* doc/md.texi: Update documentation of MIPS constraints.

[pf3gnuchains/gcc-fork.git] / gcc / config / mips / mips.md

;;  Mips.md          Machine Description for MIPS based processors
;;  Copyright (C) 1989, 1990, 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
;;  1999, 2000, 2001, 2002, 2003, 2004, 2005 Free Software Foundation, Inc.
;;  Contributed by   A. Lichnewsky, lich@inria.inria.fr
;;  Changes by       Michael Meissner, meissner@osf.org
;;  64 bit r4000 support by Ian Lance Taylor, ian@cygnus.com, and
;;  Brendan Eich, brendan@microunity.com.

;; This file is part of GCC.

;; GCC is free software; you can redistribute it and/or modify
;; it under the terms of the GNU General Public License as published by
;; the Free Software Foundation; either version 2, or (at your option)
;; any later version.

;; GCC is distributed in the hope that it will be useful,
;; but WITHOUT ANY WARRANTY; without even the implied warranty of
;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
;; GNU General Public License for more details.

;; You should have received a copy of the GNU General Public License
;; along with GCC; see the file COPYING.  If not, write to
;; the Free Software Foundation, 51 Franklin Street, Fifth Floor,
;; Boston, MA 02110-1301, USA.

(define_constants
  [(UNSPEC_LOAD_DF_LOW           0)
   (UNSPEC_LOAD_DF_HIGH          1)
   (UNSPEC_STORE_DF_HIGH         2)
   (UNSPEC_GET_FNADDR            3)
   (UNSPEC_BLOCKAGE              4)
   (UNSPEC_CPRESTORE             5)
   (UNSPEC_EH_RECEIVER           6)
   (UNSPEC_EH_RETURN             7)
   (UNSPEC_CONSTTABLE_INT        8)
   (UNSPEC_CONSTTABLE_FLOAT      9)
   (UNSPEC_ALIGN                14)
   (UNSPEC_HIGH                 17)
   (UNSPEC_LOAD_LEFT            18)
   (UNSPEC_LOAD_RIGHT           19)
   (UNSPEC_STORE_LEFT           20)
   (UNSPEC_STORE_RIGHT          21)
   (UNSPEC_LOADGP               22)
   (UNSPEC_LOAD_CALL            23)
   (UNSPEC_LOAD_GOT             24)
   (UNSPEC_GP                   25)
   (UNSPEC_MFHILO               26)
   (UNSPEC_TLS_LDM              27)
   (UNSPEC_TLS_GET_TP           28)

   (UNSPEC_ADDRESS_FIRST        100)

   (FAKE_CALL_REGNO             79)

   ;; For MIPS Paired-Singled Floating Point Instructions.

   (UNSPEC_MOVE_TF_PS           200)
   (UNSPEC_C                    201)

   ;; MIPS64/MIPS32R2 alnv.ps
   (UNSPEC_ALNV_PS              202)

   ;; MIPS-3D instructions
   (UNSPEC_CABS                 203)

   (UNSPEC_ADDR_PS              204)
   (UNSPEC_CVT_PW_PS            205)
   (UNSPEC_CVT_PS_PW            206)
   (UNSPEC_MULR_PS              207)

   (UNSPEC_RSQRT1               208)
   (UNSPEC_RSQRT2               209)
   (UNSPEC_RECIP1               210)
   (UNSPEC_RECIP2               211)

   ;; MIPS DSP ASE Revision 0.98 3/24/2005
   (UNSPEC_ADDQ                 300)
   (UNSPEC_ADDQ_S               301)
   (UNSPEC_SUBQ                 302)
   (UNSPEC_SUBQ_S               303)
   (UNSPEC_ADDSC                304)
   (UNSPEC_ADDWC                305)
   (UNSPEC_MODSUB               306)
   (UNSPEC_RADDU_W_QB           307)
   (UNSPEC_ABSQ_S               308)
   (UNSPEC_PRECRQ_QB_PH         309)
   (UNSPEC_PRECRQ_PH_W          310)
   (UNSPEC_PRECRQ_RS_PH_W       311)
   (UNSPEC_PRECRQU_S_QB_PH      312)
   (UNSPEC_PRECEQ_W_PHL         313)
   (UNSPEC_PRECEQ_W_PHR         314)
   (UNSPEC_PRECEQU_PH_QBL       315)
   (UNSPEC_PRECEQU_PH_QBR       316)
   (UNSPEC_PRECEQU_PH_QBLA      317)
   (UNSPEC_PRECEQU_PH_QBRA      318)
   (UNSPEC_PRECEU_PH_QBL        319)
   (UNSPEC_PRECEU_PH_QBR        320)
   (UNSPEC_PRECEU_PH_QBLA       321)
   (UNSPEC_PRECEU_PH_QBRA       322)
   (UNSPEC_SHLL                 323)
   (UNSPEC_SHLL_S               324)
   (UNSPEC_SHRL_QB              325)
   (UNSPEC_SHRA_PH              326)
   (UNSPEC_SHRA_R               327)
   (UNSPEC_MULEU_S_PH_QBL       328)
   (UNSPEC_MULEU_S_PH_QBR       329)
   (UNSPEC_MULQ_RS_PH           330)
   (UNSPEC_MULEQ_S_W_PHL        331)
   (UNSPEC_MULEQ_S_W_PHR        332)
   (UNSPEC_DPAU_H_QBL           333)
   (UNSPEC_DPAU_H_QBR           334)
   (UNSPEC_DPSU_H_QBL           335)
   (UNSPEC_DPSU_H_QBR           336)
   (UNSPEC_DPAQ_S_W_PH          337)
   (UNSPEC_DPSQ_S_W_PH          338)
   (UNSPEC_MULSAQ_S_W_PH        339)
   (UNSPEC_DPAQ_SA_L_W          340)
   (UNSPEC_DPSQ_SA_L_W          341)
   (UNSPEC_MAQ_S_W_PHL          342)
   (UNSPEC_MAQ_S_W_PHR          343)
   (UNSPEC_MAQ_SA_W_PHL         344)
   (UNSPEC_MAQ_SA_W_PHR         345)
   (UNSPEC_BITREV               346)
   (UNSPEC_INSV                 347)
   (UNSPEC_REPL_QB              348)
   (UNSPEC_REPL_PH              349)
   (UNSPEC_CMP_EQ               350)
   (UNSPEC_CMP_LT               351)
   (UNSPEC_CMP_LE               352)
   (UNSPEC_CMPGU_EQ_QB          353)
   (UNSPEC_CMPGU_LT_QB          354)
   (UNSPEC_CMPGU_LE_QB          355)
   (UNSPEC_PICK                 356)
   (UNSPEC_PACKRL_PH            357)
   (UNSPEC_EXTR_W               358)
   (UNSPEC_EXTR_R_W             359)
   (UNSPEC_EXTR_RS_W            360)
   (UNSPEC_EXTR_S_H             361)
   (UNSPEC_EXTP                 362)
   (UNSPEC_EXTPDP               363)
   (UNSPEC_SHILO                364)
   (UNSPEC_MTHLIP               365)
   (UNSPEC_WRDSP                366)
   (UNSPEC_RDDSP                367)
  ]
)

(include "predicates.md")
(include "constraints.md")
\f
;; ....................
;;
;;      Attributes
;;
;; ....................

(define_attr "got" "unset,xgot_high,load"
  (const_string "unset"))

;; For jal instructions, this attribute is DIRECT when the target address
;; is symbolic and INDIRECT when it is a register.
(define_attr "jal" "unset,direct,indirect"
  (const_string "unset"))

;; This attribute is YES if the instruction is a jal macro (not a
;; real jal instruction).
;;
;; jal is always a macro in SVR4 PIC since it includes an instruction to
;; restore $gp.  Direct jals are also macros in NewABI PIC since they
;; load the target address into $25.
(define_attr "jal_macro" "no,yes"
  (cond [(eq_attr "jal" "direct")
         (symbol_ref "TARGET_ABICALLS != 0")
         (eq_attr "jal" "indirect")
         (symbol_ref "(TARGET_ABICALLS && !TARGET_NEWABI) != 0")]
        (const_string "no")))

;; Classification of each insn.
;; branch       conditional branch
;; jump         unconditional jump
;; call         unconditional call
;; load         load instruction(s)
;; fpload       floating point load
;; fpidxload    floating point indexed load
;; store        store instruction(s)
;; fpstore      floating point store
;; fpidxstore   floating point indexed store
;; prefetch     memory prefetch (register + offset)
;; prefetchx    memory indexed prefetch (register + register)
;; condmove     conditional moves
;; xfer         transfer to/from coprocessor
;; mthilo       transfer to hi/lo registers
;; mfhilo       transfer from hi/lo registers
;; const        load constant
;; arith        integer arithmetic and logical instructions
;; shift        integer shift instructions
;; slt          set less than instructions
;; clz          the clz and clo instructions
;; trap         trap if instructions
;; imul         integer multiply 2 operands
;; imul3        integer multiply 3 operands
;; imadd        integer multiply-add
;; idiv         integer divide
;; fmove        floating point register move
;; fadd         floating point add/subtract
;; fmul         floating point multiply
;; fmadd        floating point multiply-add
;; fdiv         floating point divide
;; frdiv        floating point reciprocal divide
;; frdiv1       floating point reciprocal divide step 1
;; frdiv2       floating point reciprocal divide step 2
;; fabs         floating point absolute value
;; fneg         floating point negation
;; fcmp         floating point compare
;; fcvt         floating point convert
;; fsqrt        floating point square root
;; frsqrt       floating point reciprocal square root
;; frsqrt1      floating point reciprocal square root step1
;; frsqrt2      floating point reciprocal square root step2
;; multi        multiword sequence (or user asm statements)
;; nop          no operation
(define_attr "type"
  "unknown,branch,jump,call,load,fpload,fpidxload,store,fpstore,fpidxstore,prefetch,prefetchx,condmove,xfer,mthilo,mfhilo,const,arith,shift,slt,clz,trap,imul,imul3,imadd,idiv,fmove,fadd,fmul,fmadd,fdiv,frdiv,frdiv1,frdiv2,fabs,fneg,fcmp,fcvt,fsqrt,frsqrt,frsqrt1,frsqrt2,multi,nop"
  (cond [(eq_attr "jal" "!unset") (const_string "call")
         (eq_attr "got" "load") (const_string "load")]
        (const_string "unknown")))

;; Main data type used by the insn
(define_attr "mode" "unknown,none,QI,HI,SI,DI,SF,DF,FPSW"
  (const_string "unknown"))

;; Mode for conversion types (fcvt)
;; I2S          integer to float single (SI/DI to SF)
;; I2D          integer to float double (SI/DI to DF)
;; S2I          float to integer (SF to SI/DI)
;; D2I          float to integer (DF to SI/DI)
;; D2S          double to float single
;; S2D          float single to double

(define_attr "cnv_mode" "unknown,I2S,I2D,S2I,D2I,D2S,S2D" 
  (const_string "unknown"))

;; Is this an extended instruction in mips16 mode?
(define_attr "extended_mips16" "no,yes"
  (const_string "no"))

;; Length of instruction in bytes.
(define_attr "length" ""
   (cond [;; Direct branch instructions have a range of [-0x40000,0x3fffc].
          ;; If a branch is outside this range, we have a choice of two
          ;; sequences.  For PIC, an out-of-range branch like:
          ;;
          ;;    bne     r1,r2,target
          ;;    dslot
          ;;
          ;; becomes the equivalent of:
          ;;
          ;;    beq     r1,r2,1f
          ;;    dslot
          ;;    la      $at,target
          ;;    jr      $at
          ;;    nop
          ;; 1:
          ;;
          ;; where the load address can be up to three instructions long
          ;; (lw, nop, addiu).
          ;;
          ;; The non-PIC case is similar except that we use a direct
          ;; jump instead of an la/jr pair.  Since the target of this
          ;; jump is an absolute 28-bit bit address (the other bits
          ;; coming from the address of the delay slot) this form cannot
          ;; cross a 256MB boundary.  We could provide the option of
          ;; using la/jr in this case too, but we do not do so at
          ;; present.
          ;;
          ;; Note that this value does not account for the delay slot
          ;; instruction, whose length is added separately.  If the RTL
          ;; pattern has no explicit delay slot, mips_adjust_insn_length
          ;; will add the length of the implicit nop.  The values for
          ;; forward and backward branches will be different as well.
          (eq_attr "type" "branch")
          (cond [(and (le (minus (match_dup 1) (pc)) (const_int 131064))
                      (le (minus (pc) (match_dup 1)) (const_int 131068)))
                  (const_int 4)
                 (ne (symbol_ref "flag_pic") (const_int 0))
                 (const_int 24)
                 ] (const_int 12))

          (eq_attr "got" "load")
          (const_int 4)
          (eq_attr "got" "xgot_high")
          (const_int 8)

          (eq_attr "type" "const")
          (symbol_ref "mips_const_insns (operands[1]) * 4")
          (eq_attr "type" "load,fpload")
          (symbol_ref "mips_fetch_insns (operands[1]) * 4")
          (eq_attr "type" "store,fpstore")
          (symbol_ref "mips_fetch_insns (operands[0]) * 4")

          ;; In the worst case, a call macro will take 8 instructions:
          ;;
          ;;     lui $25,%call_hi(FOO)
          ;;     addu $25,$25,$28
          ;;     lw $25,%call_lo(FOO)($25)
          ;;     nop
          ;;     jalr $25
          ;;     nop
          ;;     lw $gp,X($sp)
          ;;     nop
          (eq_attr "jal_macro" "yes")
          (const_int 32)

          (and (eq_attr "extended_mips16" "yes")
               (ne (symbol_ref "TARGET_MIPS16") (const_int 0)))
          (const_int 8)

          ;; Various VR4120 errata require a nop to be inserted after a macc
          ;; instruction.  The assembler does this for us, so account for
          ;; the worst-case length here.
          (and (eq_attr "type" "imadd")
               (ne (symbol_ref "TARGET_FIX_VR4120") (const_int 0)))
          (const_int 8)

          ;; VR4120 errata MD(4): if there are consecutive dmult instructions,
          ;; the result of the second one is missed.  The assembler should work
          ;; around this by inserting a nop after the first dmult.
          (and (eq_attr "type" "imul,imul3")
               (and (eq_attr "mode" "DI")
                    (ne (symbol_ref "TARGET_FIX_VR4120") (const_int 0))))
          (const_int 8)

          (eq_attr "type" "idiv")
          (symbol_ref "mips_idiv_insns () * 4")
          ] (const_int 4)))

;; Attribute describing the processor.  This attribute must match exactly
;; with the processor_type enumeration in mips.h.
(define_attr "cpu"
  "r3000,4kc,4kp,5kc,5kf,20kc,24k,24kx,m4k,r3900,r6000,r4000,r4100,r4111,r4120,r4130,r4300,r4600,r4650,r5000,r5400,r5500,r7000,r8000,r9000,sb1,sr71000"
  (const (symbol_ref "mips_tune")))

;; The type of hardware hazard associated with this instruction.
;; DELAY means that the next instruction cannot read the result
;; of this one.  HILO means that the next two instructions cannot
;; write to HI or LO.
(define_attr "hazard" "none,delay,hilo"
  (cond [(and (eq_attr "type" "load,fpload,fpidxload")
              (ne (symbol_ref "ISA_HAS_LOAD_DELAY") (const_int 0)))
         (const_string "delay")

         (and (eq_attr "type" "xfer")
              (ne (symbol_ref "ISA_HAS_XFER_DELAY") (const_int 0)))
         (const_string "delay")

         (and (eq_attr "type" "fcmp")
              (ne (symbol_ref "ISA_HAS_FCMP_DELAY") (const_int 0)))
         (const_string "delay")

         ;; The r4000 multiplication patterns include an mflo instruction.
         (and (eq_attr "type" "imul")
              (ne (symbol_ref "TARGET_FIX_R4000") (const_int 0)))
         (const_string "hilo")

         (and (eq_attr "type" "mfhilo")
              (eq (symbol_ref "ISA_HAS_HILO_INTERLOCKS") (const_int 0)))
         (const_string "hilo")]
        (const_string "none")))

;; Is it a single instruction?
(define_attr "single_insn" "no,yes"
  (symbol_ref "get_attr_length (insn) == (TARGET_MIPS16 ? 2 : 4)"))

;; Can the instruction be put into a delay slot?
(define_attr "can_delay" "no,yes"
  (if_then_else (and (eq_attr "type" "!branch,call,jump")
                     (and (eq_attr "hazard" "none")
                          (eq_attr "single_insn" "yes")))
                (const_string "yes")
                (const_string "no")))

;; Attribute defining whether or not we can use the branch-likely instructions
(define_attr "branch_likely" "no,yes"
  (const
   (if_then_else (ne (symbol_ref "GENERATE_BRANCHLIKELY") (const_int 0))
                 (const_string "yes")
                 (const_string "no"))))

;; True if an instruction might assign to hi or lo when reloaded.
;; This is used by the TUNE_MACC_CHAINS code.
(define_attr "may_clobber_hilo" "no,yes"
  (if_then_else (eq_attr "type" "imul,imul3,imadd,idiv,mthilo")
                (const_string "yes")
                (const_string "no")))

;; Describe a user's asm statement.
(define_asm_attributes
  [(set_attr "type" "multi")
   (set_attr "can_delay" "no")])
\f
;; This mode macro allows 32-bit and 64-bit GPR patterns to be generated
;; from the same template.
(define_mode_macro GPR [SI (DI "TARGET_64BIT")])

;; This mode macro allows :P to be used for patterns that operate on
;; pointer-sized quantities.  Exactly one of the two alternatives will match.
(define_mode_macro P [(SI "Pmode == SImode") (DI "Pmode == DImode")])

;; This mode macro allows :MOVECC to be used anywhere that a
;; conditional-move-type condition is needed.
(define_mode_macro MOVECC [SI (DI "TARGET_64BIT") (CC "TARGET_HARD_FLOAT")])

