;; Mips.md Machine Description for MIPS based processors
;; Copyright (C) 1989-2022 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 3, 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 COPYING3. If not see
;; .
(define_enum "processor" [
r3000
4kc
4kp
5kc
5kf
20kc
24kc
24kf2_1
24kf1_1
74kc
74kf2_1
74kf1_1
74kf3_2
loongson_2e
loongson_2f
gs464
gs464e
gs264e
m4k
octeon
octeon2
octeon3
r3900
r6000
r4000
r4100
r4111
r4120
r4130
r4300
r4600
r4650
r4700
r5000
r5400
r5500
r5900
r7000
r8000
r9000
r10000
sb1
sb1a
sr71000
xlr
xlp
p5600
m5100
i6400
p6600
])
(define_c_enum "unspec" [
;; Unaligned accesses.
UNSPEC_LOAD_LEFT
UNSPEC_LOAD_RIGHT
UNSPEC_STORE_LEFT
UNSPEC_STORE_RIGHT
;; Integer operations that are too cumbersome to describe directly.
UNSPEC_WSBH
UNSPEC_DSBH
UNSPEC_DSHD
;; Floating-point moves.
UNSPEC_LOAD_LOW
UNSPEC_LOAD_HIGH
UNSPEC_STORE_WORD
UNSPEC_MFHC1
UNSPEC_MTHC1
;; Floating-point environment.
UNSPEC_GET_FCSR
UNSPEC_SET_FCSR
;; HI/LO moves.
UNSPEC_MFHI
UNSPEC_MTHI
UNSPEC_SET_HILO
;; GP manipulation.
UNSPEC_LOADGP
UNSPEC_COPYGP
UNSPEC_MOVE_GP
UNSPEC_POTENTIAL_CPRESTORE
UNSPEC_CPRESTORE
UNSPEC_RESTORE_GP
UNSPEC_EH_RETURN
UNSPEC_GP
UNSPEC_SET_GOT_VERSION
UNSPEC_UPDATE_GOT_VERSION
;; Symbolic accesses.
UNSPEC_LOAD_CALL
UNSPEC_LOAD_GOT
UNSPEC_TLS_LDM
UNSPEC_TLS_GET_TP
UNSPEC_UNSHIFTED_HIGH
;; MIPS16 constant pools.
UNSPEC_ALIGN
UNSPEC_CONSTTABLE
UNSPEC_CONSTTABLE_END
UNSPEC_CONSTTABLE_INT
UNSPEC_CONSTTABLE_FLOAT
;; Blockage and synchronisation.
UNSPEC_BLOCKAGE
UNSPEC_CLEAR_HAZARD
UNSPEC_RDHWR
UNSPEC_SYNCI
UNSPEC_SYNC
;; Cache manipulation.
UNSPEC_MIPS_CACHE
UNSPEC_R10K_CACHE_BARRIER
;; Interrupt handling.
UNSPEC_ERET
UNSPEC_DERET
UNSPEC_DI
UNSPEC_EHB
UNSPEC_RDPGPR
UNSPEC_COP0
;; Used in a call expression in place of args_size. It's present for PIC
;; indirect calls where it contains args_size and the function symbol.
UNSPEC_CALL_ATTR
;; MIPS16 casesi jump table dispatch.
UNSPEC_CASESI_DISPATCH
;; Stack checking.
UNSPEC_PROBE_STACK_RANGE
;; The `.insn' pseudo-op.
UNSPEC_INSN_PSEUDO
])
(define_constants
[(TLS_GET_TP_REGNUM 3)
(GET_FCSR_REGNUM 2)
(SET_FCSR_REGNUM 4)
(PIC_FUNCTION_ADDR_REGNUM 25)
(RETURN_ADDR_REGNUM 31)
(CPRESTORE_SLOT_REGNUM 76)
(GOT_VERSION_REGNUM 79)
;; PIC long branch sequences are never longer than 100 bytes.
(MAX_PIC_BRANCH_LENGTH 100)
]
)
(include "predicates.md")
(include "constraints.md")
;; ....................
;;
;; 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 for TARGET_CALL_CLOBBERED_GP because it includes
;; an instruction to restore $gp. Direct jals are also macros for
;; !TARGET_ABSOLUTE_JUMPS because they first load the target address
;; into a register.
(define_attr "jal_macro" "no,yes"
(cond [(eq_attr "jal" "direct")
(symbol_ref "(TARGET_CALL_CLOBBERED_GP || !TARGET_ABSOLUTE_JUMPS
? JAL_MACRO_YES : JAL_MACRO_NO)")
(eq_attr "jal" "indirect")
(symbol_ref "(TARGET_CALL_CLOBBERED_GP
? JAL_MACRO_YES : JAL_MACRO_NO)")]
(const_string "no")))
;; Classification of moves, extensions and truncations. Most values
;; are as for "type" (see below) but there are also the following
;; move-specific values:
;;
;; constN move an N-constraint integer into a MIPS16 register
;; sll0 "sll DEST,SRC,0", which on 64-bit targets is guaranteed
;; to produce a sign-extended DEST, even if SRC is not
;; properly sign-extended
;; ext_ins EXT, DEXT, INS or DINS instruction
;; andi a single ANDI instruction
;; loadpool move a constant into a MIPS16 register by loading it
;; from the pool
;; shift_shift a shift left followed by a shift right
;;
;; This attribute is used to determine the instruction's length and
;; scheduling type. For doubleword moves, the attribute always describes
;; the split instructions; in some cases, it is more appropriate for the
;; scheduling type to be "multi" instead.
(define_attr "move_type"
"unknown,load,fpload,store,fpstore,mtc,mfc,mtlo,mflo,imul,move,fmove,
const,constN,signext,ext_ins,logical,arith,sll0,andi,loadpool,
shift_shift"
(const_string "unknown"))
(define_attr "alu_type" "unknown,add,sub,not,nor,and,or,xor,simd_add"
(const_string "unknown"))
;; Main data type used by the insn
(define_attr "mode" "unknown,none,QI,HI,SI,DI,TI,SF,DF,TF,FPSW,
V2DI,V4SI,V8HI,V16QI,V2DF,V4SF"
(const_string "unknown"))
;; True if the main data type is twice the size of a word.
(define_attr "dword_mode" "no,yes"
(cond [(and (eq_attr "mode" "DI,DF")
(not (match_test "TARGET_64BIT")))
(const_string "yes")
(and (eq_attr "mode" "TI,TF")
(match_test "TARGET_64BIT"))
(const_string "yes")]
(const_string "no")))
;; True if the main data type is four times of the size of a word.
(define_attr "qword_mode" "no,yes"
(cond [(and (eq_attr "mode" "TI,TF")
(not (match_test "TARGET_64BIT")))
(const_string "yes")]
(const_string "no")))
;; Attributes describing a sync loop. These loops have the form:
;;
;; if (RELEASE_BARRIER == YES) sync
;; 1: OLDVAL = *MEM
;; if ((OLDVAL & INCLUSIVE_MASK) != REQUIRED_OLDVAL) goto 2
;; CMP = 0 [delay slot]
;; $TMP1 = OLDVAL & EXCLUSIVE_MASK
;; $TMP2 = INSN1 (OLDVAL, INSN1_OP2)
;; $TMP3 = INSN2 ($TMP2, INCLUSIVE_MASK)
;; $AT |= $TMP1 | $TMP3
;; if (!commit (*MEM = $AT)) goto 1.
;; if (INSN1 != MOVE && INSN1 != LI) NEWVAL = $TMP3 [delay slot]
;; CMP = 1
;; if (ACQUIRE_BARRIER == YES) sync
;; 2:
;;
;; where "$" values are temporaries and where the other values are
;; specified by the attributes below. Values are specified as operand
;; numbers and insns are specified as enums. If no operand number is
;; specified, the following values are used instead:
;;
;; - OLDVAL: $AT
;; - CMP: NONE
;; - NEWVAL: $AT
;; - INCLUSIVE_MASK: -1
;; - REQUIRED_OLDVAL: OLDVAL & INCLUSIVE_MASK
;; - EXCLUSIVE_MASK: 0
;;
;; MEM and INSN1_OP2 are required.
;;
;; Ideally, the operand attributes would be integers, with -1 meaning "none",
;; but the gen* programs don't yet support that.
(define_attr "sync_mem" "none,0,1,2,3,4,5" (const_string "none"))
(define_attr "sync_oldval" "none,0,1,2,3,4,5" (const_string "none"))
(define_attr "sync_cmp" "none,0,1,2,3,4,5" (const_string "none"))
(define_attr "sync_newval" "none,0,1,2,3,4,5" (const_string "none"))
(define_attr "sync_inclusive_mask" "none,0,1,2,3,4,5" (const_string "none"))
(define_attr "sync_exclusive_mask" "none,0,1,2,3,4,5" (const_string "none"))
(define_attr "sync_required_oldval" "none,0,1,2,3,4,5" (const_string "none"))
(define_attr "sync_insn1_op2" "none,0,1,2,3,4,5" (const_string "none"))
(define_attr "sync_insn1" "move,li,addu,addiu,subu,and,andi,or,ori,xor,xori"
(const_string "move"))
(define_attr "sync_insn2" "nop,and,xor,not"
(const_string "nop"))
;; Memory model specifier.
;; "0"-"9" values specify the operand that stores the memory model value.
;; "10" specifies MEMMODEL_ACQ_REL,
;; "11" specifies MEMMODEL_ACQUIRE.
(define_attr "sync_memmodel" "" (const_int 10))
;; Accumulator operand for madd patterns.
(define_attr "accum_in" "none,0,1,2,3,4,5" (const_string "none"))
;; 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
;; mtc transfer to coprocessor
;; mfc transfer from coprocessor
;; mthi transfer to a hi register
;; mtlo transfer to a lo register
;; mfhi transfer from a hi register
;; mflo transfer from a lo register
;; const load constant
;; arith integer arithmetic instructions
;; logical integer logical instructions
;; shift integer shift instructions
;; slt set less than instructions
;; signext sign extend instructions
;; clz the clz and clo instructions
;; pop the pop instruction
;; trap trap if instructions
;; imul integer multiply 2 operands
;; imul3 integer multiply 3 operands
;; imul3nc integer multiply 3 operands without clobbering HI/LO
;; imadd integer multiply-add
;; idiv integer divide 2 operands
;; idiv3 integer divide 3 operands
;; move integer register move ({,D}ADD{,U} with rt = 0)
;; 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
;; dspmac DSP MAC instructions not saturating the accumulator
;; dspmacsat DSP MAC instructions that saturate the accumulator
;; accext DSP accumulator extract instructions
;; accmod DSP accumulator modify instructions
;; dspalu DSP ALU instructions not saturating the result
;; dspalusat DSP ALU instructions that saturate the result
;; multi multiword sequence (or user asm statements)
;; atomic atomic memory update instruction
;; syncloop memory atomic operation implemented as a sync loop
;; nop no operation
;; ghost an instruction that produces no real code
;; multimem microMIPS multiword load and store
(define_attr "type"
"unknown,branch,jump,call,load,fpload,fpidxload,store,fpstore,fpidxstore,
prefetch,prefetchx,condmove,mtc,mfc,mthi,mtlo,mfhi,mflo,const,arith,logical,
shift,slt,signext,clz,pop,trap,imul,imul3,imul3nc,imadd,idiv,idiv3,move,
fmove,fadd,fmul,fmadd,fdiv,frdiv,frdiv1,frdiv2,fabs,fneg,fcmp,fcvt,fsqrt,
frsqrt,frsqrt1,frsqrt2,dspmac,dspmacsat,accext,accmod,dspalu,dspalusat,
multi,atomic,syncloop,nop,ghost,multimem,
simd_div,simd_fclass,simd_flog2,simd_fadd,simd_fcvt,simd_fmul,simd_fmadd,
simd_fdiv,simd_bitins,simd_bitmov,simd_insert,simd_sld,simd_mul,simd_fcmp,
simd_fexp2,simd_int_arith,simd_bit,simd_shift,simd_splat,simd_fill,
simd_permute,simd_shf,simd_sat,simd_pcnt,simd_copy,simd_branch,simd_cmsa,
simd_fminmax,simd_logic,simd_move,simd_load,simd_store"
(cond [(eq_attr "jal" "!unset") (const_string "call")
(eq_attr "got" "load") (const_string "load")
(eq_attr "alu_type" "add,sub") (const_string "arith")
(eq_attr "alu_type" "not,nor,and,or,xor") (const_string "logical")
;; If a doubleword move uses these expensive instructions,
;; it is usually better to schedule them in the same way
;; as the singleword form, rather than as "multi".
