r1, &r2)
632
		p2 := gc.Gbranch(obj.AJMP, nil, 0)
633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688
		gc.Patch(p1, gc.Pc)
		gmove(&r1, &r3)
		gins(x86.ASHRQ, &one, &r3)
		gmove(&r1, &r4)
		gins(x86.AANDL, &one, &r4)
		gins(x86.AORQ, &r4, &r3)
		gins(a, &r3, &r2)
		gins(optoas(gc.OADD, t.Type), &r2, &r2)
		gc.Patch(p2, gc.Pc)
		gmove(&r2, t)
		regfree(&r4)
		regfree(&r3)
		regfree(&r2)
		regfree(&r1)
		return

		/*
		 * float to float
		 */
	case gc.TFLOAT32<<16 | gc.TFLOAT32:
		a = x86.AMOVSS

	case gc.TFLOAT64<<16 | gc.TFLOAT64:
		a = x86.AMOVSD

	case gc.TFLOAT32<<16 | gc.TFLOAT64:
		a = x86.ACVTSS2SD
		goto rdst

	case gc.TFLOAT64<<16 | gc.TFLOAT32:
		a = x86.ACVTSD2SS
		goto rdst
	}

	gins(a, f, t)
	return

	// requires register destination
rdst:
	regalloc(&r1, t.Type, t)

	gins(a, f, &r1)
	gmove(&r1, t)
	regfree(&r1)
	return

	// requires register intermediate
hard:
	regalloc(&r1, cvt, t)

	gmove(f, &r1)
	gmove(&r1, t)
	regfree(&r1)
	return
}

689
func samaddr(f *gc.Node, t *gc.Node) bool {
690
	if f.Op != t.Op {
691
		return false
692 693 694 695 696 697 698
	}

	switch f.Op {
	case gc.OREGISTER:
		if f.Val.U.Reg != t.Val.U.Reg {
			break
		}
699
		return true
700 701
	}

702
	return false
703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735
}

/*
 * generate one instruction:
 *	as f, t
 */
func gins(as int, f *gc.Node, t *gc.Node) *obj.Prog {
	//	Node nod;

	//	if(f != N && f->op == OINDEX) {
	//		regalloc(&nod, &regnode, Z);
	//		v = constnode.vconst;
	//		cgen(f->right, &nod);
	//		constnode.vconst = v;
	//		idx.reg = nod.reg;
	//		regfree(&nod);
	//	}
	//	if(t != N && t->op == OINDEX) {
	//		regalloc(&nod, &regnode, Z);
	//		v = constnode.vconst;
	//		cgen(t->right, &nod);
	//		constnode.vconst = v;
	//		idx.reg = nod.reg;
	//		regfree(&nod);
	//	}

	switch as {
	case x86.AMOVB,
		x86.AMOVW,
		x86.AMOVL,
		x86.AMOVQ,
		x86.AMOVSS,
		x86.AMOVSD:
736
		if f != nil && t != nil && samaddr(f, t) {
737 738 739 740
			return nil
		}

	case x86.ALEAQ:
741
		if f != nil && gc.Isconst(f, gc.CTNIL) {
742 743 744 745
			gc.Fatal("gins LEAQ nil %v", gc.Tconv(f.Type, 0))
		}
	}

746 747
	af := obj.Addr{}
	at := obj.Addr{}
748 749 750 751 752 753
	if f != nil {
		gc.Naddr(f, &af, 1)
	}
	if t != nil {
		gc.Naddr(t, &at, 1)
	}
754
	p := gc.Prog(as)
755 756 757 758 759 760 761 762 763 764
	if f != nil {
		p.From = af
	}
	if t != nil {
		p.To = at
	}
	if gc.Debug['g'] != 0 {
		fmt.Printf("%v\n", p)
	}