;; This mode macro allows the QI and HI extension patterns to be defined from
;; the same template.
(define_mode_macro SHORT [QI HI])

;; This mode macro allows :ANYF to be used wherever a scalar or vector
;; floating-point mode is allowed.
(define_mode_macro ANYF [(SF "TARGET_HARD_FLOAT")
                         (DF "TARGET_HARD_FLOAT && TARGET_DOUBLE_FLOAT")
                         (V2SF "TARGET_PAIRED_SINGLE_FLOAT")])

;; Like ANYF, but only applies to scalar modes.
(define_mode_macro SCALARF [(SF "TARGET_HARD_FLOAT")
                            (DF "TARGET_HARD_FLOAT && TARGET_DOUBLE_FLOAT")])

;; In GPR templates, a string like "<d>subu" will expand to "subu" in the
;; 32-bit version and "dsubu" in the 64-bit version.
(define_mode_attr d [(SI "") (DI "d")])

;; This attribute gives the length suffix for a sign- or zero-extension
;; instruction.
(define_mode_attr size [(QI "b") (HI "h")])

;; This attributes gives the mode mask of a SHORT.
(define_mode_attr mask [(QI "0x00ff") (HI "0xffff")])

;; Mode attributes for GPR loads and stores.
(define_mode_attr load [(SI "lw") (DI "ld")])
(define_mode_attr store [(SI "sw") (DI "sd")])

;; Similarly for MIPS IV indexed FPR loads and stores.
(define_mode_attr loadx [(SF "lwxc1") (DF "ldxc1") (V2SF "ldxc1")])
(define_mode_attr storex [(SF "swxc1") (DF "sdxc1") (V2SF "sdxc1")])

;; The unextended ranges of the MIPS16 addiu and daddiu instructions
;; are different.  Some forms of unextended addiu have an 8-bit immediate
;; field but the equivalent daddiu has only a 5-bit field.
(define_mode_attr si8_di5 [(SI "8") (DI "5")])

;; This attribute gives the best constraint to use for registers of
;; a given mode.
(define_mode_attr reg [(SI "d") (DI "d") (CC "z")])

;; This attribute gives the format suffix for floating-point operations.
(define_mode_attr fmt [(SF "s") (DF "d") (V2SF "ps")])

;; This attribute gives the upper-case mode name for one unit of a
;; floating-point mode.
(define_mode_attr UNITMODE [(SF "SF") (DF "DF") (V2SF "SF")])

;; This attribute works around the early SB-1 rev2 core "F2" erratum:
;;
;; In certain cases, div.s and div.ps may have a rounding error
;; and/or wrong inexact flag.
;;
;; Therefore, we only allow div.s if not working around SB-1 rev2
;; errata or if a slight loss of precision is OK.
(define_mode_attr divide_condition
  [DF (SF "!TARGET_FIX_SB1 || flag_unsafe_math_optimizations")
   (V2SF "TARGET_SB1 && (!TARGET_FIX_SB1 || flag_unsafe_math_optimizations)")])

; This attribute gives the condition for which sqrt instructions exist.
(define_mode_attr sqrt_condition
  [(SF "!ISA_MIPS1") (DF "!ISA_MIPS1") (V2SF "TARGET_SB1")])

; This attribute gives the condition for which recip and rsqrt instructions
; exist.
(define_mode_attr recip_condition
  [(SF "ISA_HAS_FP4") (DF "ISA_HAS_FP4") (V2SF "TARGET_SB1")])

;; This code macro allows all branch instructions to be generated from
;; a single define_expand template.
(define_code_macro any_cond [unordered ordered unlt unge uneq ltgt unle ungt
                             eq ne gt ge lt le gtu geu ltu leu])

;; This code macro allows signed and unsigned widening multiplications
;; to use the same template.
(define_code_macro any_extend [sign_extend zero_extend])

;; This code macro allows the three shift instructions to be generated
;; from the same template.
(define_code_macro any_shift [ashift ashiftrt lshiftrt])

;; This code macro allows all native floating-point comparisons to be
;; generated from the same template.
(define_code_macro fcond [unordered uneq unlt unle eq lt le])

;; This code macro is used for comparisons that can be implemented
;; by swapping the operands.
(define_code_macro swapped_fcond [ge gt unge ungt])

;; <u> expands to an empty string when doing a signed operation and
;; "u" when doing an unsigned operation.
(define_code_attr u [(sign_extend "") (zero_extend "u")])

;; <su> is like <u>, but the signed form expands to "s" rather than "".
(define_code_attr su [(sign_extend "s") (zero_extend "u")])

;; <optab> expands to the name of the optab for a particular code.
(define_code_attr optab [(ashift "ashl")
                         (ashiftrt "ashr")
                         (lshiftrt "lshr")])

;; <insn> expands to the name of the insn that implements a particular code.
(define_code_attr insn [(ashift "sll")
                        (ashiftrt "sra")
                        (lshiftrt "srl")])

;; <fcond> is the c.cond.fmt condition associated with a particular code.
(define_code_attr fcond [(unordered "un")
                         (uneq "ueq")
                         (unlt "ult")
                         (unle "ule")
                         (eq "eq")
                         (lt "lt")
                         (le "le")])

;; Similar, but for swapped conditions.
(define_code_attr swapped_fcond [(ge "le")
                                 (gt "lt")
                                 (unge "ule")
                                 (ungt "ult")])
\f
;; .........................
;;
;;      Branch, call and jump delay slots
;;
;; .........................

(define_delay (and (eq_attr "type" "branch")
                   (eq (symbol_ref "TARGET_MIPS16") (const_int 0)))
  [(eq_attr "can_delay" "yes")
   (nil)
   (and (eq_attr "branch_likely" "yes")
        (eq_attr "can_delay" "yes"))])

(define_delay (eq_attr "type" "jump")
  [(eq_attr "can_delay" "yes")
   (nil)
   (nil)])

(define_delay (and (eq_attr "type" "call")
                   (eq_attr "jal_macro" "no"))
  [(eq_attr "can_delay" "yes")
   (nil)
   (nil)])
\f
;; Pipeline descriptions.
;;
;; generic.md provides a fallback for processors without a specific
;; pipeline description.  It is derived from the old define_function_unit
;; version and uses the "alu" and "imuldiv" units declared below.
;;
;; Some of the processor-specific files are also derived from old
;; define_function_unit descriptions and simply override the parts of
;; generic.md that don't apply.  The other processor-specific files
;; are self-contained.
(define_automaton "alu,imuldiv")

(define_cpu_unit "alu" "alu")
(define_cpu_unit "imuldiv" "imuldiv")

(include "4k.md")
(include "5k.md")
(include "24k.md")
(include "3000.md")
(include "4000.md")
(include "4100.md")
(include "4130.md")
(include "4300.md")
(include "4600.md")
(include "5000.md")
(include "5400.md")
(include "5500.md")
(include "6000.md")
(include "7000.md")
(include "9000.md")
(include "sb1.md")
(include "sr71k.md")
(include "generic.md")
\f
;;
;;  ....................
;;
;;      CONDITIONAL TRAPS
;;
;;  ....................
;;

(define_insn "trap"
  [(trap_if (const_int 1) (const_int 0))]
  ""
{
  if (ISA_HAS_COND_TRAP)
    return "teq\t$0,$0";
  else if (TARGET_MIPS16)
    return "break 0";
  else
    return "break";
}
  [(set_attr "type" "trap")])

(define_expand "conditional_trap"
  [(trap_if (match_operator 0 "comparison_operator"
                            [(match_dup 2) (match_dup 3)])
            (match_operand 1 "const_int_operand"))]
  "ISA_HAS_COND_TRAP"
{
  if (GET_MODE_CLASS (GET_MODE (cmp_operands[0])) == MODE_INT
      && operands[1] == const0_rtx)
    {
      mips_gen_conditional_trap (operands);
      DONE;
    }
  else
    FAIL;
})

(define_insn "*conditional_trap<mode>"
  [(trap_if (match_operator:GPR 0 "trap_comparison_operator"
                                [(match_operand:GPR 1 "reg_or_0_operand" "dJ")
                                 (match_operand:GPR 2 "arith_operand" "dI")])
            (const_int 0))]
  "ISA_HAS_COND_TRAP"
  "t%C0\t%z1,%2"
  [(set_attr "type" "trap")])
\f
;;
;;  ....................
;;
;;      ADDITION
;;
;;  ....................
;;

(define_insn "add<mode>3"
  [(set (match_operand:ANYF 0 "register_operand" "=f")
        (plus:ANYF (match_operand:ANYF 1 "register_operand" "f")
                   (match_operand:ANYF 2 "register_operand" "f")))]
  ""
  "add.<fmt>\t%0,%1,%2"
  [(set_attr "type" "fadd")
   (set_attr "mode" "<UNITMODE>")])

(define_expand "add<mode>3"
  [(set (match_operand:GPR 0 "register_operand")
        (plus:GPR (match_operand:GPR 1 "register_operand")
                  (match_operand:GPR 2 "arith_operand")))]
  "")

(define_insn "*add<mode>3"
  [(set (match_operand:GPR 0 "register_operand" "=d,d")
        (plus:GPR (match_operand:GPR 1 "register_operand" "d,d")
                  (match_operand:GPR 2 "arith_operand" "d,Q")))]
  "!TARGET_MIPS16"
  "@
    <d>addu\t%0,%1,%2
    <d>addiu\t%0,%1,%2"
  [(set_attr "type" "arith")
   (set_attr "mode" "<MODE>")])

;; We need to recognize MIPS16 stack pointer additions explicitly, since
;; we don't have a constraint for $sp.  These insns will be generated by
;; the save_restore_insns functions.

(define_insn "*add<mode>3_sp1"
  [(set (reg:GPR 29)
        (plus:GPR (reg:GPR 29)
                  (match_operand:GPR 0 "const_arith_operand" "")))]
  "TARGET_MIPS16"
  "<d>addiu\t%$,%$,%0"
  [(set_attr "type" "arith")
   (set_attr "mode" "<MODE>")
   (set (attr "length") (if_then_else (match_operand 0 "m16_simm8_8")
                                      (const_int 4)
                                      (const_int 8)))])

(define_insn "*add<mode>3_sp2"
  [(set (match_operand:GPR 0 "register_operand" "=d")
        (plus:GPR (reg:GPR 29)
                  (match_operand:GPR 1 "const_arith_operand" "")))]
  "TARGET_MIPS16"
  "<d>addiu\t%0,%$,%1"
  [(set_attr "type" "arith")
   (set_attr "mode" "<MODE>")
   (set (attr "length") (if_then_else (match_operand 1 "m16_uimm<si8_di5>_4")
                                      (const_int 4)
                                      (const_int 8)))])

(define_insn "*add<mode>3_mips16"
  [(set (match_operand:GPR 0 "register_operand" "=d,d,d")
        (plus:GPR (match_operand:GPR 1 "register_operand" "0,d,d")
                  (match_operand:GPR 2 "arith_operand" "Q,O,d")))]
  "TARGET_MIPS16"
  "@
    <d>addiu\t%0,%2
    <d>addiu\t%0,%1,%2
    <d>addu\t%0,%1,%2"
  [(set_attr "type" "arith")
   (set_attr "mode" "<MODE>")
   (set_attr_alternative "length"
                [(if_then_else (match_operand 2 "m16_simm<si8_di5>_1")
                               (const_int 4)
                               (const_int 8))
                 (if_then_else (match_operand 2 "m16_simm4_1")
                               (const_int 4)
                               (const_int 8))
                 (const_int 4)])])


;; On the mips16, we can sometimes split an add of a constant which is
;; a 4 byte instruction into two adds which are both 2 byte
;; instructions.  There are two cases: one where we are adding a
;; constant plus a register to another register, and one where we are
;; simply adding a constant to a register.

(define_split
  [(set (match_operand:SI 0 "register_operand")
        (plus:SI (match_dup 0)
                 (match_operand:SI 1 "const_int_operand")))]
  "TARGET_MIPS16 && reload_completed && !TARGET_DEBUG_D_MODE
   && REG_P (operands[0])
   && M16_REG_P (REGNO (operands[0]))
   && GET_CODE (operands[1]) == CONST_INT
   && ((INTVAL (operands[1]) > 0x7f
        && INTVAL (operands[1]) <= 0x7f + 0x7f)
       || (INTVAL (operands[1]) < - 0x80
           && INTVAL (operands[1]) >= - 0x80 - 0x80))"
  [(set (match_dup 0) (plus:SI (match_dup 0) (match_dup 1)))
   (set (match_dup 0) (plus:SI (match_dup 0) (match_dup 2)))]
{
  HOST_WIDE_INT val = INTVAL (operands[1]);

  if (val >= 0)
    {
      operands[1] = GEN_INT (0x7f);
      operands[2] = GEN_INT (val - 0x7f);
    }
  else
    {
      operands[1] = GEN_INT (- 0x80);
      operands[2] = GEN_INT (val + 0x80);
    }
})

(define_split
  [(set (match_operand:SI 0 "register_operand")
        (plus:SI (match_operand:SI 1 "register_operand")
                 (match_operand:SI 2 "const_int_operand")))]
  "TARGET_MIPS16 && reload_completed && !TARGET_DEBUG_D_MODE
   && REG_P (operands[0])
   && M16_REG_P (REGNO (operands[0]))
   && REG_P (operands[1])
   && M16_REG_P (REGNO (operands[1]))
   && REGNO (operands[0]) != REGNO (operands[1])
   && GET_CODE (operands[2]) == CONST_INT
   && ((INTVAL (operands[2]) > 0x7
        && INTVAL (operands[2]) <= 0x7 + 0x7f)
       || (INTVAL (operands[2]) < - 0x8
           && INTVAL (operands[2]) >= - 0x8 - 0x80))"
  [(set (match_dup 0) (plus:SI (match_dup 1) (match_dup 2)))
   (set (match_dup 0) (plus:SI (match_dup 0) (match_dup 3)))]
{
  HOST_WIDE_INT val = INTVAL (operands[2]);

  if (val >= 0)
    {
      operands[2] = GEN_INT (0x7);
      operands[3] = GEN_INT (val - 0x7);
    }
  else
    {
      operands[2] = GEN_INT (- 0x8);
      operands[3] = GEN_INT (val + 0x8);
    }
})

(define_split
  [(set (match_operand:DI 0 "register_operand")
        (plus:DI (match_dup 0)
                 (match_operand:DI 1 "const_int_operand")))]
  "TARGET_MIPS16 && TARGET_64BIT && reload_completed && !TARGET_DEBUG_D_MODE
   && REG_P (operands[0])
   && M16_REG_P (REGNO (operands[0]))
   && GET_CODE (operands[1]) == CONST_INT
   && ((INTVAL (operands[1]) > 0xf
        && INTVAL (operands[1]) <= 0xf + 0xf)
       || (INTVAL (operands[1]) < - 0x10
           && INTVAL (operands[1]) >= - 0x10 - 0x10))"
  [(set (match_dup 0) (plus:DI (match_dup 0) (match_dup 1)))
   (set (match_dup 0) (plus:DI (match_dup 0) (match_dup 2)))]
{
  HOST_WIDE_INT val = INTVAL (operands[1]);

  if (val >= 0)
    {
      operands[1] = GEN_INT (0xf);
      operands[2] = GEN_INT (val - 0xf);
    }
  else
    {
      operands[1] = GEN_INT (- 0x10);
      operands[2] = GEN_INT (val + 0x10);
    }
})

(define_split
  [(set (match_operand:DI 0 "register_operand")
        (plus:DI (match_operand:DI 1 "register_operand")
                 (match_operand:DI 2 "const_int_operand")))]
  "TARGET_MIPS16 && TARGET_64BIT && reload_completed && !TARGET_DEBUG_D_MODE
   && REG_P (operands[0])
   && M16_REG_P (REGNO (operands[0]))
   && REG_P (operands[1])
   && M16_REG_P (REGNO (operands[1]))
   && REGNO (operands[0]) != REGNO (operands[1])
   && GET_CODE (operands[2]) == CONST_INT
   && ((INTVAL (operands[2]) > 0x7
        && INTVAL (operands[2]) <= 0x7 + 0xf)
       || (INTVAL (operands[2]) < - 0x8
           && INTVAL (operands[2]) >= - 0x8 - 0x10))"
  [(set (match_dup 0) (plus:DI (match_dup 1) (match_dup 2)))
   (set (match_dup 0) (plus:DI (match_dup 0) (match_dup 3)))]
{
  HOST_WIDE_INT val = INTVAL (operands[2]);

  if (val >= 0)
    {
      operands[2] = GEN_INT (0x7);
      operands[3] = GEN_INT (val - 0x7);
    }
  else
    {
      operands[2] = GEN_INT (- 0x8);
      operands[3] = GEN_INT (val + 0x8);
    }
})

(define_insn "*addsi3_extended"
  [(set (match_operand:DI 0 "register_operand" "=d,d")
        (sign_extend:DI
             (plus:SI (match_operand:SI 1 "register_operand" "d,d")
                      (match_operand:SI 2 "arith_operand" "d,Q"))))]
  "TARGET_64BIT && !TARGET_MIPS16"
  "@
    addu\t%0,%1,%2
    addiu\t%0,%1,%2"
  [(set_attr "type" "arith")
   (set_attr "mode" "SI")])

;; Split this insn so that the addiu splitters can have a crack at it.
;; Use a conservative length estimate until the split.
(define_insn_and_split "*addsi3_extended_mips16"
  [(set (match_operand:DI 0 "register_operand" "=d,d,d")
        (sign_extend:DI
             (plus:SI (match_operand:SI 1 "register_operand" "0,d,d")
                      (match_operand:SI 2 "arith_operand" "Q,O,d"))))]
  "TARGET_64BIT && TARGET_MIPS16"
  "#"
  "&& reload_completed"
  [(set (match_dup 3) (plus:SI (match_dup 1) (match_dup 2)))]
  { operands[3] = gen_lowpart (SImode, operands[0]); }
  [(set_attr "type" "arith")
   (set_attr "mode" "SI")
   (set_attr "extended_mips16" "yes")])
\f
;;
;;  ....................
;;
;;      SUBTRACTION
;;
;;  ....................
;;

(define_insn "sub<mode>3"
  [(set (match_operand:ANYF 0 "register_operand" "=f")
        (minus:ANYF (match_operand:ANYF 1 "register_operand" "f")
                    (match_operand:ANYF 2 "register_operand" "f")))]
  ""
  "sub.<fmt>\t%0,%1,%2"
  [(set_attr "type" "fadd")
   (set_attr "mode" "<UNITMODE>")])

(define_insn "sub<mode>3"
  [(set (match_operand:GPR 0 "register_operand" "=d")
        (minus:GPR (match_operand:GPR 1 "register_operand" "d")
                   (match_operand:GPR 2 "register_operand" "d")))]
  ""
  "<d>subu\t%0,%1,%2"
  [(set_attr "type" "arith")
   (set_attr "mode" "<MODE>")])

(define_insn "*subsi3_extended"
  [(set (match_operand:DI 0 "register_operand" "=d")
        (sign_extend:DI
            (minus:SI (match_operand:SI 1 "register_operand" "d")
                      (match_operand:SI 2 "register_operand" "d"))))]
  "TARGET_64BIT"
  "subu\t%0,%1,%2"
  [(set_attr "type" "arith")
   (set_attr "mode" "DI")])
\f
;;
;;  ....................
;;
;;      MULTIPLICATION
;;
;;  ....................
;;

(define_expand "mul<mode>3"
  [(set (match_operand:SCALARF 0 "register_operand")
        (mult:SCALARF (match_operand:SCALARF 1 "register_operand")
                      (match_operand:SCALARF 2 "register_operand")))]
  ""
  "")

(define_insn "*mul<mode>3"
  [(set (match_operand:SCALARF 0 "register_operand" "=f")
        (mult:SCALARF (match_operand:SCALARF 1 "register_operand" "f")
                      (match_operand:SCALARF 2 "register_operand" "f")))]
  "!TARGET_4300_MUL_FIX"
  "mul.<fmt>\t%0,%1,%2"
  [(set_attr "type" "fmul")
   (set_attr "mode" "<MODE>")])

;; Early VR4300 silicon has a CPU bug where multiplies with certain
;; operands may corrupt immediately following multiplies. This is a
;; simple fix to insert NOPs.

