llvm-6502/lib/Target/PowerPC/PPCInstrInfo.td
Chris Lattner 54e853b8a6 Rewrite the branch selector to be correct in the face of large functions.
The algorithm it used before wasn't 100% correct, we now use an iterative
expansion model.  This fixes assembler errors when compiling 403.gcc with
tail merging enabled.

Change the way the branch selector works overall: Now, the isel generates
PPC::BCC instructions (as it used to) directly, and these BCC instructions
are emitted to the output or jitted directly if branches don't need
expansion.  Only if branches need expansion are instructions rewritten
and created.  This should make branch select faster, and eliminates the
Bxx instructions from the .td file.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@31837 91177308-0d34-0410-b5e6-96231b3b80d8
2006-11-18 00:32:03 +00:00

1127 lines
50 KiB
TableGen

//===- PPCInstrInfo.td - The PowerPC Instruction Set -------*- tablegen -*-===//
//
// The LLVM Compiler Infrastructure
//
// This file was developed by the LLVM research group and is distributed under
// the University of Illinois Open Source License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file describes the subset of the 32-bit PowerPC instruction set, as used
// by the PowerPC instruction selector.
//
//===----------------------------------------------------------------------===//
include "PPCInstrFormats.td"
//===----------------------------------------------------------------------===//
// PowerPC specific type constraints.
//
def SDT_PPCstfiwx : SDTypeProfile<0, 2, [ // stfiwx
SDTCisVT<0, f64>, SDTCisPtrTy<1>
]>;
def SDT_PPCShiftOp : SDTypeProfile<1, 2, [ // PPCshl, PPCsra, PPCsrl
SDTCisVT<0, i32>, SDTCisVT<1, i32>, SDTCisVT<2, i32>
]>;
def SDT_PPCCallSeq : SDTypeProfile<0, 1, [ SDTCisVT<0, i32> ]>;
def SDT_PPCvperm : SDTypeProfile<1, 3, [
SDTCisVT<3, v16i8>, SDTCisSameAs<0, 1>, SDTCisSameAs<0, 2>
]>;
def SDT_PPCvcmp : SDTypeProfile<1, 3, [
SDTCisSameAs<0, 1>, SDTCisSameAs<1, 2>, SDTCisVT<3, i32>
]>;
def SDT_PPCcondbr : SDTypeProfile<0, 3, [
SDTCisVT<0, i32>, SDTCisVT<2, OtherVT>
]>;
def SDT_PPClbrx : SDTypeProfile<1, 3, [
SDTCisVT<0, i32>, SDTCisPtrTy<1>, SDTCisVT<2, OtherVT>, SDTCisVT<3, OtherVT>
]>;
def SDT_PPCstbrx : SDTypeProfile<0, 4, [
SDTCisVT<0, i32>, SDTCisPtrTy<1>, SDTCisVT<2, OtherVT>, SDTCisVT<3, OtherVT>
]>;
//===----------------------------------------------------------------------===//
// PowerPC specific DAG Nodes.
//
def PPCfcfid : SDNode<"PPCISD::FCFID" , SDTFPUnaryOp, []>;
def PPCfctidz : SDNode<"PPCISD::FCTIDZ", SDTFPUnaryOp, []>;
def PPCfctiwz : SDNode<"PPCISD::FCTIWZ", SDTFPUnaryOp, []>;
def PPCstfiwx : SDNode<"PPCISD::STFIWX", SDT_PPCstfiwx, [SDNPHasChain]>;
def PPCfsel : SDNode<"PPCISD::FSEL",
// Type constraint for fsel.
SDTypeProfile<1, 3, [SDTCisSameAs<0, 2>, SDTCisSameAs<0, 3>,
SDTCisFP<0>, SDTCisVT<1, f64>]>, []>;
def PPChi : SDNode<"PPCISD::Hi", SDTIntBinOp, []>;
def PPClo : SDNode<"PPCISD::Lo", SDTIntBinOp, []>;
def PPCvmaddfp : SDNode<"PPCISD::VMADDFP", SDTFPTernaryOp, []>;
def PPCvnmsubfp : SDNode<"PPCISD::VNMSUBFP", SDTFPTernaryOp, []>;
def PPCvperm : SDNode<"PPCISD::VPERM", SDT_PPCvperm, []>;
// These nodes represent the 32-bit PPC shifts that operate on 6-bit shift
// amounts. These nodes are generated by the multi-precision shift code.
def PPCsrl : SDNode<"PPCISD::SRL" , SDT_PPCShiftOp>;
def PPCsra : SDNode<"PPCISD::SRA" , SDT_PPCShiftOp>;
def PPCshl : SDNode<"PPCISD::SHL" , SDT_PPCShiftOp>;
def PPCextsw_32 : SDNode<"PPCISD::EXTSW_32" , SDTIntUnaryOp>;
def PPCstd_32 : SDNode<"PPCISD::STD_32" , SDTStore, [SDNPHasChain]>;
// These are target-independent nodes, but have target-specific formats.
def callseq_start : SDNode<"ISD::CALLSEQ_START", SDT_PPCCallSeq,
[SDNPHasChain, SDNPOutFlag]>;
def callseq_end : SDNode<"ISD::CALLSEQ_END", SDT_PPCCallSeq,
[SDNPHasChain, SDNPOutFlag]>;
def SDT_PPCCall : SDTypeProfile<0, -1, [SDTCisInt<0>]>;
def PPCcall : SDNode<"PPCISD::CALL", SDT_PPCCall,
[SDNPHasChain, SDNPOptInFlag, SDNPOutFlag]>;
def PPCmtctr : SDNode<"PPCISD::MTCTR", SDT_PPCCall,
[SDNPHasChain, SDNPOptInFlag, SDNPOutFlag]>;
def PPCbctrl : SDNode<"PPCISD::BCTRL", SDTRet,
[SDNPHasChain, SDNPOptInFlag, SDNPOutFlag]>;
def retflag : SDNode<"PPCISD::RET_FLAG", SDTRet,
[SDNPHasChain, SDNPOptInFlag]>;
def PPCvcmp : SDNode<"PPCISD::VCMP" , SDT_PPCvcmp, []>;
def PPCvcmp_o : SDNode<"PPCISD::VCMPo", SDT_PPCvcmp, [SDNPOutFlag]>;
def PPCcondbranch : SDNode<"PPCISD::COND_BRANCH", SDT_PPCcondbr,
[SDNPHasChain, SDNPOptInFlag]>;
def PPClbrx : SDNode<"PPCISD::LBRX", SDT_PPClbrx, [SDNPHasChain]>;
def PPCstbrx : SDNode<"PPCISD::STBRX", SDT_PPCstbrx, [SDNPHasChain]>;
// Instructions to support dynamic alloca.
def SDTDynOp : SDTypeProfile<1, 2, []>;
def PPCdynalloc : SDNode<"PPCISD::DYNALLOC", SDTDynOp, [SDNPHasChain]>;
//===----------------------------------------------------------------------===//
// PowerPC specific transformation functions and pattern fragments.
//
def SHL32 : SDNodeXForm<imm, [{
// Transformation function: 31 - imm
return getI32Imm(31 - N->getValue());
}]>;
def SRL32 : SDNodeXForm<imm, [{
// Transformation function: 32 - imm
return N->getValue() ? getI32Imm(32 - N->getValue()) : getI32Imm(0);
}]>;
def LO16 : SDNodeXForm<imm, [{
// Transformation function: get the low 16 bits.
return getI32Imm((unsigned short)N->getValue());
}]>;
def HI16 : SDNodeXForm<imm, [{
// Transformation function: shift the immediate value down into the low bits.
return getI32Imm((unsigned)N->getValue() >> 16);
}]>;
def HA16 : SDNodeXForm<imm, [{
// Transformation function: shift the immediate value down into the low bits.
signed int Val = N->getValue();
return getI32Imm((Val - (signed short)Val) >> 16);
}]>;
def MB : SDNodeXForm<imm, [{
// Transformation function: get the start bit of a mask
unsigned mb, me;
(void)isRunOfOnes((unsigned)N->getValue(), mb, me);
return getI32Imm(mb);
}]>;
def ME : SDNodeXForm<imm, [{
// Transformation function: get the end bit of a mask
unsigned mb, me;
(void)isRunOfOnes((unsigned)N->getValue(), mb, me);
return getI32Imm(me);
}]>;
def maskimm32 : PatLeaf<(imm), [{
// maskImm predicate - True if immediate is a run of ones.