(eq_attr "move_type" "load") (const_string "load")
(eq_attr "move_type" "fpload") (const_string "fpload")
(eq_attr "move_type" "store") (const_string "store")
(eq_attr "move_type" "fpstore") (const_string "fpstore")
(eq_attr "move_type" "mtc") (const_string "mtc")
(eq_attr "move_type" "mfc") (const_string "mfc")
(eq_attr "move_type" "mtlo") (const_string "mtlo")
(eq_attr "move_type" "mflo") (const_string "mflo")
;; These types of move are always single insns.
(eq_attr "move_type" "imul") (const_string "imul")
(eq_attr "move_type" "fmove") (const_string "fmove")
(eq_attr "move_type" "loadpool") (const_string "load")
(eq_attr "move_type" "signext") (const_string "signext")
(eq_attr "move_type" "ext_ins") (const_string "arith")
(eq_attr "move_type" "arith") (const_string "arith")
(eq_attr "move_type" "logical") (const_string "logical")
(eq_attr "move_type" "sll0") (const_string "shift")
(eq_attr "move_type" "andi") (const_string "logical")
;; These types of move are always split.
(eq_attr "move_type" "constN,shift_shift")
(const_string "multi")
;; These types of move are split for quadword modes only.
(and (eq_attr "move_type" "move,const")
(eq_attr "qword_mode" "yes"))
(const_string "multi")
;; These types of move are split for doubleword modes only.
(and (eq_attr "move_type" "move,const")
(eq_attr "dword_mode" "yes"))
(const_string "multi")
(eq_attr "move_type" "move") (const_string "move")
(eq_attr "move_type" "const") (const_string "const")
(eq_attr "sync_mem" "!none") (const_string "syncloop")]
(const_string "unknown")))
(define_attr "compact_form" "always,maybe,never"
(cond [(eq_attr "jal" "direct")
(const_string "always")
(eq_attr "jal" "indirect")
(const_string "maybe")
(eq_attr "type" "jump")
(const_string "maybe")]
(const_string "never")))
;; 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"
(if_then_else (ior ;; In general, constant-pool loads are extended
;; instructions. We don't yet optimize for 16-bit
;; PC-relative references.
(eq_attr "move_type" "sll0,loadpool")
(eq_attr "jal" "direct")
(eq_attr "got" "load"))
(const_string "yes")
(const_string "no")))
(define_attr "compression" "none,all,micromips32,micromips"
(const_string "none"))
(define_attr "enabled" "no,yes"
(cond [;; The o32 FPXX and FP64A ABI extensions prohibit direct moves between
;; GR_REG and FR_REG for 64-bit values.
(and (eq_attr "move_type" "mtc,mfc")
(match_test "(TARGET_FLOATXX && !ISA_HAS_MXHC1)
|| TARGET_O32_FP64A_ABI")
(eq_attr "dword_mode" "yes"))
(const_string "no")
(and (eq_attr "compression" "micromips32,micromips")
(match_test "!TARGET_MICROMIPS"))
(const_string "no")]
(const_string "yes")))
;; The number of individual instructions that a non-branch pattern generates,
;; using units of BASE_INSN_LENGTH.
(define_attr "insn_count" ""
(cond [;; "Ghost" instructions occupy no space.
(eq_attr "type" "ghost")
(const_int 0)
;; Extended instructions count as 2.
(and (eq_attr "extended_mips16" "yes")
(match_test "TARGET_MIPS16"))
(const_int 2)
;; A GOT load followed by an add of $gp. This is not used for MIPS16.
(eq_attr "got" "xgot_high")
(const_int 2)
;; SHIFT_SHIFTs are decomposed into two separate instructions.
;; They are extended instructions on MIPS16 targets.
(eq_attr "move_type" "shift_shift")
(if_then_else (match_test "TARGET_MIPS16")
(const_int 4)
(const_int 2))
;; Check for doubleword moves that are decomposed into two
;; instructions. The individual instructions are unextended
;; MIPS16 ones.
(and (eq_attr "move_type" "mtc,mfc,mtlo,mflo,move")
(eq_attr "dword_mode" "yes"))
(const_int 2)
;; Check for quadword moves that are decomposed into four
;; instructions.
(and (eq_attr "move_type" "mtc,mfc,move")
(eq_attr "qword_mode" "yes"))
(const_int 4)
;; Constants, loads and stores are handled by external routines.
(and (eq_attr "move_type" "const,constN")
(eq_attr "dword_mode" "yes"))
(symbol_ref "mips_split_const_insns (operands[1])")
(eq_attr "move_type" "const,constN")
(symbol_ref "mips_const_insns (operands[1])")
(eq_attr "move_type" "load,fpload")
(symbol_ref "mips_load_store_insns (operands[1], insn)")
(eq_attr "move_type" "store,fpstore")
(symbol_ref "mips_load_store_insns (operands[0], insn)
+ (TARGET_FIX_24K ? 1 : 0)")
;; 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 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")
(match_test "TARGET_FIX_VR4120"))
(const_int 2)
;; 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")
(eq_attr "mode" "DI")
(match_test "TARGET_FIX_VR4120"))
(const_int 2)
(eq_attr "type" "idiv,idiv3")
(symbol_ref "mips_idiv_insns (GET_MODE (PATTERN (insn)))")
;; simd div have 3 instruction if TARGET_CHECK_ZERO_DIV is true.
(eq_attr "type" "simd_div")
(if_then_else (match_test "TARGET_CHECK_ZERO_DIV")
(const_int 3)
(const_int 1))
(not (eq_attr "sync_mem" "none"))
(symbol_ref "mips_sync_loop_insns (insn, operands)")]
(const_int 1)))
;; Length of instruction in bytes. The default is derived from "insn_count",
;; but there are special cases for branches (which must be handled here)
;; and for compressed single instructions.
(define_attr "length" ""
(cond [(and (ior (eq_attr "compression" "micromips,all")
(and (eq_attr "compression" "micromips32")
(eq_attr "mode" "SI,SF")))
(eq_attr "dword_mode" "no")
(match_test "TARGET_MICROMIPS"))
(const_int 2)
;; Direct microMIPS branch instructions have a range of
;; [-0x10000,0xfffe], otherwise the range is [-0x20000,0x1fffc].
;; 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:
;;
;; 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.
;;
;; The value we specify here 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 range of
;; [-0x20000, 0x1fffc] from the address of the delay slot
;; therefore translates to a range of:
;;
;; [-(0x20000 - sizeof (branch)), 0x1fffc - sizeof (slot)]
;; == [-0x1fffc, 0x1fff8]
;;
;; from the shorten_branches reference address.
(and (eq_attr "type" "branch")
(not (match_test "TARGET_MIPS16")))
(cond [;; Any variant can handle the 17-bit range.
(and (le (minus (match_dup 0) (pc)) (const_int 65532))
(le (minus (pc) (match_dup 0)) (const_int 65534)))
(const_int 4)
;; The 18-bit range is OK other than for microMIPS.
(and (not (match_test "TARGET_MICROMIPS"))
(and (le (minus (match_dup 0) (pc)) (const_int 131064))
(le (minus (pc) (match_dup 0)) (const_int 131068))))
(const_int 4)
;; The non-PIC case: branch, first delay slot, and J.
(match_test "TARGET_ABSOLUTE_JUMPS")
(const_int 12)]
;; Use MAX_PIC_BRANCH_LENGTH as a (gross) overestimate.
;; mips_adjust_insn_length substitutes the correct length.
;;
;; Note that we can't simply use (symbol_ref ...) here
;; because genattrtab needs to know the maximum length
;; of an insn.
(const_int MAX_PIC_BRANCH_LENGTH))
;; An unextended MIPS16 branch has a range of [-0x100, 0xfe]
;; from the address of the following instruction, which leads
;; to a range of:
;;
;; [-(0x100 - sizeof (branch)), 0xfe]
;; == [-0xfe, 0xfe]
;;
;; from the shorten_branches reference address. Extended branches
;; likewise have a range of [-0x10000, 0xfffe] from the address
;; of the following instruction, which leads to a range of:
;;
;; [-(0x10000 - sizeof (branch)), 0xfffe]
;; == [-0xfffc, 0xfffe]
;;
;; from the reference address.
;;
;; When a branch is out of range, mips_reorg splits it into a form
;; that uses in-range branches. There are four basic sequences:
;;
;; (1) Absolute addressing with a readable text segment
;; (32-bit addresses):
;;
;; b... foo 2 bytes
;; move $1,$2 2 bytes
;; lw $2,label 2 bytes
;; jr $2 2 bytes
;; move $2,$1 2 bytes
;; .align 2 0 or 2 bytes
;; label:
;; .word target 4 bytes
;; foo:
;; (16 bytes in the worst case)
;;
;; (2) Absolute addressing with a readable text segment
;; (64-bit addresses):
;;
;; b... foo 2 bytes
;; move $1,$2 2 bytes
;; ld $2,label 2 bytes
;; jr $2 2 bytes
;; move $2,$1 2 bytes
;; .align 3 0 to 6 bytes
;; label:
;; .dword target 8 bytes
;; foo:
;; (24 bytes in the worst case)
;;
;; (3) Absolute addressing without a readable text segment
;; (which requires 32-bit addresses at present):
;;
;; b... foo 2 bytes
;; move $1,$2 2 bytes
;; lui $2,%hi(target) 4 bytes
;; sll $2,8 2 bytes
;; sll $2,8 2 bytes
;; addiu $2,%lo(target) 4 bytes
;; jr $2 2 bytes
;; move $2,$1 2 bytes
;; foo:
;; (20 bytes)
;;
;; (4) PIC addressing (which requires 32-bit addresses at present):
;;
;; b... foo 2 bytes
;; move $1,$2 2 bytes
;; lw $2,cprestore 0, 2 or 4 bytes
;; lw $2,%got(target)($2) 4 bytes
;; addiu $2,%lo(target) 4 bytes
;; jr $2 2 bytes
;; move $2,$1 2 bytes
;; foo:
;; (20 bytes in the worst case)
(and (eq_attr "type" "branch")
(match_test "TARGET_MIPS16"))
(cond [(and (le (minus (match_dup 0) (pc)) (const_int 254))
(le (minus (pc) (match_dup 0)) (const_int 254)))
(const_int 2)
(and (le (minus (match_dup 0) (pc)) (const_int 65534))
(le (minus (pc) (match_dup 0)) (const_int 65532)))
(const_int 4)
(and (match_test "TARGET_ABICALLS")
(not (match_test "TARGET_ABSOLUTE_ABICALLS")))
(const_int 20)
(match_test "Pmode == SImode")
(const_int 16)
] (const_int 24))]
(symbol_ref "get_attr_insn_count (insn) * BASE_INSN_LENGTH")))
;; Attribute describing the processor.