(define_insn "*mul<mode>3_r4300"
  [(set (match_operand:SCALARF 0 "register_operand" "=f")
        (mult:SCALARF (match_operand:SCALARF 1 "register_operand" "f")
                      (match_operand:SCALARF 2 "register_operand" "f")))]
  "TARGET_4300_MUL_FIX"
  "mul.<fmt>\t%0,%1,%2\;nop"
  [(set_attr "type" "fmul")
   (set_attr "mode" "<MODE>")
   (set_attr "length" "8")])

(define_insn "mulv2sf3"
  [(set (match_operand:V2SF 0 "register_operand" "=f")
        (mult:V2SF (match_operand:V2SF 1 "register_operand" "f")
                   (match_operand:V2SF 2 "register_operand" "f")))]
  "TARGET_PAIRED_SINGLE_FLOAT"
  "mul.ps\t%0,%1,%2"
  [(set_attr "type" "fmul")
   (set_attr "mode" "SF")])

;; The original R4000 has a cpu bug.  If a double-word or a variable
;; shift executes while an integer multiplication is in progress, the
;; shift may give an incorrect result.  Avoid this by keeping the mflo
;; with the mult on the R4000.
;;
;; From "MIPS R4000PC/SC Errata, Processor Revision 2.2 and 3.0"
;; (also valid for MIPS R4000MC processors):
;;
;; "16. R4000PC, R4000SC: Please refer to errata 28 for an update to
;;      this errata description.
;;      The following code sequence causes the R4000 to incorrectly
;;      execute the Double Shift Right Arithmetic 32 (dsra32)
;;      instruction.  If the dsra32 instruction is executed during an
;;      integer multiply, the dsra32 will only shift by the amount in
;;      specified in the instruction rather than the amount plus 32
;;      bits.
;;      instruction 1:          mult    rs,rt           integer multiply
;;      instruction 2-12:       dsra32  rd,rt,rs        doubleword shift
;;                                                      right arithmetic + 32
;;      Workaround: A dsra32 instruction placed after an integer
;;      multiply should not be one of the 11 instructions after the
;;      multiply instruction."
;;
;; and:
;;
;; "28. R4000PC, R4000SC: The text from errata 16 should be replaced by
;;      the following description.
;;      All extended shifts (shift by n+32) and variable shifts (32 and
;;      64-bit versions) may produce incorrect results under the
;;      following conditions:
;;      1) An integer multiply is currently executing
;;      2) These types of shift instructions are executed immediately
;;         following an integer divide instruction.
;;      Workaround:
;;      1) Make sure no integer multiply is running wihen these
;;         instruction are executed.  If this cannot be predicted at
;;         compile time, then insert a "mfhi" to R0 instruction
;;         immediately after the integer multiply instruction.  This
;;         will cause the integer multiply to complete before the shift
;;         is executed.
;;      2) Separate integer divide and these two classes of shift
;;         instructions by another instruction or a noop."
;;
;; These processors have PRId values of 0x00004220 and 0x00004300,
;; respectively.

(define_expand "mul<mode>3"
  [(set (match_operand:GPR 0 "register_operand")
        (mult:GPR (match_operand:GPR 1 "register_operand")
                  (match_operand:GPR 2 "register_operand")))]
  ""
{
  if (GENERATE_MULT3_<MODE>)
    emit_insn (gen_mul<mode>3_mult3 (operands[0], operands[1], operands[2]));
  else if (!TARGET_FIX_R4000)
    emit_insn (gen_mul<mode>3_internal (operands[0], operands[1],
                                        operands[2]));
  else
    emit_insn (gen_mul<mode>3_r4000 (operands[0], operands[1], operands[2]));
  DONE;
})

(define_insn "mulsi3_mult3"
  [(set (match_operand:SI 0 "register_operand" "=d,l")
        (mult:SI (match_operand:SI 1 "register_operand" "d,d")
                 (match_operand:SI 2 "register_operand" "d,d")))
   (clobber (match_scratch:SI 3 "=h,h"))
   (clobber (match_scratch:SI 4 "=l,X"))]
  "GENERATE_MULT3_SI"
{
  if (which_alternative == 1)
    return "mult\t%1,%2";
  if (TARGET_MAD
      || TARGET_MIPS5400
      || TARGET_MIPS5500
      || TARGET_MIPS7000
      || TARGET_MIPS9000
      || ISA_MIPS32
      || ISA_MIPS32R2
      || ISA_MIPS64)
    return "mul\t%0,%1,%2";
  return "mult\t%0,%1,%2";
}
  [(set_attr "type" "imul3,imul")
   (set_attr "mode" "SI")])

(define_insn "muldi3_mult3"
  [(set (match_operand:DI 0 "register_operand" "=d")
        (mult:DI (match_operand:DI 1 "register_operand" "d")
                 (match_operand:DI 2 "register_operand" "d")))
   (clobber (match_scratch:DI 3 "=h"))
   (clobber (match_scratch:DI 4 "=l"))]
  "TARGET_64BIT && GENERATE_MULT3_DI"
  "dmult\t%0,%1,%2"
  [(set_attr "type" "imul3")
   (set_attr "mode" "DI")])

;; If a register gets allocated to LO, and we spill to memory, the reload
;; will include a move from LO to a GPR.  Merge it into the multiplication
;; if it can set the GPR directly.
;;
;; Operand 0: LO
;; Operand 1: GPR (1st multiplication operand)
;; Operand 2: GPR (2nd multiplication operand)
;; Operand 3: HI
;; Operand 4: GPR (destination)
(define_peephole2
  [(parallel
       [(set (match_operand:SI 0 "register_operand")
             (mult:SI (match_operand:SI 1 "register_operand")
                      (match_operand:SI 2 "register_operand")))
        (clobber (match_operand:SI 3 "register_operand"))
        (clobber (scratch:SI))])
   (set (match_operand:SI 4 "register_operand")
        (unspec [(match_dup 0) (match_dup 3)] UNSPEC_MFHILO))]
  "GENERATE_MULT3_SI && peep2_reg_dead_p (2, operands[0])"
  [(parallel
       [(set (match_dup 4)
             (mult:SI (match_dup 1)
                      (match_dup 2)))
        (clobber (match_dup 3))
        (clobber (match_dup 0))])])

(define_insn "mul<mode>3_internal"
  [(set (match_operand:GPR 0 "register_operand" "=l")
        (mult:GPR (match_operand:GPR 1 "register_operand" "d")
                  (match_operand:GPR 2 "register_operand" "d")))
   (clobber (match_scratch:GPR 3 "=h"))]
  "!TARGET_FIX_R4000"
  "<d>mult\t%1,%2"
  [(set_attr "type" "imul")
   (set_attr "mode" "<MODE>")])

(define_insn "mul<mode>3_r4000"
  [(set (match_operand:GPR 0 "register_operand" "=d")
        (mult:GPR (match_operand:GPR 1 "register_operand" "d")
                  (match_operand:GPR 2 "register_operand" "d")))
   (clobber (match_scratch:GPR 3 "=h"))
   (clobber (match_scratch:GPR 4 "=l"))]
  "TARGET_FIX_R4000"
  "<d>mult\t%1,%2\;mflo\t%0"
  [(set_attr "type" "imul")
   (set_attr "mode" "<MODE>")
   (set_attr "length" "8")])

;; On the VR4120 and VR4130, it is better to use "mtlo $0; macc" instead
;; of "mult; mflo".  They have the same latency, but the first form gives
;; us an extra cycle to compute the operands.

;; Operand 0: LO
;; Operand 1: GPR (1st multiplication operand)
;; Operand 2: GPR (2nd multiplication operand)
;; Operand 3: HI
;; Operand 4: GPR (destination)
(define_peephole2
  [(parallel
       [(set (match_operand:SI 0 "register_operand")
             (mult:SI (match_operand:SI 1 "register_operand")
                      (match_operand:SI 2 "register_operand")))
        (clobber (match_operand:SI 3 "register_operand"))])
   (set (match_operand:SI 4 "register_operand")
        (unspec:SI [(match_dup 0) (match_dup 3)] UNSPEC_MFHILO))]
  "ISA_HAS_MACC && !GENERATE_MULT3_SI"
  [(set (match_dup 0)
        (const_int 0))
   (parallel
       [(set (match_dup 0)
             (plus:SI (mult:SI (match_dup 1)
                               (match_dup 2))
                      (match_dup 0)))
        (set (match_dup 4)
             (plus:SI (mult:SI (match_dup 1)
                               (match_dup 2))
                      (match_dup 0)))
        (clobber (match_dup 3))])])

;; Multiply-accumulate patterns

;; For processors that can copy the output to a general register:
;;
;; The all-d alternative is needed because the combiner will find this
;; pattern and then register alloc/reload will move registers around to
;; make them fit, and we don't want to trigger unnecessary loads to LO.
;;
;; The last alternative should be made slightly less desirable, but adding
;; "?" to the constraint is too strong, and causes values to be loaded into
;; LO even when that's more costly.  For now, using "*d" mostly does the
;; trick.
(define_insn "*mul_acc_si"
  [(set (match_operand:SI 0 "register_operand" "=l,*d,*d")
        (plus:SI (mult:SI (match_operand:SI 1 "register_operand" "d,d,d")
                          (match_operand:SI 2 "register_operand" "d,d,d"))
                 (match_operand:SI 3 "register_operand" "0,l,*d")))
   (clobber (match_scratch:SI 4 "=h,h,h"))
   (clobber (match_scratch:SI 5 "=X,3,l"))
   (clobber (match_scratch:SI 6 "=X,X,&d"))]
  "(TARGET_MIPS3900
   || ISA_HAS_MADD_MSUB)
   && !TARGET_MIPS16"
{
  static const char *const madd[] = { "madd\t%1,%2", "madd\t%0,%1,%2" };
  if (which_alternative == 2)
    return "#";
  if (ISA_HAS_MADD_MSUB && which_alternative != 0)
    return "#";
  return madd[which_alternative];
}
  [(set_attr "type"     "imadd,imadd,multi")
   (set_attr "mode"     "SI")
   (set_attr "length"   "4,4,8")])

;; Split the above insn if we failed to get LO allocated.
(define_split
  [(set (match_operand:SI 0 "register_operand")
        (plus:SI (mult:SI (match_operand:SI 1 "register_operand")
                          (match_operand:SI 2 "register_operand"))
                 (match_operand:SI 3 "register_operand")))
   (clobber (match_scratch:SI 4))
   (clobber (match_scratch:SI 5))
   (clobber (match_scratch:SI 6))]
  "reload_completed && !TARGET_DEBUG_D_MODE
   && GP_REG_P (true_regnum (operands[0]))
   && GP_REG_P (true_regnum (operands[3]))"
  [(parallel [(set (match_dup 6)
                   (mult:SI (match_dup 1) (match_dup 2)))
              (clobber (match_dup 4))
              (clobber (match_dup 5))])
   (set (match_dup 0) (plus:SI (match_dup 6) (match_dup 3)))]
  "")

;; Splitter to copy result of MADD to a general register
(define_split
  [(set (match_operand:SI                   0 "register_operand")
        (plus:SI (mult:SI (match_operand:SI 1 "register_operand")
                          (match_operand:SI 2 "register_operand"))
                 (match_operand:SI          3 "register_operand")))
   (clobber (match_scratch:SI               4))
   (clobber (match_scratch:SI               5))
   (clobber (match_scratch:SI               6))]
  "reload_completed && !TARGET_DEBUG_D_MODE
   && GP_REG_P (true_regnum (operands[0]))
   && true_regnum (operands[3]) == LO_REGNUM"
  [(parallel [(set (match_dup 3)
                   (plus:SI (mult:SI (match_dup 1) (match_dup 2))
                            (match_dup 3)))
              (clobber (match_dup 4))
              (clobber (match_dup 5))
              (clobber (match_dup 6))])
   (set (match_dup 0) (unspec:SI [(match_dup 5) (match_dup 4)] UNSPEC_MFHILO))]
  "")

(define_insn "*macc"
  [(set (match_operand:SI 0 "register_operand" "=l,d")
        (plus:SI (mult:SI (match_operand:SI 1 "register_operand" "d,d")
                          (match_operand:SI 2 "register_operand" "d,d"))
                 (match_operand:SI 3 "register_operand" "0,l")))
   (clobber (match_scratch:SI 4 "=h,h"))
   (clobber (match_scratch:SI 5 "=X,3"))]
  "ISA_HAS_MACC"
{
  if (which_alternative == 1)
    return "macc\t%0,%1,%2";
  else if (TARGET_MIPS5500)
    return "madd\t%1,%2";
  else
    /* The VR4130 assumes that there is a two-cycle latency between a macc
       that "writes" to $0 and an instruction that reads from it.  We avoid
       this by assigning to $1 instead.  */
    return "%[macc\t%@,%1,%2%]";
}
  [(set_attr "type" "imadd")
   (set_attr "mode" "SI")])

(define_insn "*msac"
  [(set (match_operand:SI 0 "register_operand" "=l,d")
        (minus:SI (match_operand:SI 1 "register_operand" "0,l")
                  (mult:SI (match_operand:SI 2 "register_operand" "d,d")
                           (match_operand:SI 3 "register_operand" "d,d"))))
   (clobber (match_scratch:SI 4 "=h,h"))
   (clobber (match_scratch:SI 5 "=X,1"))]
  "ISA_HAS_MSAC"
{
  if (which_alternative == 1)
    return "msac\t%0,%2,%3";
  else if (TARGET_MIPS5500)
    return "msub\t%2,%3";
  else
    return "msac\t$0,%2,%3";
}
  [(set_attr "type"     "imadd")
   (set_attr "mode"     "SI")])

;; An msac-like instruction implemented using negation and a macc.
(define_insn_and_split "*msac_using_macc"
  [(set (match_operand:SI 0 "register_operand" "=l,d")
        (minus:SI (match_operand:SI 1 "register_operand" "0,l")
                  (mult:SI (match_operand:SI 2 "register_operand" "d,d")
                           (match_operand:SI 3 "register_operand" "d,d"))))
   (clobber (match_scratch:SI 4 "=h,h"))
   (clobber (match_scratch:SI 5 "=X,1"))
   (clobber (match_scratch:SI 6 "=d,d"))]
  "ISA_HAS_MACC && !ISA_HAS_MSAC"
  "#"
  "&& reload_completed"
  [(set (match_dup 6)
        (neg:SI (match_dup 3)))
   (parallel
       [(set (match_dup 0)
             (plus:SI (mult:SI (match_dup 2)
                               (match_dup 6))
                      (match_dup 1)))
        (clobber (match_dup 4))
        (clobber (match_dup 5))])]
  ""
  [(set_attr "type"     "imadd")
   (set_attr "length"   "8")])

;; Patterns generated by the define_peephole2 below.