765
	w := int32(0)
766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804
	switch as {
	case x86.AMOVB:
		w = 1

	case x86.AMOVW:
		w = 2

	case x86.AMOVL:
		w = 4

	case x86.AMOVQ:
		w = 8
	}

	if w != 0 && ((f != nil && af.Width < int64(w)) || (t != nil && at.Width > int64(w))) {
		gc.Dump("f", f)
		gc.Dump("t", t)
		gc.Fatal("bad width: %v (%d, %d)\n", p, af.Width, at.Width)
	}

	if p.To.Type == obj.TYPE_ADDR && w > 0 {
		gc.Fatal("bad use of addr: %v", p)
	}

	return p
}

func fixlargeoffset(n *gc.Node) {
	if n == nil {
		return
	}
	if n.Op != gc.OINDREG {
		return
	}
	if n.Val.U.Reg == x86.REG_SP { // stack offset cannot be large
		return
	}
	if n.Xoffset != int64(int32(n.Xoffset)) {
		// offset too large, add to register instead.
805
		a := *n
806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823

		a.Op = gc.OREGISTER
		a.Type = gc.Types[gc.Tptr]
		a.Xoffset = 0
		gc.Cgen_checknil(&a)
		ginscon(optoas(gc.OADD, gc.Types[gc.Tptr]), n.Xoffset, &a)
		n.Xoffset = 0
	}
}

/*
 * return Axxx for Oxxx on type t.
 */
func optoas(op int, t *gc.Type) int {
	if t == nil {
		gc.Fatal("optoas: t is nil")
	}

824
	a := obj.AXXX
825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306
	switch uint32(op)<<16 | uint32(gc.Simtype[t.Etype]) {
	default:
		gc.Fatal("optoas: no entry %v-%v", gc.Oconv(int(op), 0), gc.Tconv(t, 0))

	case gc.OADDR<<16 | gc.TPTR32:
		a = x86.ALEAL

	case gc.OADDR<<16 | gc.TPTR64:
		a = x86.ALEAQ

	case gc.OEQ<<16 | gc.TBOOL,
		gc.OEQ<<16 | gc.TINT8,
		gc.OEQ<<16 | gc.TUINT8,
		gc.OEQ<<16 | gc.TINT16,
		gc.OEQ<<16 | gc.TUINT16,
		gc.OEQ<<16 | gc.TINT32,
		gc.OEQ<<16 | gc.TUINT32,
		gc.OEQ<<16 | gc.TINT64,
		gc.OEQ<<16 | gc.TUINT64,
		gc.OEQ<<16 | gc.TPTR32,
		gc.OEQ<<16 | gc.TPTR64,
		gc.OEQ<<16 | gc.TFLOAT32,
		gc.OEQ<<16 | gc.TFLOAT64:
		a = x86.AJEQ

	case gc.ONE<<16 | gc.TBOOL,
		gc.ONE<<16 | gc.TINT8,
		gc.ONE<<16 | gc.TUINT8,
		gc.ONE<<16 | gc.TINT16,
		gc.ONE<<16 | gc.TUINT16,
		gc.ONE<<16 | gc.TINT32,
		gc.ONE<<16 | gc.TUINT32,
		gc.ONE<<16 | gc.TINT64,
		gc.ONE<<16 | gc.TUINT64,
		gc.ONE<<16 | gc.TPTR32,
		gc.ONE<<16 | gc.TPTR64,
		gc.ONE<<16 | gc.TFLOAT32,
		gc.ONE<<16 | gc.TFLOAT64:
		a = x86.AJNE

	case gc.OLT<<16 | gc.TINT8,
		gc.OLT<<16 | gc.TINT16,
		gc.OLT<<16 | gc.TINT32,
		gc.OLT<<16 | gc.TINT64:
		a = x86.AJLT

	case gc.OLT<<16 | gc.TUINT8,
		gc.OLT<<16 | gc.TUINT16,
		gc.OLT<<16 | gc.TUINT32,
		gc.OLT<<16 | gc.TUINT64:
		a = x86.AJCS

	case gc.OLE<<16 | gc.TINT8,
		gc.OLE<<16 | gc.TINT16,
		gc.OLE<<16 | gc.TINT32,
		gc.OLE<<16 | gc.TINT64:
		a = x86.AJLE