(define_enum_attr "cpu" "processor"
(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,forbidden_slot"
(cond [(and (eq_attr "type" "load,fpload,fpidxload")
(match_test "ISA_HAS_LOAD_DELAY"))
(const_string "delay")
(and (eq_attr "type" "mfc,mtc")
(match_test "ISA_HAS_XFER_DELAY"))
(const_string "delay")
(and (eq_attr "type" "fcmp")
(match_test "ISA_HAS_FCMP_DELAY"))
(const_string "delay")
;; The r4000 multiplication patterns include an mflo instruction.
(and (eq_attr "type" "imul")
(match_test "TARGET_FIX_R4000"))
(const_string "hilo")
(and (eq_attr "type" "mfhi,mflo")
(not (match_test "ISA_HAS_HILO_INTERLOCKS")))
(const_string "hilo")]
(const_string "none")))
;; 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,simd_branch")
(eq_attr "hazard" "none")
(match_test "get_attr_insn_count (insn) == 1"))
(const_string "yes")
(const_string "no")))
;; Attribute defining whether or not we can use the branch-likely
;; instructions.
(define_attr "branch_likely" "no,yes"
(if_then_else (match_test "GENERATE_BRANCHLIKELY")
(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,mthi,mtlo")
(const_string "yes")
(const_string "no")))
;; Describe a user's asm statement.
(define_asm_attributes
[(set_attr "type" "multi")
(set_attr "can_delay" "no")])
;; This mode iterator allows 32-bit and 64-bit GPR patterns to be generated
;; from the same template.
(define_mode_iterator GPR [SI (DI "TARGET_64BIT")])
;; A copy of GPR that can be used when a pattern has two independent
;; modes.
(define_mode_iterator GPR2 [SI (DI "TARGET_64BIT")])
(define_mode_iterator MOVEP1 [SI SF])
(define_mode_iterator MOVEP2 [SI SF])
(define_mode_iterator JOIN_MODE [HI
SI
(SF "TARGET_HARD_FLOAT")
(DF "TARGET_HARD_FLOAT
&& TARGET_DOUBLE_FLOAT")])
;; This mode iterator allows :HILO to be used as the mode of the
;; concatenated HI and LO registers.
(define_mode_iterator HILO [(DI "!TARGET_64BIT") (TI "TARGET_64BIT")])
;; This mode iterator allows :P to be used for patterns that operate on
;; pointer-sized quantities. Exactly one of the two alternatives will match.
(define_mode_iterator P [(SI "Pmode == SImode") (DI "Pmode == DImode")])
;; This mode iterator allows :MOVECC to be used anywhere that a
;; conditional-move-type condition is needed.
(define_mode_iterator MOVECC [SI (DI "TARGET_64BIT")
(CC "TARGET_HARD_FLOAT
&& !TARGET_LOONGSON_2EF
&& !TARGET_MIPS5900")])
;; This mode iterator allows :FPCC to be used anywhere that an FP condition
;; is needed.
(define_mode_iterator FPCC [(CC "!ISA_HAS_CCF")
(CCF "ISA_HAS_CCF")])
;; 32-bit integer moves for which we provide move patterns.
(define_mode_iterator IMOVE32
[SI
(V2HI "TARGET_DSP")
(V4QI "TARGET_DSP")
(V2HQ "TARGET_DSP")
(V2UHQ "TARGET_DSP")
(V2HA "TARGET_DSP")
(V2UHA "TARGET_DSP")
(V4QQ "TARGET_DSP")
(V4UQQ "TARGET_DSP")])
;; 64-bit modes for which we provide move patterns.
(define_mode_iterator MOVE64
[DI DF
(V2SF "TARGET_HARD_FLOAT && TARGET_PAIRED_SINGLE_FLOAT")
(V2SI "TARGET_HARD_FLOAT && TARGET_LOONGSON_MMI")
(V4HI "TARGET_HARD_FLOAT && TARGET_LOONGSON_MMI")
(V8QI "TARGET_HARD_FLOAT && TARGET_LOONGSON_MMI")])
;; 128-bit modes for which we provide move patterns on 64-bit targets.
(define_mode_iterator MOVE128 [TI TF])
;; This mode iterator allows the QI and HI extension patterns to be
;; defined from the same template.
(define_mode_iterator SHORT [QI HI])
;; Likewise the 64-bit truncate-and-shift patterns.
(define_mode_iterator SUBDI [QI HI SI])
;; This mode iterator allows :ANYF to be used wherever a scalar or vector
;; floating-point mode is allowed.
(define_mode_iterator ANYF [(SF "TARGET_HARD_FLOAT")
(DF "TARGET_HARD_FLOAT && TARGET_DOUBLE_FLOAT")
(V2SF "TARGET_HARD_FLOAT && TARGET_PAIRED_SINGLE_FLOAT")])
;; Like ANYF, but only applies to scalar modes.
(define_mode_iterator SCALARF [(SF "TARGET_HARD_FLOAT")
(DF "TARGET_HARD_FLOAT && TARGET_DOUBLE_FLOAT")])
;; A floating-point mode for which moves involving FPRs may need to be split.
(define_mode_iterator SPLITF
[(DF "!TARGET_64BIT && TARGET_DOUBLE_FLOAT")
(DI "!TARGET_64BIT && TARGET_DOUBLE_FLOAT")
(V2SF "!TARGET_64BIT && TARGET_PAIRED_SINGLE_FLOAT")
(V2SI "!TARGET_64BIT && TARGET_LOONGSON_MMI")
(V4HI "!TARGET_64BIT && TARGET_LOONGSON_MMI")
(V8QI "!TARGET_64BIT && TARGET_LOONGSON_MMI")
(TF "TARGET_64BIT && TARGET_FLOAT64")])
;; In GPR templates, a string like "subu" will expand to "subu" in the
;; 32-bit version and "dsubu" in the 64-bit version.
(define_mode_attr d [(SI "") (DI "d")
(QQ "") (HQ "") (SQ "") (DQ "d")
(UQQ "") (UHQ "") (USQ "") (UDQ "d")
(HA "") (SA "") (DA "d")
(UHA "") (USA "") (UDA "d")])
;; Same as d but upper-case.
(define_mode_attr D [(SI "") (DI "D")
(QQ "") (HQ "") (SQ "") (DQ "D")
(UQQ "") (UHQ "") (USQ "") (UDQ "D")
(HA "") (SA "") (DA "D")
(UHA "") (USA "") (UDA "D")])
;; This attribute gives the length suffix for a load or store instruction.
;; The same suffixes work for zero and sign extensions.
(define_mode_attr size [(QI "b") (HI "h") (SI "w") (DI "d")])
(define_mode_attr SIZE [(QI "B") (HI "H") (SI "W") (DI "D")])
;; This attributes gives the mode mask of a SHORT.
(define_mode_attr mask [(QI "0x00ff") (HI "0xffff")])
;; Mode attributes for GPR loads.
(define_mode_attr load [(SI "lw") (DI "ld")])
;; Instruction names for stores.
(define_mode_attr store [(QI "sb") (HI "sh") (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") (CCF "f")])
;; 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 or vector mode.
(define_mode_attr UNITMODE [(SF "SF") (DF "DF") (V2SF "SF") (V4SF "SF")
(V16QI "QI") (V8HI "HI") (V4SI "SI") (V2DI "DI")
(V2DF "DF")])
;; As above, but in lower case.
(define_mode_attr unitmode [(SF "sf") (DF "df") (V2SF "sf") (V4SF "sf")
(V16QI "qi") (V8QI "qi") (V8HI "hi") (V4HI "hi")
(V4SI "si") (V2SI "si") (V2DI "di") (V2DF "df")])
;; This attribute gives the integer mode that has the same size as a
;; fixed-point mode.
(define_mode_attr IMODE [(QQ "QI") (HQ "HI") (SQ "SI") (DQ "DI")
(UQQ "QI") (UHQ "HI") (USQ "SI") (UDQ "DI")
(HA "HI") (SA "SI") (DA "DI")
(UHA "HI") (USA "SI") (UDA "DI")
(V4UQQ "SI") (V2UHQ "SI") (V2UHA "SI")
(V2HQ "SI") (V2HA "SI")])
;; This attribute gives the integer mode that has half the size of
;; the controlling mode.
(define_mode_attr HALFMODE [(DF "SI") (DI "SI") (V2SF "SI")
(V2SI "SI") (V4HI "SI") (V8QI "SI")
(TF "DI")])
;; 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 conditions under which SQRT.fmt instructions
;; can be used.
(define_mode_attr sqrt_condition
[(SF "!ISA_MIPS1") (DF "!ISA_MIPS1") (V2SF "TARGET_SB1")])
;; This attribute provides the correct mnemonic for each FP condition mode.
(define_mode_attr fpcmp [(CC "c") (CCF "cmp")])
;; This code iterator allows signed and unsigned widening multiplications
;; to use the same template.
(define_code_iterator any_extend [sign_extend zero_extend])
;; This code iterator allows the two right shift instructions to be
;; generated from the same template.
(define_code_iterator any_shiftrt [ashiftrt lshiftrt])
;; This code iterator allows the three shift instructions to be generated
;; from the same template.
(define_code_iterator any_shift [ashift ashiftrt lshiftrt])
;; This code iterator allows unsigned and signed division to be generated
;; from the same template.
(define_code_iterator any_div [div udiv])
;; This code iterator allows unsigned and signed modulus to be generated
;; from the same template.
(define_code_iterator any_mod [mod umod])
;; This code iterator allows addition and subtraction to be generated
;; from the same template.
(define_code_iterator addsub [plus minus])
;; This code iterator allows all native floating-point comparisons to be
;; generated from the same template.
(define_code_iterator fcond [unordered uneq unlt unle eq lt le
(ordered "ISA_HAS_CCF")
(ltgt "ISA_HAS_CCF")
(ne "ISA_HAS_CCF")])
;; This code iterator is used for comparisons that can be implemented
;; by swapping the operands.
(define_code_iterator swapped_fcond [ge gt unge ungt])
;; Equality operators.
(define_code_iterator equality_op [eq ne])
;; These code iterators allow the signed and unsigned scc operations to use
;; the same template.