(define_insn "*macc2"
  [(set (match_operand:SI 0 "register_operand" "=l")
        (plus:SI (mult:SI (match_operand:SI 1 "register_operand" "d")
                          (match_operand:SI 2 "register_operand" "d"))
                 (match_dup 0)))
   (set (match_operand:SI 3 "register_operand" "=d")
        (plus:SI (mult:SI (match_dup 1)
                          (match_dup 2))
                 (match_dup 0)))
   (clobber (match_scratch:SI 4 "=h"))]
  "ISA_HAS_MACC && reload_completed"
  "macc\t%3,%1,%2"
  [(set_attr "type"     "imadd")
   (set_attr "mode"     "SI")])

(define_insn "*msac2"
  [(set (match_operand:SI 0 "register_operand" "=l")
        (minus:SI (match_dup 0)
                  (mult:SI (match_operand:SI 1 "register_operand" "d")
                           (match_operand:SI 2 "register_operand" "d"))))
   (set (match_operand:SI 3 "register_operand" "=d")
        (minus:SI (match_dup 0)
                  (mult:SI (match_dup 1)
                           (match_dup 2))))
   (clobber (match_scratch:SI 4 "=h"))]
  "ISA_HAS_MSAC && reload_completed"
  "msac\t%3,%1,%2"
  [(set_attr "type"     "imadd")
   (set_attr "mode"     "SI")])

;; Convert macc $0,<r1>,<r2> & mflo <r3> into macc <r3>,<r1>,<r2>
;; Similarly msac.
;;
;; Operand 0: LO
;; Operand 1: macc/msac
;; Operand 2: HI
;; Operand 3: GPR (destination)
(define_peephole2
  [(parallel
       [(set (match_operand:SI 0 "register_operand")
             (match_operand:SI 1 "macc_msac_operand"))
        (clobber (match_operand:SI 2 "register_operand"))
        (clobber (scratch:SI))])
   (set (match_operand:SI 3 "register_operand")
        (unspec:SI [(match_dup 0) (match_dup 2)] UNSPEC_MFHILO))]
  ""
  [(parallel [(set (match_dup 0)
                   (match_dup 1))
              (set (match_dup 3)
                   (match_dup 1))
              (clobber (match_dup 2))])]
  "")

;; When we have a three-address multiplication instruction, it should
;; be faster to do a separate multiply and add, rather than moving
;; something into LO in order to use a macc instruction.
;;
;; This peephole needs a scratch register to cater for the case when one
;; of the multiplication operands is the same as the destination.
;;
;; Operand 0: GPR (scratch)
;; Operand 1: LO
;; Operand 2: GPR (addend)
;; Operand 3: GPR (destination)
;; Operand 4: macc/msac
;; Operand 5: HI
;; Operand 6: new multiplication
;; Operand 7: new addition/subtraction
(define_peephole2
  [(match_scratch:SI 0 "d")
   (set (match_operand:SI 1 "register_operand")
        (match_operand:SI 2 "register_operand"))
   (match_dup 0)
   (parallel
       [(set (match_operand:SI 3 "register_operand")
             (match_operand:SI 4 "macc_msac_operand"))
        (clobber (match_operand:SI 5 "register_operand"))
        (clobber (match_dup 1))])]
  "GENERATE_MULT3_SI
   && true_regnum (operands[1]) == LO_REGNUM
   && peep2_reg_dead_p (2, operands[1])
   && GP_REG_P (true_regnum (operands[3]))"
  [(parallel [(set (match_dup 0)
                   (match_dup 6))
              (clobber (match_dup 5))
              (clobber (match_dup 1))])
   (set (match_dup 3)
        (match_dup 7))]
{
  operands[6] = XEXP (operands[4], GET_CODE (operands[4]) == PLUS ? 0 : 1);
  operands[7] = gen_rtx_fmt_ee (GET_CODE (operands[4]), SImode,
                                operands[2], operands[0]);
})

;; Same as above, except LO is the initial target of the macc.
;;
;; Operand 0: GPR (scratch)
;; Operand 1: LO
;; Operand 2: GPR (addend)
;; Operand 3: macc/msac
;; Operand 4: HI
;; Operand 5: GPR (destination)
;; Operand 6: new multiplication
;; Operand 7: new addition/subtraction
(define_peephole2
  [(match_scratch:SI 0 "d")
   (set (match_operand:SI 1 "register_operand")
        (match_operand:SI 2 "register_operand"))
   (match_dup 0)
   (parallel
       [(set (match_dup 1)
             (match_operand:SI 3 "macc_msac_operand"))
        (clobber (match_operand:SI 4 "register_operand"))
        (clobber (scratch:SI))])
   (match_dup 0)
   (set (match_operand:SI 5 "register_operand")
        (unspec:SI [(match_dup 1) (match_dup 4)] UNSPEC_MFHILO))]
  "GENERATE_MULT3_SI && peep2_reg_dead_p (3, operands[1])"
  [(parallel [(set (match_dup 0)
                   (match_dup 6))
              (clobber (match_dup 4))
              (clobber (match_dup 1))])
   (set (match_dup 5)
        (match_dup 7))]
{
  operands[6] = XEXP (operands[4], GET_CODE (operands[4]) == PLUS ? 0 : 1);
  operands[7] = gen_rtx_fmt_ee (GET_CODE (operands[4]), SImode,
                                operands[2], operands[0]);
})

(define_insn "*mul_sub_si"
  [(set (match_operand:SI 0 "register_operand" "=l,*d,*d")
        (minus:SI (match_operand:SI 1 "register_operand" "0,l,*d")
                  (mult:SI (match_operand:SI 2 "register_operand" "d,d,d")
                           (match_operand:SI 3 "register_operand" "d,d,d"))))
   (clobber (match_scratch:SI 4 "=h,h,h"))
   (clobber (match_scratch:SI 5 "=X,1,l"))
   (clobber (match_scratch:SI 6 "=X,X,&d"))]
  "ISA_HAS_MADD_MSUB"
  "@
   msub\t%2,%3
   #
   #"
  [(set_attr "type"     "imadd,multi,multi")
   (set_attr "mode"     "SI")
   (set_attr "length"   "4,8,8")])

;; Split the above insn if we failed to get LO allocated.
(define_split
  [(set (match_operand:SI 0 "register_operand")
        (minus:SI (match_operand:SI 1 "register_operand")
                  (mult:SI (match_operand:SI 2 "register_operand")
                           (match_operand:SI 3 "register_operand"))))
   (clobber (match_scratch:SI 4))
   (clobber (match_scratch:SI 5))
   (clobber (match_scratch:SI 6))]
  "reload_completed && !TARGET_DEBUG_D_MODE
   && GP_REG_P (true_regnum (operands[0]))
   && GP_REG_P (true_regnum (operands[1]))"
  [(parallel [(set (match_dup 6)
                   (mult:SI (match_dup 2) (match_dup 3)))
              (clobber (match_dup 4))
              (clobber (match_dup 5))])
   (set (match_dup 0) (minus:SI (match_dup 1) (match_dup 6)))]
  "")

;; Splitter to copy result of MSUB to a general register
(define_split
  [(set (match_operand:SI 0 "register_operand")
        (minus:SI (match_operand:SI 1 "register_operand")
                  (mult:SI (match_operand:SI 2 "register_operand")
                           (match_operand:SI 3 "register_operand"))))
   (clobber (match_scratch:SI 4))
   (clobber (match_scratch:SI 5))
   (clobber (match_scratch:SI 6))]
  "reload_completed && !TARGET_DEBUG_D_MODE
   && GP_REG_P (true_regnum (operands[0]))
   && true_regnum (operands[1]) == LO_REGNUM"
  [(parallel [(set (match_dup 1)
                   (minus:SI (match_dup 1)
                             (mult:SI (match_dup 2) (match_dup 3))))
              (clobber (match_dup 4))
              (clobber (match_dup 5))
              (clobber (match_dup 6))])
   (set (match_dup 0) (unspec:SI [(match_dup 5) (match_dup 4)] UNSPEC_MFHILO))]
  "")

(define_insn "*muls"
  [(set (match_operand:SI                  0 "register_operand" "=l,d")
        (neg:SI (mult:SI (match_operand:SI 1 "register_operand" "d,d")
                         (match_operand:SI 2 "register_operand" "d,d"))))
   (clobber (match_scratch:SI              3                    "=h,h"))
   (clobber (match_scratch:SI              4                    "=X,l"))]
  "ISA_HAS_MULS"
  "@
   muls\t$0,%1,%2
   muls\t%0,%1,%2"
  [(set_attr "type"     "imul,imul3")
   (set_attr "mode"     "SI")])

;; ??? We could define a mulditi3 pattern when TARGET_64BIT.

(define_expand "<u>mulsidi3"
  [(parallel
      [(set (match_operand:DI 0 "register_operand")
            (mult:DI (any_extend:DI (match_operand:SI 1 "register_operand"))
                     (any_extend:DI (match_operand:SI 2 "register_operand"))))
       (clobber (scratch:DI))
       (clobber (scratch:DI))
       (clobber (scratch:DI))])]
  "!TARGET_64BIT || !TARGET_FIX_R4000"
{
  if (!TARGET_64BIT)
    {
      if (!TARGET_FIX_R4000)
        emit_insn (gen_<u>mulsidi3_32bit_internal (operands[0], operands[1],
                                                   operands[2]));
      else
        emit_insn (gen_<u>mulsidi3_32bit_r4000 (operands[0], operands[1],
                                                operands[2]));
      DONE;
    }
})

(define_insn "<u>mulsidi3_32bit_internal"
  [(set (match_operand:DI 0 "register_operand" "=x")
        (mult:DI (any_extend:DI (match_operand:SI 1 "register_operand" "d"))
                 (any_extend:DI (match_operand:SI 2 "register_operand" "d"))))]
  "!TARGET_64BIT && !TARGET_FIX_R4000"
  "mult<u>\t%1,%2"
  [(set_attr "type" "imul")
   (set_attr "mode" "SI")])

(define_insn "<u>mulsidi3_32bit_r4000"
  [(set (match_operand:DI 0 "register_operand" "=d")
        (mult:DI (any_extend:DI (match_operand:SI 1 "register_operand" "d"))
                 (any_extend:DI (match_operand:SI 2 "register_operand" "d"))))
   (clobber (match_scratch:DI 3 "=x"))]
  "!TARGET_64BIT && TARGET_FIX_R4000"
  "mult<u>\t%1,%2\;mflo\t%L0;mfhi\t%M0"
  [(set_attr "type" "imul")
   (set_attr "mode" "SI")
   (set_attr "length" "12")])

(define_insn_and_split "*<u>mulsidi3_64bit"
  [(set (match_operand:DI 0 "register_operand" "=d")
        (mult:DI (any_extend:DI (match_operand:SI 1 "register_operand" "d"))
                 (any_extend:DI (match_operand:SI 2 "register_operand" "d"))))
   (clobber (match_scratch:DI 3 "=l"))
   (clobber (match_scratch:DI 4 "=h"))
   (clobber (match_scratch:DI 5 "=d"))]
  "TARGET_64BIT && !TARGET_FIX_R4000"
  "#"
  "&& reload_completed"
  [(parallel
       [(set (match_dup 3)
             (sign_extend:DI
                (mult:SI (match_dup 1)
                         (match_dup 2))))
        (set (match_dup 4)
             (ashiftrt:DI
                (mult:DI (any_extend:DI (match_dup 1))
                         (any_extend:DI (match_dup 2)))
                (const_int 32)))])

   ;; OP5 <- LO, OP0 <- HI
   (set (match_dup 5) (unspec:DI [(match_dup 3) (match_dup 4)] UNSPEC_MFHILO))
   (set (match_dup 0) (unspec:DI [(match_dup 4) (match_dup 3)] UNSPEC_MFHILO))

   ;; Zero-extend OP5.
   (set (match_dup 5)
        (ashift:DI (match_dup 5)
                   (const_int 32)))
   (set (match_dup 5)
        (lshiftrt:DI (match_dup 5)
                     (const_int 32)))

   ;; Shift OP0 into place.
   (set (match_dup 0)
        (ashift:DI (match_dup 0)
                   (const_int 32)))

   ;; OR the two halves together
   (set (match_dup 0)
        (ior:DI (match_dup 0)
                (match_dup 5)))]
  ""
  [(set_attr "type" "imul")
   (set_attr "mode" "SI")
   (set_attr "length" "24")])

(define_insn "*<u>mulsidi3_64bit_parts"
  [(set (match_operand:DI 0 "register_operand" "=l")
        (sign_extend:DI
           (mult:SI (match_operand:SI 2 "register_operand" "d")
                    (match_operand:SI 3 "register_operand" "d"))))
   (set (match_operand:DI 1 "register_operand" "=h")
        (ashiftrt:DI
           (mult:DI (any_extend:DI (match_dup 2))
                    (any_extend:DI (match_dup 3)))
           (const_int 32)))]
  "TARGET_64BIT && !TARGET_FIX_R4000"
  "mult<u>\t%2,%3"
  [(set_attr "type" "imul")
   (set_attr "mode" "SI")])

;; Widening multiply with negation.
(define_insn "*muls<u>_di"
  [(set (match_operand:DI 0 "register_operand" "=x")
        (neg:DI
         (mult:DI
          (any_extend:DI (match_operand:SI 1 "register_operand" "d"))
          (any_extend:DI (match_operand:SI 2 "register_operand" "d")))))]
  "!TARGET_64BIT && ISA_HAS_MULS"
  "muls<u>\t$0,%1,%2"
  [(set_attr "type" "imul")
   (set_attr "mode" "SI")])

(define_insn "*msac<u>_di"
  [(set (match_operand:DI 0 "register_operand" "=x")
        (minus:DI
           (match_operand:DI 3 "register_operand" "0")
           (mult:DI
              (any_extend:DI (match_operand:SI 1 "register_operand" "d"))
              (any_extend:DI (match_operand:SI 2 "register_operand" "d")))))]
  "!TARGET_64BIT && ISA_HAS_MSAC"
{
  if (TARGET_MIPS5500)
    return "msub<u>\t%1,%2";
  else
    return "msac<u>\t$0,%1,%2";
}
  [(set_attr "type" "imadd")
   (set_attr "mode" "SI")])

;; _highpart patterns

(define_expand "<su>mulsi3_highpart"
  [(set (match_operand:SI 0 "register_operand")
        (truncate:SI
         (lshiftrt:DI
          (mult:DI (any_extend:DI (match_operand:SI 1 "register_operand"))
                   (any_extend:DI (match_operand:SI 2 "register_operand")))
          (const_int 32))))]
  "ISA_HAS_MULHI || !TARGET_FIX_R4000"
{
  if (ISA_HAS_MULHI)
    emit_insn (gen_<su>mulsi3_highpart_mulhi_internal (operands[0],
                                                       operands[1],
                                                       operands[2]));
  else
    emit_insn (gen_<su>mulsi3_highpart_internal (operands[0], operands[1],
                                                 operands[2]));
  DONE;
})

(define_insn "<su>mulsi3_highpart_internal"
  [(set (match_operand:SI 0 "register_operand" "=h")
        (truncate:SI
         (lshiftrt:DI
          (mult:DI (any_extend:DI (match_operand:SI 1 "register_operand" "d"))
                   (any_extend:DI (match_operand:SI 2 "register_operand" "d")))
          (const_int 32))))
   (clobber (match_scratch:SI 3 "=l"))]
  "!ISA_HAS_MULHI && !TARGET_FIX_R4000"
  "mult<u>\t%1,%2"
  [(set_attr "type" "imul")
   (set_attr "mode" "SI")])

(define_insn "<su>mulsi3_highpart_mulhi_internal"
  [(set (match_operand:SI 0 "register_operand" "=h,d")
        (truncate:SI
         (lshiftrt:DI
          (mult:DI
           (any_extend:DI (match_operand:SI 1 "register_operand" "d,d"))
           (any_extend:DI (match_operand:SI 2 "register_operand" "d,d")))
          (const_int 32))))
   (clobber (match_scratch:SI 3 "=l,l"))
   (clobber (match_scratch:SI 4 "=X,h"))]
  "ISA_HAS_MULHI"
  "@
   mult<u>\t%1,%2
   mulhi<u>\t%0,%1,%2"
  [(set_attr "type" "imul,imul3")
   (set_attr "mode" "SI")])

(define_insn "*<su>mulsi3_highpart_neg_mulhi_internal"
  [(set (match_operand:SI 0 "register_operand" "=h,d")
        (truncate:SI
         (lshiftrt:DI
          (neg:DI
           (mult:DI
            (any_extend:DI (match_operand:SI 1 "register_operand" "d,d"))
            (any_extend:DI (match_operand:SI 2 "register_operand" "d,d"))))
          (const_int 32))))
   (clobber (match_scratch:SI 3 "=l,l"))
   (clobber (match_scratch:SI 4 "=X,h"))]
  "ISA_HAS_MULHI"
  "@
   mulshi<u>\t%.,%1,%2
   mulshi<u>\t%0,%1,%2"
  [(set_attr "type" "imul,imul3")
   (set_attr "mode" "SI")])

;; Disable unsigned multiplication for -mfix-vr4120.  This is for VR4120
;; errata MD(0), which says that dmultu does not always produce the
;; correct result.
(define_insn "<su>muldi3_highpart"
  [(set (match_operand:DI 0 "register_operand" "=h")
        (truncate:DI
         (lshiftrt:TI
          (mult:TI
           (any_extend:TI (match_operand:DI 1 "register_operand" "d"))
           (any_extend:TI (match_operand:DI 2 "register_operand" "d")))
          (const_int 64))))
   (clobber (match_scratch:DI 3 "=l"))]
  "TARGET_64BIT && !TARGET_FIX_R4000
   && !(<CODE> == ZERO_EXTEND && TARGET_FIX_VR4120)"
  "dmult<u>\t%1,%2"
  [(set_attr "type" "imul")
   (set_attr "mode" "DI")])

;; The R4650 supports a 32 bit multiply/ 64 bit accumulate
;; instruction.  The HI/LO registers are used as a 64 bit accumulator.