unsigned mb, me;
if (N->getValueType(0) == MVT::i32)
return isRunOfOnes((unsigned)N->getValue(), mb, me);
else
return false;
}]>;
def immSExt16 : PatLeaf<(imm), [{
// immSExt16 predicate - True if the immediate fits in a 16-bit sign extended
// field. Used by instructions like 'addi'.
if (N->getValueType(0) == MVT::i32)
return (int32_t)N->getValue() == (short)N->getValue();
else
return (int64_t)N->getValue() == (short)N->getValue();
}]>;
def immZExt16 : PatLeaf<(imm), [{
// immZExt16 predicate - True if the immediate fits in a 16-bit zero extended
// field. Used by instructions like 'ori'.
return (uint64_t)N->getValue() == (unsigned short)N->getValue();
}], LO16>;
// imm16Shifted* - These match immediates where the low 16-bits are zero. There
// are two forms: imm16ShiftedSExt and imm16ShiftedZExt. These two forms are
// identical in 32-bit mode, but in 64-bit mode, they return true if the
// immediate fits into a sign/zero extended 32-bit immediate (with the low bits
// clear).
def imm16ShiftedZExt : PatLeaf<(imm), [{
// imm16ShiftedZExt predicate - True if only bits in the top 16-bits of the
// immediate are set. Used by instructions like 'xoris'.
return (N->getValue() & ~uint64_t(0xFFFF0000)) == 0;
}], HI16>;
def imm16ShiftedSExt : PatLeaf<(imm), [{
// imm16ShiftedSExt predicate - True if only bits in the top 16-bits of the
// immediate are set. Used by instructions like 'addis'. Identical to
// imm16ShiftedZExt in 32-bit mode.
if (N->getValue() & 0xFFFF) return false;
if (N->getValueType(0) == MVT::i32)
return true;
// For 64-bit, make sure it is sext right.
return N->getValue() == (uint64_t)(int)N->getValue();
}], HI16>;
//===----------------------------------------------------------------------===//
// PowerPC Flag Definitions.
class isPPC64 { bit PPC64 = 1; }
class isDOT {
list<Register> Defs = [CR0];
bit RC = 1;
}
class RegConstraint<string C> {
string Constraints = C;
}
class NoEncode<string E> {
string DisableEncoding = E;
}
//===----------------------------------------------------------------------===//
// PowerPC Operand Definitions.
def s5imm : Operand<i32> {
let PrintMethod = "printS5ImmOperand";
}
def u5imm : Operand<i32> {
let PrintMethod = "printU5ImmOperand";
}
def u6imm : Operand<i32> {
let PrintMethod = "printU6ImmOperand";
}
def s16imm : Operand<i32> {
let PrintMethod = "printS16ImmOperand";
}
def u16imm : Operand<i32> {
let PrintMethod = "printU16ImmOperand";
}
def s16immX4 : Operand<i32> { // Multiply imm by 4 before printing.
let PrintMethod = "printS16X4ImmOperand";
}
def target : Operand<OtherVT> {
let PrintMethod = "printBranchOperand";
}
def calltarget : Operand<iPTR> {
let PrintMethod = "printCallOperand";
}
def aaddr : Operand<iPTR> {
let PrintMethod = "printAbsAddrOperand";
}
def piclabel: Operand<iPTR> {
let PrintMethod = "printPICLabel";
}
def symbolHi: Operand<i32> {
let PrintMethod = "printSymbolHi";
}
def symbolLo: Operand<i32> {
let PrintMethod = "printSymbolLo";
}
def crbitm: Operand<i8> {
let PrintMethod = "printcrbitm";
}
// Address operands
def memri : Operand<iPTR> {
let PrintMethod = "printMemRegImm";
let MIOperandInfo = (ops i32imm:$imm, ptr_rc:$reg);
}
def memrr : Operand<iPTR> {
let PrintMethod = "printMemRegReg";
let MIOperandInfo = (ops ptr_rc, ptr_rc);
}
def memrix : Operand<iPTR> { // memri where the imm is shifted 2 bits.
let PrintMethod = "printMemRegImmShifted";
let MIOperandInfo = (ops i32imm:$imm, ptr_rc:$reg);
}
// PowerPC Predicate operand. 20 = (0<<5)|20 = always, CR0 is a dummy reg
// that doesn't matter.
def pred : PredicateOperand<(ops imm, CRRC), (ops (i32 20), CR0)> {
let PrintMethod = "printPredicateOperand";
}
// Define PowerPC specific addressing mode.
def iaddr : ComplexPattern<iPTR, 2, "SelectAddrImm", [], []>;
def xaddr : ComplexPattern<iPTR, 2, "SelectAddrIdx", [], []>;
def xoaddr : ComplexPattern<iPTR, 2, "SelectAddrIdxOnly",[], []>;
def ixaddr : ComplexPattern<iPTR, 2, "SelectAddrImmShift", [], []>; // "std"
/// This is just the offset part of iaddr, used for preinc.
def iaddroff : ComplexPattern<iPTR, 1, "SelectAddrImmOffs", [], []>;
//===----------------------------------------------------------------------===//
// PowerPC Instruction Predicate Definitions.
def FPContractions : Predicate<"!NoExcessFPPrecision">;
//===----------------------------------------------------------------------===//
// PowerPC Instruction Definitions.
// Pseudo-instructions:
let hasCtrlDep = 1 in {
def ADJCALLSTACKDOWN : Pseudo<(ops u16imm:$amt),
"${:comment} ADJCALLSTACKDOWN",
[(callseq_start imm:$amt)]>, Imp<[R1],[R1]>;
def ADJCALLSTACKUP : Pseudo<(ops u16imm:$amt),
"${:comment} ADJCALLSTACKUP",
[(callseq_end imm:$amt)]>, Imp<[R1],[R1]>;
def UPDATE_VRSAVE : Pseudo<(ops GPRC:$rD, GPRC:$rS),
"UPDATE_VRSAVE $rD, $rS", []>;
}
def DYNALLOC : Pseudo<(ops GPRC:$result, GPRC:$negsize, memri:$fpsi),
"${:comment} DYNALLOC $result, $negsize, $fpsi",
[(set GPRC:$result,
(PPCdynalloc GPRC:$negsize, iaddr:$fpsi))]>,
Imp<[R1],[R1]>;
def IMPLICIT_DEF_GPRC: Pseudo<(ops GPRC:$rD),"${:comment}IMPLICIT_DEF_GPRC $rD",
[(set GPRC:$rD, (undef))]>;
def IMPLICIT_DEF_F8 : Pseudo<(ops F8RC:$rD), "${:comment} IMPLICIT_DEF_F8 $rD",
[(set F8RC:$rD, (undef))]>;
def IMPLICIT_DEF_F4 : Pseudo<(ops F4RC:$rD), "${:comment} IMPLICIT_DEF_F4 $rD",
[(set F4RC:$rD, (undef))]>;
// SELECT_CC_* - Used to implement the SELECT_CC DAG operation. Expanded by the
// scheduler into a branch sequence.
let usesCustomDAGSchedInserter = 1, // Expanded by the scheduler.
PPC970_Single = 1 in {
def SELECT_CC_I4 : Pseudo<(ops GPRC:$dst, CRRC:$cond, GPRC:$T, GPRC:$F,
i32imm:$BROPC), "${:comment} SELECT_CC PSEUDO!",
[]>;
def SELECT_CC_I8 : Pseudo<(ops G8RC:$dst, CRRC:$cond, G8RC:$T, G8RC:$F,
i32imm:$BROPC), "${:comment} SELECT_CC PSEUDO!",
[]>;
def SELECT_CC_F4 : Pseudo<(ops F4RC:$dst, CRRC:$cond, F4RC:$T, F4RC:$F,
i32imm:$BROPC), "${:comment} SELECT_CC PSEUDO!",
[]>;
def SELECT_CC_F8 : Pseudo<(ops F8RC:$dst, CRRC:$cond, F8RC:$T, F8RC:$F,
i32imm:$BROPC), "${:comment} SELECT_CC PSEUDO!",
[]>;
def SELECT_CC_VRRC: Pseudo<(ops VRRC:$dst, CRRC:$cond, VRRC:$T, VRRC:$F,
i32imm:$BROPC), "${:comment} SELECT_CC PSEUDO!",
[]>;
}
let isTerminator = 1, isBarrier = 1, noResults = 1, PPC970_Unit = 7 in {
let isReturn = 1 in
def BLR : XLForm_2_br<19, 16, 0, (ops pred:$p),
"b${p:cc}lr ${p:reg}", BrB,
[(retflag)]>;
def BCTR : XLForm_2_ext<19, 528, 20, 0, 0, (ops), "bctr", BrB, []>;
}
let Defs = [LR] in
def MovePCtoLR : Pseudo<(ops piclabel:$label), "bl $label", []>,
PPC970_Unit_BRU;
let isBranch = 1, isTerminator = 1, hasCtrlDep = 1,
noResults = 1, PPC970_Unit = 7 in {
let isBarrier = 1 in {
def B : IForm<18, 0, 0, (ops target:$dst),
"b $dst", BrB,
[(br bb:$dst)]>;
}
// BCC represents an arbitrary conditional branch on a predicate.