(define_code_iterator any_gt [gt gtu])
(define_code_iterator any_ge [ge geu])
(define_code_iterator any_lt [lt ltu])
(define_code_iterator any_le [le leu])
(define_code_iterator any_return [return simple_return])
;; 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")
(div "") (udiv "u")
(mod "") (umod "u")
(gt "") (gtu "u")
(ge "") (geu "u")
(lt "") (ltu "u")
(le "") (leu "u")])
;; is like except uppercase.
(define_code_attr U [(sign_extend "") (zero_extend "U")])
;; is like , but the signed form expands to "s" rather than "".
(define_code_attr su [(sign_extend "s") (zero_extend "u")])
;; expands to the name of the optab for a particular code.
(define_code_attr optab [(ashift "ashl")
(ashiftrt "ashr")
(lshiftrt "lshr")
(ior "ior")
(xor "xor")
(and "and")
(plus "add")
(minus "sub")
(return "return")
(simple_return "simple_return")])
;; expands to the name of the insn that implements a particular code.
(define_code_attr insn [(ashift "sll")
(ashiftrt "sra")
(lshiftrt "srl")
(ior "or")
(xor "xor")
(and "and")
(plus "addu")
(minus "subu")])
;; expands to the name of the insn that implements
;; a particular code to operate on immediate values.
(define_code_attr immediate_insn [(ior "ori")
(xor "xori")
(and "andi")])
(define_code_attr shift_compression [(ashift "micromips32")
(lshiftrt "micromips32")
(ashiftrt "none")])
;; 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")
(ordered "or")
(ltgt "ne")
(ne "une")])
;; Similar, but for swapped conditions.
(define_code_attr swapped_fcond [(ge "le")
(gt "lt")
(unge "ule")
(ungt "ult")])
;; The value of the bit when the branch is taken for branch_bit patterns.
;; Comparison is always against zero so this depends on the operator.
(define_code_attr bbv [(eq "0") (ne "1")])
;; This is the inverse value of bbv.
(define_code_attr bbinv [(eq "1") (ne "0")])
;; The sel mnemonic to use depending on the condition test.
(define_code_attr sel [(eq "seleqz") (ne "selnez")])
(define_code_attr selinv [(eq "selnez") (ne "seleqz")])
;; .........................
;;
;; Branch, call and jump delay slots
;;
;; .........................
(define_delay (and (eq_attr "type" "branch")
(not (match_test "TARGET_MIPS16"))
(eq_attr "branch_likely" "yes"))
[(eq_attr "can_delay" "yes")
(nil)
(eq_attr "can_delay" "yes")])
;; Branches that have delay slots and don't have likely variants do
;; not annul on false.
(define_delay (and (eq_attr "type" "branch,simd_branch")
(not (match_test "TARGET_MIPS16"))
(ior (match_test "TARGET_CB_NEVER")
(and (eq_attr "compact_form" "maybe")
(not (match_test "TARGET_CB_ALWAYS")))
(eq_attr "compact_form" "never"))
(eq_attr "branch_likely" "no"))
[(eq_attr "can_delay" "yes")
(nil)
(nil)])
(define_delay (and (eq_attr "type" "jump")
(ior (match_test "TARGET_CB_NEVER")
(and (eq_attr "compact_form" "maybe")
(not (match_test "TARGET_CB_ALWAYS")))
(eq_attr "compact_form" "never")))
[(eq_attr "can_delay" "yes")
(nil)
(nil)])
;; Call type instructions should never have a compact form as the
;; type is only used for MIPS16 patterns. For safety put the compact
;; branch detection condition in anyway.
(define_delay (and (eq_attr "type" "call")
(eq_attr "jal_macro" "no")
(ior (match_test "TARGET_CB_NEVER")
(and (eq_attr "compact_form" "maybe")
(not (match_test "TARGET_CB_ALWAYS")))
(eq_attr "compact_form" "never")))
[(eq_attr "can_delay" "yes")
(nil)
(nil)])
;; 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")
;; Ghost instructions produce no real code and introduce no hazards.
;; They exist purely to express an effect on dataflow.
(define_insn_reservation "ghost" 0
(eq_attr "type" "ghost")
"nothing")
(include "i6400.md")
(include "p5600.md")
(include "m5100.md")
(include "p6600.md")
(include "4k.md")
(include "5k.md")
(include "20kc.md")
(include "24k.md")
(include "74k.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 "10000.md")
(include "loongson2ef.md")
(include "gs464.md")
(include "gs464e.md")
(include "gs264e.md")
(include "octeon.md")
(include "sb1.md")
(include "sr71k.md")
(include "xlr.md")
(include "xlp.md")
(include "generic.md")
;;
;; ....................
;;
;; 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 "ctrap4"
[(trap_if (match_operator 0 "comparison_operator"
[(match_operand:GPR 1 "reg_or_0_operand")
(match_operand:GPR 2 "arith_operand")])
(match_operand 3 "const_0_operand"))]
"ISA_HAS_COND_TRAPI || ISA_HAS_COND_TRAP"
{
mips_expand_conditional_trap (operands[0]);
DONE;
})
(define_insn "*conditional_trap_reg"
[(trap_if (match_operator:GPR 0 "trap_comparison_operator"
[(match_operand:GPR 1 "reg_or_0_operand" "dJ")
(match_operand:GPR 2 "reg_or_0_operand" "dJ")])
(const_int 0))]
"ISA_HAS_COND_TRAP && !ISA_HAS_COND_TRAPI"
"t%C0\t%z1,%2"
[(set_attr "type" "trap")])
(define_insn "*conditional_trap"
[(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_TRAPI"
"t%C0\t%z1,%2"
[(set_attr "type" "trap")])
;;
;; ....................
;;
;; ADDITION
;;
;; ....................
;;
(define_insn "add3"
[(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.\t%0,%1,%2"
[(set_attr "type" "fadd")
(set_attr "mode" "")])
(define_expand "add3"
[(set (match_operand:GPR 0 "register_operand")
(plus:GPR (match_operand:GPR 1 "register_operand")
(match_operand:GPR 2 "arith_operand")))]
"")
(define_insn "*add3"
[(set (match_operand:GPR 0 "register_operand" "=!u,d,!u,!u,!ks,!d,d")
(plus:GPR (match_operand:GPR 1 "register_operand" "!u,d,!u,!ks,!ks,0,d")
(match_operand:GPR 2 "arith_operand" "!u,d,Uead,Uuw6,Uesp,Usb4,Q")))]
"!TARGET_MIPS16"
{
if (which_alternative == 0
|| which_alternative == 1)
return "addu\t%0,%1,%2";
else
return "addiu\t%0,%1,%2";
}
[(set_attr "alu_type" "add")
(set_attr "compression" "micromips32,*,micromips32,micromips32,micromips32,micromips32,*")
(set_attr "mode" "")])
(define_insn "*add3_mips16"
[(set (match_operand:GPR 0 "register_operand" "=ks,ks,d,d,d,d,d,d,d")
(plus:GPR (match_operand:GPR 1 "register_operand" "ks,ks,ks,ks,0,0,d,d,d")
(match_operand:GPR 2 "arith_operand" "Usd8,Q,Uuw,Q,Usb,Q,Usb4,O,d")))]
"TARGET_MIPS16"
"@
addiu\t%0,%2
addiu\t%0,%2
addiu\t%0,%1,%2
addiu\t%0,%1,%2
addiu\t%0,%2
addiu\t%0,%2
addiu\t%0,%1,%2
addiu\t%0,%1,%2
addu\t%0,%1,%2"
[(set_attr "alu_type" "add")
(set_attr "mode" "")
(set_attr "extended_mips16" "no,yes,no,yes,no,yes,no,yes,no")])
;; 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 "d_operand")
(plus:SI (match_dup 0)
(match_operand:SI 1 "const_int_operand")))]
"TARGET_MIPS16 && reload_completed && !TARGET_DEBUG_D_MODE
&& ((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 "d_operand")
(plus:SI (match_operand:SI 1 "d_operand")
(match_operand:SI 2 "const_int_operand")))]
"TARGET_MIPS16 && reload_completed && !TARGET_DEBUG_D_MODE
&& REGNO (operands[0]) != REGNO (operands[1])
&& ((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 "d_operand")
(plus:DI (match_dup 0)
(match_operand:DI 1 "const_int_operand")))]
"TARGET_MIPS16 && TARGET_64BIT && reload_completed && !TARGET_DEBUG_D_MODE
&& ((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 "d_operand")
(plus:DI (match_operand:DI 1 "d_operand")
(match_operand:DI 2 "const_int_operand")))]
"TARGET_MIPS16 && TARGET_64BIT && reload_completed && !TARGET_DEBUG_D_MODE
&& REGNO (operands[0]) != REGNO (operands[1])
&& ((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 "alu_type" "add")
(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 "alu_type" "add")
(set_attr "mode" "SI")
(set_attr "extended_mips16" "yes")])
;; Combiner patterns for unsigned byte-add.
(define_insn "*baddu_si_eb"
[(set (match_operand:SI 0 "register_operand" "=d")
(zero_extend:SI
(subreg:QI
(plus:SI (match_operand:SI 1 "register_operand" "d")
(match_operand:SI 2 "register_operand" "d")) 3)))]
"ISA_HAS_BADDU && BYTES_BIG_ENDIAN"
"baddu\\t%0,%1,%2"
[(set_attr "alu_type" "add")])
(define_insn "*baddu_si_el"
[(set (match_operand:SI 0 "register_operand" "=d")
(zero_extend:SI
(subreg:QI
(plus:SI (match_operand:SI 1 "register_operand" "d")
(match_operand:SI 2 "register_operand" "d")) 0)))]
"ISA_HAS_BADDU && !BYTES_BIG_ENDIAN"
"baddu\\t%0,%1,%2"
[(set_attr "alu_type" "add")])
(define_insn "*baddu_di"
[(set (match_operand:GPR 0 "register_operand" "=d")
(zero_extend:GPR
(truncate:QI
(plus:DI (match_operand:DI 1 "register_operand" "d")
(match_operand:DI 2 "register_operand" "d")))))]
"ISA_HAS_BADDU && TARGET_64BIT"
"baddu\\t%0,%1,%2"
[(set_attr "alu_type" "add")])
;;
;; ....................
;;
;; SUBTRACTION
;;
;; ....................
;;
(define_insn "sub3"
[(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.\t%0,%1,%2"
[(set_attr "type" "fadd")
(set_attr "mode" "")])
(define_insn "sub3"
[(set (match_operand:GPR 0 "register_operand" "=!u,d")
(minus:GPR (match_operand:GPR 1 "register_operand" "!u,d")
(match_operand:GPR 2 "register_operand" "!u,d")))]
""
"subu\t%0,%1,%2"
[(set_attr "alu_type" "sub")
(set_attr "compression" "micromips32,*")
(set_attr "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 "alu_type" "sub")
(set_attr "mode" "DI")])
;;
;; ....................
;;
;; MULTIPLICATION
;;
;; ....................