(define_insn "madsi"
  [(set (match_operand:SI 0 "register_operand" "+l")
        (plus:SI (mult:SI (match_operand:SI 1 "register_operand" "d")
                          (match_operand:SI 2 "register_operand" "d"))
                 (match_dup 0)))
   (clobber (match_scratch:SI 3 "=h"))]
  "TARGET_MAD"
  "mad\t%1,%2"
  [(set_attr "type"     "imadd")
   (set_attr "mode"     "SI")])

(define_insn "*<su>mul_acc_di"
  [(set (match_operand:DI 0 "register_operand" "=x")
        (plus:DI
         (mult:DI (any_extend:DI (match_operand:SI 1 "register_operand" "d"))
                  (any_extend:DI (match_operand:SI 2 "register_operand" "d")))
         (match_operand:DI 3 "register_operand" "0")))]
  "(TARGET_MAD || ISA_HAS_MACC)
   && !TARGET_64BIT"
{
  if (TARGET_MAD)
    return "mad<u>\t%1,%2";
  else if (TARGET_MIPS5500)
    return "madd<u>\t%1,%2";
  else
    /* See comment in *macc.  */
    return "%[macc<u>\t%@,%1,%2%]";
}
  [(set_attr "type" "imadd")
   (set_attr "mode" "SI")])

;; Floating point multiply accumulate instructions.

(define_insn "*madd<mode>"
  [(set (match_operand:ANYF 0 "register_operand" "=f")
        (plus:ANYF (mult:ANYF (match_operand:ANYF 1 "register_operand" "f")
                              (match_operand:ANYF 2 "register_operand" "f"))
                   (match_operand:ANYF 3 "register_operand" "f")))]
  "ISA_HAS_FP4 && TARGET_FUSED_MADD"
  "madd.<fmt>\t%0,%3,%1,%2"
  [(set_attr "type" "fmadd")
   (set_attr "mode" "<UNITMODE>")])

(define_insn "*msub<mode>"
  [(set (match_operand:ANYF 0 "register_operand" "=f")
        (minus:ANYF (mult:ANYF (match_operand:ANYF 1 "register_operand" "f")
                               (match_operand:ANYF 2 "register_operand" "f"))
                    (match_operand:ANYF 3 "register_operand" "f")))]
  "ISA_HAS_FP4 && TARGET_FUSED_MADD"
  "msub.<fmt>\t%0,%3,%1,%2"
  [(set_attr "type" "fmadd")
   (set_attr "mode" "<UNITMODE>")])

(define_insn "*nmadd<mode>"
  [(set (match_operand:ANYF 0 "register_operand" "=f")
        (neg:ANYF (plus:ANYF
                   (mult:ANYF (match_operand:ANYF 1 "register_operand" "f")
                              (match_operand:ANYF 2 "register_operand" "f"))
                   (match_operand:ANYF 3 "register_operand" "f"))))]
  "ISA_HAS_NMADD_NMSUB && TARGET_FUSED_MADD
   && HONOR_SIGNED_ZEROS (<MODE>mode)"
  "nmadd.<fmt>\t%0,%3,%1,%2"
  [(set_attr "type" "fmadd")
   (set_attr "mode" "<UNITMODE>")])

(define_insn "*nmadd<mode>_fastmath"
  [(set (match_operand:ANYF 0 "register_operand" "=f")
        (minus:ANYF
         (mult:ANYF (neg:ANYF (match_operand:ANYF 1 "register_operand" "f"))
                    (match_operand:ANYF 2 "register_operand" "f"))
         (match_operand:ANYF 3 "register_operand" "f")))]
  "ISA_HAS_NMADD_NMSUB && TARGET_FUSED_MADD
   && !HONOR_SIGNED_ZEROS (<MODE>mode)"
  "nmadd.<fmt>\t%0,%3,%1,%2"
  [(set_attr "type" "fmadd")
   (set_attr "mode" "<UNITMODE>")])

(define_insn "*nmsub<mode>"
  [(set (match_operand:ANYF 0 "register_operand" "=f")
        (neg:ANYF (minus:ANYF
                   (mult:ANYF (match_operand:ANYF 2 "register_operand" "f")
                              (match_operand:ANYF 3 "register_operand" "f"))
                   (match_operand:ANYF 1 "register_operand" "f"))))]
  "ISA_HAS_NMADD_NMSUB && TARGET_FUSED_MADD
   && HONOR_SIGNED_ZEROS (<MODE>mode)"
  "nmsub.<fmt>\t%0,%1,%2,%3"
  [(set_attr "type" "fmadd")
   (set_attr "mode" "<UNITMODE>")])

(define_insn "*nmsub<mode>_fastmath"
  [(set (match_operand:ANYF 0 "register_operand" "=f")
        (minus:ANYF
         (match_operand:ANYF 1 "register_operand" "f")
         (mult:ANYF (match_operand:ANYF 2 "register_operand" "f")
                    (match_operand:ANYF 3 "register_operand" "f"))))]
  "ISA_HAS_NMADD_NMSUB && TARGET_FUSED_MADD
   && !HONOR_SIGNED_ZEROS (<MODE>mode)"
  "nmsub.<fmt>\t%0,%1,%2,%3"
  [(set_attr "type" "fmadd")
   (set_attr "mode" "<UNITMODE>")])
\f
;;
;;  ....................
;;
;;      DIVISION and REMAINDER
;;
;;  ....................
;;

(define_expand "div<mode>3"
  [(set (match_operand:ANYF 0 "register_operand")
        (div:ANYF (match_operand:ANYF 1 "reg_or_1_operand")
                  (match_operand:ANYF 2 "register_operand")))]
  "<divide_condition>"
{
  if (const_1_operand (operands[1], <MODE>mode))
    if (!(ISA_HAS_FP4 && flag_unsafe_math_optimizations))
      operands[1] = force_reg (<MODE>mode, operands[1]);
})

;; These patterns work around the early SB-1 rev2 core "F1" erratum:
;;
;; If an mfc1 or dmfc1 happens to access the floating point register
;; file at the same time a long latency operation (div, sqrt, recip,
;; sqrt) iterates an intermediate result back through the floating
;; point register file bypass, then instead returning the correct
;; register value the mfc1 or dmfc1 operation returns the intermediate
;; result of the long latency operation.
;;
;; The workaround is to insert an unconditional 'mov' from/to the
;; long latency op destination register.

(define_insn "*div<mode>3"
  [(set (match_operand:ANYF 0 "register_operand" "=f")
        (div:ANYF (match_operand:ANYF 1 "register_operand" "f")
                  (match_operand:ANYF 2 "register_operand" "f")))]
  "<divide_condition>"
{
  if (TARGET_FIX_SB1)
    return "div.<fmt>\t%0,%1,%2\;mov.<fmt>\t%0,%0";
  else
    return "div.<fmt>\t%0,%1,%2";
}
  [(set_attr "type" "fdiv")
   (set_attr "mode" "<UNITMODE>")
   (set (attr "length")
        (if_then_else (ne (symbol_ref "TARGET_FIX_SB1") (const_int 0))
                      (const_int 8)
                      (const_int 4)))])

(define_insn "*recip<mode>3"
  [(set (match_operand:ANYF 0 "register_operand" "=f")
        (div:ANYF (match_operand:ANYF 1 "const_1_operand" "")
                  (match_operand:ANYF 2 "register_operand" "f")))]
  "<recip_condition> && flag_unsafe_math_optimizations"
{
  if (TARGET_FIX_SB1)
    return "recip.<fmt>\t%0,%2\;mov.<fmt>\t%0,%0";
  else
    return "recip.<fmt>\t%0,%2";
}
  [(set_attr "type" "frdiv")
   (set_attr "mode" "<UNITMODE>")
   (set (attr "length")
        (if_then_else (ne (symbol_ref "TARGET_FIX_SB1") (const_int 0))
                      (const_int 8)
                      (const_int 4)))])

;; VR4120 errata MD(A1): signed division instructions do not work correctly
;; with negative operands.  We use special libgcc functions instead.
(define_insn "divmod<mode>4"
  [(set (match_operand:GPR 0 "register_operand" "=l")
        (div:GPR (match_operand:GPR 1 "register_operand" "d")
                 (match_operand:GPR 2 "register_operand" "d")))
   (set (match_operand:GPR 3 "register_operand" "=h")
        (mod:GPR (match_dup 1)
                 (match_dup 2)))]
  "!TARGET_FIX_VR4120"
  { return mips_output_division ("<d>div\t$0,%1,%2", operands); }
  [(set_attr "type" "idiv")
   (set_attr "mode" "<MODE>")])

(define_insn "udivmod<mode>4"
  [(set (match_operand:GPR 0 "register_operand" "=l")
        (udiv:GPR (match_operand:GPR 1 "register_operand" "d")
                  (match_operand:GPR 2 "register_operand" "d")))
   (set (match_operand:GPR 3 "register_operand" "=h")
        (umod:GPR (match_dup 1)
                  (match_dup 2)))]
  ""
  { return mips_output_division ("<d>divu\t$0,%1,%2", operands); }
  [(set_attr "type" "idiv")
   (set_attr "mode" "<MODE>")])
\f
;;
;;  ....................
;;
;;      SQUARE ROOT
;;
;;  ....................

;; These patterns work around the early SB-1 rev2 core "F1" erratum (see
;; "*div[sd]f3" comment for details).

(define_insn "sqrt<mode>2"
  [(set (match_operand:ANYF 0 "register_operand" "=f")
        (sqrt:ANYF (match_operand:ANYF 1 "register_operand" "f")))]
  "<sqrt_condition>"
{
  if (TARGET_FIX_SB1)
    return "sqrt.<fmt>\t%0,%1\;mov.<fmt>\t%0,%0";
  else
    return "sqrt.<fmt>\t%0,%1";
}
  [(set_attr "type" "fsqrt")
   (set_attr "mode" "<UNITMODE>")
   (set (attr "length")
        (if_then_else (ne (symbol_ref "TARGET_FIX_SB1") (const_int 0))
                      (const_int 8)
                      (const_int 4)))])

(define_insn "*rsqrt<mode>a"
  [(set (match_operand:ANYF 0 "register_operand" "=f")
        (div:ANYF (match_operand:ANYF 1 "const_1_operand" "")
                  (sqrt:ANYF (match_operand:ANYF 2 "register_operand" "f"))))]
  "<recip_condition> && flag_unsafe_math_optimizations"
{
  if (TARGET_FIX_SB1)
    return "rsqrt.<fmt>\t%0,%2\;mov.<fmt>\t%0,%0";
  else
    return "rsqrt.<fmt>\t%0,%2";
}
  [(set_attr "type" "frsqrt")
   (set_attr "mode" "<UNITMODE>")
   (set (attr "length")
        (if_then_else (ne (symbol_ref "TARGET_FIX_SB1") (const_int 0))
                      (const_int 8)
                      (const_int 4)))])

(define_insn "*rsqrt<mode>b"
  [(set (match_operand:ANYF 0 "register_operand" "=f")
        (sqrt:ANYF (div:ANYF (match_operand:ANYF 1 "const_1_operand" "")
                             (match_operand:ANYF 2 "register_operand" "f"))))]
  "<recip_condition> && flag_unsafe_math_optimizations"
{
  if (TARGET_FIX_SB1)
    return "rsqrt.<fmt>\t%0,%2\;mov.<fmt>\t%0,%0";
  else
    return "rsqrt.<fmt>\t%0,%2";
}
  [(set_attr "type" "frsqrt")
   (set_attr "mode" "<UNITMODE>")
   (set (attr "length")
        (if_then_else (ne (symbol_ref "TARGET_FIX_SB1") (const_int 0))
                      (const_int 8)
                      (const_int 4)))])
\f
;;
;;  ....................
;;
;;      ABSOLUTE VALUE
;;
;;  ....................

;; Do not use the integer abs macro instruction, since that signals an
;; exception on -2147483648 (sigh).

(define_insn "abs<mode>2"
  [(set (match_operand:ANYF 0 "register_operand" "=f")
        (abs:ANYF (match_operand:ANYF 1 "register_operand" "f")))]
  ""
  "abs.<fmt>\t%0,%1"
  [(set_attr "type" "fabs")
   (set_attr "mode" "<UNITMODE>")])
\f
;;
;;  ...................
;;
;;  Count leading zeroes.
;;
;;  ...................
;;

(define_insn "clz<mode>2"
  [(set (match_operand:GPR 0 "register_operand" "=d")
        (clz:GPR (match_operand:GPR 1 "register_operand" "d")))]
  "ISA_HAS_CLZ_CLO"
  "<d>clz\t%0,%1"
  [(set_attr "type" "clz")
   (set_attr "mode" "<MODE>")])
\f
;;
;;  ....................
;;
;;      NEGATION and ONE'S COMPLEMENT
;;
;;  ....................

(define_insn "negsi2"
  [(set (match_operand:SI 0 "register_operand" "=d")
        (neg:SI (match_operand:SI 1 "register_operand" "d")))]
  ""
{
  if (TARGET_MIPS16)
    return "neg\t%0,%1";
  else
    return "subu\t%0,%.,%1";
}
  [(set_attr "type"     "arith")
   (set_attr "mode"     "SI")])

(define_insn "negdi2"
  [(set (match_operand:DI 0 "register_operand" "=d")
        (neg:DI (match_operand:DI 1 "register_operand" "d")))]
  "TARGET_64BIT && !TARGET_MIPS16"
  "dsubu\t%0,%.,%1"
  [(set_attr "type"     "arith")
   (set_attr "mode"     "DI")])

(define_insn "neg<mode>2"
  [(set (match_operand:ANYF 0 "register_operand" "=f")
        (neg:ANYF (match_operand:ANYF 1 "register_operand" "f")))]
  ""
  "neg.<fmt>\t%0,%1"
  [(set_attr "type" "fneg")
   (set_attr "mode" "<UNITMODE>")])

(define_insn "one_cmpl<mode>2"
  [(set (match_operand:GPR 0 "register_operand" "=d")
        (not:GPR (match_operand:GPR 1 "register_operand" "d")))]
  ""
{
  if (TARGET_MIPS16)
    return "not\t%0,%1";
  else
    return "nor\t%0,%.,%1";
}
  [(set_attr "type" "arith")
   (set_attr "mode" "<MODE>")])
\f
;;
;;  ....................
;;
;;      LOGICAL
;;
;;  ....................
;;

;; Many of these instructions use trivial define_expands, because we
;; want to use a different set of constraints when TARGET_MIPS16.

(define_expand "and<mode>3"
  [(set (match_operand:GPR 0 "register_operand")
        (and:GPR (match_operand:GPR 1 "register_operand")
                 (match_operand:GPR 2 "uns_arith_operand")))]
  ""
{
  if (TARGET_MIPS16)
    operands[2] = force_reg (<MODE>mode, operands[2]);
})

(define_insn "*and<mode>3"
  [(set (match_operand:GPR 0 "register_operand" "=d,d")
        (and:GPR (match_operand:GPR 1 "register_operand" "%d,d")
                 (match_operand:GPR 2 "uns_arith_operand" "d,K")))]
  "!TARGET_MIPS16"
  "@
   and\t%0,%1,%2
   andi\t%0,%1,%x2"
  [(set_attr "type" "arith")
   (set_attr "mode" "<MODE>")])

(define_insn "*and<mode>3_mips16"
  [(set (match_operand:GPR 0 "register_operand" "=d")
        (and:GPR (match_operand:GPR 1 "register_operand" "%0")
                 (match_operand:GPR 2 "register_operand" "d")))]
  "TARGET_MIPS16"
  "and\t%0,%2"
  [(set_attr "type" "arith")
   (set_attr "mode" "<MODE>")])

(define_expand "ior<mode>3"
  [(set (match_operand:GPR 0 "register_operand")
        (ior:GPR (match_operand:GPR 1 "register_operand")
                 (match_operand:GPR 2 "uns_arith_operand")))]
  ""
{
  if (TARGET_MIPS16)
    operands[2] = force_reg (<MODE>mode, operands[2]);
})

(define_insn "*ior<mode>3"
  [(set (match_operand:GPR 0 "register_operand" "=d,d")
        (ior:GPR (match_operand:GPR 1 "register_operand" "%d,d")
                 (match_operand:GPR 2 "uns_arith_operand" "d,K")))]
  "!TARGET_MIPS16"
  "@
   or\t%0,%1,%2
   ori\t%0,%1,%x2"
  [(set_attr "type" "arith")
   (set_attr "mode" "<MODE>")])

(define_insn "*ior<mode>3_mips16"
  [(set (match_operand:GPR 0 "register_operand" "=d")
        (ior:GPR (match_operand:GPR 1 "register_operand" "%0")
                 (match_operand:GPR 2 "register_operand" "d")))]
  "TARGET_MIPS16"
  "or\t%0,%2"
  [(set_attr "type" "arith")
   (set_attr "mode" "<MODE>")])

(define_expand "xor<mode>3"
  [(set (match_operand:GPR 0 "register_operand")
        (xor:GPR (match_operand:GPR 1 "register_operand")
                 (match_operand:GPR 2 "uns_arith_operand")))]
  ""
  "")

(define_insn ""
  [(set (match_operand:GPR 0 "register_operand" "=d,d")
        (xor:GPR (match_operand:GPR 1 "register_operand" "%d,d")
                 (match_operand:GPR 2 "uns_arith_operand" "d,K")))]
  "!TARGET_MIPS16"
  "@
   xor\t%0,%1,%2
   xori\t%0,%1,%x2"
  [(set_attr "type" "arith")
   (set_attr "mode" "<MODE>")])

(define_insn ""
  [(set (match_operand:GPR 0 "register_operand" "=d,t,t")
        (xor:GPR (match_operand:GPR 1 "register_operand" "%0,d,d")
                 (match_operand:GPR 2 "uns_arith_operand" "d,K,d")))]
  "TARGET_MIPS16"
  "@
   xor\t%0,%2
   cmpi\t%1,%2
   cmp\t%1,%2"
  [(set_attr "type" "arith")
   (set_attr "mode" "<MODE>")
   (set_attr_alternative "length"
                [(const_int 4)
                 (if_then_else (match_operand:VOID 2 "m16_uimm8_1")
                               (const_int 4)
                               (const_int 8))
                 (const_int 4)])])

(define_insn "*nor<mode>3"
  [(set (match_operand:GPR 0 "register_operand" "=d")
        (and:GPR (not:GPR (match_operand:GPR 1 "register_operand" "d"))
                 (not:GPR (match_operand:GPR 2 "register_operand" "d"))))]
  "!TARGET_MIPS16"
  "nor\t%0,%1,%2"
  [(set_attr "type" "arith")
   (set_attr "mode" "<MODE>")])
\f
;;
;;  ....................
;;
;;      TRUNCATION
;;
;;  ....................