	case gc.OLE<<16 | gc.TUINT8,
		gc.OLE<<16 | gc.TUINT16,
		gc.OLE<<16 | gc.TUINT32,
		gc.OLE<<16 | gc.TUINT64:
		a = x86.AJLS

	case gc.OGT<<16 | gc.TINT8,
		gc.OGT<<16 | gc.TINT16,
		gc.OGT<<16 | gc.TINT32,
		gc.OGT<<16 | gc.TINT64:
		a = x86.AJGT

	case gc.OGT<<16 | gc.TUINT8,
		gc.OGT<<16 | gc.TUINT16,
		gc.OGT<<16 | gc.TUINT32,
		gc.OGT<<16 | gc.TUINT64,
		gc.OLT<<16 | gc.TFLOAT32,
		gc.OLT<<16 | gc.TFLOAT64:
		a = x86.AJHI

	case gc.OGE<<16 | gc.TINT8,
		gc.OGE<<16 | gc.TINT16,
		gc.OGE<<16 | gc.TINT32,
		gc.OGE<<16 | gc.TINT64:
		a = x86.AJGE

	case gc.OGE<<16 | gc.TUINT8,
		gc.OGE<<16 | gc.TUINT16,
		gc.OGE<<16 | gc.TUINT32,
		gc.OGE<<16 | gc.TUINT64,
		gc.OLE<<16 | gc.TFLOAT32,
		gc.OLE<<16 | gc.TFLOAT64:
		a = x86.AJCC

	case gc.OCMP<<16 | gc.TBOOL,
		gc.OCMP<<16 | gc.TINT8,
		gc.OCMP<<16 | gc.TUINT8:
		a = x86.ACMPB

	case gc.OCMP<<16 | gc.TINT16,
		gc.OCMP<<16 | gc.TUINT16:
		a = x86.ACMPW

	case gc.OCMP<<16 | gc.TINT32,
		gc.OCMP<<16 | gc.TUINT32,
		gc.OCMP<<16 | gc.TPTR32:
		a = x86.ACMPL

	case gc.OCMP<<16 | gc.TINT64,
		gc.OCMP<<16 | gc.TUINT64,
		gc.OCMP<<16 | gc.TPTR64:
		a = x86.ACMPQ

	case gc.OCMP<<16 | gc.TFLOAT32:
		a = x86.AUCOMISS

	case gc.OCMP<<16 | gc.TFLOAT64:
		a = x86.AUCOMISD

	case gc.OAS<<16 | gc.TBOOL,
		gc.OAS<<16 | gc.TINT8,
		gc.OAS<<16 | gc.TUINT8:
		a = x86.AMOVB

	case gc.OAS<<16 | gc.TINT16,
		gc.OAS<<16 | gc.TUINT16:
		a = x86.AMOVW

	case gc.OAS<<16 | gc.TINT32,
		gc.OAS<<16 | gc.TUINT32,
		gc.OAS<<16 | gc.TPTR32:
		a = x86.AMOVL

	case gc.OAS<<16 | gc.TINT64,
		gc.OAS<<16 | gc.TUINT64,
		gc.OAS<<16 | gc.TPTR64:
		a = x86.AMOVQ

	case gc.OAS<<16 | gc.TFLOAT32:
		a = x86.AMOVSS

	case gc.OAS<<16 | gc.TFLOAT64:
		a = x86.AMOVSD

	case gc.OADD<<16 | gc.TINT8,
		gc.OADD<<16 | gc.TUINT8:
		a = x86.AADDB

	case gc.OADD<<16 | gc.TINT16,
		gc.OADD<<16 | gc.TUINT16:
		a = x86.AADDW

	case gc.OADD<<16 | gc.TINT32,
		gc.OADD<<16 | gc.TUINT32,
		gc.OADD<<16 | gc.TPTR32:
		a = x86.AADDL

	case gc.OADD<<16 | gc.TINT64,
		gc.OADD<<16 | gc.TUINT64,
		gc.OADD<<16 | gc.TPTR64:
		a = x86.AADDQ