// FIXME: should be able to write a pattern for PPCcondbranch, but can't use
// a two-value operand where a dag node expects two operands. :(
def BCC : BForm<16, 0, 0, (ops pred:$cond, target:$dst),
"b${cond:cc} ${cond:reg}, $dst"
/*[(PPCcondbranch CRRC:$crS, imm:$opc, bb:$dst)]*/>;
}
let isCall = 1, noResults = 1, PPC970_Unit = 7,
// All calls clobber the non-callee saved registers...
Defs = [R0,R2,R3,R4,R5,R6,R7,R8,R9,R10,R11,R12,
F0,F1,F2,F3,F4,F5,F6,F7,F8,F9,F10,F11,F12,F13,
V0,V1,V2,V3,V4,V5,V6,V7,V8,V9,V10,V11,V12,V13,V14,V15,V16,V17,V18,V19,
LR,CTR,
CR0,CR1,CR5,CR6,CR7] in {
// Convenient aliases for call instructions
def BL : IForm<18, 0, 1, (ops calltarget:$func, variable_ops),
"bl $func", BrB, []>; // See Pat patterns below.
def BLA : IForm<18, 1, 1, (ops aaddr:$func, variable_ops),
"bla $func", BrB, [(PPCcall (i32 imm:$func))]>;
def BCTRL : XLForm_2_ext<19, 528, 20, 0, 1, (ops variable_ops), "bctrl", BrB,
[(PPCbctrl)]>;
}
// DCB* instructions.
def DCBA : DCB_Form<758, 0, (ops memrr:$dst),
"dcba $dst", LdStDCBF, [(int_ppc_dcba xoaddr:$dst)]>,
PPC970_DGroup_Single;
def DCBF : DCB_Form<86, 0, (ops memrr:$dst),
"dcbf $dst", LdStDCBF, [(int_ppc_dcbf xoaddr:$dst)]>,
PPC970_DGroup_Single;
def DCBI : DCB_Form<470, 0, (ops memrr:$dst),
"dcbi $dst", LdStDCBF, [(int_ppc_dcbi xoaddr:$dst)]>,
PPC970_DGroup_Single;
def DCBST : DCB_Form<54, 0, (ops memrr:$dst),
"dcbst $dst", LdStDCBF, [(int_ppc_dcbst xoaddr:$dst)]>,
PPC970_DGroup_Single;
def DCBT : DCB_Form<278, 0, (ops memrr:$dst),
"dcbt $dst", LdStDCBF, [(int_ppc_dcbt xoaddr:$dst)]>,
PPC970_DGroup_Single;
def DCBTST : DCB_Form<246, 0, (ops memrr:$dst),
"dcbtst $dst", LdStDCBF, [(int_ppc_dcbtst xoaddr:$dst)]>,
PPC970_DGroup_Single;
def DCBZ : DCB_Form<1014, 0, (ops memrr:$dst),
"dcbz $dst", LdStDCBF, [(int_ppc_dcbz xoaddr:$dst)]>,
PPC970_DGroup_Single;
def DCBZL : DCB_Form<1014, 1, (ops memrr:$dst),
"dcbzl $dst", LdStDCBF, [(int_ppc_dcbzl xoaddr:$dst)]>,
PPC970_DGroup_Single;
//===----------------------------------------------------------------------===//
// PPC32 Load Instructions.
//
// Unindexed (r+i) Loads.
let isLoad = 1, PPC970_Unit = 2 in {
def LBZ : DForm_1<34, (ops GPRC:$rD, memri:$src),
"lbz $rD, $src", LdStGeneral,
[(set GPRC:$rD, (zextloadi8 iaddr:$src))]>;
def LHA : DForm_1<42, (ops GPRC:$rD, memri:$src),
"lha $rD, $src", LdStLHA,
[(set GPRC:$rD, (sextloadi16 iaddr:$src))]>,
PPC970_DGroup_Cracked;
def LHZ : DForm_1<40, (ops GPRC:$rD, memri:$src),
"lhz $rD, $src", LdStGeneral,
[(set GPRC:$rD, (zextloadi16 iaddr:$src))]>;
def LWZ : DForm_1<32, (ops GPRC:$rD, memri:$src),
"lwz $rD, $src", LdStGeneral,
[(set GPRC:$rD, (load iaddr:$src))]>;
def LFS : DForm_1<48, (ops F4RC:$rD, memri:$src),
"lfs $rD, $src", LdStLFDU,
[(set F4RC:$rD, (load iaddr:$src))]>;
def LFD : DForm_1<50, (ops F8RC:$rD, memri:$src),
"lfd $rD, $src", LdStLFD,
[(set F8RC:$rD, (load iaddr:$src))]>;
// Unindexed (r+i) Loads with Update (preinc).
def LBZU : DForm_1<35, (ops GPRC:$rD, ptr_rc:$ea_result, memri:$addr),
"lbzu $rD, $addr", LdStGeneral,
[]>, RegConstraint<"$addr.reg = $ea_result">,
NoEncode<"$ea_result">;
def LHAU : DForm_1<43, (ops GPRC:$rD, ptr_rc:$ea_result, memri:$addr),
"lhau $rD, $addr", LdStGeneral,
[]>, RegConstraint<"$addr.reg = $ea_result">,
NoEncode<"$ea_result">;
def LHZU : DForm_1<41, (ops GPRC:$rD, ptr_rc:$ea_result, memri:$addr),
"lhzu $rD, $addr", LdStGeneral,
[]>, RegConstraint<"$addr.reg = $ea_result">,
NoEncode<"$ea_result">;
def LWZU : DForm_1<33, (ops GPRC:$rD, ptr_rc:$ea_result, memri:$addr),
"lwzu $rD, $addr", LdStGeneral,
[]>, RegConstraint<"$addr.reg = $ea_result">,
NoEncode<"$ea_result">;
def LFSU : DForm_1<49, (ops F4RC:$rD, ptr_rc:$ea_result, memri:$addr),
"lfs $rD, $addr", LdStLFDU,
[]>, RegConstraint<"$addr.reg = $ea_result">,
NoEncode<"$ea_result">;
def LFDU : DForm_1<51, (ops F8RC:$rD, ptr_rc:$ea_result, memri:$addr),
"lfd $rD, $addr", LdStLFD,
[]>, RegConstraint<"$addr.reg = $ea_result">,
NoEncode<"$ea_result">;
}
// Indexed (r+r) Loads.
//
let isLoad = 1, PPC970_Unit = 2 in {
def LBZX : XForm_1<31, 87, (ops GPRC:$rD, memrr:$src),
"lbzx $rD, $src", LdStGeneral,
[(set GPRC:$rD, (zextloadi8 xaddr:$src))]>;
def LHAX : XForm_1<31, 343, (ops GPRC:$rD, memrr:$src),
"lhax $rD, $src", LdStLHA,
[(set GPRC:$rD, (sextloadi16 xaddr:$src))]>,
PPC970_DGroup_Cracked;
def LHZX : XForm_1<31, 279, (ops GPRC:$rD, memrr:$src),
"lhzx $rD, $src", LdStGeneral,
[(set GPRC:$rD, (zextloadi16 xaddr:$src))]>;
def LWZX : XForm_1<31, 23, (ops GPRC:$rD, memrr:$src),
"lwzx $rD, $src", LdStGeneral,
[(set GPRC:$rD, (load xaddr:$src))]>;
def LHBRX : XForm_1<31, 790, (ops GPRC:$rD, memrr:$src),
"lhbrx $rD, $src", LdStGeneral,
[(set GPRC:$rD, (PPClbrx xoaddr:$src, srcvalue:$sv, i16))]>;
def LWBRX : XForm_1<31, 534, (ops GPRC:$rD, memrr:$src),
"lwbrx $rD, $src", LdStGeneral,
[(set GPRC:$rD, (PPClbrx xoaddr:$src, srcvalue:$sv, i32))]>;
def LFSX : XForm_25<31, 535, (ops F4RC:$frD, memrr:$src),
"lfsx $frD, $src", LdStLFDU,
[(set F4RC:$frD, (load xaddr:$src))]>;
def LFDX : XForm_25<31, 599, (ops F8RC:$frD, memrr:$src),
"lfdx $frD, $src", LdStLFDU,
[(set F8RC:$frD, (load xaddr:$src))]>;
}
//===----------------------------------------------------------------------===//
// PPC32 Store Instructions.