;;
(define_expand "mul3"
[(set (match_operand:SCALARF 0 "register_operand")
(mult:SCALARF (match_operand:SCALARF 1 "register_operand")
(match_operand:SCALARF 2 "register_operand")))]
""
"")
(define_insn "*mul3"
[(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.\t%0,%1,%2"
[(set_attr "type" "fmul")
(set_attr "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 "*mul3_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.\t%0,%1,%2\;nop"
[(set_attr "type" "fmul")
(set_attr "mode" "")
(set_attr "insn_count" "2")])
(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_HARD_FLOAT && 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 "mul3"
[(set (match_operand:GPR 0 "register_operand")
(mult:GPR (match_operand:GPR 1 "register_operand")
(match_operand:GPR 2 "register_operand")))]
"ISA_HAS_MULT || ISA_HAS_R6MUL"
{
rtx lo;
if (TARGET_LOONGSON_2EF || TARGET_LOONGSON_EXT || ISA_HAS_R6MUL)
emit_insn (gen_mul3_mul3_nohilo (operands[0], operands[1],
operands[2]));
else if (ISA_HAS_MUL3)
emit_insn (gen_mul3_mul3 (operands[0], operands[1], operands[2]));
else if (TARGET_MIPS16)
{
lo = gen_rtx_REG (mode, LO_REGNUM);
emit_insn (gen_mul3_internal (lo, operands[1], operands[2]));
emit_move_insn (operands[0], lo);
}
else if (TARGET_FIX_R4000)
emit_insn (gen_mul3_r4000 (operands[0], operands[1], operands[2]));
else
emit_insn
(gen_mul3_internal (operands[0], operands[1], operands[2]));
DONE;
})
(define_insn "mul3_mul3_nohilo"
[(set (match_operand:GPR 0 "register_operand" "=d")
(mult:GPR (match_operand:GPR 1 "register_operand" "d")
(match_operand:GPR 2 "register_operand" "d")))]
"TARGET_LOONGSON_2EF || TARGET_LOONGSON_EXT || ISA_HAS_R6MUL"
{
if (TARGET_LOONGSON_2EF)
return "multu.g\t%0,%1,%2";
else if (TARGET_LOONGSON_EXT)
return "gsmultu\t%0,%1,%2";
else
return "mul\t%0,%1,%2";
}
[(set_attr "type" "imul3nc")
(set_attr "mode" "")])
(define_insn "mul3_mul3"
[(set (match_operand:GPR 0 "register_operand" "=d,l")
(mult:GPR (match_operand:GPR 1 "register_operand" "d,d")
(match_operand:GPR 2 "register_operand" "d,d")))
(clobber (match_scratch:GPR 3 "=l,X"))]
"ISA_HAS_MUL3"
{
if (which_alternative == 1)
return "mult\t%1,%2";
if (mode == SImode && (TARGET_MIPS3900 || TARGET_MIPS5900))
return "mult\t%0,%1,%2";
return "mul\t%0,%1,%2";
}
[(set_attr "type" "imul3,imul")
(set_attr "mode" "")])
;; 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: GPR (destination)
(define_peephole2
[(parallel
[(set (match_operand:SI 0 "lo_operand")
(mult:SI (match_operand:SI 1 "d_operand")
(match_operand:SI 2 "d_operand")))
(clobber (scratch:SI))])
(set (match_operand:SI 3 "d_operand")
(match_dup 0))]
"ISA_HAS_MUL3 && peep2_reg_dead_p (2, operands[0])"
[(parallel
[(set (match_dup 3)
(mult:SI (match_dup 1)
(match_dup 2)))
(clobber (match_dup 0))])])
(define_insn "mul3_internal"
[(set (match_operand:GPR 0 "muldiv_target_operand" "=l")
(mult:GPR (match_operand:GPR 1 "register_operand" "d")
(match_operand:GPR 2 "register_operand" "d")))]
"ISA_HAS_MULT && !TARGET_FIX_R4000"
"mult\t%1,%2"
[(set_attr "type" "imul")
(set_attr "mode" "")])
(define_insn "mul3_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 "=l"))]
"ISA_HAS_MULT && TARGET_FIX_R4000"
"mult\t%1,%2\;mflo\t%0"
[(set_attr "type" "imul")
(set_attr "mode" "")
(set_attr "insn_count" "2")])
;; 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: GPR (destination)
(define_peephole2
[(set (match_operand:SI 0 "lo_operand")
(mult:SI (match_operand:SI 1 "d_operand")
(match_operand:SI 2 "d_operand")))
(set (match_operand:SI 3 "d_operand")
(match_dup 0))]
"ISA_HAS_MACC && !ISA_HAS_MUL3"
[(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 3)
(plus:SI (mult:SI (match_dup 1)
(match_dup 2))
(match_dup 0)))])])
;; Multiply-accumulate patterns
;; This pattern is first matched by combine, which tries to use the
;; pattern wherever it can. We don't know until later whether it
;; is actually profitable to use MADD over a "MUL; ADDIU" sequence,
;; so we need to keep both options open.
;;
;; The second alternative has a "?" marker because it is generally
;; one instruction more costly than the first alternative. This "?"
;; marker is enough to convey the relative costs to the register
;; allocator.
;;
;; However, reload counts reloads of operands 4 and 5 in the same way as
;; reloads of the other operands, even though operands 4 and 5 need no
;; copy instructions. Reload therefore thinks that the second alternative
;; is two reloads more costly than the first. We add "*?*?" to the first
;; alternative as a counterweight.
;;
;; LRA simulates reload but the cost of reloading scratches is lower
;; than of the classic reload. For the time being, removing the counterweight
;; for LRA is more profitable.
(define_insn "*mul_acc_si"
[(set (match_operand:SI 0 "register_operand" "=l*?*?,l,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" "l,l,d")))
(clobber (match_scratch:SI 4 "=X,X,l"))
(clobber (match_scratch:SI 5 "=X,X,&d"))]
"GENERATE_MADD_MSUB && !TARGET_MIPS16"
"@
madd\t%1,%2
madd\t%1,%2
#"
[(set_attr "type" "imadd")
(set_attr "accum_in" "3")
(set_attr "mode" "SI")
(set_attr "insn_count" "1,1,2")
(set (attr "enabled")
(cond [(and (eq_attr "alternative" "0")
(match_test "!mips_lra_flag"))
(const_string "yes")
(and (eq_attr "alternative" "1")
(match_test "mips_lra_flag"))
(const_string "yes")
(eq_attr "alternative" "2")
(const_string "yes")]
(const_string "no")))])
;; The same idea applies here. The middle alternative needs one less
;; clobber than the final alternative, so we add "*?" as a counterweight.
(define_insn "*mul_acc_si_r3900"
[(set (match_operand:SI 0 "register_operand" "=l*?*?,l,d*?,d?")
(plus:SI (mult:SI (match_operand:SI 1 "register_operand" "d,d,d,d")
(match_operand:SI 2 "register_operand" "d,d,d,d"))
(match_operand:SI 3 "register_operand" "l,l,l,d")))
(clobber (match_scratch:SI 4 "=X,X,3,l"))
(clobber (match_scratch:SI 5 "=X,X,X,&d"))]
"TARGET_MIPS3900 && !TARGET_MIPS16"
"@
madd\t%1,%2
madd\t%1,%2
madd\t%0,%1,%2
#"
[(set_attr "type" "imadd")
(set_attr "accum_in" "3")
(set_attr "mode" "SI")
(set_attr "insn_count" "1,1,1,2")
(set (attr "enabled")
(cond [(and (eq_attr "alternative" "0")
(match_test "!mips_lra_flag"))
(const_string "yes")
(and (eq_attr "alternative" "1")
(match_test "mips_lra_flag"))
(const_string "yes")
(eq_attr "alternative" "2,3")
(const_string "yes")]
(const_string "no")))])
;; Split *mul_acc_si if both the source and destination accumulator
;; values are GPRs.
(define_split
[(set (match_operand:SI 0 "d_operand")
(plus:SI (mult:SI (match_operand:SI 1 "d_operand")
(match_operand:SI 2 "d_operand"))
(match_operand:SI 3 "d_operand")))
(clobber (match_operand:SI 4 "lo_operand"))
(clobber (match_operand:SI 5 "d_operand"))]
"reload_completed"
[(parallel [(set (match_dup 5)
(mult:SI (match_dup 1) (match_dup 2)))
(clobber (match_dup 4))])
(set (match_dup 0) (plus:SI (match_dup 5) (match_dup 3)))]
"")
(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" "l,l")))
(clobber (match_scratch:SI 4 "=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 "accum_in" "3")
(set_attr "mode" "SI")])
(define_insn "*msac"
[(set (match_operand:SI 0 "register_operand" "=l,d")
(minus:SI (match_operand:SI 1 "register_operand" "l,l")
(mult:SI (match_operand:SI 2 "register_operand" "d,d")
(match_operand:SI 3 "register_operand" "d,d"))))
(clobber (match_scratch:SI 4 "=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 "accum_in" "1")
(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" "l,l")
(mult:SI (match_operand:SI 2 "register_operand" "d,d")
(match_operand:SI 3 "register_operand" "d,d"))))
(clobber (match_scratch:SI 4 "=X,1"))
(clobber (match_scratch:SI 5 "=d,d"))]
"ISA_HAS_MACC && !ISA_HAS_MSAC"
"#"
"&& reload_completed"
[(set (match_dup 5)
(neg:SI (match_dup 3)))
(parallel
[(set (match_dup 0)
(plus:SI (mult:SI (match_dup 2)
(match_dup 5))
(match_dup 1)))
(clobber (match_dup 4))])]
""
[(set_attr "type" "imadd")
(set_attr "accum_in" "1")
(set_attr "insn_count" "2")])
;; Patterns generated by the define_peephole2 below.
(define_insn "*macc2"
[(set (match_operand:SI 0 "muldiv_target_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)))]
"ISA_HAS_MACC && reload_completed"
"macc\t%3,%1,%2"
[(set_attr "type" "imadd")
(set_attr "accum_in" "0")
(set_attr "mode" "SI")])
(define_insn "*msac2"
[(set (match_operand:SI 0 "muldiv_target_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))))]
"ISA_HAS_MSAC && reload_completed"
"msac\t%3,%1,%2"
[(set_attr "type" "imadd")
(set_attr "accum_in" "0")
(set_attr "mode" "SI")])
;; Convert macc $0,, & mflo into macc ,,
;; Similarly msac.
;;
;; Operand 0: LO
;; Operand 1: macc/msac
;; Operand 2: GPR (destination)
(define_peephole2
[(parallel
[(set (match_operand:SI 0 "lo_operand")
(match_operand:SI 1 "macc_msac_operand"))
(clobber (scratch:SI))])
(set (match_operand:SI 2 "d_operand")
(match_dup 0))]
""
[(parallel [(set (match_dup 0)
(match_dup 1))
(set (match_dup 2)
(match_dup 1))])])
;; 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: new multiplication
;; Operand 6: new addition/subtraction
(define_peephole2
[(match_scratch:SI 0 "d")
(set (match_operand:SI 1 "lo_operand")
(match_operand:SI 2 "d_operand"))
(match_dup 0)
(parallel
[(set (match_operand:SI 3 "d_operand")
(match_operand:SI 4 "macc_msac_operand"))
(clobber (match_dup 1))])]
"ISA_HAS_MUL3 && peep2_reg_dead_p (2, operands[1])"
[(parallel [(set (match_dup 0)
(match_dup 5))
(clobber (match_dup 1))])
(set (match_dup 3)
(match_dup 6))]
{
operands[5] = XEXP (operands[4], GET_CODE (operands[4]) == PLUS ? 0 : 1);
operands[6] = 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: GPR (destination)
;; Operand 5: new multiplication
;; Operand 6: new addition/subtraction
(define_peephole2
[(match_scratch:SI 0 "d")
(set (match_operand:SI 1 "lo_operand")
(match_operand:SI 2 "d_operand"))
(match_dup 0)
(parallel
[(set (match_dup 1)
(match_operand:SI 3 "macc_msac_operand"))
(clobber (scratch:SI))])
(match_dup 0)
(set (match_operand:SI 4 "d_operand")
(match_dup 1))]
"ISA_HAS_MUL3 && peep2_reg_dead_p (3, operands[1])"
[(parallel [(set (match_dup 0)
(match_dup 5))
(clobber (match_dup 1))])
(set (match_dup 4)
(match_dup 6))]
{
operands[5] = XEXP (operands[3], GET_CODE (operands[3]) == PLUS ? 0 : 1);
operands[6] = gen_rtx_fmt_ee (GET_CODE (operands[3]), SImode,
operands[2], operands[0]);
})
;; See the comment above *mul_add_si for details.