	case gc.OADD<<16 | gc.TFLOAT32:
		a = x86.AADDSS

	case gc.OADD<<16 | gc.TFLOAT64:
		a = x86.AADDSD

	case gc.OSUB<<16 | gc.TINT8,
		gc.OSUB<<16 | gc.TUINT8:
		a = x86.ASUBB

	case gc.OSUB<<16 | gc.TINT16,
		gc.OSUB<<16 | gc.TUINT16:
		a = x86.ASUBW

	case gc.OSUB<<16 | gc.TINT32,
		gc.OSUB<<16 | gc.TUINT32,
		gc.OSUB<<16 | gc.TPTR32:
		a = x86.ASUBL

	case gc.OSUB<<16 | gc.TINT64,
		gc.OSUB<<16 | gc.TUINT64,
		gc.OSUB<<16 | gc.TPTR64:
		a = x86.ASUBQ

	case gc.OSUB<<16 | gc.TFLOAT32:
		a = x86.ASUBSS

	case gc.OSUB<<16 | gc.TFLOAT64:
		a = x86.ASUBSD

	case gc.OINC<<16 | gc.TINT8,
		gc.OINC<<16 | gc.TUINT8:
		a = x86.AINCB

	case gc.OINC<<16 | gc.TINT16,
		gc.OINC<<16 | gc.TUINT16:
		a = x86.AINCW

	case gc.OINC<<16 | gc.TINT32,
		gc.OINC<<16 | gc.TUINT32,
		gc.OINC<<16 | gc.TPTR32:
		a = x86.AINCL

	case gc.OINC<<16 | gc.TINT64,
		gc.OINC<<16 | gc.TUINT64,
		gc.OINC<<16 | gc.TPTR64:
		a = x86.AINCQ

	case gc.ODEC<<16 | gc.TINT8,
		gc.ODEC<<16 | gc.TUINT8:
		a = x86.ADECB

	case gc.ODEC<<16 | gc.TINT16,
		gc.ODEC<<16 | gc.TUINT16:
		a = x86.ADECW

	case gc.ODEC<<16 | gc.TINT32,
		gc.ODEC<<16 | gc.TUINT32,
		gc.ODEC<<16 | gc.TPTR32:
		a = x86.ADECL

	case gc.ODEC<<16 | gc.TINT64,
		gc.ODEC<<16 | gc.TUINT64,
		gc.ODEC<<16 | gc.TPTR64:
		a = x86.ADECQ

	case gc.OMINUS<<16 | gc.TINT8,
		gc.OMINUS<<16 | gc.TUINT8:
		a = x86.ANEGB

	case gc.OMINUS<<16 | gc.TINT16,
		gc.OMINUS<<16 | gc.TUINT16:
		a = x86.ANEGW

	case gc.OMINUS<<16 | gc.TINT32,
		gc.OMINUS<<16 | gc.TUINT32,
		gc.OMINUS<<16 | gc.TPTR32:
		a = x86.ANEGL

	case gc.OMINUS<<16 | gc.TINT64,
		gc.OMINUS<<16 | gc.TUINT64,
		gc.OMINUS<<16 | gc.TPTR64:
		a = x86.ANEGQ

	case gc.OAND<<16 | gc.TINT8,
		gc.OAND<<16 | gc.TUINT8:
		a = x86.AANDB

	case gc.OAND<<16 | gc.TINT16,
		gc.OAND<<16 | gc.TUINT16:
		a = x86.AANDW

	case gc.OAND<<16 | gc.TINT32,
		gc.OAND<<16 | gc.TUINT32,
		gc.OAND<<16 | gc.TPTR32:
		a = x86.AANDL

	case gc.OAND<<16 | gc.TINT64,
		gc.OAND<<16 | gc.TUINT64,
		gc.OAND<<16 | gc.TPTR64:
		a = x86.AANDQ

	case gc.OOR<<16 | gc.TINT8,
		gc.OOR<<16 | gc.TUINT8:
		a = x86.AORB

	case gc.OOR<<16 | gc.TINT16,
		gc.OOR<<16 | gc.TUINT16:
		a = x86.AORW

	case gc.OOR<<16 | gc.TINT32,
		gc.OOR<<16 | gc.TUINT32,
		gc.OOR<<16 | gc.TPTR32:
		a = x86.AORL