//
// Unindexed (r+i) Stores.
let isStore = 1, noResults = 1, PPC970_Unit = 2 in {
def STB : DForm_1<38, (ops GPRC:$rS, memri:$src),
"stb $rS, $src", LdStGeneral,
[(truncstorei8 GPRC:$rS, iaddr:$src)]>;
def STH : DForm_1<44, (ops GPRC:$rS, memri:$src),
"sth $rS, $src", LdStGeneral,
[(truncstorei16 GPRC:$rS, iaddr:$src)]>;
def STW : DForm_1<36, (ops GPRC:$rS, memri:$src),
"stw $rS, $src", LdStGeneral,
[(store GPRC:$rS, iaddr:$src)]>;
def STFS : DForm_1<52, (ops F4RC:$rS, memri:$dst),
"stfs $rS, $dst", LdStUX,
[(store F4RC:$rS, iaddr:$dst)]>;
def STFD : DForm_1<54, (ops F8RC:$rS, memri:$dst),
"stfd $rS, $dst", LdStUX,
[(store F8RC:$rS, iaddr:$dst)]>;
}
// Unindexed (r+i) Stores with Update (preinc).
let isStore = 1, PPC970_Unit = 2 in {
def STBU : DForm_1<39, (ops ptr_rc:$ea_res, GPRC:$rS,
symbolLo:$ptroff, ptr_rc:$ptrreg),
"stbu $rS, $ptroff($ptrreg)", LdStGeneral,
[(set ptr_rc:$ea_res,
(pre_truncsti8 GPRC:$rS, ptr_rc:$ptrreg,
iaddroff:$ptroff))]>,
RegConstraint<"$ptrreg = $ea_res">, NoEncode<"$ea_res">;
def STHU : DForm_1<45, (ops ptr_rc:$ea_res, GPRC:$rS,
symbolLo:$ptroff, ptr_rc:$ptrreg),
"sthu $rS, $ptroff($ptrreg)", LdStGeneral,
[(set ptr_rc:$ea_res,
(pre_truncsti16 GPRC:$rS, ptr_rc:$ptrreg,
iaddroff:$ptroff))]>,
RegConstraint<"$ptrreg = $ea_res">, NoEncode<"$ea_res">;
def STWU : DForm_1<37, (ops ptr_rc:$ea_res, GPRC:$rS,
symbolLo:$ptroff, ptr_rc:$ptrreg),
"stwu $rS, $ptroff($ptrreg)", LdStGeneral,
[(set ptr_rc:$ea_res, (pre_store GPRC:$rS, ptr_rc:$ptrreg,
iaddroff:$ptroff))]>,
RegConstraint<"$ptrreg = $ea_res">, NoEncode<"$ea_res">;
def STFSU : DForm_1<37, (ops ptr_rc:$ea_res, F4RC:$rS,
symbolLo:$ptroff, ptr_rc:$ptrreg),
"stfsu $rS, $ptroff($ptrreg)", LdStGeneral,
[(set ptr_rc:$ea_res, (pre_store F4RC:$rS, ptr_rc:$ptrreg,
iaddroff:$ptroff))]>,
RegConstraint<"$ptrreg = $ea_res">, NoEncode<"$ea_res">;
def STFDU : DForm_1<37, (ops ptr_rc:$ea_res, F8RC:$rS,
symbolLo:$ptroff, ptr_rc:$ptrreg),
"stfdu $rS, $ptroff($ptrreg)", LdStGeneral,
[(set ptr_rc:$ea_res, (pre_store F8RC:$rS, ptr_rc:$ptrreg,
iaddroff:$ptroff))]>,
RegConstraint<"$ptrreg = $ea_res">, NoEncode<"$ea_res">;
}
// Indexed (r+r) Stores.
//
let isStore = 1, noResults = 1, PPC970_Unit = 2 in {
def STBX : XForm_8<31, 215, (ops GPRC:$rS, memrr:$dst),
"stbx $rS, $dst", LdStGeneral,
[(truncstorei8 GPRC:$rS, xaddr:$dst)]>,
PPC970_DGroup_Cracked;
def STHX : XForm_8<31, 407, (ops GPRC:$rS, memrr:$dst),
"sthx $rS, $dst", LdStGeneral,
[(truncstorei16 GPRC:$rS, xaddr:$dst)]>,
PPC970_DGroup_Cracked;
def STWX : XForm_8<31, 151, (ops GPRC:$rS, memrr:$dst),
"stwx $rS, $dst", LdStGeneral,
[(store GPRC:$rS, xaddr:$dst)]>,
PPC970_DGroup_Cracked;
def STWUX : XForm_8<31, 183, (ops GPRC:$rS, GPRC:$rA, GPRC:$rB),
"stwux $rS, $rA, $rB", LdStGeneral,
[]>;
def STHBRX: XForm_8<31, 918, (ops GPRC:$rS, memrr:$dst),
"sthbrx $rS, $dst", LdStGeneral,
[(PPCstbrx GPRC:$rS, xoaddr:$dst, srcvalue:$dummy, i16)]>,
PPC970_DGroup_Cracked;
def STWBRX: XForm_8<31, 662, (ops GPRC:$rS, memrr:$dst),
"stwbrx $rS, $dst", LdStGeneral,
[(PPCstbrx GPRC:$rS, xoaddr:$dst, srcvalue:$dummy, i32)]>,
PPC970_DGroup_Cracked;
def STFIWX: XForm_28<31, 983, (ops F8RC:$frS, memrr:$dst),
"stfiwx $frS, $dst", LdStUX,
[(PPCstfiwx F8RC:$frS, xoaddr:$dst)]>;
def STFSX : XForm_28<31, 663, (ops F4RC:$frS, memrr:$dst),
"stfsx $frS, $dst", LdStUX,
[(store F4RC:$frS, xaddr:$dst)]>;
def STFDX : XForm_28<31, 727, (ops F8RC:$frS, memrr:$dst),
"stfdx $frS, $dst", LdStUX,
[(store F8RC:$frS, xaddr:$dst)]>;
}
//===----------------------------------------------------------------------===//
// PPC32 Arithmetic Instructions.
//
let PPC970_Unit = 1 in { // FXU Operations.
def ADDI : DForm_2<14, (ops GPRC:$rD, GPRC:$rA, s16imm:$imm),
"addi $rD, $rA, $imm", IntGeneral,
[(set GPRC:$rD, (add GPRC:$rA, immSExt16:$imm))]>;
def ADDIC : DForm_2<12, (ops GPRC:$rD, GPRC:$rA, s16imm:$imm),
"addic $rD, $rA, $imm", IntGeneral,
[(set GPRC:$rD, (addc GPRC:$rA, immSExt16:$imm))]>,
PPC970_DGroup_Cracked;
def ADDICo : DForm_2<13, (ops GPRC:$rD, GPRC:$rA, s16imm:$imm),
"addic. $rD, $rA, $imm", IntGeneral,
[]>;
def ADDIS : DForm_2<15, (ops GPRC:$rD, GPRC:$rA, symbolHi:$imm),
"addis $rD, $rA, $imm", IntGeneral,
[(set GPRC:$rD, (add GPRC:$rA, imm16ShiftedSExt:$imm))]>;
def LA : DForm_2<14, (ops GPRC:$rD, GPRC:$rA, symbolLo:$sym),
"la $rD, $sym($rA)", IntGeneral,
[(set GPRC:$rD, (add GPRC:$rA,
(PPClo tglobaladdr:$sym, 0)))]>;
def MULLI : DForm_2< 7, (ops GPRC:$rD, GPRC:$rA, s16imm:$imm),
"mulli $rD, $rA, $imm", IntMulLI,
[(set GPRC:$rD, (mul GPRC:$rA, immSExt16:$imm))]>;
def SUBFIC : DForm_2< 8, (ops GPRC:$rD, GPRC:$rA, s16imm:$imm),
"subfic $rD, $rA, $imm", IntGeneral,
[(set GPRC:$rD, (subc immSExt16:$imm, GPRC:$rA))]>;
def LI : DForm_2_r0<14, (ops GPRC:$rD, symbolLo:$imm),
"li $rD, $imm", IntGeneral,
[(set GPRC:$rD, immSExt16:$imm)]>;
def LIS : DForm_2_r0<15, (ops GPRC:$rD, symbolHi:$imm),
"lis $rD, $imm", IntGeneral,
[(set GPRC:$rD, imm16ShiftedSExt:$imm)]>;
}
let PPC970_Unit = 1 in { // FXU Operations.
