X-Git-Url: http://demsky.eecs.uci.edu/git/?a=blobdiff_plain;f=lib%2FCodeGen%2FSelectionDAG%2FTargetLowering.cpp;h=1e7e847b8adba831b856087144f6a0fb927bf946;hb=b3bc6352defdf1a5c6b1b0770d0c4d603f6524a8;hp=3da06f9974980b4589318a972b38164176075bbb;hpb=895c4ab564c145d16a585201ea49b91541d806b6;p=oota-llvm.git diff --git a/lib/CodeGen/SelectionDAG/TargetLowering.cpp b/lib/CodeGen/SelectionDAG/TargetLowering.cpp index 3da06f99749..1e7e847b8ad 100644 --- a/lib/CodeGen/SelectionDAG/TargetLowering.cpp +++ b/lib/CodeGen/SelectionDAG/TargetLowering.cpp @@ -2,8 +2,8 @@ // // 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 is distributed under the University of Illinois Open Source +// License. See LICENSE.TXT for details. // //===----------------------------------------------------------------------===// // @@ -11,13 +11,18 @@ // //===----------------------------------------------------------------------===// +#include "llvm/Target/TargetAsmInfo.h" #include "llvm/Target/TargetLowering.h" +#include "llvm/Target/TargetSubtarget.h" #include "llvm/Target/TargetData.h" #include "llvm/Target/TargetMachine.h" -#include "llvm/Target/MRegisterInfo.h" +#include "llvm/Target/TargetRegisterInfo.h" +#include "llvm/GlobalVariable.h" #include "llvm/DerivedTypes.h" +#include "llvm/CodeGen/MachineFrameInfo.h" #include "llvm/CodeGen/SelectionDAG.h" #include "llvm/ADT/StringExtras.h" +#include "llvm/ADT/STLExtras.h" #include "llvm/Support/MathExtras.h" using namespace llvm; @@ -26,60 +31,164 @@ using namespace llvm; static void InitLibcallNames(const char **Names) { Names[RTLIB::SHL_I32] = "__ashlsi3"; Names[RTLIB::SHL_I64] = "__ashldi3"; + Names[RTLIB::SHL_I128] = "__ashlti3"; Names[RTLIB::SRL_I32] = "__lshrsi3"; Names[RTLIB::SRL_I64] = "__lshrdi3"; + Names[RTLIB::SRL_I128] = "__lshrti3"; Names[RTLIB::SRA_I32] = "__ashrsi3"; Names[RTLIB::SRA_I64] = "__ashrdi3"; + Names[RTLIB::SRA_I128] = "__ashrti3"; Names[RTLIB::MUL_I32] = "__mulsi3"; Names[RTLIB::MUL_I64] = "__muldi3"; + Names[RTLIB::MUL_I128] = "__multi3"; Names[RTLIB::SDIV_I32] = "__divsi3"; Names[RTLIB::SDIV_I64] = "__divdi3"; + Names[RTLIB::SDIV_I128] = "__divti3"; Names[RTLIB::UDIV_I32] = "__udivsi3"; Names[RTLIB::UDIV_I64] = "__udivdi3"; + Names[RTLIB::UDIV_I128] = "__udivti3"; Names[RTLIB::SREM_I32] = "__modsi3"; Names[RTLIB::SREM_I64] = "__moddi3"; + Names[RTLIB::SREM_I128] = "__modti3"; Names[RTLIB::UREM_I32] = "__umodsi3"; Names[RTLIB::UREM_I64] = "__umoddi3"; + Names[RTLIB::UREM_I128] = "__umodti3"; Names[RTLIB::NEG_I32] = "__negsi2"; Names[RTLIB::NEG_I64] = "__negdi2"; Names[RTLIB::ADD_F32] = "__addsf3"; Names[RTLIB::ADD_F64] = "__adddf3"; + Names[RTLIB::ADD_F80] = "__addxf3"; + Names[RTLIB::ADD_PPCF128] = "__gcc_qadd"; Names[RTLIB::SUB_F32] = "__subsf3"; Names[RTLIB::SUB_F64] = "__subdf3"; + Names[RTLIB::SUB_F80] = "__subxf3"; + Names[RTLIB::SUB_PPCF128] = "__gcc_qsub"; Names[RTLIB::MUL_F32] = "__mulsf3"; Names[RTLIB::MUL_F64] = "__muldf3"; + Names[RTLIB::MUL_F80] = "__mulxf3"; + Names[RTLIB::MUL_PPCF128] = "__gcc_qmul"; Names[RTLIB::DIV_F32] = "__divsf3"; Names[RTLIB::DIV_F64] = "__divdf3"; + Names[RTLIB::DIV_F80] = "__divxf3"; + Names[RTLIB::DIV_PPCF128] = "__gcc_qdiv"; Names[RTLIB::REM_F32] = "fmodf"; Names[RTLIB::REM_F64] = "fmod"; - Names[RTLIB::NEG_F32] = "__negsf2"; - Names[RTLIB::NEG_F64] = "__negdf2"; + Names[RTLIB::REM_F80] = "fmodl"; + Names[RTLIB::REM_PPCF128] = "fmodl"; Names[RTLIB::POWI_F32] = "__powisf2"; Names[RTLIB::POWI_F64] = "__powidf2"; + Names[RTLIB::POWI_F80] = "__powixf2"; + Names[RTLIB::POWI_PPCF128] = "__powitf2"; Names[RTLIB::SQRT_F32] = "sqrtf"; Names[RTLIB::SQRT_F64] = "sqrt"; + Names[RTLIB::SQRT_F80] = "sqrtl"; + Names[RTLIB::SQRT_PPCF128] = "sqrtl"; + Names[RTLIB::LOG_F32] = "logf"; + Names[RTLIB::LOG_F64] = "log"; + Names[RTLIB::LOG_F80] = "logl"; + Names[RTLIB::LOG_PPCF128] = "logl"; + Names[RTLIB::LOG2_F32] = "log2f"; + Names[RTLIB::LOG2_F64] = "log2"; + Names[RTLIB::LOG2_F80] = "log2l"; + Names[RTLIB::LOG2_PPCF128] = "log2l"; + Names[RTLIB::LOG10_F32] = "log10f"; + Names[RTLIB::LOG10_F64] = "log10"; + Names[RTLIB::LOG10_F80] = "log10l"; + Names[RTLIB::LOG10_PPCF128] = "log10l"; + Names[RTLIB::EXP_F32] = "expf"; + Names[RTLIB::EXP_F64] = "exp"; + Names[RTLIB::EXP_F80] = "expl"; + Names[RTLIB::EXP_PPCF128] = "expl"; + Names[RTLIB::EXP2_F32] = "exp2f"; + Names[RTLIB::EXP2_F64] = "exp2"; + Names[RTLIB::EXP2_F80] = "exp2l"; + Names[RTLIB::EXP2_PPCF128] = "exp2l"; Names[RTLIB::SIN_F32] = "sinf"; Names[RTLIB::SIN_F64] = "sin"; + Names[RTLIB::SIN_F80] = "sinl"; + Names[RTLIB::SIN_PPCF128] = "sinl"; Names[RTLIB::COS_F32] = "cosf"; Names[RTLIB::COS_F64] = "cos"; + Names[RTLIB::COS_F80] = "cosl"; + Names[RTLIB::COS_PPCF128] = "cosl"; + Names[RTLIB::POW_F32] = "powf"; + Names[RTLIB::POW_F64] = "pow"; + Names[RTLIB::POW_F80] = "powl"; + Names[RTLIB::POW_PPCF128] = "powl"; + Names[RTLIB::CEIL_F32] = "ceilf"; + Names[RTLIB::CEIL_F64] = "ceil"; + Names[RTLIB::CEIL_F80] = "ceill"; + Names[RTLIB::CEIL_PPCF128] = "ceill"; + Names[RTLIB::TRUNC_F32] = "truncf"; + Names[RTLIB::TRUNC_F64] = "trunc"; + Names[RTLIB::TRUNC_F80] = "truncl"; + Names[RTLIB::TRUNC_PPCF128] = "truncl"; + Names[RTLIB::RINT_F32] = "rintf"; + Names[RTLIB::RINT_F64] = "rint"; + Names[RTLIB::RINT_F80] = "rintl"; + Names[RTLIB::RINT_PPCF128] = "rintl"; + Names[RTLIB::NEARBYINT_F32] = "nearbyintf"; + Names[RTLIB::NEARBYINT_F64] = "nearbyint"; + Names[RTLIB::NEARBYINT_F80] = "nearbyintl"; + Names[RTLIB::NEARBYINT_PPCF128] = "nearbyintl"; + Names[RTLIB::FLOOR_F32] = "floorf"; + Names[RTLIB::FLOOR_F64] = "floor"; + Names[RTLIB::FLOOR_F80] = "floorl"; + Names[RTLIB::FLOOR_PPCF128] = "floorl"; Names[RTLIB::FPEXT_F32_F64] = "__extendsfdf2"; Names[RTLIB::FPROUND_F64_F32] = "__truncdfsf2"; + Names[RTLIB::FPROUND_F80_F32] = "__truncxfsf2"; + Names[RTLIB::FPROUND_PPCF128_F32] = "__trunctfsf2"; + Names[RTLIB::FPROUND_F80_F64] = "__truncxfdf2"; + Names[RTLIB::FPROUND_PPCF128_F64] = "__trunctfdf2"; Names[RTLIB::FPTOSINT_F32_I32] = "__fixsfsi"; Names[RTLIB::FPTOSINT_F32_I64] = "__fixsfdi"; + Names[RTLIB::FPTOSINT_F32_I128] = "__fixsfti"; Names[RTLIB::FPTOSINT_F64_I32] = "__fixdfsi"; Names[RTLIB::FPTOSINT_F64_I64] = "__fixdfdi"; + Names[RTLIB::FPTOSINT_F64_I128] = "__fixdfti"; + Names[RTLIB::FPTOSINT_F80_I32] = "__fixxfsi"; + Names[RTLIB::FPTOSINT_F80_I64] = "__fixxfdi"; + Names[RTLIB::FPTOSINT_F80_I128] = "__fixxfti"; + Names[RTLIB::FPTOSINT_PPCF128_I32] = "__fixtfsi"; + Names[RTLIB::FPTOSINT_PPCF128_I64] = "__fixtfdi"; + Names[RTLIB::FPTOSINT_PPCF128_I128] = "__fixtfti"; Names[RTLIB::FPTOUINT_F32_I32] = "__fixunssfsi"; Names[RTLIB::FPTOUINT_F32_I64] = "__fixunssfdi"; + Names[RTLIB::FPTOUINT_F32_I128] = "__fixunssfti"; Names[RTLIB::FPTOUINT_F64_I32] = "__fixunsdfsi"; Names[RTLIB::FPTOUINT_F64_I64] = "__fixunsdfdi"; + Names[RTLIB::FPTOUINT_F64_I128] = "__fixunsdfti"; + Names[RTLIB::FPTOUINT_F80_I32] = "__fixunsxfsi"; + Names[RTLIB::FPTOUINT_F80_I64] = "__fixunsxfdi"; + Names[RTLIB::FPTOUINT_F80_I128] = "__fixunsxfti"; + Names[RTLIB::FPTOUINT_PPCF128_I32] = "__fixunstfsi"; + Names[RTLIB::FPTOUINT_PPCF128_I64] = "__fixunstfdi"; + Names[RTLIB::FPTOUINT_PPCF128_I128] = "__fixunstfti"; Names[RTLIB::SINTTOFP_I32_F32] = "__floatsisf"; Names[RTLIB::SINTTOFP_I32_F64] = "__floatsidf"; + Names[RTLIB::SINTTOFP_I32_F80] = "__floatsixf"; + Names[RTLIB::SINTTOFP_I32_PPCF128] = "__floatsitf"; Names[RTLIB::SINTTOFP_I64_F32] = "__floatdisf"; Names[RTLIB::SINTTOFP_I64_F64] = "__floatdidf"; + Names[RTLIB::SINTTOFP_I64_F80] = "__floatdixf"; + Names[RTLIB::SINTTOFP_I64_PPCF128] = "__floatditf"; + Names[RTLIB::SINTTOFP_I128_F32] = "__floattisf"; + Names[RTLIB::SINTTOFP_I128_F64] = "__floattidf"; + Names[RTLIB::SINTTOFP_I128_F80] = "__floattixf"; + Names[RTLIB::SINTTOFP_I128_PPCF128] = "__floattitf"; Names[RTLIB::UINTTOFP_I32_F32] = "__floatunsisf"; Names[RTLIB::UINTTOFP_I32_F64] = "__floatunsidf"; + Names[RTLIB::UINTTOFP_I32_F80] = "__floatunsixf"; + Names[RTLIB::UINTTOFP_I32_PPCF128] = "__floatunsitf"; Names[RTLIB::UINTTOFP_I64_F32] = "__floatundisf"; Names[RTLIB::UINTTOFP_I64_F64] = "__floatundidf"; + Names[RTLIB::UINTTOFP_I64_F80] = "__floatundixf"; + Names[RTLIB::UINTTOFP_I64_PPCF128] = "__floatunditf"; + Names[RTLIB::UINTTOFP_I128_F32] = "__floatuntisf"; + Names[RTLIB::UINTTOFP_I128_F64] = "__floatuntidf"; + Names[RTLIB::UINTTOFP_I128_F80] = "__floatuntixf"; + Names[RTLIB::UINTTOFP_I128_PPCF128] = "__floatuntitf"; Names[RTLIB::OEQ_F32] = "__eqsf2"; Names[RTLIB::OEQ_F64] = "__eqdf2"; Names[RTLIB::UNE_F32] = "__nesf2"; @@ -98,6 +207,173 @@ static void InitLibcallNames(const char **Names) { Names[RTLIB::O_F64] = "__unorddf2"; } +/// getFPEXT - Return the FPEXT_*_* value for the given types, or +/// UNKNOWN_LIBCALL if there is none. +RTLIB::Libcall RTLIB::getFPEXT(MVT OpVT, MVT RetVT) { + if (OpVT == MVT::f32) { + if (RetVT == MVT::f64) + return FPEXT_F32_F64; + } + return UNKNOWN_LIBCALL; +} + +/// getFPROUND - Return the FPROUND_*_* value for the given types, or +/// UNKNOWN_LIBCALL if there is none. +RTLIB::Libcall RTLIB::getFPROUND(MVT OpVT, MVT RetVT) { + if (RetVT == MVT::f32) { + if (OpVT == MVT::f64) + return FPROUND_F64_F32; + if (OpVT == MVT::f80) + return FPROUND_F80_F32; + if (OpVT == MVT::ppcf128) + return FPROUND_PPCF128_F32; + } else if (RetVT == MVT::f64) { + if (OpVT == MVT::f80) + return FPROUND_F80_F64; + if (OpVT == MVT::ppcf128) + return FPROUND_PPCF128_F64; + } + return UNKNOWN_LIBCALL; +} + +/// getFPTOSINT - Return the FPTOSINT_*_* value for the given types, or +/// UNKNOWN_LIBCALL if there is none. +RTLIB::Libcall RTLIB::getFPTOSINT(MVT OpVT, MVT RetVT) { + if (OpVT == MVT::f32) { + if (RetVT == MVT::i32) + return FPTOSINT_F32_I32; + if (RetVT == MVT::i64) + return FPTOSINT_F32_I64; + if (RetVT == MVT::i128) + return FPTOSINT_F32_I128; + } else if (OpVT == MVT::f64) { + if (RetVT == MVT::i32) + return FPTOSINT_F64_I32; + if (RetVT == MVT::i64) + return FPTOSINT_F64_I64; + if (RetVT == MVT::i128) + return FPTOSINT_F64_I128; + } else if (OpVT == MVT::f80) { + if (RetVT == MVT::i32) + return FPTOSINT_F80_I32; + if (RetVT == MVT::i64) + return FPTOSINT_F80_I64; + if (RetVT == MVT::i128) + return FPTOSINT_F80_I128; + } else if (OpVT == MVT::ppcf128) { + if (RetVT == MVT::i32) + return FPTOSINT_PPCF128_I32; + if (RetVT == MVT::i64) + return FPTOSINT_PPCF128_I64; + if (RetVT == MVT::i128) + return FPTOSINT_PPCF128_I128; + } + return UNKNOWN_LIBCALL; +} + +/// getFPTOUINT - Return the FPTOUINT_*_* value for the given types, or +/// UNKNOWN_LIBCALL if there is none. +RTLIB::Libcall RTLIB::getFPTOUINT(MVT OpVT, MVT RetVT) { + if (OpVT == MVT::f32) { + if (RetVT == MVT::i32) + return FPTOUINT_F32_I32; + if (RetVT == MVT::i64) + return FPTOUINT_F32_I64; + if (RetVT == MVT::i128) + return FPTOUINT_F32_I128; + } else if (OpVT == MVT::f64) { + if (RetVT == MVT::i32) + return FPTOUINT_F64_I32; + if (RetVT == MVT::i64) + return FPTOUINT_F64_I64; + if (RetVT == MVT::i128) + return FPTOUINT_F64_I128; + } else if (OpVT == MVT::f80) { + if (RetVT == MVT::i32) + return FPTOUINT_F80_I32; + if (RetVT == MVT::i64) + return FPTOUINT_F80_I64; + if (RetVT == MVT::i128) + return FPTOUINT_F80_I128; + } else if (OpVT == MVT::ppcf128) { + if (RetVT == MVT::i32) + return FPTOUINT_PPCF128_I32; + if (RetVT == MVT::i64) + return FPTOUINT_PPCF128_I64; + if (RetVT == MVT::i128) + return FPTOUINT_PPCF128_I128; + } + return UNKNOWN_LIBCALL; +} + +/// getSINTTOFP - Return the SINTTOFP_*_* value for the given types, or +/// UNKNOWN_LIBCALL if there is none. +RTLIB::Libcall RTLIB::getSINTTOFP(MVT OpVT, MVT RetVT) { + if (OpVT == MVT::i32) { + if (RetVT == MVT::f32) + return SINTTOFP_I32_F32; + else if (RetVT == MVT::f64) + return SINTTOFP_I32_F64; + else if (RetVT == MVT::f80) + return SINTTOFP_I32_F80; + else if (RetVT == MVT::ppcf128) + return SINTTOFP_I32_PPCF128; + } else if (OpVT == MVT::i64) { + if (RetVT == MVT::f32) + return SINTTOFP_I64_F32; + else if (RetVT == MVT::f64) + return SINTTOFP_I64_F64; + else if (RetVT == MVT::f80) + return SINTTOFP_I64_F80; + else if (RetVT == MVT::ppcf128) + return SINTTOFP_I64_PPCF128; + } else if (OpVT == MVT::i128) { + if (RetVT == MVT::f32) + return SINTTOFP_I128_F32; + else if (RetVT == MVT::f64) + return SINTTOFP_I128_F64; + else if (RetVT == MVT::f80) + return SINTTOFP_I128_F80; + else if (RetVT == MVT::ppcf128) + return SINTTOFP_I128_PPCF128; + } + return UNKNOWN_LIBCALL; +} + +/// getUINTTOFP - Return the UINTTOFP_*_* value for the given types, or +/// UNKNOWN_LIBCALL if there is none. +RTLIB::Libcall RTLIB::getUINTTOFP(MVT OpVT, MVT RetVT) { + if (OpVT == MVT::i32) { + if (RetVT == MVT::f32) + return UINTTOFP_I32_F32; + else if (RetVT == MVT::f64) + return UINTTOFP_I32_F64; + else if (RetVT == MVT::f80) + return UINTTOFP_I32_F80; + else if (RetVT == MVT::ppcf128) + return UINTTOFP_I32_PPCF128; + } else if (OpVT == MVT::i64) { + if (RetVT == MVT::f32) + return UINTTOFP_I64_F32; + else if (RetVT == MVT::f64) + return UINTTOFP_I64_F64; + else if (RetVT == MVT::f80) + return UINTTOFP_I64_F80; + else if (RetVT == MVT::ppcf128) + return UINTTOFP_I64_PPCF128; + } else if (OpVT == MVT::i128) { + if (RetVT == MVT::f32) + return UINTTOFP_I128_F32; + else if (RetVT == MVT::f64) + return UINTTOFP_I128_F64; + else if (RetVT == MVT::f80) + return UINTTOFP_I128_F80; + else if (RetVT == MVT::ppcf128) + return UINTTOFP_I128_PPCF128; + } + return