(define_insn "*mul_sub_si"
[(set (match_operand:SI 0 "register_operand" "=l*?*?,l,d?")
(minus:SI (match_operand:SI 1 "register_operand" "l,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 "=X,X,l"))
(clobber (match_scratch:SI 5 "=X,X,&d"))]
"GENERATE_MADD_MSUB"
"@
msub\t%2,%3
msub\t%2,%3
#"
[(set_attr "type" "imadd")
(set_attr "accum_in" "1")
(set_attr "mode" "SI")
(set_attr "insn_count" "1,1,2")
(set (attr "enabled")
(cond [(and (eq_attr "alternative" "0")
(match_test "!mips_lra_flag"))
(const_string "yes")
(and (eq_attr "alternative" "1")
(match_test "mips_lra_flag"))
(const_string "yes")
(eq_attr "alternative" "2")
(const_string "yes")]
(const_string "no")))])
;; Split *mul_sub_si if both the source and destination accumulator
;; values are GPRs.
(define_split
[(set (match_operand:SI 0 "d_operand")
(minus:SI (match_operand:SI 1 "d_operand")
(mult:SI (match_operand:SI 2 "d_operand")
(match_operand:SI 3 "d_operand"))))
(clobber (match_operand:SI 4 "lo_operand"))
(clobber (match_operand:SI 5 "d_operand"))]
"reload_completed"
[(parallel [(set (match_dup 5)
(mult:SI (match_dup 2) (match_dup 3)))
(clobber (match_dup 4))])
(set (match_dup 0) (minus:SI (match_dup 1) (match_dup 5)))]
"")
(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 "=X,l"))]
"ISA_HAS_MULS"
"@
muls\t$0,%1,%2
muls\t%0,%1,%2"
[(set_attr "type" "imul,imul3")
(set_attr "mode" "SI")])
(define_expand "mulsidi3"
[(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"))))]
"mips_mulsidi3_gen_fn () != NULL"
{
mulsidi3_gen_fn fn = mips_mulsidi3_gen_fn ();
emit_insn (fn (operands[0], operands[1], operands[2]));
DONE;
})
(define_expand "mulsidi3_32bit_r6"
[(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"))))]
"!TARGET_64BIT && ISA_HAS_R6MUL"
{
rtx dest = gen_reg_rtx (DImode);
rtx low = mips_subword (dest, 0);
rtx high = mips_subword (dest, 1);
emit_insn (gen_mulsi3_mul3_nohilo (low, operands[1], operands[2]));
emit_insn (gen_mulsi3_highpart_r6 (high, operands[1], operands[2]));
emit_move_insn (mips_subword (operands[0], 0), low);
emit_move_insn (mips_subword (operands[0], 1), high);
DONE;
})
(define_expand "mulsidi3_32bit_mips16"
[(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"))))]
"!TARGET_64BIT && TARGET_MIPS16"
{
rtx hilo;
hilo = gen_rtx_REG (DImode, MD_REG_FIRST);
emit_insn (gen_mulsidi3_32bit (hilo, operands[1], operands[2]));
emit_move_insn (operands[0], hilo);
DONE;
})
;; As well as being named patterns, these instructions are used by the
;; __builtin_mips_mult() functions. We must always make those functions
;; available if !TARGET_64BIT && ISA_HAS_DSP.
(define_insn "mulsidi3_32bit"
[(set (match_operand:DI 0 "muldiv_target_operand" "=ka")
(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 || ISA_HAS_DSP) && ISA_HAS_MULT"
{
if (ISA_HAS_DSP_MULT)
return "mult\t%q0,%1,%2";
else
return "mult\t%1,%2";
}
[(set_attr "type" "imul")
(set_attr "mode" "SI")])
(define_insn "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 && !ISA_HAS_DSP && ISA_HAS_MULT"
"mult\t%1,%2\;mflo\t%L0\;mfhi\t%M0"
[(set_attr "type" "imul")
(set_attr "mode" "SI")
(set_attr "insn_count" "3")])
(define_insn_and_split "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:TI 3 "=x"))
(clobber (match_scratch:DI 4 "=d"))]
"TARGET_64BIT && !TARGET_FIX_R4000 && !ISA_HAS_DMUL3
&& !TARGET_MIPS16 && ISA_HAS_MULT"
"#"
"&& reload_completed"
[(const_int 0)]
{
emit_insn (gen_mulsidi3_64bit_split (operands[0], operands[1],
operands[2], operands[4]));
DONE;
}
[(set_attr "type" "imul")
(set_attr "mode" "SI")
(set (attr "insn_count")
(if_then_else (match_test "ISA_HAS_EXT_INS")
(const_int 4)
(const_int 7)))])
(define_expand "mulsidi3_64bit_mips16"
[(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"))))]
"TARGET_64BIT && TARGET_MIPS16"
{
emit_insn (gen_mulsidi3_64bit_split (operands[0], operands[1],
operands[2], gen_reg_rtx (DImode)));
DONE;
})
(define_expand "mulsidi3_64bit_split"
[(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 (match_operand:DI 3 "register_operand"))]
""
{
rtx hilo;
hilo = gen_rtx_REG (TImode, MD_REG_FIRST);
emit_insn (gen_mulsidi3_64bit_hilo (hilo, operands[1], operands[2]));
emit_move_insn (operands[0], gen_rtx_REG (DImode, LO_REGNUM));
emit_insn (gen_mfhidi_ti (operands[3], hilo));
if (ISA_HAS_EXT_INS)
emit_insn (gen_insvdi (operands[0], GEN_INT (32), GEN_INT (32),
operands[3]));
else
{
/* Zero-extend the low part. */
mips_emit_binary (ASHIFT, operands[0], operands[0], GEN_INT (32));
mips_emit_binary (LSHIFTRT, operands[0], operands[0], GEN_INT (32));
/* Shift the high part into place. */
mips_emit_binary (ASHIFT, operands[3], operands[3], GEN_INT (32));
/* OR the two halves together. */
mips_emit_binary (IOR, operands[0], operands[0], operands[3]);
}
DONE;
})
(define_insn "mulsidi3_64bit_hilo"
[(set (match_operand:TI 0 "muldiv_target_operand" "=x")
(unspec:TI
[(mult:DI
(any_extend:DI (match_operand:SI 1 "register_operand" "d"))
(any_extend:DI (match_operand:SI 2 "register_operand" "d")))]
UNSPEC_SET_HILO))]
"TARGET_64BIT && !TARGET_FIX_R4000"
"mult\t%1,%2"
[(set_attr "type" "imul")
(set_attr "mode" "SI")])
;; See comment before the ISA_HAS_DMUL3 case in mips_mulsidi3_gen_fn.
(define_insn "mulsidi3_64bit_dmul"
[(set (match_operand:DI 0 "register_operand" "=d")
(mult:DI (sign_extend:DI (match_operand:SI 1 "register_operand" "d"))
(sign_extend:DI (match_operand:SI 2 "register_operand" "d"))))
(clobber (match_scratch:DI 3 "=l"))]
"ISA_HAS_DMUL3"
"dmul\t%0,%1,%2"
[(set_attr "type" "imul3")
(set_attr "mode" "DI")])
(define_insn "mulsidi3_64bit_r6dmul"
[(set (match_operand:DI 0 "register_operand" "=d")
(mult:DI (sign_extend:DI (match_operand:SI 1 "register_operand" "d"))
(sign_extend:DI (match_operand:SI 2 "register_operand" "d"))))]
"ISA_HAS_R6DMUL"
"dmul\t%0,%1,%2"
[(set_attr "type" "imul3nc")
(set_attr "mode" "DI")])
;; Widening multiply with negation.
(define_insn "*muls_di"
[(set (match_operand:DI 0 "muldiv_target_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\t$0,%1,%2"
[(set_attr "type" "imul")
(set_attr "mode" "SI")])
;; As well as being named patterns, these instructions are used by the
;; __builtin_mips_msub() functions. We must always make those functions
;; available if !TARGET_64BIT && ISA_HAS_DSP.
;;
;; This leads to a slight inconsistency. We honor any tuning overrides
;; in GENERATE_MADD_MSUB for -mno-dsp, but always ignore them for -mdsp,
;; even if !ISA_HAS_DSP_MULT.