def ANDIo : DForm_4<28, (ops GPRC:$dst, GPRC:$src1, u16imm:$src2),
"andi. $dst, $src1, $src2", IntGeneral,
[(set GPRC:$dst, (and GPRC:$src1, immZExt16:$src2))]>,
isDOT;
def ANDISo : DForm_4<29, (ops GPRC:$dst, GPRC:$src1, u16imm:$src2),
"andis. $dst, $src1, $src2", IntGeneral,
[(set GPRC:$dst, (and GPRC:$src1,imm16ShiftedZExt:$src2))]>,
isDOT;
def ORI : DForm_4<24, (ops GPRC:$dst, GPRC:$src1, u16imm:$src2),
"ori $dst, $src1, $src2", IntGeneral,
[(set GPRC:$dst, (or GPRC:$src1, immZExt16:$src2))]>;
def ORIS : DForm_4<25, (ops GPRC:$dst, GPRC:$src1, u16imm:$src2),
"oris $dst, $src1, $src2", IntGeneral,
[(set GPRC:$dst, (or GPRC:$src1, imm16ShiftedZExt:$src2))]>;
def XORI : DForm_4<26, (ops GPRC:$dst, GPRC:$src1, u16imm:$src2),
"xori $dst, $src1, $src2", IntGeneral,
[(set GPRC:$dst, (xor GPRC:$src1, immZExt16:$src2))]>;
def XORIS : DForm_4<27, (ops GPRC:$dst, GPRC:$src1, u16imm:$src2),
"xoris $dst, $src1, $src2", IntGeneral,
[(set GPRC:$dst, (xor GPRC:$src1,imm16ShiftedZExt:$src2))]>;
def NOP : DForm_4_zero<24, (ops), "nop", IntGeneral,
[]>;
def CMPWI : DForm_5_ext<11, (ops CRRC:$crD, GPRC:$rA, s16imm:$imm),
"cmpwi $crD, $rA, $imm", IntCompare>;
def CMPLWI : DForm_6_ext<10, (ops CRRC:$dst, GPRC:$src1, u16imm:$src2),
"cmplwi $dst, $src1, $src2", IntCompare>;
}
let PPC970_Unit = 1 in { // FXU Operations.
def NAND : XForm_6<31, 476, (ops GPRC:$rA, GPRC:$rS, GPRC:$rB),
"nand $rA, $rS, $rB", IntGeneral,
[(set GPRC:$rA, (not (and GPRC:$rS, GPRC:$rB)))]>;
def AND : XForm_6<31, 28, (ops GPRC:$rA, GPRC:$rS, GPRC:$rB),
"and $rA, $rS, $rB", IntGeneral,
[(set GPRC:$rA, (and GPRC:$rS, GPRC:$rB))]>;
def ANDC : XForm_6<31, 60, (ops GPRC:$rA, GPRC:$rS, GPRC:$rB),
"andc $rA, $rS, $rB", IntGeneral,
[(set GPRC:$rA, (and GPRC:$rS, (not GPRC:$rB)))]>;
def OR : XForm_6<31, 444, (ops GPRC:$rA, GPRC:$rS, GPRC:$rB),
"or $rA, $rS, $rB", IntGeneral,
[(set GPRC:$rA, (or GPRC:$rS, GPRC:$rB))]>;
def NOR : XForm_6<31, 124, (ops GPRC:$rA, GPRC:$rS, GPRC:$rB),
"nor $rA, $rS, $rB", IntGeneral,
[(set GPRC:$rA, (not (or GPRC:$rS, GPRC:$rB)))]>;
def ORC : XForm_6<31, 412, (ops GPRC:$rA, GPRC:$rS, GPRC:$rB),
"orc $rA, $rS, $rB", IntGeneral,
[(set GPRC:$rA, (or GPRC:$rS, (not GPRC:$rB)))]>;
def EQV : XForm_6<31, 284, (ops GPRC:$rA, GPRC:$rS, GPRC:$rB),
"eqv $rA, $rS, $rB", IntGeneral,
[(set GPRC:$rA, (not (xor GPRC:$rS, GPRC:$rB)))]>;
def XOR : XForm_6<31, 316, (ops GPRC:$rA, GPRC:$rS, GPRC:$rB),
"xor $rA, $rS, $rB", IntGeneral,
[(set GPRC:$rA, (xor GPRC:$rS, GPRC:$rB))]>;
def SLW : XForm_6<31, 24, (ops GPRC:$rA, GPRC:$rS, GPRC:$rB),
"slw $rA, $rS, $rB", IntGeneral,
[(set GPRC:$rA, (PPCshl GPRC:$rS, GPRC:$rB))]>;
def SRW : XForm_6<31, 536, (ops GPRC:$rA, GPRC:$rS, GPRC:$rB),
"srw $rA, $rS, $rB", IntGeneral,
[(set GPRC:$rA, (PPCsrl GPRC:$rS, GPRC:$rB))]>;
def SRAW : XForm_6<31, 792, (ops GPRC:$rA, GPRC:$rS, GPRC:$rB),
"sraw $rA, $rS, $rB", IntShift,
[(set GPRC:$rA, (PPCsra GPRC:$rS, GPRC:$rB))]>;
}
let PPC970_Unit = 1 in { // FXU Operations.
def SRAWI : XForm_10<31, 824, (ops GPRC:$rA, GPRC:$rS, u5imm:$SH),
"srawi $rA, $rS, $SH", IntShift,
[(set GPRC:$rA, (sra GPRC:$rS, (i32 imm:$SH)))]>;
def CNTLZW : XForm_11<31, 26, (ops GPRC:$rA, GPRC:$rS),
"cntlzw $rA, $rS", IntGeneral,
[(set GPRC:$rA, (ctlz GPRC:$rS))]>;
def EXTSB : XForm_11<31, 954, (ops GPRC:$rA, GPRC:$rS),
"extsb $rA, $rS", IntGeneral,
[(set GPRC:$rA, (sext_inreg GPRC:$rS, i8))]>;
def EXTSH : XForm_11<31, 922, (ops GPRC:$rA, GPRC:$rS),
"extsh $rA, $rS", IntGeneral,
[(set GPRC:$rA, (sext_inreg GPRC:$rS, i16))]>;
def CMPW : XForm_16_ext<31, 0, (ops CRRC:$crD, GPRC:$rA, GPRC:$rB),
"cmpw $crD, $rA, $rB", IntCompare>;
def CMPLW : XForm_16_ext<31, 32, (ops CRRC:$crD, GPRC:$rA, GPRC:$rB),
"cmplw $crD, $rA, $rB", IntCompare>;
}
let PPC970_Unit = 3 in { // FPU Operations.