	case gc.OOR<<16 | gc.TINT64,
		gc.OOR<<16 | gc.TUINT64,
		gc.OOR<<16 | gc.TPTR64:
		a = x86.AORQ

	case gc.OXOR<<16 | gc.TINT8,
		gc.OXOR<<16 | gc.TUINT8:
		a = x86.AXORB

	case gc.OXOR<<16 | gc.TINT16,
		gc.OXOR<<16 | gc.TUINT16:
		a = x86.AXORW

	case gc.OXOR<<16 | gc.TINT32,
		gc.OXOR<<16 | gc.TUINT32,
		gc.OXOR<<16 | gc.TPTR32:
		a = x86.AXORL

	case gc.OXOR<<16 | gc.TINT64,
		gc.OXOR<<16 | gc.TUINT64,
		gc.OXOR<<16 | gc.TPTR64:
		a = x86.AXORQ

	case gc.OLROT<<16 | gc.TINT8,
		gc.OLROT<<16 | gc.TUINT8:
		a = x86.AROLB

	case gc.OLROT<<16 | gc.TINT16,
		gc.OLROT<<16 | gc.TUINT16:
		a = x86.AROLW

	case gc.OLROT<<16 | gc.TINT32,
		gc.OLROT<<16 | gc.TUINT32,
		gc.OLROT<<16 | gc.TPTR32:
		a = x86.AROLL

	case gc.OLROT<<16 | gc.TINT64,
		gc.OLROT<<16 | gc.TUINT64,
		gc.OLROT<<16 | gc.TPTR64:
		a = x86.AROLQ

	case gc.OLSH<<16 | gc.TINT8,
		gc.OLSH<<16 | gc.TUINT8:
		a = x86.ASHLB

	case gc.OLSH<<16 | gc.TINT16,
		gc.OLSH<<16 | gc.TUINT16:
		a = x86.ASHLW

	case gc.OLSH<<16 | gc.TINT32,
		gc.OLSH<<16 | gc.TUINT32,
		gc.OLSH<<16 | gc.TPTR32:
		a = x86.ASHLL

	case gc.OLSH<<16 | gc.TINT64,
		gc.OLSH<<16 | gc.TUINT64,
		gc.OLSH<<16 | gc.TPTR64:
		a = x86.ASHLQ

	case gc.ORSH<<16 | gc.TUINT8:
		a = x86.ASHRB

	case gc.ORSH<<16 | gc.TUINT16:
		a = x86.ASHRW

	case gc.ORSH<<16 | gc.TUINT32,
		gc.ORSH<<16 | gc.TPTR32:
		a = x86.ASHRL

	case gc.ORSH<<16 | gc.TUINT64,
		gc.ORSH<<16 | gc.TPTR64:
		a = x86.ASHRQ

	case gc.ORSH<<16 | gc.TINT8:
		a = x86.ASARB

	case gc.ORSH<<16 | gc.TINT16:
		a = x86.ASARW

	case gc.ORSH<<16 | gc.TINT32:
		a = x86.ASARL

	case gc.ORSH<<16 | gc.TINT64:
		a = x86.ASARQ

	case gc.ORROTC<<16 | gc.TINT8,
		gc.ORROTC<<16 | gc.TUINT8:
		a = x86.ARCRB

	case gc.ORROTC<<16 | gc.TINT16,
		gc.ORROTC<<16 | gc.TUINT16:
		a = x86.ARCRW

	case gc.ORROTC<<16 | gc.TINT32,
		gc.ORROTC<<16 | gc.TUINT32:
		a = x86.ARCRL

	case gc.ORROTC<<16 | gc.TINT64,
		gc.ORROTC<<16 | gc.TUINT64:
		a = x86.ARCRQ

	case gc.OHMUL<<16 | gc.TINT8,
		gc.OMUL<<16 | gc.TINT8,
		gc.OMUL<<16 | gc.TUINT8:
		a = x86.AIMULB

	case gc.OHMUL<<16 | gc.TINT16,
		gc.OMUL<<16 | gc.TINT16,
		gc.OMUL<<16 | gc.TUINT16:
		a = x86.AIMULW