(define_insn "msubsidi4"
[(set (match_operand:DI 0 "muldiv_target_operand" "=ka")
(minus:DI
(match_operand:DI 3 "muldiv_target_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 || GENERATE_MADD_MSUB || ISA_HAS_DSP)"
{
if (ISA_HAS_DSP_MULT)
return "msub\t%q0,%1,%2";
else if (TARGET_MIPS5500 || GENERATE_MADD_MSUB)
return "msub\t%1,%2";
else
return "msac\t$0,%1,%2";
}
[(set_attr "type" "imadd")
(set_attr "accum_in" "3")
(set_attr "mode" "SI")])
;; _highpart patterns
(define_expand "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))))]
""
{
if (ISA_HAS_MULHI)
emit_insn (gen_mulsi3_highpart_mulhi_internal (operands[0],
operands[1],
operands[2]));
else if (TARGET_MIPS16)
emit_insn (gen_mulsi3_highpart_split (operands[0], operands[1],
operands[2]));
else if (ISA_HAS_R6MUL)
emit_insn (gen_mulsi3_highpart_r6 (operands[0], operands[1],
operands[2]));
else
emit_insn (gen_mulsi3_highpart_internal (operands[0], operands[1],
operands[2]));
DONE;
})
(define_insn "mulsi3_highpart_r6"
[(set (match_operand:SI 0 "register_operand" "=d")
(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))))]
"ISA_HAS_R6MUL"
"muh\t%0,%1,%2"
[(set_attr "type" "imul3nc")
(set_attr "mode" "SI")])
(define_insn_and_split "mulsi3_highpart_internal"
[(set (match_operand:SI 0 "register_operand" "=d")
(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_MULT && !ISA_HAS_MULHI && !TARGET_MIPS16"
{ return TARGET_FIX_R4000 ? "mult\t%1,%2\n\tmfhi\t%0" : "#"; }
"&& reload_completed && !TARGET_FIX_R4000"
[(const_int 0)]
{
emit_insn (gen_mulsi3_highpart_split (operands[0], operands[1],
operands[2]));
DONE;
}
[(set_attr "type" "imul")
(set_attr "mode" "SI")
(set_attr "insn_count" "2")])
(define_expand "mulsi3_highpart_split"
[(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))))]
""
{
rtx hilo;
if (TARGET_64BIT)
{
hilo = gen_rtx_REG (TImode, MD_REG_FIRST);
emit_insn (gen_mulsidi3_64bit_hilo (hilo, operands[1], operands[2]));
emit_insn (gen_mfhisi_ti (operands[0], hilo));
}
else
{
hilo = gen_rtx_REG (DImode, MD_REG_FIRST);
emit_insn (gen_mulsidi3_32bit (hilo, operands[1], operands[2]));
emit_insn (gen_mfhisi_di (operands[0], hilo));
}
DONE;
})
(define_insn "mulsi3_highpart_mulhi_internal"
[(set (match_operand:SI 0 "register_operand" "=d")
(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"
"mulhi\t%0,%1,%2"
[(set_attr "type" "imul3")
(set_attr "mode" "SI")])
(define_insn "*mulsi3_highpart_neg_mulhi_internal"
[(set (match_operand:SI 0 "register_operand" "=d")
(truncate:SI
(lshiftrt:DI
(neg: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"
"mulshi\t%0,%1,%2"
[(set_attr "type" "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_expand "muldi3_highpart"
[(set (match_operand:DI 0 "register_operand")
(truncate:DI
(lshiftrt:TI
(mult:TI (any_extend:TI (match_operand:DI 1 "register_operand"))
(any_extend:TI (match_operand:DI 2 "register_operand")))
(const_int 64))))]
"ISA_HAS_R6DMUL
|| (ISA_HAS_DMULT
&& !( == ZERO_EXTEND && TARGET_FIX_VR4120))"
{
if (TARGET_MIPS16)
emit_insn (gen_muldi3_highpart_split (operands[0], operands[1],
operands[2]));
else if (ISA_HAS_R6DMUL)
emit_insn (gen_muldi3_highpart_r6 (operands[0], operands[1],
operands[2]));
else
emit_insn (gen_muldi3_highpart_internal (operands[0], operands[1],
operands[2]));
DONE;
})
(define_insn "muldi3_highpart_r6"
[(set (match_operand:DI 0 "register_operand" "=d")
(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))))]
"ISA_HAS_R6DMUL"
"dmuh\t%0,%1,%2"
[(set_attr "type" "imul3nc")
(set_attr "mode" "DI")])
(define_insn_and_split "muldi3_highpart_internal"
[(set (match_operand:DI 0 "register_operand" "=d")
(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"))]
"ISA_HAS_DMULT
&& !TARGET_MIPS16
&& !( == ZERO_EXTEND && TARGET_FIX_VR4120)"
{ return TARGET_FIX_R4000 ? "dmult\t%1,%2\n\tmfhi\t%0" : "#"; }
"&& reload_completed && !TARGET_FIX_R4000"
[(const_int 0)]
{
emit_insn (gen_muldi3_highpart_split (operands[0], operands[1],
operands[2]));
DONE;
}
[(set_attr "type" "imul")
(set_attr "mode" "DI")
(set_attr "insn_count" "2")])
(define_expand "muldi3_highpart_split"
[(set (match_operand:DI 0 "register_operand")
(truncate:DI
(lshiftrt:TI
(mult:TI (any_extend:TI (match_operand:DI 1 "register_operand"))
(any_extend:TI (match_operand:DI 2 "register_operand")))
(const_int 64))))]
""
{
rtx hilo;
hilo = gen_rtx_REG (TImode, MD_REG_FIRST);
emit_insn (gen_mulditi3_internal (hilo, operands[1], operands[2]));
emit_insn (gen_mfhidi_ti (operands[0], hilo));
DONE;
})
(define_expand "mulditi3"
[(set (match_operand:TI 0 "register_operand")
(mult:TI (any_extend:TI (match_operand:DI 1 "register_operand"))
(any_extend:TI (match_operand:DI 2 "register_operand"))))]
"ISA_HAS_R6DMUL
|| (ISA_HAS_DMULT
&& !( == ZERO_EXTEND && TARGET_FIX_VR4120))"
{
rtx hilo, hi, lo;
if (TARGET_MIPS16)
{
hilo = gen_rtx_REG (TImode, MD_REG_FIRST);
emit_insn (gen_mulditi3_internal (hilo, operands[1], operands[2]));
emit_move_insn (operands[0], hilo);
}
else if (TARGET_FIX_R4000)
emit_insn (gen_mulditi3_r4000 (operands[0], operands[1], operands[2]));
else if (ISA_HAS_DMULT)
emit_insn (gen_mulditi3_internal (operands[0], operands[1],
operands[2]));
else
{
hi = mips_subword (operands[0], 1);
lo = mips_subword (operands[0], 0);
emit_insn (gen_muldi3_mul3_nohilo (lo, operands[1], operands[2]));
emit_insn (gen_muldi3_highpart_r6 (hi, operands[1], operands[2]));
}
DONE;
})
(define_insn "mulditi3_internal"
[(set (match_operand:TI 0 "muldiv_target_operand" "=x")
(mult:TI (any_extend:TI (match_operand:DI 1 "register_operand" "d"))
(any_extend:TI (match_operand:DI 2 "register_operand" "d"))))]
"ISA_HAS_DMULT
&& !TARGET_FIX_R4000
&& !( == ZERO_EXTEND && TARGET_FIX_VR4120)"
"dmult\t%1,%2"
[(set_attr "type" "imul")
(set_attr "mode" "DI")])
(define_insn "mulditi3_r4000"
[(set (match_operand:TI 0 "register_operand" "=d")
(mult:TI (any_extend:TI (match_operand:DI 1 "register_operand" "d"))
(any_extend:TI (match_operand:DI 2 "register_operand" "d"))))
(clobber (match_scratch:TI 3 "=x"))]
"ISA_HAS_DMULT
&& TARGET_FIX_R4000
&& !( == ZERO_EXTEND && TARGET_FIX_VR4120)"
"dmult\t%1,%2\;mflo\t%L0\;mfhi\t%M0"
[(set_attr "type" "imul")
(set_attr "mode" "DI")
(set_attr "insn_count" "3")])
;; 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)))]
"TARGET_MAD"
"mad\t%1,%2"
[(set_attr "type" "imadd")
(set_attr "accum_in" "0")
(set_attr "mode" "SI")])
;; See the comment above msubsidi4 for the relationship between
;; ISA_HAS_DSP and ISA_HAS_DSP_MULT.
(define_insn "maddsidi4"
[(set (match_operand:DI 0 "muldiv_target_operand" "=ka")
(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 "muldiv_target_operand" "0")))]
"(TARGET_MAD || ISA_HAS_MACC || GENERATE_MADD_MSUB || ISA_HAS_DSP)
&& !TARGET_64BIT"
{
if (TARGET_MAD)
return "mad\t%1,%2";
else if (ISA_HAS_DSP_MULT)
return "madd\t%q0,%1,%2";
else if (GENERATE_MADD_MSUB || TARGET_MIPS5500)
return "madd\t%1,%2";
else
/* See comment in *macc. */
return "%[macc\t%@,%1,%2%]";
}
[(set_attr "type" "imadd")
(set_attr "accum_in" "3")
(set_attr "mode" "SI")])
;; Floating point multiply accumulate instructions.
(define_expand "fma4"
[(set (match_operand:ANYF 0 "register_operand")
(fma:ANYF (match_operand:ANYF 1 "register_operand")
(match_operand:ANYF 2 "register_operand")
(match_operand:ANYF 3 "register_operand")))]
"ISA_HAS_FUSED_MADDF || ISA_HAS_FUSED_MADD3 || ISA_HAS_FUSED_MADD4")
(define_insn "*fma4_madd3"
[(set (match_operand:ANYF 0 "register_operand" "=f")
(fma:ANYF (match_operand:ANYF 1 "register_operand" "f")
(match_operand:ANYF 2 "register_operand" "f")
(match_operand:ANYF 3 "register_operand" "0")))]
"ISA_HAS_FUSED_MADD3"
"madd.\t%0,%1,%2"
[(set_attr "type" "fmadd")
(set_attr "mode" "")])
(define_insn "*fma4_madd4"
[(set (match_operand:ANYF 0 "register_operand" "=f")
(fma:ANYF (match_operand:ANYF 1 "register_operand" "f")
(match_operand:ANYF 2 "register_operand" "f")
(match_operand:ANYF 3 "register_operand" "f")))]
"ISA_HAS_FUSED_MADD4"
"madd.\t%0,%3,%1,%2"
[(set_attr "type" "fmadd")
(set_attr "mode" "")])
(define_insn "*fma4_maddf"
[(set (match_operand:ANYF 0 "register_operand" "=f")
(fma:ANYF (match_operand:ANYF 1 "register_operand" "f")
(match_operand:ANYF 2 "register_operand" "f")
(match_operand:ANYF 3 "register_operand" "0")))]
"ISA_HAS_FUSED_MADDF"
"maddf.\t%0,%1,%2"
[(set_attr "type" "fmadd")
(set_attr "mode" "")])
;; The fms, fnma, and fnms instructions can be used even when HONOR_NANS
;; is true because while IEEE 754-2008 requires the negate operation to
;; negate the sign of a NAN and the MIPS neg instruction does not do this,
;; the fma part of the instruction has no requirement on how the sign of
;; a NAN is handled and so the final sign bit of the entire operation is
;; undefined.
(define_expand "fms4"
[(set (match_operand:ANYF 0 "register_operand")
(fma:ANYF (match_operand:ANYF 1 "register_operand")
(match_operand:ANYF 2 "register_operand")
(neg:ANYF (match_operand:ANYF 3 "register_operand"))))]
"(ISA_HAS_FUSED_MADD3 || ISA_HAS_FUSED_MADD4)")
(define_insn "*fms4_msub3"
[(set (match_operand:ANYF 0 "register_operand" "=f")
(fma:ANYF (match_operand:ANYF 1 "register_operand" "f")
(match_operand:ANYF 2 "register_operand" "f")
(neg:ANYF (match_operand:ANYF 3 "register_operand" "0"))))]
"ISA_HAS_FUSED_MADD3"
"msub.\t%0,%1,%2"
[(set_attr "type" "fmadd")
(set_attr "mode" "")])
(define_insn "*fms4_msub4"
[(set (match_operand:ANYF 0 "register_operand" "=f")
(fma:ANYF (match_operand:ANYF 1 "register_operand" "f")
(match_operand:ANYF 2 "register_operand" "f")
(neg:ANYF (match_operand:ANYF 3 "register_operand" "f"))))]
"ISA_HAS_FUSED_MADD4"
"msub.\t%0,%3,%1,%2"
[(set_attr "type" "fmadd")
(set_attr "mode" "")])
;; fnma is defined in GCC as (fma (neg op1) op2 op3)
;; (-op1 * op2) + op3 ==> -(op1 * op2) + op3 ==> -((op1 * op2) - op3)
;; The mips nmsub instructions implement -((op1 * op2) - op3)
;; This transformation means we may return the wrong signed zero
;; so we check HONOR_SIGNED_ZEROS.