//def FCMPO : XForm_17<63, 32, (ops CRRC:$crD, FPRC:$fA, FPRC:$fB),
// "fcmpo $crD, $fA, $fB", FPCompare>;
def FCMPUS : XForm_17<63, 0, (ops CRRC:$crD, F4RC:$fA, F4RC:$fB),
"fcmpu $crD, $fA, $fB", FPCompare>;
def FCMPUD : XForm_17<63, 0, (ops CRRC:$crD, F8RC:$fA, F8RC:$fB),
"fcmpu $crD, $fA, $fB", FPCompare>;
def FCTIWZ : XForm_26<63, 15, (ops F8RC:$frD, F8RC:$frB),
"fctiwz $frD, $frB", FPGeneral,
[(set F8RC:$frD, (PPCfctiwz F8RC:$frB))]>;
def FRSP : XForm_26<63, 12, (ops F4RC:$frD, F8RC:$frB),
"frsp $frD, $frB", FPGeneral,
[(set F4RC:$frD, (fround F8RC:$frB))]>;
def FSQRT : XForm_26<63, 22, (ops F8RC:$frD, F8RC:$frB),
"fsqrt $frD, $frB", FPSqrt,
[(set F8RC:$frD, (fsqrt F8RC:$frB))]>;
def FSQRTS : XForm_26<59, 22, (ops F4RC:$frD, F4RC:$frB),
"fsqrts $frD, $frB", FPSqrt,
[(set F4RC:$frD, (fsqrt F4RC:$frB))]>;
}
/// FMR is split into 3 versions, one for 4/8 byte FP, and one for extending.
///
/// Note that these are defined as pseudo-ops on the PPC970 because they are
/// often coalesced away and we don't want the dispatch group builder to think
/// that they will fill slots (which could cause the load of a LSU reject to
/// sneak into a d-group with a store).
def FMRS : XForm_26<63, 72, (ops F4RC:$frD, F4RC:$frB),
"fmr $frD, $frB", FPGeneral,
[]>, // (set F4RC:$frD, F4RC:$frB)
PPC970_Unit_Pseudo;
def FMRD : XForm_26<63, 72, (ops F8RC:$frD, F8RC:$frB),
"fmr $frD, $frB", FPGeneral,
[]>, // (set F8RC:$frD, F8RC:$frB)
PPC970_Unit_Pseudo;
def FMRSD : XForm_26<63, 72, (ops F8RC:$frD, F4RC:$frB),
"fmr $frD, $frB", FPGeneral,
[(set F8RC:$frD, (fextend F4RC:$frB))]>,
PPC970_Unit_Pseudo;
let PPC970_Unit = 3 in { // FPU Operations.
// These are artificially split into two different forms, for 4/8 byte FP.
def FABSS : XForm_26<63, 264, (ops F4RC:$frD, F4RC:$frB),
"fabs $frD, $frB", FPGeneral,
[(set F4RC:$frD, (fabs F4RC:$frB))]>;
def FABSD : XForm_26<63, 264, (ops F8RC:$frD, F8RC:$frB),
"fabs $frD, $frB", FPGeneral,
[(set F8RC:$frD, (fabs F8RC:$frB))]>;
def FNABSS : XForm_26<63, 136, (ops F4RC:$frD, F4RC:$frB),
"fnabs $frD, $frB", FPGeneral,
[(set F4RC:$frD, (fneg (fabs F4RC:$frB)))]>;
def FNABSD : XForm_26<63, 136, (ops F8RC:$frD, F8RC:$frB),
"fnabs $frD, $frB", FPGeneral,
[(set F8RC:$frD, (fneg (fabs F8RC:$frB)))]>;
def FNEGS : XForm_26<63, 40, (ops F4RC:$frD, F4RC:$frB),
"fneg $frD, $frB", FPGeneral,
[(set F4RC:$frD, (fneg F4RC:$frB))]>;
def FNEGD : XForm_26<63, 40, (ops F8RC:$frD, F8RC:$frB),
"fneg $frD, $frB", FPGeneral,
[(set F8RC:$frD, (fneg F8RC:$frB))]>;
}
// XL-Form instructions. condition register logical ops.
//
def MCRF : XLForm_3<19, 0, (ops CRRC:$BF, CRRC:$BFA),
"mcrf $BF, $BFA", BrMCR>,
PPC970_DGroup_First, PPC970_Unit_CRU;
// XFX-Form instructions. Instructions that deal with SPRs.
//
def MFCTR : XFXForm_1_ext<31, 339, 9, (ops GPRC:$rT), "mfctr $rT", SprMFSPR>,
PPC970_DGroup_First, PPC970_Unit_FXU;
let Pattern = [(PPCmtctr GPRC:$rS)] in {
def MTCTR : XFXForm_7_ext<31, 467, 9, (ops GPRC:$rS), "mtctr $rS", SprMTSPR>,
PPC970_DGroup_First, PPC970_Unit_FXU;
}
def MTLR : XFXForm_7_ext<31, 467, 8, (ops GPRC:$rS), "mtlr $rS", SprMTSPR>,
PPC970_DGroup_First, PPC970_Unit_FXU;
def MFLR : XFXForm_1_ext<31, 339, 8, (ops GPRC:$rT), "mflr $rT", SprMFSPR>,
PPC970_DGroup_First, PPC970_Unit_FXU;
// Move to/from VRSAVE: despite being a SPR, the VRSAVE register is renamed like
// a GPR on the PPC970. As such, copies in and out have the same performance
// characteristics as an OR instruction.
def MTVRSAVE : XFXForm_7_ext<31, 467, 256, (ops GPRC:$rS),
"mtspr 256, $rS", IntGeneral>,
PPC970_DGroup_Single, PPC970_Unit_FXU;
def MFVRSAVE : XFXForm_1_ext<31, 339, 256, (ops GPRC:$rT),
"mfspr $rT, 256", IntGeneral>,
PPC970_DGroup_First, PPC970_Unit_FXU;
def MTCRF : XFXForm_5<31, 144, (ops crbitm:$FXM, GPRC:$rS),
"mtcrf $FXM, $rS", BrMCRX>,
PPC970_MicroCode, PPC970_Unit_CRU;
def MFCR : XFXForm_3<31, 19, (ops GPRC:$rT), "mfcr $rT", SprMFCR>,
PPC970_MicroCode, PPC970_Unit_CRU;
def MFOCRF: XFXForm_5a<31, 19, (ops GPRC:$rT, crbitm:$FXM),
"mfcr $rT, $FXM", SprMFCR>,
PPC970_DGroup_First, PPC970_Unit_CRU;
let PPC970_Unit = 1 in { // FXU Operations.
// XO-Form instructions. Arithmetic instructions that can set overflow bit
//
def ADD4 : XOForm_1<31, 266, 0, (ops GPRC:$rT, GPRC:$rA, GPRC:$rB),
"add $rT, $rA, $rB", IntGeneral,
[(set GPRC:$rT, (add GPRC:$rA, GPRC:$rB))]>;
def ADDC : XOForm_1<31, 10, 0, (ops GPRC:$rT, GPRC:$rA, GPRC:$rB),
"addc $rT, $rA, $rB", IntGeneral,
[(set GPRC:$rT, (addc GPRC:$rA, GPRC:$rB))]>,
PPC970_DGroup_Cracked;
def ADDE : XOForm_1<31, 138, 0, (ops GPRC:$rT, GPRC:$rA, GPRC:$rB),
"adde $rT, $rA, $rB", IntGeneral,
[(set GPRC:$rT, (adde GPRC:$rA, GPRC:$rB))]>;
def DIVW : XOForm_1<31, 491, 0, (ops GPRC:$rT, GPRC:$rA, GPRC:$rB),
"divw $rT, $rA, $rB", IntDivW,
[(set GPRC:$rT, (sdiv GPRC:$rA, GPRC:$rB))]>,
PPC970_DGroup_First, PPC970_DGroup_Cracked;
def DIVWU : XOForm_1<31, 459, 0, (ops GPRC:$rT, GPRC:$rA, GPRC:$rB),
"divwu $rT, $rA, $rB", IntDivW,
[(set GPRC:$rT, (udiv GPRC:$rA, GPRC:$rB))]>,
PPC970_DGroup_First, PPC970_DGroup_Cracked;
def MULHW : XOForm_1<31, 75, 0, (ops GPRC:$rT, GPRC:$rA, GPRC:$rB),
"mulhw $rT, $rA, $rB", IntMulHW,
[(set GPRC:$rT, (mulhs GPRC:$rA, GPRC:$rB))]>;
def MULHWU : XOForm_1<31, 11, 0, (ops GPRC:$rT, GPRC:$rA, GPRC:$rB),
"mulhwu $rT, $rA, $rB", IntMulHWU,
[(set GPRC:$rT, (mulhu GPRC:$rA, GPRC:$rB))]>;
def MULLW : XOForm_1<31, 235, 0, (ops GPRC:$rT, GPRC:$rA, GPRC:$rB),
"mullw $rT, $rA, $rB", IntMulHW,
[(set GPRC:$rT, (mul GPRC:$rA, GPRC:$rB))]>;
def SUBF : XOForm_1<31, 40, 0, (ops GPRC:$rT, GPRC:$rA, GPRC:$rB),
"subf $rT, $rA, $rB", IntGeneral,
[(set GPRC:$rT, (sub GPRC:$rB, GPRC:$rA))]>;
def SUBFC : XOForm_1<31, 8, 0, (ops GPRC:$rT, GPRC:$rA, GPRC:$rB),
"subfc $rT, $rA, $rB", IntGeneral,
[(set GPRC:$rT, (subc GPRC:$rB, GPRC:$rA))]>,
PPC970_DGroup_Cracked;
def SUBFE : XOForm_1<31, 136, 0, (ops GPRC:$rT, GPRC:$rA, GPRC:$rB),
"subfe $rT, $rA, $rB", IntGeneral,
[(set GPRC:$rT, (sube GPRC:$rB, GPRC:$rA))]>;
def ADDME : XOForm_3<31, 234, 0, (ops GPRC:$rT, GPRC:$rA),
"addme $rT, $rA", IntGeneral,
[(set GPRC:$rT, (adde GPRC:$rA, immAllOnes))]>;
def ADDZE : XOForm_3<31, 202, 0, (ops GPRC:$rT, GPRC:$rA),
"addze $rT, $rA", IntGeneral,
[(set GPRC:$rT, (adde GPRC:$rA, 0))]>;
def NEG : XOForm_3<31, 104, 0, (ops GPRC:$rT, GPRC:$rA),
"neg $rT, $rA", IntGeneral,
[(set GPRC:$rT, (ineg GPRC:$rA))]>;
def SUBFME : XOForm_3<31, 232, 0, (ops GPRC:$rT, GPRC:$rA),
"subfme $rT, $rA", IntGeneral,
[(set GPRC:$rT, (sube immAllOnes, GPRC:$rA))]>;
def SUBFZE : XOForm_3<31, 200, 0, (ops GPRC:$rT, GPRC:$rA),
"subfze $rT, $rA", IntGeneral,
[(set GPRC:$rT, (sube 0, GPRC:$rA))]>;
}
// A-Form instructions. Most of the instructions executed in the FPU are of
// this type.