	case gc.OHMUL<<16 | gc.TINT32,
		gc.OMUL<<16 | gc.TINT32,
		gc.OMUL<<16 | gc.TUINT32,
		gc.OMUL<<16 | gc.TPTR32:
		a = x86.AIMULL

	case gc.OHMUL<<16 | gc.TINT64,
		gc.OMUL<<16 | gc.TINT64,
		gc.OMUL<<16 | gc.TUINT64,
		gc.OMUL<<16 | gc.TPTR64:
		a = x86.AIMULQ

	case gc.OHMUL<<16 | gc.TUINT8:
		a = x86.AMULB

	case gc.OHMUL<<16 | gc.TUINT16:
		a = x86.AMULW

	case gc.OHMUL<<16 | gc.TUINT32,
		gc.OHMUL<<16 | gc.TPTR32:
		a = x86.AMULL

	case gc.OHMUL<<16 | gc.TUINT64,
		gc.OHMUL<<16 | gc.TPTR64:
		a = x86.AMULQ

	case gc.OMUL<<16 | gc.TFLOAT32:
		a = x86.AMULSS

	case gc.OMUL<<16 | gc.TFLOAT64:
		a = x86.AMULSD

	case gc.ODIV<<16 | gc.TINT8,
		gc.OMOD<<16 | gc.TINT8:
		a = x86.AIDIVB

	case gc.ODIV<<16 | gc.TUINT8,
		gc.OMOD<<16 | gc.TUINT8:
		a = x86.ADIVB

	case gc.ODIV<<16 | gc.TINT16,
		gc.OMOD<<16 | gc.TINT16:
		a = x86.AIDIVW

	case gc.ODIV<<16 | gc.TUINT16,
		gc.OMOD<<16 | gc.TUINT16:
		a = x86.ADIVW

	case gc.ODIV<<16 | gc.TINT32,
		gc.OMOD<<16 | gc.TINT32:
		a = x86.AIDIVL

	case gc.ODIV<<16 | gc.TUINT32,
		gc.ODIV<<16 | gc.TPTR32,
		gc.OMOD<<16 | gc.TUINT32,
		gc.OMOD<<16 | gc.TPTR32:
		a = x86.ADIVL

	case gc.ODIV<<16 | gc.TINT64,
		gc.OMOD<<16 | gc.TINT64:
		a = x86.AIDIVQ

	case gc.ODIV<<16 | gc.TUINT64,
		gc.ODIV<<16 | gc.TPTR64,
		gc.OMOD<<16 | gc.TUINT64,
		gc.OMOD<<16 | gc.TPTR64:
		a = x86.ADIVQ

	case gc.OEXTEND<<16 | gc.TINT16:
		a = x86.ACWD

	case gc.OEXTEND<<16 | gc.TINT32:
		a = x86.ACDQ

	case gc.OEXTEND<<16 | gc.TINT64:
		a = x86.ACQO

	case gc.ODIV<<16 | gc.TFLOAT32:
		a = x86.ADIVSS

	case gc.ODIV<<16 | gc.TFLOAT64:
		a = x86.ADIVSD
	}

	return a
}

const (
	ODynam   = 1 << 0
	OAddable = 1 << 1
)

var clean [20]gc.Node

var cleani int = 0

1307
func xgen(n *gc.Node, a *gc.Node, o int) bool {
1308 1309 1310 1311 1312 1313
	regalloc(a, gc.Types[gc.Tptr], nil)

	if o&ODynam != 0 {
		if n.Addable != 0 {
			if n.Op != gc.OINDREG {
				if n.Op != gc.OREGISTER {
1314
					return true
1315 1316 1317 1318 1319 1320
				}
			}
		}
	}

	agen(n, a)
1321
	return false
1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344
}

func sudoclean() {
	if clean[cleani-1].Op != gc.OEMPTY {
		regfree(&clean[cleani-1])
	}
	if clean[cleani-2].Op != gc.OEMPTY {
		regfree(&clean[cleani-2])
	}
	cleani -= 2
}