UNKNOWN_LIBCALL; +} + /// InitCmpLibcallCCs - Set default comparison libcall CC. /// static void InitCmpLibcallCCs(ISD::CondCode *CCs) { @@ -122,28 +398,60 @@ static void InitCmpLibcallCCs(ISD::CondCode *CCs) { TargetLowering::TargetLowering(TargetMachine &tm) : TM(tm), TD(TM.getTargetData()) { - assert(ISD::BUILTIN_OP_END <= 156 && + assert(ISD::BUILTIN_OP_END <= OpActionsCapacity && "Fixed size array in TargetLowering is not large enough!"); // All operations default to being supported. memset(OpActions, 0, sizeof(OpActions)); - memset(LoadXActions, 0, sizeof(LoadXActions)); - memset(&StoreXActions, 0, sizeof(StoreXActions)); - // Initialize all indexed load / store to expand. + memset(LoadExtActions, 0, sizeof(LoadExtActions)); + memset(TruncStoreActions, 0, sizeof(TruncStoreActions)); + memset(IndexedModeActions, 0, sizeof(IndexedModeActions)); + memset(ConvertActions, 0, sizeof(ConvertActions)); + memset(CondCodeActions, 0, sizeof(CondCodeActions)); + + // Set default actions for various operations. for (unsigned VT = 0; VT != (unsigned)MVT::LAST_VALUETYPE; ++VT) { + // Default all indexed load / store to expand. for (unsigned IM = (unsigned)ISD::PRE_INC; IM != (unsigned)ISD::LAST_INDEXED_MODE; ++IM) { - setIndexedLoadAction(IM, (MVT::ValueType)VT, Expand); - setIndexedStoreAction(IM, (MVT::ValueType)VT, Expand); + setIndexedLoadAction(IM, (MVT::SimpleValueType)VT, Expand); + setIndexedStoreAction(IM, (MVT::SimpleValueType)VT, Expand); } + + // These operations default to expand. + setOperationAction(ISD::FGETSIGN, (MVT::SimpleValueType)VT, Expand); } + // Most targets ignore the @llvm.prefetch intrinsic. + setOperationAction(ISD::PREFETCH, MVT::Other, Expand); + + // ConstantFP nodes default to expand. Targets can either change this to + // Legal, in which case all fp constants are legal, or use addLegalFPImmediate + // to optimize expansions for certain constants. + setOperationAction(ISD::ConstantFP, MVT::f32, Expand); + setOperationAction(ISD::ConstantFP, MVT::f64, Expand); + setOperationAction(ISD::ConstantFP, MVT::f80, Expand); + + // These library functions default to expand. + setOperationAction(ISD::FLOG , MVT::f64, Expand); + setOperationAction(ISD::FLOG2, MVT::f64, Expand); + setOperationAction(ISD::FLOG10,MVT::f64, Expand); + setOperationAction(ISD::FEXP , MVT::f64, Expand); + setOperationAction(ISD::FEXP2, MVT::f64, Expand); + setOperationAction(ISD::FLOG , MVT::f32, Expand); + setOperationAction(ISD::FLOG2, MVT::f32, Expand); + setOperationAction(ISD::FLOG10,MVT::f32, Expand); + setOperationAction(ISD::FEXP , MVT::f32, Expand); + setOperationAction(ISD::FEXP2, MVT::f32, Expand); + + // Default ISD::TRAP to expand (which turns it into abort). + setOperationAction(ISD::TRAP, MVT::Other, Expand); + IsLittleEndian = TD->isLittleEndian(); UsesGlobalOffsetTable = false; - ShiftAmountTy = SetCCResultTy = PointerTy = getValueType(TD->getIntPtrType()); + ShiftAmountTy = PointerTy = getValueType(TD->getIntPtrType()); ShiftAmtHandling = Undefined; memset(RegClassForVT, 0,MVT::LAST_VALUETYPE*sizeof(TargetRegisterClass*)); - memset(TargetDAGCombineArray, 0, - sizeof(TargetDAGCombineArray)/sizeof(TargetDAGCombineArray[0])); + memset(TargetDAGCombineArray, 0, array_lengthof(TargetDAGCombineArray)); maxStoresPerMemset = maxStoresPerMemcpy = maxStoresPerMemmove = 8; allowUnalignedMemoryAccesses = false; UseUnderscoreSetJmp = false; @@ -154,58 +462,24 @@ TargetLowering::TargetLowering(TargetMachine &tm) StackPointerRegisterToSaveRestore = 0; ExceptionPointerRegister = 0; ExceptionSelectorRegister = 0; + SetCCResultContents = UndefinedSetCCResult; SchedPreferenceInfo = SchedulingForLatency; JumpBufSize = 0; JumpBufAlignment = 0; + IfCvtBlockSizeLimit = 2; + IfCvtDupBlockSizeLimit = 0; + PrefLoopAlignment = 0; InitLibcallNames(LibcallRoutineNames); InitCmpLibcallCCs(CmpLibcallCCs); -} - -TargetLowering::~TargetLowering() {} -/// setValueTypeAction - Set the action for a particular value type. This -/// assumes an action has not already been set for this value type. -static void SetValueTypeAction(MVT::ValueType VT, - TargetLowering::LegalizeAction Action, - TargetLowering &TLI, - MVT::ValueType *TransformToType, - TargetLowering::ValueTypeActionImpl &ValueTypeActions) { - ValueTypeActions.setTypeAction(VT, Action); - if (Action == TargetLowering::Promote) { - MVT::ValueType PromoteTo; - if (VT == MVT::f32) - PromoteTo = MVT::f64; - else { - unsigned LargerReg = VT+1; - while (!TLI.isTypeLegal((MVT::ValueType)LargerReg)) { - ++LargerReg; - assert(MVT::isInteger((MVT::ValueType)LargerReg) && - "Nothing to promote to??"); - } - PromoteTo = (MVT::ValueType)LargerReg; - } - - assert(MVT::isInteger(VT) == MVT::isInteger(PromoteTo) && - MVT::isFloatingPoint(VT) == MVT::isFloatingPoint(PromoteTo) && - "Can only promote from int->int or fp->fp!"); - assert(VT < PromoteTo && "Must promote to a larger type!"); - TransformToType[VT] = PromoteTo; - } else if (Action == TargetLowering::Expand) { - // f32 and f64 is each expanded to corresponding integer type of same size. - if (VT == MVT::f32) - TransformToType[VT] = MVT::i32; - else if (VT == MVT::f64) - TransformToType[VT] = MVT::i64; - else { - assert((VT == MVT::Vector || MVT::isInteger(VT)) && VT > MVT::i8 && - "Cannot expand this type: target must support SOME integer reg!"); - // Expand to the next smaller integer type! - TransformToType[VT] = (MVT::ValueType)(VT-1); - } - } + // Tell Legalize whether the assembler supports DEBUG_LOC. + const TargetAsmInfo *TASM = TM.getTargetAsmInfo(); + if (!TASM || !TASM->hasDotLocAndDotFile()) + setOperationAction(ISD::DEBUG_LOC, MVT::Other, Expand); } +TargetLowering::~TargetLowering() {} /// computeRegisterProperties - Once all of the register classes are added, /// this allows us to compute derived properties we expose. @@ -213,67 +487,94 @@ void TargetLowering::computeRegisterProperties() { assert(MVT::LAST_VALUETYPE <= 32 && "Too many value types for ValueTypeActions to hold!"); - // Everything defaults to one. - for (unsigned i = 0; i != MVT::LAST_VALUETYPE; ++i) - NumElementsForVT[i] = 1; + // Everything defaults to needing one register. + for (unsigned i = 0; i != MVT::LAST_VALUETYPE; ++i) { + NumRegistersForVT[i] = 1; + RegisterTypeForVT[i] = TransformToType[i] = (MVT::SimpleValueType)i; + } + // ...except isVoid, which doesn't need any registers. + NumRegistersForVT[MVT::isVoid] = 0; // Find the largest integer register class. - unsigned LargestIntReg = MVT::i128; + unsigned LargestIntReg = MVT::LAST_INTEGER_VALUETYPE; for (; RegClassForVT[LargestIntReg] == 0; --LargestIntReg) assert(LargestIntReg != MVT::i1 && "No integer registers defined!"); // Every integer value type larger than this largest register takes twice as // many registers to represent as the previous ValueType. - unsigned ExpandedReg = LargestIntReg; ++LargestIntReg; - for (++ExpandedReg; MVT::isInteger((MVT::ValueType)ExpandedReg);++ExpandedReg) - NumElementsForVT[ExpandedReg] = 2*NumElementsForVT[ExpandedReg-1]; - - // Inspect all of the ValueType's possible, deciding how to process them. - for (unsigned IntReg = MVT::i1; IntReg <= MVT::i128; ++IntReg) - // If we are expanding this type, expand it! - if (getNumElements((MVT::ValueType)IntReg) != 1) - SetValueTypeAction((MVT::ValueType)IntReg, Expand, *this, TransformToType, - ValueTypeActions); - else if (!isTypeLegal((MVT::ValueType)IntReg)) - // Otherwise, if we don't have native support, we must promote to a - // larger type. - SetValueTypeAction((MVT::ValueType)IntReg, Promote, *this, - TransformToType, ValueTypeActions); - else - TransformToType[(MVT::ValueType)IntReg] = (MVT::ValueType)IntReg; - - // If the target does not have native F64 support, expand it to I64. We will - // be generating soft float library calls. If the target does not have native - // support for F32, promote it to F64 if it is legal. Otherwise, expand it to - // I32. - if (isTypeLegal(MVT::f64)) - TransformToType[MVT::f64] = MVT::f64; - else { - NumElementsForVT[MVT::f64] = NumElementsForVT[MVT::i64]; - SetValueTypeAction(MVT::f64, Expand, *this, TransformToType, - ValueTypeActions); + for (unsigned ExpandedReg = LargestIntReg + 1; ; ++ExpandedReg) { + MVT EVT = (MVT::SimpleValueType)ExpandedReg; + if (!EVT.isInteger()) + break; + NumRegistersForVT[ExpandedReg] = 2*NumRegistersForVT[ExpandedReg-1]; + RegisterTypeForVT[ExpandedReg] = (MVT::SimpleValueType)LargestIntReg; + TransformToType[ExpandedReg] = (MVT::SimpleValueType)(ExpandedReg - 1); + ValueTypeActions.setTypeAction(EVT, Expand); } - if (isTypeLegal(MVT::f32)) - TransformToType[MVT::f32] = MVT::f32; - else if (isTypeLegal(MVT::f64)) - SetValueTypeAction(MVT::f32, Promote, *this, TransformToType, - ValueTypeActions); - else { - NumElementsForVT[MVT::f32] = NumElementsForVT[MVT::i32]; - SetValueTypeAction(MVT::f32, Expand, *this, TransformToType, - ValueTypeActions); + + // Inspect all of the ValueType's smaller than the largest integer + // register to see which ones need promotion. + unsigned LegalIntReg = LargestIntReg; + for (unsigned IntReg = LargestIntReg - 1; + IntReg >= (unsigned)MVT::i1; --IntReg) { + MVT IVT = (MVT::SimpleValueType)IntReg; + if (isTypeLegal(IVT)) { + LegalIntReg = IntReg; + } else { + RegisterTypeForVT[IntReg] = TransformToType[IntReg] = + (MVT::SimpleValueType)LegalIntReg; + ValueTypeActions.setTypeAction(IVT, Promote); + } + } + + // ppcf128 type is really two f64's. + if (!isTypeLegal(MVT::ppcf128)) { + NumRegistersForVT[MVT::ppcf128] = 2*NumRegistersForVT[MVT::f64]; + RegisterTypeForVT[MVT::ppcf128] = MVT::f64; + TransformToType[MVT::ppcf128] = MVT::f64; + ValueTypeActions.setTypeAction(MVT::ppcf128, Expand); + } + + // Decide how to handle f64. If the target does not have native f64 support, + // expand it to i64 and we will be generating soft float library calls. + if (!isTypeLegal(MVT::f64)) { + NumRegistersForVT[MVT::f64] = NumRegistersForVT[MVT::i64]; + RegisterTypeForVT[MVT::f64] = RegisterTypeForVT[MVT::i64]; + TransformToType[MVT::f64] = MVT::i64; + ValueTypeActions.setTypeAction(MVT::f64, Expand); + } + + // Decide how to handle f32. If the target does not have native support for + // f32, promote it to f64 if it is legal. Otherwise, expand it to i32. + if (!isTypeLegal(MVT::f32)) { + if (isTypeLegal(MVT::f64)) { + NumRegistersForVT[MVT::f32] = NumRegistersForVT[MVT::f64]; + RegisterTypeForVT[MVT::f32] = RegisterTypeForVT[MVT::f64]; + TransformToType[MVT::f32] = MVT::f64; + ValueTypeActions.setTypeAction(MVT::f32, Promote); + } else { + NumRegistersForVT[MVT::f32] = NumRegistersForVT[MVT::i32]; + RegisterTypeForVT[MVT::f32] = RegisterTypeForVT[MVT::i32]; + TransformToType[MVT::f32] = MVT::i32; + ValueTypeActions.setTypeAction(MVT::f32, Expand); + } } - // Set MVT::Vector to always be Expanded - SetValueTypeAction(MVT::Vector, Expand, *this, TransformToType, - ValueTypeActions); - - // Loop over all of the legal vector value types, specifying an identity type - // transformation. + // Loop over all of the vector value types to see which need transformations. for (unsigned i = MVT::FIRST_VECTOR_VALUETYPE; - i <= MVT::LAST_VECTOR_VALUETYPE; ++i) { - if (isTypeLegal((MVT::ValueType)i)) - TransformToType[i] = (MVT::ValueType)i; + i <= (unsigned)MVT::LAST_VECTOR_VALUETYPE; ++i) { + MVT VT = (MVT::SimpleValueType)i; + if (!isTypeLegal(VT)) { + MVT IntermediateVT, RegisterVT; + unsigned NumIntermediates; + NumRegistersForVT[i] = + getVectorTypeBreakdown(VT, + IntermediateVT, NumIntermediates, + RegisterVT); + RegisterTypeForVT[i] = RegisterVT; + TransformToType[i] = MVT::Other; // this isn't actually used + ValueTypeActions.setTypeAction(VT, Expand); + } } } @@ -281,38 +582,57 @@ const char *TargetLowering::getTargetNodeName(unsigned Opcode) const { return NULL; } -/// getVectorTypeBreakdown - Packed types are broken down into some number of -/// legal first class types. For example, <8 x float> maps to 2 MVT::v4f32 + +MVT TargetLowering::getSetCCResultType(const SDValue &) const { + return getValueType(TD->getIntPtrType()); +} + + +/// getVectorTypeBreakdown - Vector types are broken down into some number of +/// legal first class types. For example, MVT::v8f32 maps to 2 MVT::v4f32 /// with Altivec or SSE1, or 8 promoted MVT::f64 values with the X86 FP stack. +/// Similarly, MVT::v2i64 turns into 4 MVT::i32 values with both PPC and X86. /// -/// This method returns the number and type of the resultant breakdown. +/// This method returns the number of registers needed, and the VT for each +/// register. It also returns the VT and quantity of the intermediate values +/// before they are promoted/expanded. /// -unsigned TargetLowering::getVectorTypeBreakdown(const VectorType *PTy, - MVT::ValueType &PTyElementVT, - MVT::ValueType &PTyLegalElementVT) const { +unsigned TargetLowering::getVectorTypeBreakdown(MVT VT, + MVT &IntermediateVT, + unsigned &NumIntermediates, + MVT &RegisterVT) const { // Figure out the right, legal destination reg to copy into. - unsigned NumElts = PTy->getNumElements(); - MVT::ValueType EltTy = getValueType(PTy->getElementType()); + unsigned NumElts = VT.getVectorNumElements(); + MVT EltTy = VT.getVectorElementType(); unsigned NumVectorRegs = 1; + // FIXME: We don't support non-power-of-2-sized vectors for now. Ideally we + // could break down into LHS/RHS