//
let PPC970_Unit = 3 in { // FPU Operations.
def FMADD : AForm_1<63, 29,
(ops F8RC:$FRT, F8RC:$FRA, F8RC:$FRC, F8RC:$FRB),
"fmadd $FRT, $FRA, $FRC, $FRB", FPFused,
[(set F8RC:$FRT, (fadd (fmul F8RC:$FRA, F8RC:$FRC),
F8RC:$FRB))]>,
Requires<[FPContractions]>;
def FMADDS : AForm_1<59, 29,
(ops F4RC:$FRT, F4RC:$FRA, F4RC:$FRC, F4RC:$FRB),
"fmadds $FRT, $FRA, $FRC, $FRB", FPGeneral,
[(set F4RC:$FRT, (fadd (fmul F4RC:$FRA, F4RC:$FRC),
F4RC:$FRB))]>,
Requires<[FPContractions]>;
def FMSUB : AForm_1<63, 28,
(ops F8RC:$FRT, F8RC:$FRA, F8RC:$FRC, F8RC:$FRB),
"fmsub $FRT, $FRA, $FRC, $FRB", FPFused,
[(set F8RC:$FRT, (fsub (fmul F8RC:$FRA, F8RC:$FRC),
F8RC:$FRB))]>,
Requires<[FPContractions]>;
def FMSUBS : AForm_1<59, 28,
(ops F4RC:$FRT, F4RC:$FRA, F4RC:$FRC, F4RC:$FRB),
"fmsubs $FRT, $FRA, $FRC, $FRB", FPGeneral,
[(set F4RC:$FRT, (fsub (fmul F4RC:$FRA, F4RC:$FRC),
F4RC:$FRB))]>,
Requires<[FPContractions]>;
def FNMADD : AForm_1<63, 31,
(ops F8RC:$FRT, F8RC:$FRA, F8RC:$FRC, F8RC:$FRB),
"fnmadd $FRT, $FRA, $FRC, $FRB", FPFused,
[(set F8RC:$FRT, (fneg (fadd (fmul F8RC:$FRA, F8RC:$FRC),
F8RC:$FRB)))]>,
Requires<[FPContractions]>;
def FNMADDS : AForm_1<59, 31,
(ops F4RC:$FRT, F4RC:$FRA, F4RC:$FRC, F4RC:$FRB),
"fnmadds $FRT, $FRA, $FRC, $FRB", FPGeneral,
[(set F4RC:$FRT, (fneg (fadd (fmul F4RC:$FRA, F4RC:$FRC),
F4RC:$FRB)))]>,
Requires<[FPContractions]>;
def FNMSUB : AForm_1<63, 30,
(ops F8RC:$FRT, F8RC:$FRA, F8RC:$FRC, F8RC:$FRB),
"fnmsub $FRT, $FRA, $FRC, $FRB", FPFused,
[(set F8RC:$FRT, (fneg (fsub (fmul F8RC:$FRA, F8RC:$FRC),
F8RC:$FRB)))]>,
Requires<[FPContractions]>;
def FNMSUBS : AForm_1<59, 30,
(ops F4RC:$FRT, F4RC:$FRA, F4RC:$FRC, F4RC:$FRB),
"fnmsubs $FRT, $FRA, $FRC, $FRB", FPGeneral,
[(set F4RC:$FRT, (fneg (fsub (fmul F4RC:$FRA, F4RC:$FRC),
F4RC:$FRB)))]>,
Requires<[FPContractions]>;
// FSEL is artificially split into 4 and 8-byte forms for the result. To avoid
// having 4 of these, force the comparison to always be an 8-byte double (code
// should use an FMRSD if the input comparison value really wants to be a float)
// and 4/8 byte forms for the result and operand type..
def FSELD : AForm_1<63, 23,
(ops F8RC:$FRT, F8RC:$FRA, F8RC:$FRC, F8RC:$FRB),
"fsel $FRT, $FRA, $FRC, $FRB", FPGeneral,
[(set F8RC:$FRT, (PPCfsel F8RC:$FRA,F8RC:$FRC,F8RC:$FRB))]>;
def FSELS : AForm_1<63, 23,
(ops F4RC:$FRT, F8RC:$FRA, F4RC:$FRC, F4RC:$FRB),
"fsel $FRT, $FRA, $FRC, $FRB", FPGeneral,
[(set F4RC:$FRT, (PPCfsel F8RC:$FRA,F4RC:$FRC,F4RC:$FRB))]>;
def FADD : AForm_2<63, 21,
(ops F8RC:$FRT, F8RC:$FRA, F8RC:$FRB),
"fadd $FRT, $FRA, $FRB", FPGeneral,
[(set F8RC:$FRT, (fadd F8RC:$FRA, F8RC:$FRB))]>;
def FADDS : AForm_2<59, 21,
(ops F4RC:$FRT, F4RC:$FRA, F4RC:$FRB),
"fadds $FRT, $FRA, $FRB", FPGeneral,
[(set F4RC:$FRT, (fadd F4RC:$FRA, F4RC:$FRB))]>;
def FDIV : AForm_2<63, 18,
(ops F8RC:$FRT, F8RC:$FRA, F8RC:$FRB),
"fdiv $FRT, $FRA, $FRB", FPDivD,
[(set F8RC:$FRT, (fdiv F8RC:$FRA, F8RC:$FRB))]>;
def FDIVS : AForm_2<59, 18,
(ops F4RC:$FRT, F4RC:$FRA, F4RC:$FRB),
"fdivs $FRT, $FRA, $FRB", FPDivS,
[(set F4RC:$FRT, (fdiv F4RC:$FRA, F4RC:$FRB))]>;
def FMUL : AForm_3<63, 25,
(ops F8RC:$FRT, F8RC:$FRA, F8RC:$FRB),
"fmul $FRT, $FRA, $FRB", FPFused,
[(set F8RC:$FRT, (fmul F8RC:$FRA, F8RC:$FRB))]>;
def FMULS : AForm_3<59, 25,
(ops F4RC:$FRT, F4RC:$FRA, F4RC:$FRB),
"fmuls $FRT, $FRA, $FRB", FPGeneral,
[(set F4RC:$FRT, (fmul F4RC:$FRA, F4RC:$FRB))]>;
def FSUB : AForm_2<63, 20,
(ops F8RC:$FRT, F8RC:$FRA, F8RC:$FRB),
"fsub $FRT, $FRA, $FRB", FPGeneral,
[(set F8RC:$FRT, (fsub F8RC:$FRA, F8RC:$FRB))]>;
def FSUBS : AForm_2<59, 20,
(ops F4RC:$FRT, F4RC:$FRA, F4RC:$FRB),
"fsubs $FRT, $FRA, $FRB", FPGeneral,
[(set F4RC:$FRT, (fsub F4RC:$FRA, F4RC:$FRB))]>;
}
let PPC970_Unit = 1 in { // FXU Operations.