(define_insn "truncdfsf2"
  [(set (match_operand:SF 0 "register_operand" "=f")
        (float_truncate:SF (match_operand:DF 1 "register_operand" "f")))]
  "TARGET_HARD_FLOAT && TARGET_DOUBLE_FLOAT"
  "cvt.s.d\t%0,%1"
  [(set_attr "type"     "fcvt")
   (set_attr "cnv_mode" "D2S")   
   (set_attr "mode"     "SF")])

;; Integer truncation patterns.  Truncating SImode values to smaller
;; modes is a no-op, as it is for most other GCC ports.  Truncating
;; DImode values to SImode is not a no-op for TARGET_64BIT since we
;; need to make sure that the lower 32 bits are properly sign-extended
;; (see TRULY_NOOP_TRUNCATION).  Truncating DImode values into modes
;; smaller than SImode is equivalent to two separate truncations:
;;
;;                        A       B
;;    DI ---> HI  ==  DI ---> SI ---> HI
;;    DI ---> QI  ==  DI ---> SI ---> QI
;;
;; Step A needs a real instruction but step B does not.

(define_insn "truncdisi2"
  [(set (match_operand:SI 0 "nonimmediate_operand" "=d,m")
        (truncate:SI (match_operand:DI 1 "register_operand" "d,d")))]
  "TARGET_64BIT"
  "@
    sll\t%0,%1,0
    sw\t%1,%0"
  [(set_attr "type" "shift,store")
   (set_attr "mode" "SI")
   (set_attr "extended_mips16" "yes,*")])

(define_insn "truncdihi2"
  [(set (match_operand:HI 0 "nonimmediate_operand" "=d,m")
        (truncate:HI (match_operand:DI 1 "register_operand" "d,d")))]
  "TARGET_64BIT"
  "@
    sll\t%0,%1,0
    sh\t%1,%0"
  [(set_attr "type" "shift,store")
   (set_attr "mode" "SI")
   (set_attr "extended_mips16" "yes,*")])

(define_insn "truncdiqi2"
  [(set (match_operand:QI 0 "nonimmediate_operand" "=d,m")
        (truncate:QI (match_operand:DI 1 "register_operand" "d,d")))]
  "TARGET_64BIT"
  "@
    sll\t%0,%1,0
    sb\t%1,%0"
  [(set_attr "type" "shift,store")
   (set_attr "mode" "SI")
   (set_attr "extended_mips16" "yes,*")])

;; Combiner patterns to optimize shift/truncate combinations.

(define_insn ""
  [(set (match_operand:SI 0 "register_operand" "=d")
        (truncate:SI
          (ashiftrt:DI (match_operand:DI 1 "register_operand" "d")
                       (match_operand:DI 2 "const_arith_operand" ""))))]
  "TARGET_64BIT && !TARGET_MIPS16 && INTVAL (operands[2]) >= 32"
  "dsra\t%0,%1,%2"
  [(set_attr "type" "shift")
   (set_attr "mode" "SI")])

(define_insn ""
  [(set (match_operand:SI 0 "register_operand" "=d")
        (truncate:SI (lshiftrt:DI (match_operand:DI 1 "register_operand" "d")
                                  (const_int 32))))]
  "TARGET_64BIT && !TARGET_MIPS16"
  "dsra\t%0,%1,32"
  [(set_attr "type" "shift")
   (set_attr "mode" "SI")])


;; Combiner patterns for truncate/sign_extend combinations.  They use
;; the shift/truncate patterns above.

(define_insn_and_split ""
  [(set (match_operand:SI 0 "register_operand" "=d")
        (sign_extend:SI
            (truncate:HI (match_operand:DI 1 "register_operand" "d"))))]
  "TARGET_64BIT && !TARGET_MIPS16"
  "#"
  "&& reload_completed"
  [(set (match_dup 2)
        (ashift:DI (match_dup 1)
                   (const_int 48)))
   (set (match_dup 0)
        (truncate:SI (ashiftrt:DI (match_dup 2)
                                  (const_int 48))))]
  { operands[2] = gen_lowpart (DImode, operands[0]); })

(define_insn_and_split ""
  [(set (match_operand:SI 0 "register_operand" "=d")
        (sign_extend:SI
            (truncate:QI (match_operand:DI 1 "register_operand" "d"))))]
  "TARGET_64BIT && !TARGET_MIPS16"
  "#"
  "&& reload_completed"
  [(set (match_dup 2)
        (ashift:DI (match_dup 1)
                   (const_int 56)))
   (set (match_dup 0)
        (truncate:SI (ashiftrt:DI (match_dup 2)
                                  (const_int 56))))]
  { operands[2] = gen_lowpart (DImode, operands[0]); })


;; Combiner patterns to optimize truncate/zero_extend combinations.

(define_insn ""
  [(set (match_operand:SI 0 "register_operand" "=d")
        (zero_extend:SI (truncate:HI
                         (match_operand:DI 1 "register_operand" "d"))))]
  "TARGET_64BIT && !TARGET_MIPS16"
  "andi\t%0,%1,0xffff"
  [(set_attr "type"     "arith")
   (set_attr "mode"     "SI")])

(define_insn ""
  [(set (match_operand:SI 0 "register_operand" "=d")
        (zero_extend:SI (truncate:QI
                         (match_operand:DI 1 "register_operand" "d"))))]
  "TARGET_64BIT && !TARGET_MIPS16"
  "andi\t%0,%1,0xff"
  [(set_attr "type"     "arith")
   (set_attr "mode"     "SI")])

(define_insn ""
  [(set (match_operand:HI 0 "register_operand" "=d")
        (zero_extend:HI (truncate:QI
                         (match_operand:DI 1 "register_operand" "d"))))]
  "TARGET_64BIT && !TARGET_MIPS16"
  "andi\t%0,%1,0xff"
  [(set_attr "type"     "arith")
   (set_attr "mode"     "HI")])
\f
;;
;;  ....................
;;
;;      ZERO EXTENSION
;;
;;  ....................

;; Extension insns.

(define_insn_and_split "zero_extendsidi2"
  [(set (match_operand:DI 0 "register_operand" "=d,d")
        (zero_extend:DI (match_operand:SI 1 "nonimmediate_operand" "d,W")))]
  "TARGET_64BIT"
  "@
   #
   lwu\t%0,%1"
  "&& reload_completed && REG_P (operands[1])"
  [(set (match_dup 0)
        (ashift:DI (match_dup 1) (const_int 32)))
   (set (match_dup 0)
        (lshiftrt:DI (match_dup 0) (const_int 32)))]
  { operands[1] = gen_lowpart (DImode, operands[1]); }
  [(set_attr "type" "multi,load")
   (set_attr "mode" "DI")
   (set_attr "length" "8,*")])

;; Combine is not allowed to convert this insn into a zero_extendsidi2
;; because of TRULY_NOOP_TRUNCATION.

(define_insn_and_split "*clear_upper32"
  [(set (match_operand:DI 0 "register_operand" "=d,d")
        (and:DI (match_operand:DI 1 "nonimmediate_operand" "d,o")
                (const_int 4294967295)))]
  "TARGET_64BIT"
{
  if (which_alternative == 0)
    return "#";

  operands[1] = gen_lowpart (SImode, operands[1]);
  return "lwu\t%0,%1";
}
  "&& reload_completed && REG_P (operands[1])"
  [(set (match_dup 0)
        (ashift:DI (match_dup 1) (const_int 32)))
   (set (match_dup 0)
        (lshiftrt:DI (match_dup 0) (const_int 32)))]
  ""
  [(set_attr "type" "multi,load")
   (set_attr "mode" "DI")
   (set_attr "length" "8,*")])

(define_expand "zero_extend<SHORT:mode><GPR:mode>2"
  [(set (match_operand:GPR 0 "register_operand")
        (zero_extend:GPR (match_operand:SHORT 1 "nonimmediate_operand")))]
  ""
{
  if (TARGET_MIPS16 && !GENERATE_MIPS16E
      && !memory_operand (operands[1], <SHORT:MODE>mode))
    {
      emit_insn (gen_and<GPR:mode>3 (operands[0],
                                     gen_lowpart (<GPR:MODE>mode, operands[1]),
                                     force_reg (<GPR:MODE>mode,
                                                GEN_INT (<SHORT:mask>))));
      DONE;
    }
})

(define_insn "*zero_extend<SHORT:mode><GPR:mode>2"
  [(set (match_operand:GPR 0 "register_operand" "=d,d")
        (zero_extend:GPR
             (match_operand:SHORT 1 "nonimmediate_operand" "d,m")))]
  "!TARGET_MIPS16"
  "@
   andi\t%0,%1,<SHORT:mask>
   l<SHORT:size>u\t%0,%1"
  [(set_attr "type" "arith,load")
   (set_attr "mode" "<GPR:MODE>")])

(define_insn "*zero_extend<SHORT:mode><GPR:mode>2_mips16e"
  [(set (match_operand:GPR 0 "register_operand" "=d")
        (zero_extend:GPR (match_operand:SHORT 1 "register_operand" "0")))]
  "GENERATE_MIPS16E"
  "ze<SHORT:size>\t%0"
  [(set_attr "type" "arith")
   (set_attr "mode" "<GPR:MODE>")])

(define_insn "*zero_extend<SHORT:mode><GPR:mode>2_mips16"
  [(set (match_operand:GPR 0 "register_operand" "=d")
        (zero_extend:GPR (match_operand:SHORT 1 "memory_operand" "m")))]
  "TARGET_MIPS16"
  "l<SHORT:size>u\t%0,%1"
  [(set_attr "type" "load")
   (set_attr "mode" "<GPR:MODE>")])

(define_expand "zero_extendqihi2"
  [(set (match_operand:HI 0 "register_operand")
        (zero_extend:HI (match_operand:QI 1 "nonimmediate_operand")))]
  ""
{
  if (TARGET_MIPS16 && !memory_operand (operands[1], QImode))
    {
      emit_insn (gen_zero_extendqisi2 (gen_lowpart (SImode, operands[0]),
                                       operands[1]));
      DONE;
    }
})

(define_insn "*zero_extendqihi2"
  [(set (match_operand:HI 0 "register_operand" "=d,d")
        (zero_extend:HI (match_operand:QI 1 "nonimmediate_operand" "d,m")))]
  "!TARGET_MIPS16"
  "@
   andi\t%0,%1,0x00ff
   lbu\t%0,%1"
  [(set_attr "type" "arith,load")
   (set_attr "mode" "HI")])

(define_insn "*zero_extendqihi2_mips16"
  [(set (match_operand:HI 0 "register_operand" "=d")
        (zero_extend:HI (match_operand:QI 1 "memory_operand" "m")))]
  "TARGET_MIPS16"
  "lbu\t%0,%1"
  [(set_attr "type" "load")
   (set_attr "mode" "HI")])
\f
;;
;;  ....................
;;
;;      SIGN EXTENSION
;;
;;  ....................

;; Extension insns.
;; Those for integer source operand are ordered widest source type first.

;; When TARGET_64BIT, all SImode integer registers should already be in
;; sign-extended form (see TRULY_NOOP_TRUNCATION and truncdisi2).  We can
;; therefore get rid of register->register instructions if we constrain
;; the source to be in the same register as the destination.
;;
;; The register alternative has type "arith" so that the pre-reload
;; scheduler will treat it as a move.  This reflects what happens if
;; the register alternative needs a reload.
(define_insn_and_split "extendsidi2"
  [(set (match_operand:DI 0 "register_operand" "=d,d")
        (sign_extend:DI (match_operand:SI 1 "nonimmediate_operand" "0,m")))]
  "TARGET_64BIT"
  "@
   #
   lw\t%0,%1"
  "&& reload_completed && register_operand (operands[1], VOIDmode)"
  [(const_int 0)]
{
  emit_note (NOTE_INSN_DELETED);
  DONE;
}
  [(set_attr "type" "arith,load")
   (set_attr "mode" "DI")])

(define_expand "extend<SHORT:mode><GPR:mode>2"
  [(set (match_operand:GPR 0 "register_operand")
        (sign_extend:GPR (match_operand:SHORT 1 "nonimmediate_operand")))]
  "")

(define_insn "*extend<SHORT:mode><GPR:mode>2_mips16e"
  [(set (match_operand:GPR 0 "register_operand" "=d,d")
        (sign_extend:GPR (match_operand:SHORT 1 "nonimmediate_operand" "0,m")))]
  "GENERATE_MIPS16E"
  "@
   se<SHORT:size>\t%0
   l<SHORT:size>\t%0,%1"
  [(set_attr "type" "arith,load")
   (set_attr "mode" "<GPR:MODE>")])

(define_insn_and_split "*extend<SHORT:mode><GPR:mode>2"
  [(set (match_operand:GPR 0 "register_operand" "=d,d")
        (sign_extend:GPR
             (match_operand:SHORT 1 "nonimmediate_operand" "d,m")))]
  "!ISA_HAS_SEB_SEH && !GENERATE_MIPS16E"
  "@
   #
   l<SHORT:size>\t%0,%1"
  "&& reload_completed && REG_P (operands[1])"
  [(set (match_dup 0) (ashift:GPR (match_dup 1) (match_dup 2)))
   (set (match_dup 0) (ashiftrt:GPR (match_dup 0) (match_dup 2)))]
{
  operands[1] = gen_lowpart (<GPR:MODE>mode, operands[1]);
  operands[2] = GEN_INT (GET_MODE_BITSIZE (<GPR:MODE>mode)
                         - GET_MODE_BITSIZE (<SHORT:MODE>mode));
}
  [(set_attr "type" "arith,load")
   (set_attr "mode" "<GPR:MODE>")
   (set_attr "length" "8,*")])

(define_insn "*extend<SHORT:mode><GPR:mode>2_se<SHORT:size>"
  [(set (match_operand:GPR 0 "register_operand" "=d,d")
        (sign_extend:GPR
             (match_operand:SHORT 1 "nonimmediate_operand" "d,m")))]
  "ISA_HAS_SEB_SEH"
  "@
   se<SHORT:size>\t%0,%1
   l<SHORT:size>\t%0,%1"
  [(set_attr "type" "arith,load")
   (set_attr "mode" "<GPR:MODE>")])

;; This pattern generates the same code as extendqisi2; split it into
;; that form after reload.
(define_insn_and_split "extendqihi2"
  [(set (match_operand:HI 0 "register_operand" "=d,d")
        (sign_extend:HI (match_operand:QI 1 "nonimmediate_operand" "d,m")))]
  ""
  "#"
  "reload_completed"
  [(set (match_dup 0) (sign_extend:SI (match_dup 1)))]
  { operands[0] = gen_lowpart (SImode, operands[0]); }
  [(set_attr "type" "arith,load")
   (set_attr "mode" "SI")
   (set_attr "length" "8,*")])

(define_insn "extendsfdf2"
  [(set (match_operand:DF 0 "register_operand" "=f")
        (float_extend:DF (match_operand:SF 1 "register_operand" "f")))]
  "TARGET_HARD_FLOAT && TARGET_DOUBLE_FLOAT"
  "cvt.d.s\t%0,%1"
  [(set_attr "type"     "fcvt")
   (set_attr "cnv_mode" "S2D")   
   (set_attr "mode"     "DF")])
\f
;;
;;  ....................
;;
;;      CONVERSIONS
;;
;;  ....................