like LegalizeDAG does. + if (!isPowerOf2_32(NumElts)) { + NumVectorRegs = NumElts; + NumElts = 1; + } + // Divide the input until we get to a supported size. This will always // end with a scalar if the target doesn't support vectors. - while (NumElts > 1 && !isTypeLegal(getVectorType(EltTy, NumElts))) { + while (NumElts > 1 && !isTypeLegal(MVT::getVectorVT(EltTy, NumElts))) { NumElts >>= 1; NumVectorRegs <<= 1; } + + NumIntermediates = NumVectorRegs; - MVT::ValueType VT = getVectorType(EltTy, NumElts); - if (!isTypeLegal(VT)) - VT = EltTy; - PTyElementVT = VT; - - MVT::ValueType DestVT = getTypeToTransformTo(VT); - PTyLegalElementVT = DestVT; - if (DestVT < VT) { + MVT NewVT = MVT::getVectorVT(EltTy, NumElts); + if (!isTypeLegal(NewVT)) + NewVT = EltTy; + IntermediateVT = NewVT; + + MVT DestVT = getTypeToTransformTo(NewVT); + RegisterVT = DestVT; + if (DestVT.bitsLT(NewVT)) { // Value is expanded, e.g. i64 -> i16. - return NumVectorRegs*(MVT::getSizeInBits(VT)/MVT::getSizeInBits(DestVT)); + return NumVectorRegs*(NewVT.getSizeInBits()/DestVT.getSizeInBits()); } else { // Otherwise, promotion or legal types use the same number of registers as // the vector decimated to the appropriate level. @@ -322,6 +642,37 @@ unsigned TargetLowering::getVectorTypeBreakdown(const VectorType *PTy, return 1; } +/// getByValTypeAlignment - Return the desired alignment for ByVal aggregate +/// function arguments in the caller parameter area. This is the actual +/// alignment, not its logarithm. +unsigned TargetLowering::getByValTypeAlignment(const Type *Ty) const { + return TD->getCallFrameTypeAlignment(Ty); +} + +SDValue TargetLowering::getPICJumpTableRelocBase(SDValue Table, + SelectionDAG &DAG) const { + if (usesGlobalOffsetTable()) + return DAG.getNode(ISD::GLOBAL_OFFSET_TABLE, getPointerTy()); + return Table; +} + +bool +TargetLowering::isOffsetFoldingLegal(const GlobalAddressSDNode *GA) const { + // Assume that everything is safe in static mode. + if (getTargetMachine().getRelocationModel() == Reloc::Static) + return true; + + // In dynamic-no-pic mode, assume that known defined values are safe. + if (getTargetMachine().getRelocationModel() == Reloc::DynamicNoPIC && + GA && + !GA->getGlobal()->isDeclaration() && + !GA->getGlobal()->mayBeOverridden()) + return true; + + // Otherwise assume nothing is safe. + return false; +} + //===----------------------------------------------------------------------===// // Optimization Methods //===----------------------------------------------------------------------===// @@ -330,8 +681,8 @@ unsigned TargetLowering::getVectorTypeBreakdown(const VectorType *PTy, /// specified instruction is a constant integer. If so, check to see if there /// are any bits set in the constant that are not demanded. If so, shrink the /// constant and return true. -bool TargetLowering::TargetLoweringOpt::ShrinkDemandedConstant(SDOperand Op, - uint64_t Demanded) { +bool TargetLowering::TargetLoweringOpt::ShrinkDemandedConstant(SDValue Op, + const APInt &Demanded) { // FIXME: ISD::SELECT, ISD::SELECT_CC switch(Op.getOpcode()) { default: break; @@ -339,10 +690,11 @@ bool TargetLowering::TargetLoweringOpt::ShrinkDemandedConstant(SDOperand Op, case ISD::OR: case ISD::XOR: if (ConstantSDNode *C = dyn_cast(Op.getOperand(1))) - if ((~Demanded & C->getValue()) != 0) { - MVT::ValueType VT = Op.getValueType(); - SDOperand New = DAG.getNode(Op.getOpcode(), VT, Op.getOperand(0), - DAG.getConstant(Demanded & C->getValue(), + if (C->getAPIntValue().intersects(~Demanded)) { + MVT VT = Op.getValueType(); + SDValue New = DAG.getNode(Op.getOpcode(), VT, Op.getOperand(0), + DAG.getConstant(Demanded & + C->getAPIntValue(), VT)); return CombineTo(Op, New); } @@ -358,23 +710,31 @@ bool TargetLowering::TargetLoweringOpt::ShrinkDemandedConstant(SDOperand Op, /// analyze the expression and return a mask of KnownOne and KnownZero bits for /// the expression (used to simplify the caller). The KnownZero/One bits may /// only be accurate for those bits in the DemandedMask. -bool TargetLowering::SimplifyDemandedBits(SDOperand Op, uint64_t DemandedMask, - uint64_t &KnownZero, - uint64_t &KnownOne, +bool TargetLowering::SimplifyDemandedBits(SDValue Op, + const APInt &DemandedMask, + APInt &KnownZero, + APInt &KnownOne, TargetLoweringOpt &TLO, unsigned Depth) const { - KnownZero = KnownOne = 0; // Don't know anything. + unsigned BitWidth = DemandedMask.getBitWidth(); + assert(Op.getValueSizeInBits() == BitWidth && + "Mask size mismatches value type size!"); + APInt NewMask = DemandedMask; + + // Don't know anything. + KnownZero = KnownOne = APInt(BitWidth, 0); + // Other users may use these bits. - if (!Op.Val->hasOneUse()) { + if (!Op.getNode()->hasOneUse()) { if (Depth != 0) { // If not at the root, Just compute the KnownZero/KnownOne bits to // simplify things downstream. - ComputeMaskedBits(Op, DemandedMask, KnownZero, KnownOne, Depth); + TLO.DAG.ComputeMaskedBits(Op, DemandedMask, KnownZero, KnownOne, Depth); return false; } // If this is the root being simplified, allow it to have multiple uses, - // just set the DemandedMask to all bits. - DemandedMask = MVT::getIntVTBitMask(Op.getValueType()); + // just set the NewMask to all bits. + NewMask = APInt::getAllOnesValue(BitWidth); } else if (DemandedMask == 0) { // Not demanding any bits from Op. if (Op.getOpcode() != ISD::UNDEF) @@ -384,12 +744,12 @@ bool TargetLowering::SimplifyDemandedBits(SDOperand Op, uint64_t DemandedMask, return false; } - uint64_t KnownZero2, KnownOne2, KnownZeroOut, KnownOneOut; + APInt KnownZero2, KnownOne2, KnownZeroOut, KnownOneOut; switch (Op.getOpcode()) { case ISD::Constant: // We know all of the bits for a constant! - KnownOne = cast(Op)->getValue() & DemandedMask; - KnownZero = ~KnownOne & DemandedMask; + KnownOne = cast(Op)->getAPIntValue() & NewMask; + KnownZero = ~KnownOne & NewMask; return false; // Don't fall through, will infinitely loop. case ISD::AND: // If the RHS is a constant, check to see if the LHS would be zero without @@ -397,38 +757,38 @@ bool TargetLowering::SimplifyDemandedBits(SDOperand Op, uint64_t DemandedMask, // simplify the LHS, here we're using information from the LHS to simplify // the RHS. if (ConstantSDNode *RHSC = dyn_cast(Op.getOperand(1))) { - uint64_t LHSZero, LHSOne; - ComputeMaskedBits(Op.getOperand(0), DemandedMask, - LHSZero, LHSOne, Depth+1); + APInt LHSZero, LHSOne; + TLO.DAG.ComputeMaskedBits(Op.getOperand(0), NewMask, + LHSZero, LHSOne, Depth+1); // If the LHS already has zeros where RHSC does, this and is dead. - if ((LHSZero & DemandedMask) == (~RHSC->getValue() & DemandedMask)) + if ((LHSZero & NewMask) == (~RHSC->getAPIntValue() & NewMask)) return TLO.CombineTo(Op, Op.getOperand(0)); // If any of the set bits in the RHS are known zero on the LHS, shrink // the constant. - if (TLO.ShrinkDemandedConstant(Op, ~LHSZero & DemandedMask)) + if (TLO.ShrinkDemandedConstant(Op, ~LHSZero & NewMask)) return true; } - if (SimplifyDemandedBits(Op.getOperand(1), DemandedMask, KnownZero, + if (SimplifyDemandedBits(Op.getOperand(1), NewMask, KnownZero, KnownOne, TLO, Depth+1)) return true; assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?"); - if (SimplifyDemandedBits(Op.getOperand(0), DemandedMask & ~KnownZero, + if (SimplifyDemandedBits(Op.getOperand(0), ~KnownZero & NewMask, KnownZero2, KnownOne2, TLO, Depth+1)) return true; assert((KnownZero2 & KnownOne2) == 0 && "Bits known to be one AND zero?"); // If all of the demanded bits are known one on one side, return the other. // These bits cannot contribute to the result of the 'and'. - if ((DemandedMask & ~KnownZero2 & KnownOne)==(DemandedMask & ~KnownZero2)) + if ((NewMask & ~KnownZero2 & KnownOne) == (~KnownZero2 & NewMask)) return TLO.CombineTo(Op, Op.getOperand(0)); - if ((DemandedMask & ~KnownZero & KnownOne2)==(DemandedMask & ~KnownZero)) + if ((NewMask & ~KnownZero & KnownOne2) == (~KnownZero & NewMask)) return TLO.CombineTo(Op, Op.getOperand(1)); // If all of the demanded bits in the inputs are known zeros, return zero. - if ((DemandedMask & (KnownZero|KnownZero2)) == DemandedMask) + if ((NewMask & (KnownZero|KnownZero2)) == NewMask) return TLO.CombineTo(Op, TLO.DAG.getConstant(0, Op.getValueType())); // If the RHS is a constant, see if we can simplify it. - if (TLO.ShrinkDemandedConstant(Op, DemandedMask & ~KnownZero2)) + if (TLO.ShrinkDemandedConstant(Op, ~KnownZero2 & NewMask)) return true; // Output known-1 bits are only known if set in both the LHS & RHS. @@ -437,31 +797,29 @@ bool TargetLowering::SimplifyDemandedBits(SDOperand Op, uint64_t DemandedMask, KnownZero |= KnownZero2; break; case ISD::OR: - if (SimplifyDemandedBits(Op.getOperand(1), DemandedMask, KnownZero, + if (SimplifyDemandedBits(Op.getOperand(1), NewMask, KnownZero, KnownOne, TLO, Depth+1)) return true; assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?"); - if (SimplifyDemandedBits(Op.getOperand(0), DemandedMask & ~KnownOne, + if (SimplifyDemandedBits(Op.getOperand(0), ~KnownOne & NewMask, KnownZero2, KnownOne2, TLO, Depth+1)) return true; assert((KnownZero2 & KnownOne2) == 0 && "Bits known to be one AND zero?"); // If all of the demanded bits are known zero on one side, return the other. // These bits cannot contribute to the result of the 'or'. - if ((DemandedMask & ~KnownOne2 & KnownZero) == (DemandedMask & ~KnownOne2)) + if ((NewMask & ~KnownOne2 & KnownZero) == (~KnownOne2 & NewMask)) return TLO.CombineTo(Op, Op.getOperand(0)); - if ((DemandedMask & ~KnownOne & KnownZero2) == (DemandedMask & ~KnownOne)) + if ((NewMask & ~KnownOne & KnownZero2) == (~KnownOne & NewMask)) return TLO.CombineTo(Op, Op.getOperand(1)); // If all of the potentially set bits on one side are known to be set on // the other side, just use the 'other' side. - if ((DemandedMask & (~KnownZero) & KnownOne2) == - (DemandedMask & (~KnownZero))) + if ((NewMask & ~KnownZero & KnownOne2) == (~KnownZero & NewMask)) return TLO.CombineTo(Op, Op.getOperand(0)); - if ((DemandedMask & (~KnownZero2) & KnownOne) == - (DemandedMask & (~KnownZero2))) + if ((NewMask & ~KnownZero2 & KnownOne) == (~KnownZero2 & NewMask)) return TLO.CombineTo(Op, Op.getOperand(1)); // If the RHS is a constant, see if we can simplify it. - if (TLO.ShrinkDemandedConstant(Op, DemandedMask)) + if (TLO.ShrinkDemandedConstant(Op, NewMask)) return true; // Output known-0 bits are only known if clear in both the LHS & RHS. @@ -470,26 +828,26 @@ bool TargetLowering::SimplifyDemandedBits(SDOperand Op, uint64_t DemandedMask, KnownOne |= KnownOne2; break; case ISD::XOR: - if (SimplifyDemandedBits(Op.getOperand(1), DemandedMask, KnownZero, + if (SimplifyDemandedBits(Op.getOperand(1), NewMask, KnownZero, KnownOne, TLO, Depth+1)) return true; assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?"); - if (SimplifyDemandedBits(Op.getOperand(0), DemandedMask, KnownZero2, + if (SimplifyDemandedBits(Op.getOperand(0), NewMask, KnownZero2, KnownOne2, TLO, Depth+1)) return true; assert((KnownZero2 & KnownOne2) == 0 && "Bits known to be one AND zero?"); // If all of the demanded bits are known zero on one side, return the other. // These bits cannot contribute to the result of the 'xor'. - if ((DemandedMask & KnownZero) == DemandedMask) + if ((KnownZero & NewMask) == NewMask) return TLO.CombineTo(Op, Op.getOperand(0)); - if ((DemandedMask & KnownZero2) == DemandedMask) + if ((KnownZero2 & NewMask) == NewMask) return TLO.CombineTo(Op, Op.getOperand(1)); // If all of the unknown bits are known to be zero on one side or the other // (but not both) turn this into an *inclusive* or. // e.g. (A & C1)^(B & C2) -> (A & C1)|(B & C2) iff C1&C2 == 0 - if ((DemandedMask & ~KnownZero & ~KnownZero2) == 0) + if ((NewMask & ~KnownZero & ~KnownZero2) == 0) return TLO.CombineTo(Op, TLO.DAG.getNode(ISD::OR, Op.getValueType(), Op.getOperand(0), Op.getOperand(1))); @@ -503,40 +861,50 @@ bool TargetLowering::SimplifyDemandedBits(SDOperand Op, uint64_t DemandedMask, // bits on that side are also known to be set on the other side, turn this // into an AND, as we know the bits will be cleared. // e.g. (X | C1) ^ C2 --> (X | C1) & ~C2 iff (C1&C2) == C2 - if ((DemandedMask & (KnownZero|KnownOne)) == DemandedMask) { // all known + if ((NewMask & (KnownZero|KnownOne)) == NewMask) { // all known if ((KnownOne & KnownOne2) == KnownOne) { - MVT::ValueType VT = Op.getValueType(); - SDOperand ANDC = TLO.DAG.getConstant(~KnownOne & DemandedMask, VT); + MVT VT = Op.getValueType(); + SDValue ANDC = TLO.DAG.getConstant(~KnownOne & NewMask, VT); return TLO.CombineTo(Op, TLO.DAG.getNode(ISD::AND, VT, Op.getOperand(0), ANDC)); } } // If the RHS is a constant, see if we can simplify it. - // FIXME: for XOR, we prefer to force bits to 1 if they will make a -1. - if (TLO.ShrinkDemandedConstant(Op, DemandedMask)) - return true; - + // for XOR, we prefer to force bits to 1 if they will make a -1. + // if we can't force bits, try to shrink constant + if (ConstantSDNode *C = dyn_cast(Op.getOperand(1))) { + APInt Expanded = C->getAPIntValue() | (~NewMask); + // if we can expand it to have all bits set, do it + if (Expanded.isAllOnesValue()) { + if (Expanded != C->getAPIntValue()) { + MVT VT = Op.getValueType(); + SDValue New = TLO.DAG.getNode(Op.getOpcode(), VT, Op.getOperand(0), + TLO.DAG.getConstant(Expanded, VT)); + return