// M-Form instructions. rotate and mask instructions.
//
let isCommutable = 1 in {
// RLWIMI can be commuted if the rotate amount is zero.
def RLWIMI : MForm_2<20,
(ops GPRC:$rA, GPRC:$rSi, GPRC:$rS, u5imm:$SH, u5imm:$MB,
u5imm:$ME), "rlwimi $rA, $rS, $SH, $MB, $ME", IntRotate,
[]>, PPC970_DGroup_Cracked, RegConstraint<"$rSi = $rA">,
NoEncode<"$rSi">;
}
def RLWINM : MForm_2<21,
(ops GPRC:$rA, GPRC:$rS, u5imm:$SH, u5imm:$MB, u5imm:$ME),
"rlwinm $rA, $rS, $SH, $MB, $ME", IntGeneral,
[]>;
def RLWINMo : MForm_2<21,
(ops GPRC:$rA, GPRC:$rS, u5imm:$SH, u5imm:$MB, u5imm:$ME),
"rlwinm. $rA, $rS, $SH, $MB, $ME", IntGeneral,
[]>, isDOT, PPC970_DGroup_Cracked;
def RLWNM : MForm_2<23,
(ops GPRC:$rA, GPRC:$rS, GPRC:$rB, u5imm:$MB, u5imm:$ME),
"rlwnm $rA, $rS, $rB, $MB, $ME", IntGeneral,
[]>;
}
//===----------------------------------------------------------------------===//
// DWARF Pseudo Instructions
//
def DWARF_LOC : Pseudo<(ops i32imm:$line, i32imm:$col, i32imm:$file),
"${:comment} .loc $file, $line, $col",
[(dwarf_loc (i32 imm:$line), (i32 imm:$col),
(i32 imm:$file))]>;
def DWARF_LABEL : Pseudo<(ops i32imm:$id),
"\n${:private}debug_loc$id:",
[(dwarf_label (i32 imm:$id))]>;
//===----------------------------------------------------------------------===//
// PowerPC Instruction Patterns
//
// Arbitrary immediate support. Implement in terms of LIS/ORI.
def : Pat<(i32 imm:$imm),
(ORI (LIS (HI16 imm:$imm)), (LO16 imm:$imm))>;
// Implement the 'not' operation with the NOR instruction.
def NOT : Pat<(not GPRC:$in),
(NOR GPRC:$in, GPRC:$in)>;
// ADD an arbitrary immediate.
def : Pat<(add GPRC:$in, imm:$imm),
(ADDIS (ADDI GPRC:$in, (LO16 imm:$imm)), (HA16 imm:$imm))>;
// OR an arbitrary immediate.
def : Pat<(or GPRC:$in, imm:$imm),
(ORIS (ORI GPRC:$in, (LO16 imm:$imm)), (HI16 imm:$imm))>;
// XOR an arbitrary immediate.
def : Pat<(xor GPRC:$in, imm:$imm),
(XORIS (XORI GPRC:$in, (LO16 imm:$imm)), (HI16 imm:$imm))>;
// SUBFIC
def : Pat<(sub immSExt16:$imm, GPRC:$in),
(SUBFIC GPRC:$in, imm:$imm)>;
// Return void support.
def : Pat<(ret), (BLR)>;
// SHL/SRL
def : Pat<(shl GPRC:$in, (i32 imm:$imm)),
(RLWINM GPRC:$in, imm:$imm, 0, (SHL32 imm:$imm))>;
def : Pat<(srl GPRC:$in, (i32 imm:$imm)),
(RLWINM GPRC:$in, (SRL32 imm:$imm), imm:$imm, 31)>;
// ROTL
def : Pat<(rotl GPRC:$in, GPRC:$sh),
(RLWNM GPRC:$in, GPRC:$sh, 0, 31)>;
def : Pat<(rotl GPRC:$in, (i32 imm:$imm)),
(RLWINM GPRC:$in, imm:$imm, 0, 31)>;
// RLWNM
def : Pat<(and (rotl GPRC:$in, GPRC:$sh), maskimm32:$imm),
(RLWNM GPRC:$in, GPRC:$sh, (MB maskimm32:$imm), (ME maskimm32:$imm))>;
// Calls
def : Pat<(PPCcall (i32 tglobaladdr:$dst)),
(BL tglobaladdr:$dst)>;
def : Pat<(PPCcall (i32 texternalsym:$dst)),
(BL texternalsym:$dst)>;
// Hi and Lo for Darwin Global Addresses.
def : Pat<(PPChi tglobaladdr:$in, 0), (LIS tglobaladdr:$in)>;
def : Pat<(PPClo tglobaladdr:$in, 0), (LI tglobaladdr:$in)>;
def : Pat<(PPChi tconstpool:$in, 0), (LIS tconstpool:$in)>;
def : Pat<(PPClo tconstpool:$in, 0), (LI tconstpool:$in)>;
def : Pat<(PPChi tjumptable:$in, 0), (LIS tjumptable:$in)>;
def : Pat<(PPClo tjumptable:$in, 0), (LI tjumptable:$in)>;
def : Pat<(add GPRC:$in, (PPChi tglobaladdr:$g, 0)),
(ADDIS GPRC:$in, tglobaladdr:$g)>;
def : Pat<(add GPRC:$in, (PPChi tconstpool:$g, 0)),
(ADDIS GPRC:$in, tconstpool:$g)>;
def : Pat<(add GPRC:$in, (PPChi tjumptable:$g, 0)),
(ADDIS GPRC:$in, tjumptable:$g)>;
// Fused negative multiply subtract, alternate pattern
def : Pat<(fsub F8RC:$B, (fmul F8RC:$A, F8RC:$C)),
(FNMSUB F8RC:$A, F8RC:$C, F8RC:$B)>,
Requires<[FPContractions]>;
def : Pat<(fsub F4RC:$B, (fmul F4RC:$A, F4RC:$C)),
(FNMSUBS F4RC:$A, F4RC:$C, F4RC:$B)>,
Requires<[FPContractions]>;
// Standard shifts. These are represented separately from the real shifts above
// so that we can distinguish between shifts that allow 5-bit and 6-bit shift
// amounts.
def : Pat<(sra GPRC:$rS, GPRC:$rB),
(SRAW GPRC:$rS, GPRC:$rB)>;
def : Pat<(srl GPRC:$rS, GPRC:$rB),
(SRW GPRC:$rS, GPRC:$rB)>;
def : Pat<(shl GPRC:$rS, GPRC:$rB),
(SLW GPRC:$rS, GPRC:$rB)>;
def : Pat<(zextloadi1 iaddr:$src),
(LBZ iaddr:$src)>;
def : Pat<(zextloadi1 xaddr:$src),
(LBZX xaddr:$src)>;
def : Pat<(extloadi1 iaddr:$src),
(LBZ iaddr:$src)>;
def : Pat<(extloadi1 xaddr:$src),
(LBZX xaddr:$src)>;
def : Pat<(extloadi8 iaddr:$src),
(LBZ iaddr:$src)>;
def : Pat<(extloadi8 xaddr:$src),
(LBZX xaddr:$src)>;
def : Pat<(extloadi16 iaddr:$src),
(LHZ iaddr:$src)>;
def : Pat<(extloadi16 xaddr:$src),
(LHZX xaddr:$src)>;
def : Pat<(extloadf32 iaddr:$src),
(FMRSD (LFS iaddr:$src))>;
def : Pat<(extloadf32 xaddr:$src),
(FMRSD (LFSX xaddr:$src))>;
include "PPCInstrAltivec.td"
include "PPCInstr64Bit.td"