(define_expand "fix_truncdfsi2"
  [(set (match_operand:SI 0 "register_operand")
        (fix:SI (match_operand:DF 1 "register_operand")))]
  "TARGET_HARD_FLOAT && TARGET_DOUBLE_FLOAT"
{
  if (!ISA_HAS_TRUNC_W)
    {
      emit_insn (gen_fix_truncdfsi2_macro (operands[0], operands[1]));
      DONE;
    }
})

(define_insn "fix_truncdfsi2_insn"
  [(set (match_operand:SI 0 "register_operand" "=f")
        (fix:SI (match_operand:DF 1 "register_operand" "f")))]
  "TARGET_HARD_FLOAT && TARGET_DOUBLE_FLOAT && ISA_HAS_TRUNC_W"
  "trunc.w.d %0,%1"
  [(set_attr "type"     "fcvt")
   (set_attr "mode"     "DF")
   (set_attr "cnv_mode" "D2I")
   (set_attr "length"   "4")])

(define_insn "fix_truncdfsi2_macro"
  [(set (match_operand:SI 0 "register_operand" "=f")
        (fix:SI (match_operand:DF 1 "register_operand" "f")))
   (clobber (match_scratch:DF 2 "=d"))]
  "TARGET_HARD_FLOAT && TARGET_DOUBLE_FLOAT && !ISA_HAS_TRUNC_W"
{
  if (set_nomacro)
    return ".set\tmacro\;trunc.w.d %0,%1,%2\;.set\tnomacro";
  else
    return "trunc.w.d %0,%1,%2";
}
  [(set_attr "type"     "fcvt")
   (set_attr "mode"     "DF")
   (set_attr "cnv_mode" "D2I")
   (set_attr "length"   "36")])

(define_expand "fix_truncsfsi2"
  [(set (match_operand:SI 0 "register_operand")
        (fix:SI (match_operand:SF 1 "register_operand")))]
  "TARGET_HARD_FLOAT"
{
  if (!ISA_HAS_TRUNC_W)
    {
      emit_insn (gen_fix_truncsfsi2_macro (operands[0], operands[1]));
      DONE;
    }
})

(define_insn "fix_truncsfsi2_insn"
  [(set (match_operand:SI 0 "register_operand" "=f")
        (fix:SI (match_operand:SF 1 "register_operand" "f")))]
  "TARGET_HARD_FLOAT && ISA_HAS_TRUNC_W"
  "trunc.w.s %0,%1"
  [(set_attr "type"     "fcvt")
   (set_attr "mode"     "SF")
   (set_attr "cnv_mode" "S2I")
   (set_attr "length"   "4")])

(define_insn "fix_truncsfsi2_macro"
  [(set (match_operand:SI 0 "register_operand" "=f")
        (fix:SI (match_operand:SF 1 "register_operand" "f")))
   (clobber (match_scratch:SF 2 "=d"))]
  "TARGET_HARD_FLOAT && !ISA_HAS_TRUNC_W"
{
  if (set_nomacro)
    return ".set\tmacro\;trunc.w.s %0,%1,%2\;.set\tnomacro";
  else
    return "trunc.w.s %0,%1,%2";
}
  [(set_attr "type"     "fcvt")
   (set_attr "mode"     "SF")
   (set_attr "cnv_mode" "S2I")
   (set_attr "length"   "36")])


(define_insn "fix_truncdfdi2"
  [(set (match_operand:DI 0 "register_operand" "=f")
        (fix:DI (match_operand:DF 1 "register_operand" "f")))]
  "TARGET_HARD_FLOAT && TARGET_FLOAT64 && TARGET_DOUBLE_FLOAT"
  "trunc.l.d %0,%1"
  [(set_attr "type"     "fcvt")
   (set_attr "mode"     "DF")
   (set_attr "cnv_mode" "D2I")
   (set_attr "length"   "4")])


(define_insn "fix_truncsfdi2"
  [(set (match_operand:DI 0 "register_operand" "=f")
        (fix:DI (match_operand:SF 1 "register_operand" "f")))]
  "TARGET_HARD_FLOAT && TARGET_FLOAT64 && TARGET_DOUBLE_FLOAT"
  "trunc.l.s %0,%1"
  [(set_attr "type"     "fcvt")
   (set_attr "mode"     "SF")
   (set_attr "cnv_mode" "S2I")
   (set_attr "length"   "4")])


(define_insn "floatsidf2"
  [(set (match_operand:DF 0 "register_operand" "=f")
        (float:DF (match_operand:SI 1 "register_operand" "f")))]
  "TARGET_HARD_FLOAT && TARGET_DOUBLE_FLOAT"
  "cvt.d.w\t%0,%1"
  [(set_attr "type"     "fcvt")
   (set_attr "mode"     "DF")
   (set_attr "cnv_mode" "I2D")   
   (set_attr "length"   "4")])


(define_insn "floatdidf2"
  [(set (match_operand:DF 0 "register_operand" "=f")
        (float:DF (match_operand:DI 1 "register_operand" "f")))]
  "TARGET_HARD_FLOAT && TARGET_FLOAT64 && TARGET_DOUBLE_FLOAT"
  "cvt.d.l\t%0,%1"
  [(set_attr "type"     "fcvt")
   (set_attr "mode"     "DF")
   (set_attr "cnv_mode" "I2D")   
   (set_attr "length"   "4")])


(define_insn "floatsisf2"
  [(set (match_operand:SF 0 "register_operand" "=f")
        (float:SF (match_operand:SI 1 "register_operand" "f")))]
  "TARGET_HARD_FLOAT"
  "cvt.s.w\t%0,%1"
  [(set_attr "type"     "fcvt")
   (set_attr "mode"     "SF")
   (set_attr "cnv_mode" "I2S")   
   (set_attr "length"   "4")])


(define_insn "floatdisf2"
  [(set (match_operand:SF 0 "register_operand" "=f")
        (float:SF (match_operand:DI 1 "register_operand" "f")))]
  "TARGET_HARD_FLOAT && TARGET_FLOAT64 && TARGET_DOUBLE_FLOAT"
  "cvt.s.l\t%0,%1"
  [(set_attr "type"     "fcvt")
   (set_attr "mode"     "SF")
   (set_attr "cnv_mode" "I2S")   
   (set_attr "length"   "4")])


(define_expand "fixuns_truncdfsi2"
  [(set (match_operand:SI 0 "register_operand")
        (unsigned_fix:SI (match_operand:DF 1 "register_operand")))]
  "TARGET_HARD_FLOAT && TARGET_DOUBLE_FLOAT"
{
  rtx reg1 = gen_reg_rtx (DFmode);
  rtx reg2 = gen_reg_rtx (DFmode);
  rtx reg3 = gen_reg_rtx (SImode);
  rtx label1 = gen_label_rtx ();
  rtx label2 = gen_label_rtx ();
  REAL_VALUE_TYPE offset;

  real_2expN (&offset, 31);

  if (reg1)                     /* Turn off complaints about unreached code.  */
    {
      emit_move_insn (reg1, CONST_DOUBLE_FROM_REAL_VALUE (offset, DFmode));
      do_pending_stack_adjust ();

      emit_insn (gen_cmpdf (operands[1], reg1));
      emit_jump_insn (gen_bge (label1));

      emit_insn (gen_fix_truncdfsi2 (operands[0], operands[1]));
      emit_jump_insn (gen_rtx_SET (VOIDmode, pc_rtx,
                                   gen_rtx_LABEL_REF (VOIDmode, label2)));
      emit_barrier ();

      emit_label (label1);
      emit_move_insn (reg2, gen_rtx_MINUS (DFmode, operands[1], reg1));
      emit_move_insn (reg3, GEN_INT (trunc_int_for_mode
                                     (BITMASK_HIGH, SImode)));

      emit_insn (gen_fix_truncdfsi2 (operands[0], reg2));
      emit_insn (gen_iorsi3 (operands[0], operands[0], reg3));

      emit_label (label2);

      /* Allow REG_NOTES to be set on last insn (labels don't have enough
         fields, and can't be used for REG_NOTES anyway).  */
      emit_insn (gen_rtx_USE (VOIDmode, stack_pointer_rtx));
      DONE;
    }
})


(define_expand "fixuns_truncdfdi2"
  [(set (match_operand:DI 0 "register_operand")
        (unsigned_fix:DI (match_operand:DF 1 "register_operand")))]
  "TARGET_HARD_FLOAT && TARGET_64BIT && TARGET_DOUBLE_FLOAT"
{
  rtx reg1 = gen_reg_rtx (DFmode);
  rtx reg2 = gen_reg_rtx (DFmode);
  rtx reg3 = gen_reg_rtx (DImode);
  rtx label1 = gen_label_rtx ();
  rtx label2 = gen_label_rtx ();
  REAL_VALUE_TYPE offset;

  real_2expN (&offset, 63);

  emit_move_insn (reg1, CONST_DOUBLE_FROM_REAL_VALUE (offset, DFmode));
  do_pending_stack_adjust ();

  emit_insn (gen_cmpdf (operands[1], reg1));
  emit_jump_insn (gen_bge (label1));

  emit_insn (gen_fix_truncdfdi2 (operands[0], operands[1]));
  emit_jump_insn (gen_rtx_SET (VOIDmode, pc_rtx,
                               gen_rtx_LABEL_REF (VOIDmode, label2)));
  emit_barrier ();

  emit_label (label1);
  emit_move_insn (reg2, gen_rtx_MINUS (DFmode, operands[1], reg1));
  emit_move_insn (reg3, GEN_INT (BITMASK_HIGH));
  emit_insn (gen_ashldi3 (reg3, reg3, GEN_INT (32)));

  emit_insn (gen_fix_truncdfdi2 (operands[0], reg2));
  emit_insn (gen_iordi3 (operands[0], operands[0], reg3));

  emit_label (label2);

  /* Allow REG_NOTES to be set on last insn (labels don't have enough
     fields, and can't be used for REG_NOTES anyway).  */
  emit_insn (gen_rtx_USE (VOIDmode, stack_pointer_rtx));
  DONE;
})


(define_expand "fixuns_truncsfsi2"
  [(set (match_operand:SI 0 "register_operand")
        (unsigned_fix:SI (match_operand:SF 1 "register_operand")))]
  "TARGET_HARD_FLOAT"
{
  rtx reg1 = gen_reg_rtx (SFmode);
  rtx reg2 = gen_reg_rtx (SFmode);
  rtx reg3 = gen_reg_rtx (SImode);
  rtx label1 = gen_label_rtx ();
  rtx label2 = gen_label_rtx ();
  REAL_VALUE_TYPE offset;

  real_2expN (&offset, 31);

  emit_move_insn (reg1, CONST_DOUBLE_FROM_REAL_VALUE (offset, SFmode));
  do_pending_stack_adjust ();

  emit_insn (gen_cmpsf (operands[1], reg1));
  emit_jump_insn (gen_bge (label1));

  emit_insn (gen_fix_truncsfsi2 (operands[0], operands[1]));
  emit_jump_insn (gen_rtx_SET (VOIDmode, pc_rtx,
                               gen_rtx_LABEL_REF (VOIDmode, label2)));
  emit_barrier ();

  emit_label (label1);
  emit_move_insn (reg2, gen_rtx_MINUS (SFmode, operands[1], reg1));
  emit_move_insn (reg3, GEN_INT (trunc_int_for_mode
                                 (BITMASK_HIGH, SImode)));

  emit_insn (gen_fix_truncsfsi2 (operands[0], reg2));
  emit_insn (gen_iorsi3 (operands[0], operands[0], reg3));

  emit_label (label2);

  /* Allow REG_NOTES to be set on last insn (labels don't have enough
     fields, and can't be used for REG_NOTES anyway).  */
  emit_insn (gen_rtx_USE (VOIDmode, stack_pointer_rtx));
  DONE;
})


(define_expand "fixuns_truncsfdi2"
  [(set (match_operand:DI 0 "register_operand")
        (unsigned_fix:DI (match_operand:SF 1 "register_operand")))]
  "TARGET_HARD_FLOAT && TARGET_64BIT && TARGET_DOUBLE_FLOAT"
{
  rtx reg1 = gen_reg_rtx (SFmode);
  rtx reg2 = gen_reg_rtx (SFmode);
  rtx reg3 = gen_reg_rtx (DImode);
  rtx label1 = gen_label_rtx ();
  rtx label2 = gen_label_rtx ();
  REAL_VALUE_TYPE offset;

  real_2expN (&offset, 63);

  emit_move_insn (reg1, CONST_DOUBLE_FROM_REAL_VALUE (offset, SFmode));
  do_pending_stack_adjust ();

  emit_insn (gen_cmpsf (operands[1], reg1));
  emit_jump_insn (gen_bge (label1));

  emit_insn (gen_fix_truncsfdi2 (operands[0], operands[1]));
  emit_jump_insn (gen_rtx_SET (VOIDmode, pc_rtx,
                               gen_rtx_LABEL_REF (VOIDmode, label2)));
  emit_barrier ();

  emit_label (label1);
  emit_move_insn (reg2, gen_rtx_MINUS (SFmode, operands[1], reg1));
  emit_move_insn (reg3, GEN_INT (BITMASK_HIGH));
  emit_insn (gen_ashldi3 (reg3, reg3, GEN_INT (32)));

  emit_insn (gen_fix_truncsfdi2 (operands[0], reg2));
  emit_insn (gen_iordi3 (operands[0], operands[0], reg3));

  emit_label (label2);

  /* Allow REG_NOTES to be set on last insn (labels don't have enough
     fields, and can't be used for REG_NOTES anyway).  */
  emit_insn (gen_rtx_USE (VOIDmode, stack_pointer_rtx));
  DONE;
})
\f
;;
;;  ....................
;;
;;      DATA MOVEMENT
;;
;;  ....................

;; Bit field extract patterns which use lwl/lwr or ldl/ldr.

(define_expand "extv"
  [(set (match_operand 0 "register_operand")
        (sign_extract (match_operand:QI 1 "memory_operand")
                      (match_operand 2 "immediate_operand")
                      (match_operand 3 "immediate_operand")))]
  "!TARGET_MIPS16"
{
  if (mips_expand_unaligned_load (operands[0], operands[1],
                                  INTVAL (operands[2]),
                                  INTVAL (operands[3])))
    DONE;
  else
    FAIL;
})

(define_expand "extzv"
  [(set (match_operand 0 "register_operand")
        (zero_extract (match_operand 1 "nonimmediate_operand")
                      (match_operand 2 "immediate_operand")
                      (match_operand 3 "immediate_operand")))]
  "!TARGET_MIPS16"
{
  if (mips_expand_unaligned_load (operands[0], operands[1],
                                  INTVAL (operands[2]),
                                  INTVAL (operands[3])))
    DONE;
  else if (mips_use_ins_ext_p (operands[1], operands[2], operands[3]))
    {
      if (GET_MODE (operands[0]) == DImode)
        emit_insn (gen_extzvdi (operands[0], operands[1], operands[2],
                                operands[3]));
      else
        emit_insn (gen_extzvsi (operands[0], operands[1], operands[2],
                                operands[3]));
      DONE;
    }
  else
    FAIL;
})

(define_insn "extzv<mode>"
  [(set (match_operand:GPR 0 "register_operand" "=d")
        (zero_extract:GPR (match_operand:GPR 1 "register_operand" "d")
                          (match_operand:SI 2 "immediate_operand" "I")
                          (match_operand:SI 3 "immediate_operand" "I")))]
  "mips_use_ins_ext_p (operands[1], operands[2], operands[3])"
  "<d>ext\t%0,%1,%3,%2"
  [(set_attr "type"     "arith")
   (set_attr "mode"     "<MODE>")])


(define_expand "insv"
  [(set (zero_extract (match_operand 0 "nonimmediate_operand")
                      (match_operand 1 "immediate_operand")
                      (match_operand 2 "immediate_operand"))
        (match_operand 3 "reg_or_0_operand"))]
  "!TARGET_MIPS16"
{
  if (mips_expand_unaligned_store (operands[0], operands[3],
                                   INTVAL (operands[1]),
                                   INTVAL (operands[2])))
    DONE;
  else if (mips_use_ins_ext_p (operands[0], operands[1], operands[2]))
    {
      if (GET_MODE (operands[0]) == DImode)
        emit_insn (gen_insvdi (operands[0], operands[1], operands[2],
                               operands[3]));
      else
        emit_insn (gen_insvsi (operands[0], operands[1], operands[2],
                               operands[3]));
      DONE;
   }
   else
     FAIL;
})

(define_insn "insv<mode>"
  [(set (zero_extract:GPR (match_operand:GPR 0 "register_operand" "+d")
                          (match_operand:SI 1 "immediate_operand" "I")
                          (match_operand:SI 2 "immediate_operand" "I"))
        (match_operand:GPR 3 "reg_or_0_operand" "dJ"))]
  "mips_use_ins_ext_p (operands[0], operands[1], operands[2])"
  "<d>ins\t%0,%z3,%2,%1"
  [(set_attr "type"     "arith")
   (set_attr "mode"     "<MODE>")])

;; Unaligned word moves generated by the bit field patterns.
;;
;; As far as the rtl is concerned, both the left-part and right-part
;; instructions can access the whole field.  However, the real operand
;; refers to just the first or the last byte (depending on endianness).
;; We therefore use two memory operands to each instruction, one to
;; describe the rtl effect and one to use in the assembly output.
;;
;; Operands 0 and 1 are the rtl-level target and source respectively.
;; This allows us to use the standard length calculations for the "load"
;; and "store" type attributes.