TLO.CombineTo(Op, New); + } + // if it already has all the bits set, nothing to change + // but don't shrink either! + } else if (TLO.ShrinkDemandedConstant(Op, NewMask)) { + return true; + } + } + KnownZero = KnownZeroOut; KnownOne = KnownOneOut; break; - case ISD::SETCC: - // If we know the result of a setcc has the top bits zero, use this info. - if (getSetCCResultContents() == TargetLowering::ZeroOrOneSetCCResult) - KnownZero |= (MVT::getIntVTBitMask(Op.getValueType()) ^ 1ULL); - break; case ISD::SELECT: - if (SimplifyDemandedBits(Op.getOperand(2), DemandedMask, KnownZero, + if (SimplifyDemandedBits(Op.getOperand(2), NewMask, KnownZero, KnownOne, TLO, Depth+1)) return true; - if (SimplifyDemandedBits(Op.getOperand(1), DemandedMask, KnownZero2, + if (SimplifyDemandedBits(Op.getOperand(1), NewMask, KnownZero2, KnownOne2, TLO, Depth+1)) return true; assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?"); assert((KnownZero2 & KnownOne2) == 0 && "Bits known to be one AND zero?"); // If the operands are constants, see if we can simplify them. - if (TLO.ShrinkDemandedConstant(Op, DemandedMask)) + if (TLO.ShrinkDemandedConstant(Op, NewMask)) return true; // Only known if known in both the LHS and RHS. @@ -544,17 +912,17 @@ bool TargetLowering::SimplifyDemandedBits(SDOperand Op, uint64_t DemandedMask, KnownZero &= KnownZero2; break; case ISD::SELECT_CC: - if (SimplifyDemandedBits(Op.getOperand(3), DemandedMask, KnownZero, + if (SimplifyDemandedBits(Op.getOperand(3), NewMask, KnownZero, KnownOne, TLO, Depth+1)) return true; - if (SimplifyDemandedBits(Op.getOperand(2), DemandedMask, KnownZero2, + if (SimplifyDemandedBits(Op.getOperand(2), NewMask, KnownZero2, KnownOne2, TLO, Depth+1)) return true; assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?"); assert((KnownZero2 & KnownOne2) == 0 && "Bits known to be one AND zero?"); // If the operands are constants, see if we can simplify them. - if (TLO.ShrinkDemandedConstant(Op, DemandedMask)) + if (TLO.ShrinkDemandedConstant(Op, NewMask)) return true; // Only known if known in both the LHS and RHS. @@ -563,16 +931,20 @@ bool TargetLowering::SimplifyDemandedBits(SDOperand Op, uint64_t DemandedMask, break; case ISD::SHL: if (ConstantSDNode *SA = dyn_cast(Op.getOperand(1))) { - unsigned ShAmt = SA->getValue(); - SDOperand InOp = Op.getOperand(0); + unsigned ShAmt = SA->getZExtValue(); + SDValue InOp = Op.getOperand(0); + + // If the shift count is an invalid immediate, don't do anything. + if (ShAmt >= BitWidth) + break; // If this is ((X >>u C1) << ShAmt), see if we can simplify this into a // single shift. We can do this if the bottom bits (which are shifted // out) are never demanded. if (InOp.getOpcode() == ISD::SRL && isa(InOp.getOperand(1))) { - if (ShAmt && (DemandedMask & ((1ULL << ShAmt)-1)) == 0) { - unsigned C1 = cast(InOp.getOperand(1))->getValue(); + if (ShAmt && (NewMask & APInt::getLowBitsSet(BitWidth, ShAmt)) == 0) { + unsigned C1= cast(InOp.getOperand(1))->getZExtValue(); unsigned Opc = ISD::SHL; int Diff = ShAmt-C1; if (Diff < 0) { @@ -580,37 +952,41 @@ bool TargetLowering::SimplifyDemandedBits(SDOperand Op, uint64_t DemandedMask, Opc = ISD::SRL; } - SDOperand NewSA = - TLO.DAG.getConstant(ShAmt-C1, Op.getOperand(0).getValueType()); - MVT::ValueType VT = Op.getValueType(); - return TLO.CombineTo(Op, TLO.DAG.getNode(ISD::SRL, VT, + SDValue NewSA = + TLO.DAG.getConstant(Diff, Op.getOperand(1).getValueType()); + MVT VT = Op.getValueType(); + return TLO.CombineTo(Op, TLO.DAG.getNode(Opc, VT, InOp.getOperand(0), NewSA)); } } - if (SimplifyDemandedBits(Op.getOperand(0), DemandedMask >> ShAmt, + if (SimplifyDemandedBits(Op.getOperand(0), NewMask.lshr(ShAmt), KnownZero, KnownOne, TLO, Depth+1)) return true; - KnownZero <<= SA->getValue(); - KnownOne <<= SA->getValue(); - KnownZero |= (1ULL << SA->getValue())-1; // low bits known zero. + KnownZero <<= SA->getZExtValue(); + KnownOne <<= SA->getZExtValue(); + // low bits known zero. + KnownZero |= APInt::getLowBitsSet(BitWidth, SA->getZExtValue()); } break; case ISD::SRL: if (ConstantSDNode *SA = dyn_cast(Op.getOperand(1))) { - MVT::ValueType VT = Op.getValueType(); - unsigned ShAmt = SA->getValue(); - uint64_t TypeMask = MVT::getIntVTBitMask(VT); - unsigned VTSize = MVT::getSizeInBits(VT); - SDOperand InOp = Op.getOperand(0); + MVT VT = Op.getValueType(); + unsigned ShAmt = SA->getZExtValue(); + unsigned VTSize = VT.getSizeInBits(); + SDValue InOp = Op.getOperand(0); + // If the shift count is an invalid immediate, don't do anything. + if (ShAmt >= BitWidth) + break; + // If this is ((X << C1) >>u ShAmt), see if we can simplify this into a // single shift. We can do this if the top bits (which are shifted out) // are never demanded. if (InOp.getOpcode() == ISD::SHL && isa(InOp.getOperand(1))) { - if (ShAmt && (DemandedMask & (~0ULL << (VTSize-ShAmt))) == 0) { - unsigned C1 = cast(InOp.getOperand(1))->getValue(); + if (ShAmt && (NewMask & APInt::getHighBitsSet(VTSize, ShAmt)) == 0) { + unsigned C1= cast(InOp.getOperand(1))->getZExtValue(); unsigned Opc = ISD::SRL; int Diff = ShAmt-C1; if (Diff < 0) { @@ -618,81 +994,80 @@ bool TargetLowering::SimplifyDemandedBits(SDOperand Op, uint64_t DemandedMask, Opc = ISD::SHL; } - SDOperand NewSA = - TLO.DAG.getConstant(Diff, Op.getOperand(0).getValueType()); + SDValue NewSA = + TLO.DAG.getConstant(Diff, Op.getOperand(1).getValueType()); return TLO.CombineTo(Op, TLO.DAG.getNode(Opc, VT, InOp.getOperand(0), NewSA)); } } // Compute the new bits that are at the top now. - if (SimplifyDemandedBits(InOp, (DemandedMask << ShAmt) & TypeMask, + if (SimplifyDemandedBits(InOp, (NewMask << ShAmt), KnownZero, KnownOne, TLO, Depth+1)) return true; assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?"); - KnownZero &= TypeMask; - KnownOne &= TypeMask; - KnownZero >>= ShAmt; - KnownOne >>= ShAmt; + KnownZero = KnownZero.lshr(ShAmt); + KnownOne = KnownOne.lshr(ShAmt); - uint64_t HighBits = (1ULL << ShAmt)-1; - HighBits <<= VTSize - ShAmt; + APInt HighBits = APInt::getHighBitsSet(BitWidth, ShAmt); KnownZero |= HighBits; // High bits known zero. } break; case ISD::SRA: if (ConstantSDNode *SA = dyn_cast(Op.getOperand(1))) { - MVT::ValueType VT = Op.getValueType(); - unsigned ShAmt = SA->getValue(); - - // Compute the new bits that are at the top now. - uint64_t TypeMask = MVT::getIntVTBitMask(VT); + MVT VT = Op.getValueType(); + unsigned ShAmt = SA->getZExtValue(); - uint64_t InDemandedMask = (DemandedMask << ShAmt) & TypeMask; + // If the shift count is an invalid immediate, don't do anything. + if (ShAmt >= BitWidth) + break; + + APInt InDemandedMask = (NewMask << ShAmt); // If any of the demanded bits are produced by the sign extension, we also // demand the input sign bit. - uint64_t HighBits = (1ULL << ShAmt)-1; - HighBits <<= MVT::getSizeInBits(VT) - ShAmt; - if (HighBits & DemandedMask) - InDemandedMask |= MVT::getIntVTSignBit(VT); + APInt HighBits = APInt::getHighBitsSet(BitWidth, ShAmt); + if (HighBits.intersects(NewMask)) + InDemandedMask |= APInt::getSignBit(VT.getSizeInBits()); if (SimplifyDemandedBits(Op.getOperand(0), InDemandedMask, KnownZero, KnownOne, TLO, Depth+1)) return true; assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?"); - KnownZero &= TypeMask; - KnownOne &= TypeMask; - KnownZero >>= ShAmt; - KnownOne >>= ShAmt; + KnownZero = KnownZero.lshr(ShAmt); + KnownOne = KnownOne.lshr(ShAmt); - // Handle the sign bits. - uint64_t SignBit = MVT::getIntVTSignBit(VT); - SignBit >>= ShAmt; // Adjust to where it is now in the mask. + // Handle the sign bit, adjusted to where it is now in the mask. + APInt SignBit = APInt::getSignBit(BitWidth).lshr(ShAmt); // If the input sign bit is known to be zero, or if none of the top bits // are demanded, turn this into an unsigned shift right. - if ((KnownZero & SignBit) || (HighBits & ~DemandedMask) == HighBits) { + if (KnownZero.intersects(SignBit) || (HighBits & ~NewMask) == HighBits) { return TLO.CombineTo(Op, TLO.DAG.getNode(ISD::SRL, VT, Op.getOperand(0), Op.getOperand(1))); - } else if (KnownOne & SignBit) { // New bits are known one. + } else if (KnownOne.intersects(SignBit)) { // New bits are known one. KnownOne |= HighBits; } } break; case ISD::SIGN_EXTEND_INREG: { - MVT::ValueType EVT = cast(Op.getOperand(1))->getVT(); + MVT EVT = cast(Op.getOperand(1))->getVT(); // Sign extension. Compute the demanded bits in the result that are not // present in the input. - uint64_t NewBits = ~MVT::getIntVTBitMask(EVT) & DemandedMask; + APInt NewBits = APInt::getHighBitsSet(BitWidth, + BitWidth - EVT.getSizeInBits()) & + NewMask; // If none of the extended bits are demanded, eliminate the sextinreg. if (NewBits == 0) return TLO.CombineTo(Op, Op.getOperand(0)); - uint64_t InSignBit = MVT::getIntVTSignBit(EVT); - int64_t InputDemandedBits = DemandedMask & MVT::getIntVTBitMask(EVT); + APInt InSignBit = APInt::getSignBit(EVT.getSizeInBits()); + InSignBit.zext(BitWidth); + APInt InputDemandedBits = APInt::getLowBitsSet(BitWidth, + EVT.getSizeInBits()) & + NewMask; // Since the sign extended bits are demanded, we know that the sign // bit is demanded. @@ -707,11 +1082,11 @@ bool TargetLowering::SimplifyDemandedBits(SDOperand Op, uint64_t DemandedMask, // top bits of the result. // If the input sign bit is known zero, convert this into a zero extension. - if (KnownZero & InSignBit) + if (KnownZero.intersects(InSignBit)) return TLO.CombineTo(Op, TLO.DAG.getZeroExtendInReg(Op.getOperand(0), EVT)); - if (KnownOne & InSignBit) { // Input sign bit known set + if (KnownOne.intersects(InSignBit)) { // Input sign bit known set KnownOne |= NewBits; KnownZero &= ~NewBits; } else { // Input sign bit unknown @@ -720,45 +1095,34 @@ bool TargetLowering::SimplifyDemandedBits(SDOperand Op, uint64_t DemandedMask, } break; } - case ISD::CTTZ: - case ISD::CTLZ: - case ISD::CTPOP: { - MVT::ValueType VT = Op.getValueType(); - unsigned LowBits = Log2_32(MVT::getSizeInBits(VT))+1; - KnownZero = ~((1ULL << LowBits)-1) & MVT::getIntVTBitMask(VT); - KnownOne = 0; - break; - } - case ISD::LOAD: { - if (ISD::isZEXTLoad(Op.Val)) { - LoadSDNode *LD = cast(Op); - MVT::ValueType VT = LD->getLoadedVT(); - KnownZero |= ~MVT::getIntVTBitMask(VT) & DemandedMask; - } - break; - } case ISD::ZERO_EXTEND: { - uint64_t InMask = MVT::getIntVTBitMask(Op.getOperand(0).getValueType()); + unsigned OperandBitWidth = Op.getOperand(0).getValueSizeInBits(); + APInt InMask = NewMask; + InMask.trunc(OperandBitWidth); // If none of the top bits are demanded, convert this into an any_extend. - uint64_t NewBits = (~InMask) & DemandedMask; - if (NewBits == 0) + APInt NewBits = + APInt::getHighBitsSet(BitWidth, BitWidth - OperandBitWidth) & NewMask; + if (!NewBits.intersects(NewMask)) return TLO.CombineTo(Op, TLO.DAG.getNode(ISD::ANY_EXTEND, Op.getValueType(), Op.getOperand(0))); - if (SimplifyDemandedBits(Op.getOperand(0), DemandedMask & InMask, + if (SimplifyDemandedBits(Op.getOperand(0), InMask, KnownZero, KnownOne, TLO, Depth+1)) return true; assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?"); + KnownZero.zext(BitWidth); + KnownOne.zext(BitWidth); KnownZero |= NewBits; break; } case ISD::SIGN_EXTEND: { - MVT::ValueType InVT = Op.getOperand(0).getValueType(); - uint64_t InMask = MVT::getIntVTBitMask(InVT); - uint64_t InSignBit = MVT::getIntVTSignBit(InVT); - uint64_t NewBits = (~InMask) & DemandedMask; + MVT InVT = Op.getOperand(0).getValueType(); + unsigned InBits = InVT.getSizeInBits(); + APInt InMask = APInt::getLowBitsSet(BitWidth, InBits); + APInt InSignBit = APInt::getBitsSet(BitWidth, InBits - 1, InBits); + APInt NewBits = ~InMask & NewMask; // If none of the top bits are demanded, convert this into an any_extend. if (NewBits == 0) @@ -767,21 +1131,24 @@ bool TargetLowering::SimplifyDemandedBits(SDOperand Op, uint64_t DemandedMask, // Since some of the sign extended bits are demanded, we know that the sign // bit is demanded. - uint64_t InDemandedBits = DemandedMask & InMask; + APInt InDemandedBits = InMask & NewMask; InDemandedBits |= InSignBit; + InDemandedBits.trunc(InBits); if (SimplifyDemandedBits(Op.getOperand(0), InDemandedBits, KnownZero, KnownOne, TLO, Depth+1)) return true; + KnownZero.zext(BitWidth); + KnownOne.zext(BitWidth); // If the sign bit is known zero, convert this to a zero extend. - if (KnownZero & InSignBit) + if (KnownZero.intersects(InSignBit)) return TLO.CombineTo(Op, TLO.DAG.getNode(ISD::ZERO_EXTEND, Op.getValueType(), Op.getOperand(0))); // If the sign bit is known one, the top bits match. - if (KnownOne & InSignBit) { + if (KnownOne.intersects(InSignBit)) { KnownOne |= NewBits; KnownZero &= ~NewBits; } else { // Otherwise, top bits aren't known. @@ -791,39 +1158,48 @@ bool TargetLowering::SimplifyDemandedBits(SDOperand Op, uint64_t DemandedMask, break; } case ISD::ANY_EXTEND: { - uint64_t InMask = MVT::getIntVTBitMask(Op.getOperand(0).getValueType()); - if (SimplifyDemandedBits(Op.getOperand(0), DemandedMask & InMask, + unsigned OperandBitWidth = Op.getOperand(0).getValueSizeInBits(); + APInt InMask = NewMask; + InMask.trunc(OperandBitWidth); + if (SimplifyDemandedBits(Op.getOperand(0), InMask, KnownZero, KnownOne, TLO, Depth+1)) return true; assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?"); + KnownZero.zext(BitWidth); + KnownOne.zext(BitWidth); break; } case ISD::TRUNCATE: { // Simplify the input, using demanded bit information, and compute the known // zero/one bits live out. - if (SimplifyDemandedBits(Op.getOperand(0), DemandedMask, + APInt TruncMask = NewMask; + TruncMask.zext(Op.getOperand(0).getValueSizeInBits()); + if (SimplifyDemandedBits(Op.getOperand(0), TruncMask, KnownZero, KnownOne, TLO, Depth+1)) return true; + KnownZero.trunc(BitWidth); + KnownOne.trunc(BitWidth); // If the input is only used by this truncate, see if we can shrink it based // on the known demanded bits. - if (Op.getOperand(0).Val->hasOneUse()) { - SDOperand In = Op.getOperand(0); + if (Op.getOperand(0).getNode()->hasOneUse()) { + SDValue In = Op.getOperand(0); + unsigned InBitWidth = In.getValueSizeInBits(); switch (In.getOpcode()) { default: break; case ISD::SRL: // Shrink SRL by a constant if none of the high bits shifted in are // demanded. if (ConstantSDNode *ShAmt = dyn_cast(In.getOperand(1))){ - uint64_t HighBits = MVT::getIntVTBitMask(In.getValueType()); - HighBits &= ~MVT::getIntVTBitMask(Op.getValueType()); - HighBits >>= ShAmt->getValue(); + APInt HighBits = APInt::getHighBitsSet(InBitWidth, + InBitWidth - BitWidth); + HighBits = HighBits.lshr(ShAmt->getZExtValue()); + HighBits.trunc(BitWidth); - if (ShAmt->getValue() < MVT::getSizeInBits(Op.getValueType()) && - (DemandedMask & HighBits) == 0) { + if (ShAmt->getZExtValue() < BitWidth && !