/*
 * generate code to compute address of n,
 * a reference to a (perhaps nested) field inside
 * an array or struct.
 * return 0 on failure, 1 on success.
 * on success, leaves usable address in a.
 *
 * caller is responsible for calling sudoclean
 * after successful sudoaddable,
 * to release the register used for a.
 */
1345
func sudoaddable(as int, n *gc.Node, a *obj.Addr) bool {
1346
	if n.Type == nil {
1347
		return false
1348 1349 1350 1351
	}

	*a = obj.Addr{}

1352 1353 1354 1355 1356 1357
	var o int
	var n1 gc.Node
	var oary [10]int64
	var nn *gc.Node
	var reg *gc.Node
	var reg1 *gc.Node
1358 1359
	switch n.Op {
	case gc.OLITERAL:
1360
		if !gc.Isconst(n, gc.CTINT) {
1361 1362
			break
		}
1363
		v := gc.Mpgetfix(n.Val.U.Xval)
1364 1365 1366 1367 1368 1369 1370 1371 1372
		if v >= 32000 || v <= -32000 {
			break
		}
		goto lit

	case gc.ODOT,
		gc.ODOTPTR:
		cleani += 2
		reg = &clean[cleani-1]
1373
		reg1 := &clean[cleani-2]
1374 1375 1376 1377 1378
		reg.Op = gc.OEMPTY
		reg1.Op = gc.OEMPTY
		goto odot

	case gc.OINDEX:
1379
		return false
1380 1381
	}

1382
	return false
1383 1384 1385 1386

lit:
	switch as {
	default:
1387
		return false
1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439

	case x86.AADDB,
		x86.AADDW,
		x86.AADDL,
		x86.AADDQ,
		x86.ASUBB,
		x86.ASUBW,
		x86.ASUBL,
		x86.ASUBQ,
		x86.AANDB,
		x86.AANDW,
		x86.AANDL,
		x86.AANDQ,
		x86.AORB,
		x86.AORW,
		x86.AORL,
		x86.AORQ,
		x86.AXORB,
		x86.AXORW,
		x86.AXORL,
		x86.AXORQ,
		x86.AINCB,
		x86.AINCW,
		x86.AINCL,
		x86.AINCQ,
		x86.ADECB,
		x86.ADECW,
		x86.ADECL,
		x86.ADECQ,
		x86.AMOVB,
		x86.AMOVW,
		x86.AMOVL,
		x86.AMOVQ:
		break
	}

	cleani += 2
	reg = &clean[cleani-1]
	reg1 = &clean[cleani-2]
	reg.Op = gc.OEMPTY
	reg1.Op = gc.OEMPTY
	gc.Naddr(n, a, 1)
	goto yes

odot:
	o = gc.Dotoffset(n, oary[:], &nn)
	if nn == nil {
		goto no
	}

	if nn.Addable != 0 && o == 1 && oary[0] >= 0 {
		// directly addressable set of DOTs
1440
		n1 := *nn
1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459

		n1.Type = n.Type
		n1.Xoffset += oary[0]
		gc.Naddr(&n1, a, 1)
		goto yes
	}

	regalloc(reg, gc.Types[gc.Tptr], nil)
	n1 = *reg
	n1.Op = gc.OINDREG
	if oary[0] >= 0 {
		agen(nn, reg)
		n1.Xoffset = oary[0]
	} else {
		cgen(nn, reg)
		gc.Cgen_checknil(reg)
		n1.Xoffset = -(oary[0] + 1)
	}

1460
	for i := 1; i < o; i++ {
1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475
		if oary[i] >= 0 {
			gc.Fatal("can't happen")
		}
		gins(movptr, &n1, reg)
		gc.Cgen_checknil(reg)
		n1.Xoffset = -(oary[i] + 1)
	}

	a.Type = obj.TYPE_NONE
	a.Index = obj.TYPE_NONE
	fixlargeoffset(&n1)
	gc.Naddr(&n1, a, 1)
	goto yes

yes:
1476
	return true
1477 1478 1479

no:
	sudoclean()
1480
	return false
1481
}