(define_insn "mov_<load>l"
  [(set (match_operand:GPR 0 "register_operand" "=d")
        (unspec:GPR [(match_operand:BLK 1 "memory_operand" "m")
                     (match_operand:QI 2 "memory_operand" "m")]
                    UNSPEC_LOAD_LEFT))]
  "!TARGET_MIPS16"
  "<load>l\t%0,%2"
  [(set_attr "type" "load")
   (set_attr "mode" "<MODE>")])

(define_insn "mov_<load>r"
  [(set (match_operand:GPR 0 "register_operand" "=d")
        (unspec:GPR [(match_operand:BLK 1 "memory_operand" "m")
                     (match_operand:QI 2 "memory_operand" "m")
                     (match_operand:GPR 3 "register_operand" "0")]
                    UNSPEC_LOAD_RIGHT))]
  "!TARGET_MIPS16"
  "<load>r\t%0,%2"
  [(set_attr "type" "load")
   (set_attr "mode" "<MODE>")])

(define_insn "mov_<store>l"
  [(set (match_operand:BLK 0 "memory_operand" "=m")
        (unspec:BLK [(match_operand:GPR 1 "reg_or_0_operand" "dJ")
                     (match_operand:QI 2 "memory_operand" "m")]
                    UNSPEC_STORE_LEFT))]
  "!TARGET_MIPS16"
  "<store>l\t%z1,%2"
  [(set_attr "type" "store")
   (set_attr "mode" "<MODE>")])

(define_insn "mov_<store>r"
  [(set (match_operand:BLK 0 "memory_operand" "+m")
        (unspec:BLK [(match_operand:GPR 1 "reg_or_0_operand" "dJ")
                     (match_operand:QI 2 "memory_operand" "m")
                     (match_dup 0)]
                    UNSPEC_STORE_RIGHT))]
  "!TARGET_MIPS16"
  "<store>r\t%z1,%2"
  [(set_attr "type" "store")
   (set_attr "mode" "<MODE>")])

;; An instruction to calculate the high part of a 64-bit SYMBOL_GENERAL.
;; The required value is:
;;
;;      (%highest(op1) << 48) + (%higher(op1) << 32) + (%hi(op1) << 16)
;;
;; which translates to:
;;
;;      lui     op0,%highest(op1)
;;      daddiu  op0,op0,%higher(op1)
;;      dsll    op0,op0,16
;;      daddiu  op0,op0,%hi(op1)
;;      dsll    op0,op0,16
;;
;; The split is deferred until after flow2 to allow the peephole2 below
;; to take effect.
(define_insn_and_split "*lea_high64"
  [(set (match_operand:DI 0 "register_operand" "=d")
        (high:DI (match_operand:DI 1 "general_symbolic_operand" "")))]
  "TARGET_EXPLICIT_RELOCS && ABI_HAS_64BIT_SYMBOLS"
  "#"
  "&& flow2_completed"
  [(set (match_dup 0) (high:DI (match_dup 2)))
   (set (match_dup 0) (lo_sum:DI (match_dup 0) (match_dup 2)))
   (set (match_dup 0) (ashift:DI (match_dup 0) (const_int 16)))
   (set (match_dup 0) (lo_sum:DI (match_dup 0) (match_dup 3)))
   (set (match_dup 0) (ashift:DI (match_dup 0) (const_int 16)))]
{
  operands[2] = mips_unspec_address (operands[1], SYMBOL_64_HIGH);
  operands[3] = mips_unspec_address (operands[1], SYMBOL_64_MID);
}
  [(set_attr "length" "20")])

;; Use a scratch register to reduce the latency of the above pattern
;; on superscalar machines.  The optimized sequence is:
;;
;;      lui     op1,%highest(op2)
;;      lui     op0,%hi(op2)
;;      daddiu  op1,op1,%higher(op2)
;;      dsll32  op1,op1,0
;;      daddu   op1,op1,op0
(define_peephole2
  [(set (match_operand:DI 1 "register_operand")
        (high:DI (match_operand:DI 2 "general_symbolic_operand")))
   (match_scratch:DI 0 "d")]
  "TARGET_EXPLICIT_RELOCS && ABI_HAS_64BIT_SYMBOLS"
  [(set (match_dup 1) (high:DI (match_dup 3)))
   (set (match_dup 0) (high:DI (match_dup 4)))
   (set (match_dup 1) (lo_sum:DI (match_dup 1) (match_dup 3)))
   (set (match_dup 1) (ashift:DI (match_dup 1) (const_int 32)))
   (set (match_dup 1) (plus:DI (match_dup 1) (match_dup 0)))]
{
  operands[3] = mips_unspec_address (operands[2], SYMBOL_64_HIGH);
  operands[4] = mips_unspec_address (operands[2], SYMBOL_64_LOW);
})

;; On most targets, the expansion of (lo_sum (high X) X) for a 64-bit
;; SYMBOL_GENERAL X will take 6 cycles.  This next pattern allows combine
;; to merge the HIGH and LO_SUM parts of a move if the HIGH part is only
;; used once.  We can then use the sequence:
;;
;;      lui     op0,%highest(op1)
;;      lui     op2,%hi(op1)
;;      daddiu  op0,op0,%higher(op1)
;;      daddiu  op2,op2,%lo(op1)
;;      dsll32  op0,op0,0
;;      daddu   op0,op0,op2
;;
;; which takes 4 cycles on most superscalar targets.
(define_insn_and_split "*lea64"
  [(set (match_operand:DI 0 "register_operand" "=d")
        (match_operand:DI 1 "general_symbolic_operand" ""))
   (clobber (match_scratch:DI 2 "=&d"))]
  "TARGET_EXPLICIT_RELOCS && ABI_HAS_64BIT_SYMBOLS && cse_not_expected"
  "#"
  "&& reload_completed"
  [(set (match_dup 0) (high:DI (match_dup 3)))
   (set (match_dup 2) (high:DI (match_dup 4)))
   (set (match_dup 0) (lo_sum:DI (match_dup 0) (match_dup 3)))
   (set (match_dup 2) (lo_sum:DI (match_dup 2) (match_dup 4)))
   (set (match_dup 0) (ashift:DI (match_dup 0) (const_int 32)))
   (set (match_dup 0) (plus:DI (match_dup 0) (match_dup 2)))]
{
  operands[3] = mips_unspec_address (operands[1], SYMBOL_64_HIGH);
  operands[4] = mips_unspec_address (operands[1], SYMBOL_64_LOW);
}
  [(set_attr "length" "24")])

;; Insns to fetch a global symbol from a big GOT.

(define_insn_and_split "*xgot_hi<mode>"
  [(set (match_operand:P 0 "register_operand" "=d")
        (high:P (match_operand:P 1 "global_got_operand" "")))]
  "TARGET_EXPLICIT_RELOCS && TARGET_XGOT"
  "#"
  "&& reload_completed"
  [(set (match_dup 0) (high:P (match_dup 2)))
   (set (match_dup 0) (plus:P (match_dup 0) (match_dup 3)))]
{
  operands[2] = mips_unspec_address (operands[1], SYMBOL_GOTOFF_GLOBAL);
  operands[3] = pic_offset_table_rtx;
}
  [(set_attr "got" "xgot_high")
   (set_attr "mode" "<MODE>")])

(define_insn_and_split "*xgot_lo<mode>"
  [(set (match_operand:P 0 "register_operand" "=d")
        (lo_sum:P (match_operand:P 1 "register_operand" "d")
                  (match_operand:P 2 "global_got_operand" "")))]
  "TARGET_EXPLICIT_RELOCS && TARGET_XGOT"
  "#"
  "&& reload_completed"
  [(set (match_dup 0)
        (unspec:P [(match_dup 1) (match_dup 3)] UNSPEC_LOAD_GOT))]
  { operands[3] = mips_unspec_address (operands[2], SYMBOL_GOTOFF_GLOBAL); }
  [(set_attr "got" "load")
   (set_attr "mode" "<MODE>")])

;; Insns to fetch a global symbol from a normal GOT.

(define_insn_and_split "*got_disp<mode>"
  [(set (match_operand:P 0 "register_operand" "=d")
        (match_operand:P 1 "global_got_operand" ""))]
  "TARGET_EXPLICIT_RELOCS && !TARGET_XGOT"
  "#"
  "&& reload_completed"
  [(set (match_dup 0)
        (unspec:P [(match_dup 2) (match_dup 3)] UNSPEC_LOAD_GOT))]
{
  operands[2] = pic_offset_table_rtx;
  operands[3] = mips_unspec_address (operands[1], SYMBOL_GOTOFF_GLOBAL);
}
  [(set_attr "got" "load")
   (set_attr "mode" "<MODE>")])

;; Insns for loading the high part of a local symbol.

(define_insn_and_split "*got_page<mode>"
  [(set (match_operand:P 0 "register_operand" "=d")
        (high:P (match_operand:P 1 "local_got_operand" "")))]
  "TARGET_EXPLICIT_RELOCS"
  "#"
  "&& reload_completed"
  [(set (match_dup 0)
        (unspec:P [(match_dup 2) (match_dup 3)] UNSPEC_LOAD_GOT))]
{
  operands[2] = pic_offset_table_rtx;
  operands[3] = mips_unspec_address (operands[1], SYMBOL_GOTOFF_PAGE);
}
  [(set_attr "got" "load")
   (set_attr "mode" "<MODE>")])

;; Lower-level instructions for loading an address from the GOT.
;; We could use MEMs, but an unspec gives more optimization
;; opportunities.

(define_insn "load_got<mode>"
  [(set (match_operand:P 0 "register_operand" "=d")
        (unspec:P [(match_operand:P 1 "register_operand" "d")
                   (match_operand:P 2 "immediate_operand" "")]
                  UNSPEC_LOAD_GOT))]
  ""
  "<load>\t%0,%R2(%1)"
  [(set_attr "type" "load")
   (set_attr "mode" "<MODE>")
   (set_attr "length" "4")])

;; Instructions for adding the low 16 bits of an address to a register.
;; Operand 2 is the address: print_operand works out which relocation
;; should be applied.

(define_insn "*low<mode>"
  [(set (match_operand:P 0 "register_operand" "=d")
        (lo_sum:P (match_operand:P 1 "register_operand" "d")
                  (match_operand:P 2 "immediate_operand" "")))]
  "!TARGET_MIPS16"
  "<d>addiu\t%0,%1,%R2"
  [(set_attr "type" "arith")
   (set_attr "mode" "<MODE>")])

(define_insn "*low<mode>_mips16"
  [(set (match_operand:P 0 "register_operand" "=d")
        (lo_sum:P (match_operand:P 1 "register_operand" "0")
                  (match_operand:P 2 "immediate_operand" "")))]
  "TARGET_MIPS16"
  "<d>addiu\t%0,%R2"
  [(set_attr "type" "arith")
   (set_attr "mode" "<MODE>")
   (set_attr "length" "8")])

;; 64-bit integer moves

;; Unlike most other insns, the move insns can't be split with
;; different predicates, because register spilling and other parts of
;; the compiler, have memoized the insn number already.

(define_expand "movdi"
  [(set (match_operand:DI 0 "")
        (match_operand:DI 1 ""))]
  ""
{
  if (mips_legitimize_move (DImode, operands[0], operands[1]))
    DONE;
})

;; For mips16, we need a special case to handle storing $31 into
;; memory, since we don't have a constraint to match $31.  This
;; instruction can be generated by save_restore_insns.

(define_insn "*mov<mode>_ra"
  [(set (match_operand:GPR 0 "stack_operand" "=m")
        (reg:GPR 31))]
  "TARGET_MIPS16"
  "<store>\t$31,%0"
  [(set_attr "type" "store")
   (set_attr "mode" "<MODE>")])

(define_insn "*movdi_32bit"
  [(set (match_operand:DI 0 "nonimmediate_operand" "=d,d,d,m,*a,*d,*B*C*D,*B*C*D,*d,*m")
        (match_operand:DI 1 "move_operand" "d,i,m,d,*J*d,*a,*d,*m,*B*C*D,*B*C*D"))]
  "!TARGET_64BIT && !TARGET_MIPS16
   && (register_operand (operands[0], DImode)
       || reg_or_0_operand (operands[1], DImode))"
  { return mips_output_move (operands[0], operands[1]); }
  [(set_attr "type"     "arith,arith,load,store,mthilo,mfhilo,xfer,load,xfer,store")
   (set_attr "mode"     "DI")
   (set_attr "length"   "8,16,*,*,8,8,8,*,8,*")])

(define_insn "*movdi_32bit_mips16"
  [(set (match_operand:DI 0 "nonimmediate_operand" "=d,y,d,d,d,d,m,*d")
        (match_operand:DI 1 "move_operand" "d,d,y,K,N,m,d,*x"))]
  "!TARGET_64BIT && TARGET_MIPS16
   && (register_operand (operands[0], DImode)
       || register_operand (operands[1], DImode))"
  { return mips_output_move (operands[0], operands[1]); }
  [(set_attr "type"     "arith,arith,arith,arith,arith,load,store,mfhilo")
   (set_attr "mode"     "DI")
   (set_attr "length"   "8,8,8,8,12,*,*,8")])

(define_insn "*movdi_64bit"
  [(set (match_operand:DI 0 "nonimmediate_operand" "=d,d,e,d,m,*f,*f,*f,*d,*m,*x,*B*C*D,*B*C*D,*d,*m")
        (match_operand:DI 1 "move_operand" "d,U,T,m,dJ,*f,*d*J,*m,*f,*f,*J*d,*d,*m,*B*C*D,*B*C*D"))]
  "TARGET_64BIT && !TARGET_MIPS16
   && (register_operand (operands[0], DImode)
       || reg_or_0_operand (operands[1], DImode))"
  { return mips_output_move (operands[0], operands[1]); }
  [(set_attr "type"     "arith,const,const,load,store,fmove,xfer,fpload,xfer,fpstore,mthilo,xfer,load,xfer,store")
   (set_attr "mode"     "DI")
   (set_attr "length"   "4,*,*,*,*,4,4,*,4,*,4,8,*,8,*")])

(define_insn "*movdi_64bit_mips16"
  [(set (match_operand:DI 0 "nonimmediate_operand" "=d,y,d,d,d,d,d,m")
        (match_operand:DI 1 "move_operand" "d,d,y,K,N,U,m,d"))]
  "TARGET_64BIT && TARGET_MIPS16
   && (register_operand (operands[0], DImode)
       || register_operand (operands[1], DImode))"
  { return mips_output_move (operands[0], operands[1]); }
  [(set_attr "type"     "arith,arith,arith,arith,arith,const,load,store")
   (set_attr "mode"     "DI")
   (set_attr_alternative "length"
                [(const_int 4)
                 (const_int 4)
                 (const_int 4)
                 (if_then_else (match_operand:VOID 1 "m16_uimm8_1")
                               (const_int 4)
                               (const_int 8))
                 (if_then_else (match_operand:VOID 1 "m16_nuimm8_1")
                               (const_int 8)
                               (const_int 12))
                 (const_string "*")
                 (const_string "*")
                 (const_string "*")])])


;; On the mips16, we can split ld $r,N($r) into an add and a load,
;; when the original load is a 4 byte instruction but the add and the
;; load are 2 2 byte instructions.

(define_split
  [(set (match_operand:DI 0 "register_operand")
        (mem:DI (plus:DI (match_dup 0)
                         (match_operand:DI 1 "const_int_operand"))))]
  "TARGET_64BIT && TARGET_MIPS16 && reload_completed
   && !TARGET_DEBUG_D_MODE
   && REG_P (operands[0])
   && M16_REG_P (REGNO (operands[0]))
   && GET_CODE (operands[1]) == CONST_INT
   && ((INTVAL (operands[1]) < 0
        && INTVAL (operands[1]) >= -0x10)
       || (INTVAL (operands[1]) >= 32 * 8
           && INTVAL (operands[1]) <= 31 * 8 + 0x8)
       || (INTVAL (operands[1]) >= 0
           && INTVAL (operands[1]) < 32 * 8
           && (INTVAL (operands[1]) & 7) != 0))"
  [(set (match_dup 0) (plus:DI (match_dup 0) (match_dup 1)))
   (set (match_dup 0) (mem:DI (plus:DI (match_dup 0) (match_dup 2))))]
{
  HOST_WIDE_INT val = INTVAL (operands[1]);

  if (val < 0)
    operands[2] = const0_rtx;
  else if (val >= 32 * 8)
    {
      int off = val & 7;

      operands[1] = GEN_INT (0x8 + off);
      operands[2] = GEN_INT (val - off - 0x8);
    }
  else
    {
      int off = val & 7;

      operands[1] = GEN_INT (off);
      operands[2] = GEN_INT (val - off);
    }
})

;; 32-bit Integer moves

;; Unlike most other insns, the move insns can't be split with
;; different predicates, because register spilling and other parts of
;; the compiler, have memoized the insn number already.

(define_expand "movsi"
  [(set (match_operand:SI 0 "")
        (match_operand:SI 1 ""))]
  ""
{
  if (mips_legitimize_move (SImode, operands[0], operands[1]))
    DONE;
})

;; The difference between these two is whether or not ints are allowed
;; in FP registers (off by default, use -mdebugh to enable).

(define_insn "*movsi_internal"
  [(set (match_operand:SI 0 "nonimmediate_operand" "=d,d,e,d,m,*f,*f,*f,*d,*m,*d,*z,*a,*d,*B*C*D,*B*C*D,*d,*m")
        (match_operand:SI 1 "move_operand" "d,U,T,m,dJ,*f,*d*J,*m,*f,*f,*z,*d,*J*d,*A,*d,*m,*B*C*D,*B*C*D"))]
  "!TARGET_MIPS16
   && (register_operand (operands[0], SImode)
       || reg_or_0_operand (operands[1], SImode))"
  { return mips_output_move (operands[0], operands[1]); }
  [(set_attr "type"     "arith,const,const,load,store,fmove,xfer,fpload,xfer,fpstore,xfer,xfer,mthilo,mfhilo,xfer,load,xfer,store")
   (set_attr "mode"     "SI")
   (set_attr "length"   "4,*,*,*,*,4,4,*,4,*,4,4,4,4,4,*,4,*")])

(define_insn "*movsi_mips16"
  [(set (match_operand:SI 0 "nonimmediate_operand" "=d,y,d,d,d,d,d,m")
        (match_operand:SI 1 "move_operand" "d,d,y,K,N,U,m,d"))]
  "TARGET_MIPS16
   && (register_operand (operands[0], SImode)
       || register_operand (operands[1], SImode))"
  { return mips_output_move (operands[0], operands[1]); }
  [(set_attr "type"     "arith,arith,arith,arith,arith,const,load,store")
   (set_attr "mode"     "SI")
   (set_attr_alternative "length"
                [(const_int 4)
                 (const_int 4)
                 (const_int 4)
                 (if_then_else (match_operand:VOID 1 "m16_uimm8_1")
                               (const_int 4)
                               (const_int 8))
                 (if_then_else (match_operand:VOID 1 "m16_nuimm8_1")
                               (const_int 8)
                               (const_int 12))
                 (const_string "*")
                 (const_string "*")
                 (const_string "*")])])

;; On the mips16, we can split lw $r,N($r) into an add and a load,
;; when the original load is a 4 byte instruction but the add and the
;; load are 2 2 byte instructions.

(define_split
  [(set (match_operand:SI 0 "register_operand")
        (mem:SI (plus:SI (match_dup 0)
                         (match_operand:SI 1 "const_int_operand"))))]
  "TARGET_MIPS16 && reload_completed && !TARGET_DEBUG_D_MODE
   && REG_P (operands[0])
   && M16_REG_P (REGNO (operands[0]))
   && GET_CODE (operands[1]) == CONST_INT
   && ((INTVAL (operands[1]) < 0
        && INTVAL (operands[1]) >= -0x80)
       || (INTVAL (operands[1]) >= 32 * 4
           && INTVAL (operands[1]) <= 31 * 4 + 0x7c)
       || (INTVAL (operands[1]) >= 0
           && INTVAL (operands[1]) < 32 * 4
           && (INTVAL (operands[1]) & 3) != 0))"
  [(set (match_dup 0) (plus:SI (match_dup 0) (match_dup 1)))
   (set (match_dup 0) (mem:SI (plus:SI (match_dup 0) (match_dup 2))))]
{
  HOST_WIDE_INT val = INTVAL (operands[1]);

  if (val < 0)