(HighBits & NewMask)) { // None of the shifted in bits are needed. Add a truncate of the // shift input, then shift it. - SDOperand NewTrunc = TLO.DAG.getNode(ISD::TRUNCATE, + SDValue NewTrunc = TLO.DAG.getNode(ISD::TRUNCATE, Op.getValueType(), In.getOperand(0)); return TLO.CombineTo(Op, TLO.DAG.getNode(ISD::SRL,Op.getValueType(), @@ -835,369 +1211,63 @@ bool TargetLowering::SimplifyDemandedBits(SDOperand Op, uint64_t DemandedMask, } assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?"); - uint64_t OutMask = MVT::getIntVTBitMask(Op.getValueType()); - KnownZero &= OutMask; - KnownOne &= OutMask; break; } case ISD::AssertZext: { - MVT::ValueType VT = cast(Op.getOperand(1))->getVT(); - uint64_t InMask = MVT::getIntVTBitMask(VT); - if (SimplifyDemandedBits(Op.getOperand(0), DemandedMask & InMask, + MVT VT = cast(Op.getOperand(1))->getVT(); + APInt InMask = APInt::getLowBitsSet(BitWidth, + VT.getSizeInBits()); + if (SimplifyDemandedBits(Op.getOperand(0), InMask & NewMask, KnownZero, KnownOne, TLO, Depth+1)) return true; assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?"); - KnownZero |= ~InMask & DemandedMask; + KnownZero |= ~InMask & NewMask; break; } - case ISD::ADD: - case ISD::SUB: - case ISD::INTRINSIC_WO_CHAIN: - case ISD::INTRINSIC_W_CHAIN: - case ISD::INTRINSIC_VOID: + case ISD::BIT_CONVERT: +#if 0 + // If this is an FP->Int bitcast and if the sign bit is the only thing that + // is demanded, turn this into a FGETSIGN. + if (NewMask == MVT::getIntegerVTSignBit(Op.getValueType()) && + MVT::isFloatingPoint(Op.getOperand(0).getValueType()) && + !MVT::isVector(Op.getOperand(0).getValueType())) { + // Only do this xform if FGETSIGN is valid or if before legalize. + if (!TLO.AfterLegalize || + isOperationLegal(ISD::FGETSIGN, Op.getValueType())) { + // Make a FGETSIGN + SHL to move the sign bit into the appropriate + // place. We expect the SHL to be eliminated by other optimizations. + SDValue Sign = TLO.DAG.getNode(ISD::FGETSIGN, Op.getValueType(), + Op.getOperand(0)); + unsigned ShVal = Op.getValueType().getSizeInBits()-1; + SDValue ShAmt = TLO.DAG.getConstant(ShVal, getShiftAmountTy()); + return TLO.CombineTo(Op, TLO.DAG.getNode(ISD::SHL, Op.getValueType(), + Sign, ShAmt)); + } + } +#endif + break; + default: // Just use ComputeMaskedBits to compute output bits. - ComputeMaskedBits(Op, DemandedMask, KnownZero, KnownOne, Depth); + TLO.DAG.ComputeMaskedBits(Op, NewMask, KnownZero, KnownOne, Depth); break; } // If we know the value of all of the demanded bits, return this as a // constant. - if ((DemandedMask & (KnownZero|KnownOne)) == DemandedMask) + if ((NewMask & (KnownZero|KnownOne)) == NewMask) return TLO.CombineTo(Op, TLO.DAG.getConstant(KnownOne, Op.getValueType())); return false; } -/// MaskedValueIsZero - Return true if 'V & Mask' is known to be zero. We use -/// this predicate to simplify operations downstream. Mask is known to be zero -/// for bits that V cannot have. -bool TargetLowering::MaskedValueIsZero(SDOperand Op, uint64_t Mask, - unsigned Depth) const { - uint64_t KnownZero, KnownOne; - ComputeMaskedBits(Op, Mask, KnownZero, KnownOne, Depth); - assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?"); - return (KnownZero & Mask) == Mask; -} - -/// ComputeMaskedBits - Determine which of the bits specified in Mask are -/// known to be either zero or one and return them in the KnownZero/KnownOne -/// bitsets. This code only analyzes bits in Mask, in order to short-circuit -/// processing. -void TargetLowering::ComputeMaskedBits(SDOperand Op, uint64_t Mask, - uint64_t &KnownZero, uint64_t &KnownOne, - unsigned Depth) const { - KnownZero = KnownOne = 0; // Don't know anything. - if (Depth == 6 || Mask == 0) - return; // Limit search depth. - - uint64_t KnownZero2, KnownOne2; - - switch (Op.getOpcode()) { - case ISD::Constant: - // We know all of the bits for a constant! - KnownOne = cast(Op)->getValue() & Mask; - KnownZero = ~KnownOne & Mask; - return; - case ISD::AND: - // If either the LHS or the RHS are Zero, the result is zero. - ComputeMaskedBits(Op.getOperand(1), Mask, KnownZero, KnownOne, Depth+1); - Mask &= ~KnownZero; - ComputeMaskedBits(Op.getOperand(0), Mask, KnownZero2, KnownOne2, Depth+1); - assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?"); - assert((KnownZero2 & KnownOne2) == 0 && "Bits known to be one AND zero?"); - - // Output known-1 bits are only known if set in both the LHS & RHS. - KnownOne &= KnownOne2; - // Output known-0 are known to be clear if zero in either the LHS | RHS. - KnownZero |= KnownZero2; - return; - case ISD::OR: - ComputeMaskedBits(Op.getOperand(1), Mask, KnownZero, KnownOne, Depth+1); - Mask &= ~KnownOne; - ComputeMaskedBits(Op.getOperand(0), Mask, KnownZero2, KnownOne2, Depth+1); - assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?"); - assert((KnownZero2 & KnownOne2) == 0 && "Bits known to be one AND zero?"); - - // Output known-0 bits are only known if clear in both the LHS & RHS. - KnownZero &= KnownZero2; - // Output known-1 are known to be set if set in either the LHS | RHS. - KnownOne |= KnownOne2; - return; - case ISD::XOR: { - ComputeMaskedBits(Op.getOperand(1), Mask, KnownZero, KnownOne, Depth+1); - ComputeMaskedBits(Op.getOperand(0), Mask, KnownZero2, KnownOne2, Depth+1); - assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?"); - assert((KnownZero2 & KnownOne2) == 0 && "Bits known to be one AND zero?"); - - // Output known-0 bits are known if clear or set in both the LHS & RHS. - uint64_t KnownZeroOut = (KnownZero & KnownZero2) | (KnownOne & KnownOne2); - // Output known-1 are known to be set if set in only one of the LHS, RHS. - KnownOne = (KnownZero & KnownOne2) | (KnownOne & KnownZero2); - KnownZero = KnownZeroOut; - return; - } - case ISD::SELECT: - ComputeMaskedBits(Op.getOperand(2), Mask, KnownZero, KnownOne, Depth+1); - ComputeMaskedBits(Op.getOperand(1), Mask, KnownZero2, KnownOne2, Depth+1); - assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?"); - assert((KnownZero2 & KnownOne2) == 0 && "Bits known to be one AND zero?"); - - // Only known if known in both the LHS and RHS. - KnownOne &= KnownOne2; - KnownZero &= KnownZero2; - return; - case ISD::SELECT_CC: - ComputeMaskedBits(Op.getOperand(3), Mask, KnownZero, KnownOne, Depth+1); - ComputeMaskedBits(Op.getOperand(2), Mask, KnownZero2, KnownOne2, Depth+1); - assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?"); - assert((KnownZero2 & KnownOne2) == 0 && "Bits known to be one AND zero?"); - - // Only known if known in both the LHS and RHS. - KnownOne &= KnownOne2; - KnownZero &= KnownZero2; - return; - case ISD::SETCC: - // If we know the result of a setcc has the top bits zero, use this info. - if (getSetCCResultContents() == TargetLowering::ZeroOrOneSetCCResult) - KnownZero |= (MVT::getIntVTBitMask(Op.getValueType()) ^ 1ULL); - return; - case ISD::SHL: - // (shl X, C1) & C2 == 0 iff (X & C2 >>u C1) == 0 - if (ConstantSDNode *SA = dyn_cast(Op.getOperand(1))) { - ComputeMaskedBits(Op.getOperand(0), Mask >> SA->getValue(), - KnownZero, KnownOne, Depth+1); - assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?"); - KnownZero <<= SA->getValue(); - KnownOne <<= SA->getValue(); - KnownZero |= (1ULL << SA->getValue())-1; // low bits known zero. - } - return; - case ISD::SRL: - // (ushr X, C1) & C2 == 0 iff (-1 >> C1) & C2 == 0 - if (ConstantSDNode *SA = dyn_cast(Op.getOperand(1))) { - MVT::ValueType VT = Op.getValueType(); - unsigned ShAmt = SA->getValue(); - - uint64_t TypeMask = MVT::getIntVTBitMask(VT); - ComputeMaskedBits(Op.getOperand(0), (Mask << ShAmt) & TypeMask, - KnownZero, KnownOne, Depth+1); - assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?"); - KnownZero &= TypeMask; - KnownOne &= TypeMask; - KnownZero >>= ShAmt; - KnownOne >>= ShAmt; - - uint64_t HighBits = (1ULL << ShAmt)-1; - HighBits <<= MVT::getSizeInBits(VT)-ShAmt; - KnownZero |= HighBits; // High bits known zero. - } - return; - case ISD::SRA: - if (ConstantSDNode *SA = dyn_cast(Op.getOperand(1))) { - MVT::ValueType VT = Op.getValueType(); - unsigned ShAmt = SA->getValue(); - - // Compute the new bits that are at the top now. - uint64_t TypeMask = MVT::getIntVTBitMask(VT); - - uint64_t InDemandedMask = (Mask << ShAmt) & TypeMask; - // If any of the demanded bits are produced by the sign extension, we also - // demand the input sign bit. - uint64_t HighBits = (1ULL << ShAmt)-1; - HighBits <<= MVT::getSizeInBits(VT) - ShAmt; - if (HighBits & Mask) - InDemandedMask |= MVT::getIntVTSignBit(VT); - - ComputeMaskedBits(Op.getOperand(0), InDemandedMask, KnownZero, KnownOne, - Depth+1); - assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?"); - KnownZero &= TypeMask; - KnownOne &= TypeMask; - KnownZero >>= ShAmt; - KnownOne >>= ShAmt; - - // Handle the sign bits. - uint64_t SignBit = MVT::getIntVTSignBit(VT); - SignBit >>= ShAmt; // Adjust to where it is now in the mask. - - if (KnownZero & SignBit) { - KnownZero |= HighBits; // New bits are known zero. - } else if (KnownOne & SignBit) { - KnownOne |= HighBits; // New bits are known one. - } - } - return; - case ISD::SIGN_EXTEND_INREG: { - MVT::ValueType EVT = cast(Op.getOperand(1))->getVT(); - - // Sign extension. Compute the demanded bits in the result that are not - // present in the input. - uint64_t NewBits = ~MVT::getIntVTBitMask(EVT) & Mask; - - uint64_t InSignBit = MVT::getIntVTSignBit(EVT); - int64_t InputDemandedBits = Mask & MVT::getIntVTBitMask(EVT); - - // If the sign extended bits are demanded, we know that the sign - // bit is demanded. - if (NewBits) - InputDemandedBits |= InSignBit; - - ComputeMaskedBits(Op.getOperand(0), InputDemandedBits, - KnownZero, KnownOne, Depth+1); - assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?"); - - // If the sign bit of the input is known set or clear, then we know the - // top bits of the result. - if (KnownZero & InSignBit) { // Input sign bit known clear - KnownZero |= NewBits; - KnownOne &= ~NewBits; - } else if (KnownOne & InSignBit) { // Input sign bit known set - KnownOne |= NewBits; - KnownZero &= ~NewBits; - } else { // Input sign bit unknown - KnownZero &= ~NewBits; - KnownOne &= ~NewBits; - } - return; - } - case ISD::CTTZ: - case ISD::CTLZ: - case ISD::CTPOP: { - MVT::ValueType VT = Op.getValueType(); - unsigned LowBits = Log2_32(MVT::getSizeInBits(VT))+1; - KnownZero = ~((1ULL << LowBits)-1) & MVT::getIntVTBitMask(VT); - KnownOne = 0; - return; - } - case ISD::LOAD: { - if (ISD::isZEXTLoad(Op.Val)) { - LoadSDNode *LD = cast(Op); - MVT::ValueType VT = LD->getLoadedVT(); - KnownZero |= ~MVT::getIntVTBitMask(VT) & Mask; - } - return; - } - case ISD::ZERO_EXTEND: { - uint64_t InMask = MVT::getIntVTBitMask(Op.getOperand(0).getValueType()); - uint64_t NewBits = (~InMask) & Mask; - ComputeMaskedBits(Op.getOperand(0), Mask & InMask, KnownZero, - KnownOne, Depth+1); - KnownZero |= NewBits & Mask; - KnownOne &= ~NewBits; - return; - } - case ISD::SIGN_EXTEND: { - MVT::ValueType InVT = Op.getOperand(0).getValueType(); - unsigned InBits = MVT::getSizeInBits(InVT); - uint64_t InMask = MVT::getIntVTBitMask(InVT); - uint64_t InSignBit = 1ULL << (InBits-1); - uint64_t NewBits = (~InMask) & Mask; - uint64_t InDemandedBits = Mask & InMask; - - // If any of the sign extended bits are demanded, we know that the sign - // bit is demanded. - if (NewBits & Mask) - InDemandedBits |= InSignBit; - - ComputeMaskedBits(Op.getOperand(0), InDemandedBits, KnownZero, - KnownOne, Depth+1); - // If the sign bit is known zero or one, the top bits match. - if (KnownZero & InSignBit) { - KnownZero |= NewBits; - KnownOne &= ~NewBits; - } else if (KnownOne & InSignBit) { - KnownOne |= NewBits; - KnownZero &= ~NewBits; - } else { // Otherwise, top bits aren't known. - KnownOne &= ~NewBits; - KnownZero &= ~NewBits; - } - return; - } - case ISD::ANY_EXTEND: { - MVT::ValueType VT = Op.getOperand(0).getValueType(); - ComputeMaskedBits(Op.getOperand(0), Mask & MVT::getIntVTBitMask(VT), - KnownZero, KnownOne, Depth+1); - return; - } - case ISD::TRUNCATE: { - ComputeMaskedBits(Op.getOperand(0), Mask, KnownZero, KnownOne, Depth+1); - assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?"); - uint64_t OutMask = MVT::getIntVTBitMask(Op.getValueType()); - KnownZero &= OutMask; - KnownOne &= OutMask; - break; - } - case ISD::AssertZext: { - MVT::ValueType VT = cast(Op.getOperand(1))->getVT(); - uint64_t InMask = MVT::getIntVTBitMask(VT); - ComputeMaskedBits(Op.getOperand(0), Mask & InMask, KnownZero, - KnownOne, Depth+1); - KnownZero |= (~InMask) & Mask; - return; - } - case ISD::ADD: { - // If either the LHS or the RHS are Zero, the result is zero. - ComputeMaskedBits(Op.getOperand(1), Mask, KnownZero, KnownOne, Depth+1); - ComputeMaskedBits(Op.getOperand(0), Mask, KnownZero2, KnownOne2, Depth+1); - assert((KnownZero & KnownOne) == 0 && "Bits known to be one AND zero?"); - assert((KnownZero2 & KnownOne2) == 0 && "Bits known to be one AND zero?"); - - // Output known-0 bits are known if clear or set in both the low clear bits - // common to both LHS & RHS. For example, 8+(X<<3) is known to have the - // low 3 bits clear. - uint64_t KnownZeroOut = std::min(CountTrailingZeros_64(~KnownZero), - CountTrailingZeros_64(~KnownZero2)); - - KnownZero = (1ULL << KnownZeroOut) - 1; - KnownOne = 0; - return; - } - case ISD::SUB: { - ConstantSDNode *CLHS = dyn_cast(Op.getOperand(0)); - if (!CLHS) return; - - // We know that the top bits of C-X are clear if X contains less bits - // than C (i.e. no wrap-around can happen). For example, 20-X is - // positive if we can prove that X is >= 0 and < 16. - MVT::ValueType VT = CLHS->getValueType(0); - if ((CLHS->getValue() & MVT::getIntVTSignBit(VT)) == 0) { // sign bit clear - unsigned NLZ = CountLeadingZeros_64(CLHS->getValue()+1); - uint64_t MaskV = (1ULL << (63-NLZ))-1; // NLZ can't be 64 with no sign bit - MaskV = ~MaskV & MVT::getIntVTBitMask(VT); - ComputeMaskedBits(Op.getOperand(1), MaskV, KnownZero, KnownOne, Depth+1); - - // If all of the MaskV bits are known to be zero, then we know the output - // top bits are zero, because we now know that the output is from [0-C]. - if ((KnownZero & MaskV) == MaskV) { - unsigned NLZ2 = CountLeadingZeros_64(CLHS->getValue()); - KnownZero = ~((1ULL << (64-NLZ2))-1) & Mask; // Top bits known zero. - KnownOne = 0; // No one bits known. - } else { - KnownZero = KnownOne = 0; // Otherwise, nothing known. - } - } - return; - } - default: - // Allow the target to implement this method for its nodes. - if (Op.getOpcode() >= ISD::BUILTIN_OP_END) { - case ISD::INTRINSIC_WO_CHAIN: - case ISD::INTRINSIC_W_CHAIN: - case ISD::INTRINSIC_VOID: - computeMaskedBitsForTargetNode(Op, Mask, KnownZero, KnownOne); - } - return; - } -} - /// computeMaskedBitsForTargetNode - Determine which of the bits specified /// in Mask are known to be either zero or one and return them in the /// KnownZero/KnownOne bitsets. -void TargetLowering::computeMaskedBitsForTargetNode(const SDOperand Op, - uint64_t Mask, - uint64_t &KnownZero, - uint64_t &KnownOne, +void TargetLowering::computeMaskedBitsForTargetNode(const SDValue Op, + const APInt &Mask, + APInt &KnownZero, + APInt &KnownOne, + const SelectionDAG &DAG, unsigned Depth) const { assert((Op.getOpcode() >= ISD::BUILTIN_OP_END || Op.getOpcode() == ISD::INTRINSIC_WO_CHAIN || @@ -1205,230 +1275,13 @@ void TargetLowering::computeMaskedBitsForTargetNode(const SDOperand Op, Op.getOpcode() == ISD::INTRINSIC_VOID) && "Should use MaskedValueIsZero if you don't know whether Op" " is a target node!"); - KnownZero = 0; - KnownOne = 0; -} - -/// ComputeNumSignBits - Return the number of times the sign bit of the -/// register is replicated into the other bits. We know that at least 1 bit -/// is always equal to the sign bit (itself), but other cases can give us -/// information. For example, immediately after an "SRA X, 2", we know that -/// the top 3 bits are all equal to each other, so we return 3. -unsigned TargetLowering::ComputeNumSignBits(SDOperand Op, unsigned Depth) const{ - MVT::ValueType VT = Op.getValueType(); - assert(MVT::isInteger(VT) && "Invalid VT!"); - unsigned VTBits = MVT::getSizeInBits(VT); - unsigned Tmp, Tmp2; - - if (Depth == 6) - return 1; // Limit search depth. - - switch (Op.getOpcode()) { - default: break; - case ISD::AssertSext: - Tmp = MVT::getSizeInBits(cast(Op.getOperand(1))->getVT()); - return VTBits-Tmp+1; - case ISD::AssertZext: - Tmp = MVT::getSizeInBits(cast(Op.getOperand(1))->getVT()); - return VTBits-Tmp; - - case ISD::Constant: { - uint64_t Val = cast(Op)->getValue(); - // If negative, invert the bits, then look at it. - if (Val & MVT::getIntVTSignBit(VT)) - Val = ~Val; - - // Shift the bits so they are the leading bits in the int64_t. - Val <<= 64-VTBits; - - // Return # leading zeros. We use 'min' here in case Val was zero before - // shifting. We don't want to return '64' as for an i32 "0". - return std::min(VTBits, CountLeadingZeros_64(Val)); - } - - case ISD::SIGN_EXTEND: - Tmp = VTBits-MVT::getSizeInBits(Op.getOperand(0).getValueType()); - return ComputeNumSignBits(Op.getOperand(0), Depth+1) + Tmp; - - case ISD::SIGN_EXTEND_INREG: - // Max of the input and what this extends. - Tmp = MVT::getSizeInBits(cast(Op.getOperand(1))->getVT()); - Tmp = VTBits-Tmp+1; - - Tmp2 = ComputeNumSignBits(Op.getOperand(0), Depth+1); - return std::max(Tmp, Tmp2); - - case ISD::SRA: - Tmp = ComputeNumSignBits(Op.getOperand(0), Depth+1); - // SRA X, C -> adds C sign bits. - if (ConstantSDNode *C = dyn_cast(Op.getOperand(1))) { - Tmp += C->getValue(); - if (Tmp > VTBits) Tmp = VTBits; - } - return Tmp; - case ISD::SHL: - if (ConstantSDNode *C = dyn_cast(Op.getOperand(1))) { - // shl destroys sign bits. - Tmp = ComputeNumSignBits(Op.getOperand(0), Depth+1); - if (C->getValue() >= VTBits || // Bad shift. - C->getValue() >= Tmp) break; // Shifted all sign bits out. - return Tmp - C->getValue(); - } - break; - case ISD::AND: - case ISD::OR: - case ISD::XOR: // NOT is handled here. - // Logical binary ops preserve the number of sign bits. - Tmp = ComputeNumSignBits(Op.getOperand(0), Depth+1); - if (Tmp == 1) return 1; // Early out. - Tmp2 = ComputeNumSignBits(Op.getOperand(1), Depth+1); - return std::min(Tmp, Tmp2); - - case ISD::SELECT: - Tmp = ComputeNumSignBits(Op.getOperand(0), Depth+1); - if (Tmp == 1) return 1; // Early out. - Tmp2 = ComputeNumSignBits(Op.getOperand(1), Depth+1); - return std::min(Tmp, Tmp2); - - case ISD::SETCC: - // If setcc returns 0/-1, all bits are sign bits. - if (getSetCCResultContents() == ZeroOrNegativeOneSetCCResult) - return VTBits; - break; - case ISD::ROTL: - case ISD::ROTR: - if (ConstantSDNode *C = dyn_cast(Op.getOperand(1))) { - unsigned RotAmt = C->getValue() & (VTBits-1); - - // Handle rotate right by N like a rotate left by 32-N. - if (Op.getOpcode() == ISD::ROTR) - RotAmt = (VTBits-RotAmt) & (VTBits-1); - - // If we aren't rotating out all of the known-in sign bits, return the - // number that are left. This handles rotl(sext(x), 1) for example. - Tmp = ComputeNumSignBits(Op.getOperand(0), Depth+1); - if (Tmp > RotAmt+1) return Tmp-RotAmt; - } - break; - case ISD::ADD: - // Add can have at most one carry bit. Thus we know that the output - // is, at worst, one more bit than the inputs. - Tmp = ComputeNumSignBits(Op.getOperand(0), Depth+1); - if (Tmp == 1) return 1; // Early out. - - // Special case decrementing a value (ADD X, -1): - if (ConstantSDNode *CRHS = dyn_cast(Op.getOperand(0))) - if (CRHS->isAllOnesValue()) { - uint64_t KnownZero, KnownOne; - uint64_t Mask = MVT::getIntVTBitMask(VT); - ComputeMaskedBits(Op.getOperand(0), Mask, KnownZero, KnownOne, Depth+1); - - // If the input is known to be 0 or 1, the output is 0/-1, which is all - // sign bits set. - if ((KnownZero|1) == Mask) - return VTBits; - - // If we are subtracting one from a positive number, there is no carry - // out of the result. - if (KnownZero & MVT::getIntVTSignBit(VT)) - return Tmp; - } - - Tmp2 = ComputeNumSignBits(Op.getOperand(1), Depth+1); - if (Tmp2 == 1) return 1; - return std::min(Tmp, Tmp2)-1; - break; - - case ISD::SUB: - Tmp2 = ComputeNumSignBits(Op.getOperand(1), Depth+1); - if (Tmp2 == 1) return 1; - - // Handle NEG. - if (ConstantSDNode *CLHS = dyn_cast(Op.getOperand(0))) - if (CLHS->getValue() == 0) { - uint64_t KnownZero, KnownOne; - uint64_t Mask = MVT::getIntVTBitMask(VT); - ComputeMaskedBits(Op.getOperand(1), Mask, KnownZero, KnownOne, Depth+1); - // If the input is known to be 0 or 1, the output is 0/-1, which is all - // sign bits set. - if ((KnownZero|1) == Mask) - return VTBits; - - // If the input is known to be positive (the sign bit is known clear), - // the output of the NEG has the same number of sign bits as the input. - if (KnownZero & MVT::getIntVTSignBit(VT)) - return Tmp2; - - // Otherwise, we treat this like a SUB. - } - - // Sub can have at most one carry bit. Thus we know that the output - // is, at worst, one more bit than the inputs. - Tmp = ComputeNumSignBits(Op.getOperand(0), Depth+1); - if (Tmp == 1) return 1; // Early out. - return std::min(Tmp, Tmp2)-1; - break; - case ISD::TRUNCATE: - // FIXME: it's tricky to do anything useful for this, but it is an important - // case for targets like X86. - break; - } - - // Handle LOADX separately here. EXTLOAD case will fallthrough. - if (Op.getOpcode() == ISD::LOAD) { - LoadSDNode *LD = cast(Op); - unsigned ExtType = LD->getExtensionType(); - switch (ExtType) { - default: break; - case ISD::SEXTLOAD: // '17' bits known - Tmp = MVT::getSizeInBits(LD->getLoadedVT()); - return VTBits-Tmp+1; - case ISD::ZEXTLOAD: // '16' bits known - Tmp = MVT::getSizeInBits(LD->getLoadedVT()); - return VTBits-Tmp; - } - } - - // Allow the target to implement this method for its nodes. - if (Op.getOpcode() >= ISD::BUILTIN_OP_END || - Op.getOpcode() == ISD::INTRINSIC_WO_CHAIN || - Op.getOpcode() == ISD::INTRINSIC_W_CHAIN || - Op.getOpcode() == ISD::INTRINSIC_VOID) { - unsigned NumBits = ComputeNumSignBitsForTargetNode(Op, Depth); - if (NumBits > 1) return NumBits; - } - - // Finally, if we can prove that the top bits of the result are 0's or 1's, - // use this information. - uint64_t KnownZero, KnownOne; - uint64_t Mask = MVT::getIntVTBitMask(VT); - ComputeMaskedBits(Op, Mask, KnownZero, KnownOne, Depth); - - uint64_t SignBit = MVT::getIntVTSignBit(VT); - if (KnownZero & SignBit) { // SignBit is 0 - Mask = KnownZero; - } else if (KnownOne & SignBit) { // SignBit is 1; - Mask = KnownOne; - } else { - // Nothing known. - return 1; - } - - // Okay, we know that the sign bit in Mask is set. Use CLZ to determine - // the number of identical bits in the top of the input value. - Mask ^= ~0ULL; - Mask <<= 64-VTBits; - // Return # leading zeros. We use 'min' here in case Val was zero before - // shifting. We don't want to return '64' as for an i32 "0". - return std::min(VTBits, CountLeadingZeros_64(Mask)); + KnownZero = KnownOne = APInt(Mask.getBitWidth(), 0); } - - /// ComputeNumSignBitsForTargetNode - This method can be implemented by /// targets that want to expose additional information about sign bits to the /// DAG Combiner. -unsigned TargetLowering::ComputeNumSignBitsForTargetNode(SDOperand Op, +unsigned TargetLowering::ComputeNumSignBitsForTargetNode(SDValue Op, unsigned Depth) const { assert((Op.getOpcode() >= ISD::BUILTIN_OP_END || Op.getOpcode() == ISD::INTRINSIC_WO_CHAIN || @@ -1441,9 +1294,9 @@ unsigned TargetLowering::ComputeNumSignBitsForTargetNode(SDOperand Op, /// SimplifySetCC - Try to simplify a setcc built with the specified operands -/// and cc. If it is unable to simplify it, return a null SDOperand. -SDOperand -TargetLowering::SimplifySetCC(MVT::ValueType VT, SDOperand N0, SDOperand N1, +/// and cc. If it is unable to simplify it, return a null SDValue. +SDValue +TargetLowering::SimplifySetCC(MVT VT, SDValue N0, SDValue N1, ISD::CondCode Cond, bool foldBooleans, DAGCombinerInfo &DCI) const { SelectionDAG &DAG = DCI.DAG; @@ -1457,9 +1310,9 @@ TargetLowering::SimplifySetCC(MVT::ValueType VT, SDOperand N0, SDOperand N1, case ISD::SETTRUE2: return DAG.getConstant(1, VT); } - if (ConstantSDNode *N1C = dyn_cast(N1.Val)) { - uint64_t C1 = N1C->getValue(); - if (isa(N0.Val)) { + if (ConstantSDNode *N1C = dyn_cast(N1.getNode())) { + const APInt &C1 = N1C->getAPIntValue(); + if (isa(N0.getNode())) { return DAG.FoldSetCC(VT, N0, N1, Cond); } else { // If the LHS is '(srl (ctlz x), 5)', the RHS is 0/1, and this is an @@ -1468,9 +1321,9 @@ TargetLowering::SimplifySetCC(MVT::ValueType VT, SDOperand N0, SDOperand N1, if (N0.getOpcode() == ISD::SRL && (C1 == 0 || C1 == 1) && N0.getOperand(0).getOpcode() == ISD::CTLZ && N0.getOperand(1).getOpcode() == ISD::Constant) { - unsigned ShAmt = cast(N0.getOperand(1))->getValue(); + unsigned ShAmt = cast(N0.getOperand(1))->getZExtValue(); if ((Cond == ISD::SETEQ || Cond == ISD::SETNE) && - ShAmt == Log2_32(MVT::getSizeInBits(N0.getValueType()))) { + ShAmt == Log2_32(N0.getValueType().getSizeInBits())) { if ((C1 == 0) == (Cond == ISD::SETEQ)) { // (srl (ctlz x), 5) == 0 -> X != 0 // (srl (ctlz x), 5) != 1 -> X != 0 @@ -1480,7 +1333,7 @@ TargetLowering::SimplifySetCC(MVT::ValueType VT, SDOperand N0, SDOperand N1, // (srl (ctlz x), 5) == 1 -> X == 0 Cond = ISD::SETEQ; } - SDOperand Zero = DAG.getConstant(0, N0.getValueType()); + SDValue Zero = DAG.getConstant(0, N0.getValueType()); return DAG.getSetCC(VT, N0.getOperand(0).getOperand(0), Zero, Cond); } @@ -1488,12 +1341,12 @@ TargetLowering::SimplifySetCC(MVT::ValueType VT, SDOperand N0, SDOperand N1, // If the LHS is a ZERO_EXTEND, perform the comparison on the input. if (N0.getOpcode() == ISD::ZERO_EXTEND) { - unsigned InSize = MVT::getSizeInBits(N0.getOperand(0).getValueType()); + unsigned InSize = N0.getOperand(0).getValueType().getSizeInBits(); // If the comparison constant has bits in the upper part, the // zero-extended value could never match. - if (C1 & (~0ULL << InSize)) { - unsigned VSize = MVT::getSizeInBits(N0.getValueType()); + if (C1.intersects(APInt::getHighBitsSet(C1.getBitWidth(), + C1.getBitWidth() - InSize))) { switch (Cond) { case ISD::SETUGT: case ISD::SETUGE: @@ -1504,11 +1357,11 @@ TargetLowering::SimplifySetCC(MVT::ValueType VT, SDOperand N0, SDOperand N1, case ISD::SETGT: case ISD::SETGE: // True if the sign bit of C1 is set. - return DAG.getConstant((C1 & (1ULL << (VSize-1))) != 0, VT); + return DAG.getConstant(C1.isNegative(), VT); case ISD::SETLT: case ISD::SETLE: // True if the sign bit of C1 isn't set. - return DAG.getConstant((C1 & (1ULL << (VSize-1))) == 0, VT); + return DAG.getConstant(C1.isNonNegative(), VT); default: break; } @@ -1523,55 +1376,58 @@ TargetLowering::SimplifySetCC(MVT::ValueType VT, SDOperand N0, SDOperand N1, case ISD::SETULT: case ISD::SETULE: return DAG.getSetCC(VT, N0.getOperand(0), - DAG.getConstant(C1, N0.getOperand(0).getValueType()), + DAG.getConstant(APInt(C1).trunc(InSize), + N0.getOperand(0).getValueType()), Cond); default: break; // todo, be more careful with signed comparisons } } else if (N0.getOpcode() == ISD::SIGN_EXTEND_INREG && (Cond == ISD::SETEQ || Cond == ISD::SETNE)) { - MVT::ValueType ExtSrcTy = cast(N0.getOperand(1))->getVT(); - unsigned ExtSrcTyBits = MVT::getSizeInBits(ExtSrcTy); - MVT::ValueType ExtDstTy = N0.getValueType(); - unsigned ExtDstTyBits = MVT::getSizeInBits(ExtDstTy); + MVT ExtSrcTy = cast(N0.getOperand(1))->getVT(); + unsigned ExtSrcTyBits = ExtSrcTy.getSizeInBits(); + MVT ExtDstTy = N0.getValueType(); + unsigned ExtDstTyBits = ExtDstTy.getSizeInBits(); // If the extended part has any inconsistent bits, it cannot ever // compare equal. In other words, they have to be all ones or all // zeros. - uint64_t ExtBits = - (~0ULL >> (64-ExtSrcTyBits)) & (~0ULL << (ExtDstTyBits-1)); + APInt ExtBits = + APInt::getHighBitsSet(ExtDstTyBits, ExtDstTyBits - ExtSrcTyBits); if ((C1 & ExtBits) != 0 && (C1 & ExtBits) != ExtBits) return DAG.getConstant(Cond == ISD::SETNE, VT); - SDOperand ZextOp; - MVT::ValueType Op0Ty = N0.getOperand(0).getValueType(); + SDValue ZextOp; + MVT Op0Ty = N0.getOperand(0).getValueType(); if (Op0Ty == ExtSrcTy) { ZextOp = N0.getOperand(0); } else { - int64_t Imm = ~0ULL >> (64-ExtSrcTyBits); + APInt Imm = APInt::getLowBitsSet(ExtDstTyBits, ExtSrcTyBits); ZextOp = DAG.getNode(ISD::AND, Op0Ty, N0.getOperand(0), DAG.getConstant(Imm, Op0Ty)); } if (!DCI.isCalledByLegalizer()) - DCI.AddToWorklist(ZextOp.Val); + DCI.AddToWorklist(ZextOp.getNode()); // Otherwise, make this a use of a zext. return DAG.getSetCC(VT, ZextOp, - DAG.getConstant(C1 & (~0ULL>>(64-ExtSrcTyBits)), + DAG.getConstant(C1 & APInt::getLowBitsSet( + ExtDstTyBits, + ExtSrcTyBits), ExtDstTy), Cond); - } else if ((N1C->getValue() == 0 || N1C->getValue() == 1) && + } else if ((N1C->isNullValue() || N1C->getAPIntValue() == 1) && (Cond == ISD::SETEQ || Cond == ISD::SETNE)) { // SETCC (SETCC), [0|1], [EQ|NE] -> SETCC if (N0.getOpcode() == ISD::SETCC) { - bool TrueWhenTrue = (Cond == ISD::SETEQ) ^ (N1C->getValue() != 1); + bool TrueWhenTrue = (Cond == ISD::SETEQ) ^ (N1C->getZExtValue() != 1); if (TrueWhenTrue) return N0; // Invert the condition. ISD::CondCode CC = cast(N0.getOperand(2))->get(); CC = ISD::getSetCCInverse(CC, - MVT::isInteger(N0.getOperand(0).getValueType())); + N0.getOperand(0).getValueType().isInteger()); return DAG.getSetCC(VT, N0.getOperand(0), N0.getOperand(1), CC); } @@ -1580,12 +1436,15 @@ TargetLowering::SimplifySetCC(MVT::ValueType VT, SDOperand N0, SDOperand N1, N0.getOperand(0).getOpcode() == ISD::XOR && N0.getOperand(1) == N0.getOperand(0).getOperand(1))) && isa(N0.getOperand(1)) && - cast(N0.getOperand(1))->getValue() == 1) { + cast(N0.getOperand(1))->getAPIntValue() == 1) { // If this is (X^1) == 0/1, swap the RHS and eliminate the xor. We // can only do this if the top bits are known zero. - if (MaskedValueIsZero(N0, MVT::getIntVTBitMask(N0.getValueType())-1)){ + unsigned BitWidth = N0.getValueSizeInBits(); + if (DAG.MaskedValueIsZero(N0, + APInt::getHighBitsSet(BitWidth, + BitWidth-1))) { // Okay, get the un-inverted input value. - SDOperand Val; + SDValue Val; if (N0.getOpcode() == ISD::XOR) Val = N0.getOperand(0); else { @@ -1602,31 +1461,28 @@ TargetLowering::SimplifySetCC(MVT::ValueType VT, SDOperand N0, SDOperand N1, } } - uint64_t MinVal, MaxVal; - unsigned OperandBitSize = MVT::getSizeInBits(N1C->getValueType(0)); + APInt MinVal, MaxVal; + unsigned OperandBitSize = N1C->getValueType(0).getSizeInBits(); if (ISD::isSignedIntSetCC(Cond)) { - MinVal = 1ULL << (OperandBitSize-1); - if (OperandBitSize != 1) // Avoid X >> 64, which is undefined. - MaxVal = ~0ULL >> (65-OperandBitSize); - else - MaxVal = 0; + MinVal = APInt::getSignedMinValue(OperandBitSize); + MaxVal = APInt::getSignedMaxValue(OperandBitSize); } else { - MinVal = 0; - MaxVal = ~0ULL >> (64-OperandBitSize); + MinVal = APInt::getMinValue(OperandBitSize); + MaxVal = APInt::getMaxValue(OperandBitSize); } // Canonicalize GE/LE comparisons to use GT/LT comparisons. if (Cond == ISD::SETGE || Cond == ISD::SETUGE) { if (C1 == MinVal) return DAG.getConstant(1, VT); // X >= MIN --> true - --C1; // X >= C0 --> X > (C0-1) - return DAG.getSetCC(VT, N0, DAG.getConstant(C1, N1.getValueType()), + // X >= C0 --> X > (C0-1) + return DAG.getSetCC(VT, N0, DAG.getConstant(C1-1, N1.getValueType()), (Cond == ISD::SETGE) ? ISD::SETGT : ISD::SETUGT); } if (Cond == ISD::SETLE || Cond == ISD::SETULE) { if (C1 == MaxVal) return DAG.getConstant(1, VT); // X <= MAX --> true - ++C1; // X <= C0 --> X < (C0+1) - return DAG.getSetCC(VT, N0, DAG.getConstant(C1, N1.getValueType()), + // X <= C0 --> X < (C0+1) + return DAG.getSetCC(VT, N0, DAG.getConstant(C1+1, N1.getValueType()), (Cond == ISD::SETLE) ? ISD::SETLT : ISD::SETULT); } @@ -1673,35 +1529,57 @@ TargetLowering::SimplifySetCC(MVT::ValueType VT, SDOperand N0, SDOperand N1, dyn_cast(N0.getOperand(1))) { if (Cond == ISD::SETNE && C1 == 0) {// (X & 8) != 0 --> (X & 8) >> 3 // Perform the xform if the AND RHS is a single bit. - if (isPowerOf2_64(AndRHS->getValue())) { + if (isPowerOf2_64(AndRHS->getZExtValue())) { return DAG.getNode(ISD::SRL, VT, N0, - DAG.getConstant(Log2_64(AndRHS->getValue()), + DAG.getConstant(Log2_64(AndRHS->getZExtValue()), getShiftAmountTy())); } - } else if (Cond == ISD::SETEQ && C1 == AndRHS->getValue()) { + } else if (Cond == ISD::SETEQ && C1 == AndRHS->getZExtValue()) { // (X & 8) == 8 --> (X & 8) >> 3 // Perform the xform if C1 is a single bit. - if (isPowerOf2_64(C1)) { + if (C1.isPowerOf2()) { return DAG.getNode(ISD::SRL, VT, N0, - DAG.getConstant(Log2_64(C1), getShiftAmountTy())); + DAG.getConstant(C1.logBase2(), getShiftAmountTy())); } } } } - } else if (isa(N0.Val)) { + } else if (isa(N0.getNode())) { // Ensure that the constant occurs on the RHS. return DAG.getSetCC(VT, N1, N0, ISD::getSetCCSwappedOperands(Cond)); } - if (isa(N0.Val)) { + if (isa(N0.getNode())) { // Constant fold or commute setcc. - SDOperand O = DAG.FoldSetCC(VT, N0, N1, Cond); - if (O.Val) return O; + SDValue O = DAG.FoldSetCC(VT, N0, N1, Cond); + if (O.getNode()) return O; + } else if (ConstantFPSDNode *CFP = dyn_cast(N1.getNode())) { + // If the RHS of an FP comparison is a constant, simplify it away in + // some cases. + if (CFP->getValueAPF().isNaN()) { + // If an operand is known to be a nan, we can fold it. + switch (ISD::getUnorderedFlavor(Cond)) { + default: assert(0 && "Unknown flavor!"); + case 0: // Known false. + return DAG.getConstant(0, VT); + case 1: // Known true. + return DAG.getConstant(1, VT); + case 2: // Undefined. + return DAG.getNode(ISD::UNDEF, VT); + } + } + + // Otherwise, we know the RHS is not a NaN. Simplify the node to drop the + // constant if knowing that the operand is non-nan is enough. We prefer to + // have SETO(x,x) instead of SETO(x, 0.0) because this avoids having to + // materialize 0.0. + if (Cond == ISD::SETO || Cond == ISD::SETUO) + return DAG.getSetCC(VT, N0, N0, Cond); } if (N0 == N1) { // We can always fold X == X for integer setcc's. - if (MVT::isInteger(N0.getValueType())) + if (N0.getValueType().isInteger()) return DAG.getConstant(ISD::isTrueWhenEqual(Cond), VT); unsigned UOF = ISD::getUnorderedFlavor(Cond); if (UOF == 2) // FP operators that are undefined on NaNs. @@ -1716,7 +1594,7 @@ TargetLowering::SimplifySetCC(MVT::ValueType VT, SDOperand N0, SDOperand N1, } if ((Cond == ISD::SETEQ || Cond == ISD::SETNE) && - MVT::isInteger(N0.getValueType())) { + N0.getValueType().isInteger()) { if (N0.getOpcode() == ISD::ADD || N0.getOpcode() == ISD::SUB || N0.getOpcode() == ISD::XOR) { // Simplify (X+Y) == (X+Z) --> Y == Z @@ -1737,9 +1615,10 @@ TargetLowering::SimplifySetCC(MVT::ValueType VT, SDOperand N0, SDOperand N1, if (ConstantSDNode *RHSC = dyn_cast(N1)) { if (ConstantSDNode *LHSR = dyn_cast(N0.getOperand(1))) { // Turn (X+C1) == C2 --> X == C2-C1 - if (N0.getOpcode() == ISD::ADD && N0.Val->hasOneUse()) { + if (N0.getOpcode() == ISD::ADD && N0.getNode()->hasOneUse()) { return DAG.getSetCC(VT, N0.getOperand(0), - DAG.getConstant(RHSC->getValue()-LHSR->getValue(), + DAG.getConstant(RHSC->getAPIntValue()- + LHSR->getAPIntValue(), N0.getValueType()), Cond); } @@ -1747,18 +1626,24 @@ TargetLowering::SimplifySetCC(MVT::ValueType VT, SDOperand N0, SDOperand N1, if (N0.getOpcode() == ISD::XOR) // If we know that all of the inverted bits are zero, don't bother // performing the inversion. - if (MaskedValueIsZero(N0.getOperand(0), ~LHSR->getValue())) - return DAG.getSetCC(VT, N0.getOperand(0), - DAG.getConstant(LHSR->getValue()^RHSC->getValue(), - N0.getValueType()), Cond); + if (DAG.MaskedValueIsZero(N0.getOperand(0), ~LHSR->getAPIntValue())) + return + DAG.getSetCC(VT, N0.getOperand(0), + DAG.getConstant(LHSR->getAPIntValue() ^ + RHSC->getAPIntValue(), + N0.getValueType()), + Cond); } // Turn (C1-X) == C2 --> X == C1-C2 if (ConstantSDNode *SUBC = dyn_cast(N0.getOperand(0))) { - if (N0.getOpcode() == ISD::SUB && N0.Val->hasOneUse()) { - return DAG.getSetCC(VT, N0.getOperand(1), - DAG.getConstant(SUBC->getValue()-RHSC->getValue(), - N0.getValueType()), Cond); + if (N0.getOpcode() == ISD::SUB && N0.getNode()->hasOneUse()) { + return + DAG.getSetCC(VT, N0.getOperand(1), + DAG.getConstant(SUBC->getAPIntValue() - + RHSC->getAPIntValue(), + N0.getValueType()), + Cond); } } } @@ -1771,14 +1656,14 @@ TargetLowering::SimplifySetCC(MVT::ValueType VT, SDOperand N0, SDOperand N1, if (DAG.isCommutativeBinOp(N0.getOpcode())) return DAG.getSetCC(VT, N0.getOperand(0), DAG.getConstant(0, N0.getValueType()), Cond); - else { + else if (N0.getNode()->hasOneUse()) { assert(N0.getOpcode() == ISD::SUB && "Unexpected operation!"); // (Z-X) == X --> Z == X<<1 - SDOperand SH = DAG.getNode(ISD::SHL, N1.getValueType(), + SDValue SH = DAG.getNode(ISD::SHL, N1.getValueType(), N1, DAG.getConstant(1, getShiftAmountTy())); if (!DCI.isCalledByLegalizer()) - DCI.AddToWorklist(SH.Val); + DCI.AddToWorklist(SH.getNode()); return DAG.getSetCC(VT, N0.getOperand(0), SH, Cond); } } @@ -1794,13 +1679,13 @@ TargetLowering::SimplifySetCC(MVT::ValueType VT, SDOperand N0, SDOperand N1, if (DAG.isCommutativeBinOp(N1.getOpcode())) { return DAG.getSetCC(VT, N1.getOperand(0), DAG.getConstant(0, N1.getValueType()), Cond); - } else { + } else if (N1.getNode()->hasOneUse()) { assert(N1.getOpcode() == ISD::SUB && "Unexpected operation!"); // X == (Z-X) --> X<<1 == Z - SDOperand SH = DAG.getNode(ISD::SHL, N1.getValueType(), N0, + SDValue SH = DAG.getNode(ISD::SHL, N1.getValueType(), N0, DAG.getConstant(1, getShiftAmountTy())); if (!DCI.isCalledByLegalizer()) - DCI.AddToWorklist(SH.Val); + DCI.AddToWorklist(SH.getNode()); return DAG.getSetCC(VT, SH, N1.getOperand(0), Cond); } } @@ -1808,7 +1693,7 @@ TargetLowering::SimplifySetCC(MVT::ValueType VT, SDOperand N0, SDOperand N1, } // Fold away ALL boolean setcc's. - SDOperand Temp; + SDValue Temp; if (N0.getValueType() == MVT::i1 && foldBooleans) { switch (Cond) { default: assert(0 && "Unknown integer setcc!"); @@ -1816,7 +1701,7 @@ TargetLowering::SimplifySetCC(MVT::ValueType VT, SDOperand N0, SDOperand N1, Temp = DAG.getNode(ISD::XOR, MVT::i1, N0, N1); N0 = DAG.getNode(ISD::XOR, MVT::i1, Temp, DAG.getConstant(1, MVT::i1)); if (!DCI.isCalledByLegalizer()) - DCI.AddToWorklist(Temp.Val); + DCI.AddToWorklist(Temp.getNode()); break; case ISD::SETNE: // X != Y --> (X^Y) N0 = DAG.getNode(ISD::XOR, MVT::i1, N0, N1); @@ -1826,21 +1711,21 @@ TargetLowering::SimplifySetCC(MVT::ValueType VT, SDOperand N0, SDOperand N1, Temp = DAG.getNode(ISD::XOR, MVT::i1, N0, DAG.getConstant(1, MVT::i1)); N0 = DAG.getNode(ISD::AND, MVT::i1, N1, Temp); if (!DCI.isCalledByLegalizer()) - DCI.AddToWorklist(Temp.Val); + DCI.AddToWorklist(Temp.getNode()); break; case