(define_expand "fnma4"
[(set (match_operand:ANYF 0 "register_operand")
(fma:ANYF (neg:ANYF (match_operand:ANYF 1 "register_operand"))
(match_operand:ANYF 2 "register_operand")
(match_operand:ANYF 3 "register_operand")))]
"(ISA_HAS_FUSED_MADD3 || ISA_HAS_FUSED_MADD4)
&& !HONOR_SIGNED_ZEROS (mode)")
(define_insn "*fnma4_nmsub3"
[(set (match_operand:ANYF 0 "register_operand" "=f")
(fma:ANYF (neg:ANYF (match_operand:ANYF 1 "register_operand" "f"))
(match_operand:ANYF 2 "register_operand" "f")
(match_operand:ANYF 3 "register_operand" "0")))]
"ISA_HAS_FUSED_MADD3 && !HONOR_SIGNED_ZEROS (mode)"
"nmsub.\t%0,%1,%2"
[(set_attr "type" "fmadd")
(set_attr "mode" "")])
(define_insn "*fnma4_nmsub4"
[(set (match_operand:ANYF 0 "register_operand" "=f")
(fma: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_FUSED_MADD4 && !HONOR_SIGNED_ZEROS (mode)"
"nmsub.\t%0,%3,%1,%2"
[(set_attr "type" "fmadd")
(set_attr "mode" "")])
;; fnms is defined as: (fma (neg op1) op2 (neg op3))
;; ((-op1) * op2) - op3 ==> -(op1 * op2) - op3 ==> -((op1 * op2) + op3)
;; The mips nmadd instructions implement -((op1 * op2) + op3)
;; This transformation means we may return the wrong signed zero
;; so we check HONOR_SIGNED_ZEROS.
(define_expand "fnms4"
[(set (match_operand:ANYF 0 "register_operand")
(fma:ANYF
(neg:ANYF (match_operand:ANYF 1 "register_operand"))
(match_operand:ANYF 2 "register_operand")
(neg:ANYF (match_operand:ANYF 3 "register_operand"))))]
"(ISA_HAS_FUSED_MADD3 || ISA_HAS_FUSED_MADD4)
&& !HONOR_SIGNED_ZEROS (mode)")
(define_insn "*fnms4_nmadd3"
[(set (match_operand:ANYF 0 "register_operand" "=f")
(fma:ANYF
(neg:ANYF (match_operand:ANYF 1 "register_operand" "f"))
(match_operand:ANYF 2 "register_operand" "f")
(neg:ANYF (match_operand:ANYF 3 "register_operand" "0"))))]
"ISA_HAS_FUSED_MADD3 && !HONOR_SIGNED_ZEROS (mode)"
"nmadd.\t%0,%1,%2"
[(set_attr "type" "fmadd")
(set_attr "mode" "")])
(define_insn "*fnms4_nmadd4"
[(set (match_operand:ANYF 0 "register_operand" "=f")
(fma:ANYF
(neg:ANYF (match_operand:ANYF 1 "register_operand" "f"))
(match_operand:ANYF 2 "register_operand" "f")
(neg:ANYF (match_operand:ANYF 3 "register_operand" "f"))))]
"ISA_HAS_FUSED_MADD4 && !HONOR_SIGNED_ZEROS (mode)"
"nmadd.\t%0,%3,%1,%2"
[(set_attr "type" "fmadd")
(set_attr "mode" "")])
;; Non-fused Floating point multiply accumulate instructions.
;; These instructions are not fused and round in between the multiply
;; and the add (or subtract) so they are equivalent to the separate
;; multiply and add/sub instructions.
(define_insn "*madd4"
[(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_UNFUSED_MADD4"
"madd.\t%0,%3,%1,%2"
[(set_attr "type" "fmadd")
(set_attr "mode" "")])
(define_insn "*msub4"
[(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_UNFUSED_MADD4"
"msub.\t%0,%3,%1,%2"
[(set_attr "type" "fmadd")
(set_attr "mode" "")])
;; Like with the fused fms, fnma, and fnms instructions, these unfused
;; instructions can be used even if HONOR_NANS is set because while
;; IEEE 754-2008 requires the negate operation to negate the sign of a
;; NAN and the MIPS neg instruction does not do this, the multiply and
;; add (or subtract) part of the instruction has no requirement on how
;; the sign of a NAN is handled and so the final sign bit of the entire
;; operation is undefined.
(define_insn "*nmadd4"
[(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_UNFUSED_MADD4"
"nmadd.\t%0,%3,%1,%2"
[(set_attr "type" "fmadd")
(set_attr "mode" "")])
(define_insn "*nmsub4"
[(set (match_operand:ANYF 0 "register_operand" "=f")
(neg:ANYF (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_UNFUSED_MADD4"
"nmsub.\t%0,%3,%1,%2"
[(set_attr "type" "fmadd")
(set_attr "mode" "")])
;; Fast-math Non-fused Floating point multiply accumulate instructions.
;; These instructions are not fused but the expressions they match are
;; not exactly what the instruction implements in the sense that they
;; may not generate the properly signed zeros.
;; This instruction recognizes ((-op1) * op2) - op3 and generates an
;; nmadd which is really -((op1 * op2) + op3). They are equivalent
;; except for the sign bit when the result is zero or NaN.
(define_insn "*nmadd4_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_UNFUSED_MADD4
&& !HONOR_SIGNED_ZEROS (mode)"
"nmadd.\t%0,%3,%1,%2"
[(set_attr "type" "fmadd")
(set_attr "mode" "")])
;; This instruction recognizes (op1 - (op2 * op3) and generates an
;; nmsub which is really -((op2 * op3) - op1). They are equivalent
;; except for the sign bit when the result is zero or NaN.
(define_insn "*nmsub4_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_UNFUSED_MADD4
&& !HONOR_SIGNED_ZEROS (mode)"
"nmsub.\t%0,%1,%2,%3"
[(set_attr "type" "fmadd")
(set_attr "mode" "")])
;;
;; ....................
;;
;; DIVISION and REMAINDER
;;
;; ....................
;;
(define_expand "div3"
[(set (match_operand:ANYF 0 "register_operand")
(div:ANYF (match_operand:ANYF 1 "reg_or_1_operand")
(match_operand:ANYF 2 "register_operand")))]
""
{
if (const_1_operand (operands[1], mode))
if (!(ISA_HAS_FP_RECIP_RSQRT (mode)
&& flag_unsafe_math_optimizations))
operands[1] = force_reg (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 "*div3"
[(set (match_operand:ANYF 0 "register_operand" "=f")
(div:ANYF (match_operand:ANYF 1 "register_operand" "f")
(match_operand:ANYF 2 "register_operand" "f")))]
""
{
if (TARGET_FIX_SB1)
return "div.\t%0,%1,%2\;mov.\t%0,%0";
else
return "div.\t%0,%1,%2";
}
[(set_attr "type" "fdiv")
(set_attr "mode" "")
(set (attr "insn_count")
(if_then_else (match_test "TARGET_FIX_SB1")
(const_int 2)
(const_int 1)))])
(define_insn "*recip3"
[(set (match_operand:ANYF 0 "register_operand" "=f")
(div:ANYF (match_operand:ANYF 1 "const_1_operand" "")
(match_operand:ANYF 2 "register_operand" "f")))]
"ISA_HAS_FP_RECIP_RSQRT (mode) && flag_unsafe_math_optimizations"
{
if (TARGET_FIX_SB1)
return "recip.\t%0,%2\;mov.\t%0,%0";
else
return "recip.\t%0,%2";
}
[(set_attr "type" "frdiv")
(set_attr "mode" "")
(set (attr "insn_count")
(if_then_else (match_test "TARGET_FIX_SB1")
(const_int 2)
(const_int 1)))])
;; VR4120 errata MD(A1): signed division instructions do not work correctly
;; with negative operands. We use special libgcc functions instead.
(define_expand "divmod4"
[(parallel
[(set (match_operand:GPR 0 "register_operand")
(div:GPR (match_operand:GPR 1 "register_operand")
(match_operand:GPR 2 "register_operand")))
(set (match_operand:GPR 3 "register_operand")
(mod:GPR (match_dup 1)
(match_dup 2)))])]
"ISA_HAS_DIV && !TARGET_FIX_VR4120"
{
if (TARGET_MIPS16)
{
rtx lo = gen_rtx_REG (mode, LO_REGNUM);
emit_insn (gen_divmod4_mips16 (operands[0], operands[1],
operands[2], operands[3], lo));
DONE;
}
})
(define_insn_and_split "*divmod4"
[(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" "=d")
(mod:GPR (match_dup 1)
(match_dup 2)))]
"ISA_HAS_DIV && !TARGET_FIX_VR4120 && !TARGET_MIPS16"
"#"
"&& reload_completed"
[(const_int 0)]
{
emit_insn (gen_divmod4_split (operands[3], operands[1], operands[2]));
DONE;
}
[(set_attr "type" "idiv")
(set_attr "mode" "")
(set_attr "insn_count" "2")])
;; Expand generates divmod instructions for individual division and modulus
;; operations. We then rely on CSE to reuse earlier divmods where possible.
;; This means that, when generating MIPS16 code, it is better not to expose
;; the fixed LO register until after CSE has finished. However, it's still
;; better to split before register allocation, so that we don't allocate
;; one of the scarce MIPS16 registers to an unused result.
(define_insn_and_split "divmod4_mips16"
[(set (match_operand:GPR 0 "register_operand" "=d")
(div:GPR (match_operand:GPR 1 "register_operand" "d")
(match_operand:GPR 2 "register_operand" "d")))
(set (match_operand:GPR 3 "register_operand" "=d")
(mod:GPR (match_dup 1)
(match_dup 2)))
(clobber (match_operand:GPR 4 "lo_operand" "=l"))]
"ISA_HAS_DIV && !TARGET_FIX_VR4120 && TARGET_MIPS16"
"#"
"&& cse_not_expected"
[(const_int 0)]
{
emit_insn (gen_divmod4_split (operands[3], operands[1], operands[2]));
emit_move_insn (operands[0], operands[4]);
DONE;
}
[(set_attr "type" "idiv")
(set_attr "mode" "")
(set_attr "insn_count" "3")])
(define_expand "udivmod4"
[(parallel
[(set (match_operand:GPR 0 "register_operand")
(udiv:GPR (match_operand:GPR 1 "register_operand")
(match_operand:GPR 2 "register_operand")))
(set (match_operand:GPR 3 "register_operand")
(umod:GPR (match_dup 1)
(match_dup 2)))])]
"ISA_HAS_DIV && !TARGET_FIX_VR4120"
{
if (TARGET_MIPS16)
{
rtx lo = gen_rtx_REG (mode, LO_REGNUM);
emit_insn (gen_udivmod4_mips16 (operands[0], operands[1],
operands[2], operands[3], lo));
DONE;
}
})
(define_insn_and_split "*udivmod4"
[(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" "=d")
(umod:GPR (match_dup 1)
(match_dup 2)))]
"ISA_HAS_DIV && !TARGET_MIPS16"
"#"
"reload_completed"
[(const_int 0)]
{
emit_insn (gen_udivmod4_split (operands[3], operands[1], operands[2]));
DONE;
}
[(set_attr "type" "idiv")
(set_attr "mode" "