ISD::SETLT: // X X == 1 & Y == 0 --> Y^1 & X case ISD::SETUGT: // X >u Y --> X == 1 & Y == 0 --> Y^1 & X Temp = DAG.getNode(ISD::XOR, MVT::i1, N1, DAG.getConstant(1, MVT::i1)); N0 = DAG.getNode(ISD::AND, MVT::i1, N0, Temp); if (!DCI.isCalledByLegalizer()) - DCI.AddToWorklist(Temp.Val); + DCI.AddToWorklist(Temp.getNode()); break; case ISD::SETULE: // X <=u Y --> X == 0 | Y == 1 --> X^1 | Y case ISD::SETGE: // X >=s Y --> X == 0 | Y == 1 --> X^1 | Y Temp = DAG.getNode(ISD::XOR, MVT::i1, N0, DAG.getConstant(1, MVT::i1)); N0 = DAG.getNode(ISD::OR, MVT::i1, N1, Temp); if (!DCI.isCalledByLegalizer()) - DCI.AddToWorklist(Temp.Val); + DCI.AddToWorklist(Temp.getNode()); break; case ISD::SETUGE: // X >=u Y --> X == 1 | Y == 0 --> Y^1 | X case ISD::SETLE: // X <=s Y --> X == 1 | Y == 0 --> Y^1 | X @@ -1850,7 +1735,7 @@ TargetLowering::SimplifySetCC(MVT::ValueType VT, SDOperand N0, SDOperand N1, } if (VT != MVT::i1) { if (!DCI.isCalledByLegalizer()) - DCI.AddToWorklist(N0.Val); + DCI.AddToWorklist(N0.getNode()); // FIXME: If running after legalize, we probably can't do this. N0 = DAG.getNode(ISD::ZERO_EXTEND, VT, N0); } @@ -1858,19 +1743,89 @@ TargetLowering::SimplifySetCC(MVT::ValueType VT, SDOperand N0, SDOperand N1, } // Could not fold it. - return SDOperand(); + return SDValue(); +} + +/// isGAPlusOffset - Returns true (and the GlobalValue and the offset) if the +/// node is a GlobalAddress + offset. +bool TargetLowering::isGAPlusOffset(SDNode *N, GlobalValue* &GA, + int64_t &Offset) const { + if (isa(N)) { + GlobalAddressSDNode *GASD = cast(N); + GA = GASD->getGlobal(); + Offset += GASD->getOffset(); + return true; + } + + if (N->getOpcode() == ISD::ADD) { + SDValue N1 = N->getOperand(0); + SDValue N2 = N->getOperand(1); + if (isGAPlusOffset(N1.getNode(), GA, Offset)) { + ConstantSDNode *V = dyn_cast(N2); + if (V) { + Offset += V->getSExtValue(); + return true; + } + } else if (isGAPlusOffset(N2.getNode(), GA, Offset)) { + ConstantSDNode *V = dyn_cast(N1); + if (V) { + Offset += V->getSExtValue(); + return true; + } + } + } + return false; +} + + +/// isConsecutiveLoad - Return true if LD (which must be a LoadSDNode) is +/// loading 'Bytes' bytes from a location that is 'Dist' units away from the +/// location that the 'Base' load is loading from. +bool TargetLowering::isConsecutiveLoad(SDNode *LD, SDNode *Base, + unsigned Bytes, int Dist, + const MachineFrameInfo *MFI) const { + if (LD->getOperand(0).getNode() != Base->getOperand(0).getNode()) + return false; + MVT VT = LD->getValueType(0); + if (VT.getSizeInBits() / 8 != Bytes) + return false; + + SDValue Loc = LD->getOperand(1); + SDValue BaseLoc = Base->getOperand(1); + if (Loc.getOpcode() == ISD::FrameIndex) { + if (BaseLoc.getOpcode() != ISD::FrameIndex) + return false; + int FI = cast(Loc)->getIndex(); + int BFI = cast(BaseLoc)->getIndex(); + int FS = MFI->getObjectSize(FI); + int BFS = MFI->getObjectSize(BFI); + if (FS != BFS || FS != (int)Bytes) return false; + return MFI->getObjectOffset(FI) == (MFI->getObjectOffset(BFI) + Dist*Bytes); + } + + GlobalValue *GV1 = NULL; + GlobalValue *GV2 = NULL; + int64_t Offset1 = 0; + int64_t Offset2 = 0; + bool isGA1 = isGAPlusOffset(Loc.getNode(), GV1, Offset1); + bool isGA2 = isGAPlusOffset(BaseLoc.getNode(), GV2, Offset2); + if (isGA1 && isGA2 && GV1 == GV2) + return Offset1 == (Offset2 + Dist*Bytes); + return false; } -SDOperand TargetLowering:: + +SDValue TargetLowering:: PerformDAGCombine(SDNode *N, DAGCombinerInfo &DCI) const { // Default implementation: no optimization. - return SDOperand(); + return SDValue(); } //===----------------------------------------------------------------------===// // Inline Assembler Implementation Methods //===----------------------------------------------------------------------===// + TargetLowering::ConstraintType TargetLowering::getConstraintType(const std::string &Constraint) const { // FIXME: lots more standard ones to handle. @@ -1904,46 +1859,87 @@ TargetLowering::getConstraintType(const std::string &Constraint) const { return C_Unknown; } -/// isOperandValidForConstraint - Return the specified operand (possibly -/// modified) if the specified SDOperand is valid for the specified target -/// constraint letter, otherwise return null. -SDOperand TargetLowering::isOperandValidForConstraint(SDOperand Op, - char ConstraintLetter, - SelectionDAG &DAG) { +/// LowerXConstraint - try to replace an X constraint, which matches anything, +/// with another that has more specific requirements based on the type of the +/// corresponding operand. +const char *TargetLowering::LowerXConstraint(MVT ConstraintVT) const{ + if (ConstraintVT.isInteger()) + return "r"; + if (ConstraintVT.isFloatingPoint()) + return "f"; // works for many targets + return 0; +} + +/// LowerAsmOperandForConstraint - Lower the specified operand into the Ops +/// vector. If it is invalid, don't add anything to Ops. +void TargetLowering::LowerAsmOperandForConstraint(SDValue Op, + char ConstraintLetter, + bool hasMemory, + std::vector &Ops, + SelectionDAG &DAG) const { switch (ConstraintLetter) { default: break; + case 'X': // Allows any operand; labels (basic block) use this. + if (Op.getOpcode() == ISD::BasicBlock) { + Ops.push_back(Op); + return; + } + // fall through case 'i': // Simple Integer or Relocatable Constant case 'n': // Simple Integer - case 's': // Relocatable Constant - case 'X': // Allows any operand. - // These are okay if the operand is either a global variable address or a - // simple immediate value. If we have one of these, map to the TargetXXX - // version so that the value itself doesn't get selected. - if (ConstantSDNode *C = dyn_cast(Op)) { - // Simple constants are not allowed for 's'. - if (ConstraintLetter != 's') - return DAG.getTargetConstant(C->getValue(), Op.getValueType()); + case 's': { // Relocatable Constant + // These operands are interested in values of the form (GV+C), where C may + // be folded in as an offset of GV, or it may be explicitly added. Also, it + // is possible and fine if either GV or C are missing. + ConstantSDNode *C = dyn_cast(Op); + GlobalAddressSDNode *GA = dyn_cast(Op); + + // If we have "(add GV, C)", pull out GV/C + if (Op.getOpcode() == ISD::ADD) { + C = dyn_cast(Op.getOperand(1)); + GA = dyn_cast(Op.getOperand(0)); + if (C == 0 || GA == 0) { + C = dyn_cast(Op.getOperand(0)); + GA = dyn_cast(Op.getOperand(1)); + } + if (C == 0 || GA == 0) + C = 0, GA = 0; } - if (GlobalAddressSDNode *GA = dyn_cast(Op)) { - if (ConstraintLetter != 'n') - return DAG.getTargetGlobalAddress(GA->getGlobal(), Op.getValueType(), - GA->getOffset()); + + // If we find a valid operand, map to the TargetXXX version so that the + // value itself doesn't get selected. + if (GA) { // Either &GV or &GV+C + if (ConstraintLetter != 'n') { + int64_t Offs = GA->getOffset(); + if (C) Offs += C->getZExtValue(); + Ops.push_back(DAG.getTargetGlobalAddress(GA->getGlobal(), + Op.getValueType(), Offs)); + return; + } + } + if (C) { // just C, no GV. + // Simple constants are not allowed for 's'. + if (ConstraintLetter != 's') { + Ops.push_back(DAG.getTargetConstant(C->getAPIntValue(), + Op.getValueType())); + return; + } } break; } - return SDOperand(0,0); + } } std::vector TargetLowering:: getRegClassForInlineAsmConstraint(const std::string &Constraint, - MVT::ValueType VT) const { + MVT VT) const { return std::vector(); } std::pair TargetLowering:: getRegForInlineAsmConstraint(const std::string &Constraint, - MVT::ValueType VT) const { + MVT VT) const { if (Constraint[0] != '{') return std::pair(0, 0); assert(*(Constraint.end()-1) == '}' && "Not a brace enclosed constraint?"); @@ -1952,8 +1948,8 @@ getRegForInlineAsmConstraint(const std::string &Constraint, std::string RegName(Constraint.begin()+1, Constraint.end()-1); // Figure out which register class contains this reg. - const MRegisterInfo *RI = TM.getRegisterInfo(); - for (MRegisterInfo::regclass_iterator RCI = RI->regclass_begin(), + const TargetRegisterInfo *RI = TM.getRegisterInfo(); + for (TargetRegisterInfo::regclass_iterator RCI = RI->regclass_begin(), E = RI->regclass_end(); RCI != E; ++RCI) { const TargetRegisterClass *RC = *RCI; @@ -1972,7 +1968,7 @@ getRegForInlineAsmConstraint(const std::string &Constraint, for (TargetRegisterClass::iterator I = RC->begin(), E = RC->end(); I != E; ++I) { - if (StringsEqualNoCase(RegName, RI->get(*I).Name)) + if (StringsEqualNoCase(RegName, RI->get(*I).AsmName)) return std::make_pair(*I, RC); } } @@ -1980,6 +1976,138 @@ getRegForInlineAsmConstraint(const std::string &Constraint, return std::pair(0, 0); } +//===----------------------------------------------------------------------===// +// Constraint Selection. + +/// isMatchingInputConstraint - Return true of this is an input operand that is +/// a matching constraint like "4". +bool TargetLowering::AsmOperandInfo::isMatchingInputConstraint() const { + assert(!ConstraintCode.empty() && "No known constraint!"); + return isdigit(ConstraintCode[0]); +} + +/// getMatchedOperand - If this is an input matching constraint, this method +/// returns the output operand it matches. +unsigned TargetLowering::AsmOperandInfo::getMatchedOperand() const { + assert(!ConstraintCode.empty() && "No known constraint!"); + return atoi(ConstraintCode.c_str()); +} + + +/// getConstraintGenerality - Return an integer indicating how general CT +/// is. +static unsigned getConstraintGenerality(TargetLowering::ConstraintType CT) { + switch (CT) { + default: assert(0 && "Unknown constraint type!"); + case TargetLowering::C_Other: + case TargetLowering::C_Unknown: + return 0; + case TargetLowering::C_Register: + return 1; + case TargetLowering::C_RegisterClass: + return 2; + case TargetLowering::C_Memory: + return 3; + } +} + +/// ChooseConstraint - If there are multiple different constraints that we +/// could pick for this operand (e.g. "imr") try to pick the 'best' one. +/// This is somewhat tricky: constraints fall into four classes: +/// Other -> immediates and magic values +/// Register -> one specific register +/// RegisterClass -> a group of regs +/// Memory -> memory +/// Ideally, we would pick the most specific constraint possible: if we have +/// something that fits into a register, we would pick it. The problem here +/// is that if we have something that could either be in a register or in +/// memory that use of the register could cause selection of *other* +/// operands to fail: they might only succeed if we pick memory. Because of +/// this the heuristic we use is: +/// +/// 1) If there is an 'other' constraint, and if the operand is valid for +/// that constraint, use it. This makes us take advantage of 'i' +/// constraints when available. +/// 2) Otherwise, pick the most general constraint present. This prefers +/// 'm' over 'r', for example. +/// +static void ChooseConstraint(TargetLowering::AsmOperandInfo &OpInfo, + bool hasMemory, const TargetLowering &TLI, + SDValue Op, SelectionDAG *DAG) { + assert(OpInfo.Codes.size() > 1 && "Doesn't have multiple constraint options"); + unsigned BestIdx = 0; + TargetLowering::ConstraintType BestType = TargetLowering::C_Unknown; + int BestGenerality = -1; + + // Loop over the options, keeping track of the most general one. + for (unsigned i = 0, e = OpInfo.Codes.size(); i != e; ++i) { + TargetLowering::ConstraintType CType = + TLI.getConstraintType(OpInfo.Codes[i]); + + // If this is an 'other' constraint, see if the operand is valid for it. + // For example, on X86 we might have an 'rI' constraint. If the operand + // is an integer in the range [0..31] we want to use I (saving a load + // of a register), otherwise we must use 'r'. + if (CType == TargetLowering::C_Other && Op.getNode()) { + assert(OpInfo.Codes[i].size() == 1 && + "Unhandled multi-letter 'other' constraint"); + std::vector ResultOps; + TLI.LowerAsmOperandForConstraint(Op, OpInfo.Codes[i][0], hasMemory, + ResultOps, *DAG); + if (!ResultOps.empty()) { + BestType = CType; + BestIdx = i; + break; + } + } + + // This constraint letter is more general than the previous one, use it. + int Generality = getConstraintGenerality(CType); + if (Generality > BestGenerality) { + BestType = CType; + BestIdx = i; + BestGenerality = Generality; + } + } + + OpInfo.ConstraintCode = OpInfo.Codes[BestIdx]; + OpInfo.ConstraintType = BestType; +} + +/// ComputeConstraintToUse - Determines the constraint code and constraint +/// type to use for the specific AsmOperandInfo, setting +/// OpInfo.ConstraintCode and OpInfo.ConstraintType. +void TargetLowering::ComputeConstraintToUse(AsmOperandInfo &OpInfo, + SDValue Op, + bool hasMemory, + SelectionDAG *DAG) const { + assert(!OpInfo.Codes.empty() && "Must have at least one constraint"); + + // Single-letter constraints ('r') are very common. + if (OpInfo.Codes.size() == 1) { + OpInfo.ConstraintCode = OpInfo.Codes[0]; + OpInfo.ConstraintType = getConstraintType(OpInfo.ConstraintCode); + } else { + ChooseConstraint(OpInfo, hasMemory, *this, Op, DAG); + } + + // 'X' matches anything. + if (OpInfo.ConstraintCode == "X" && OpInfo.CallOperandVal) { + // Labels and constants are handled elsewhere ('X' is the only thing + // that matches labels). + if (isa(OpInfo.CallOperandVal) || + isa(OpInfo.CallOperandVal)) + return; + + // Otherwise, try to resolve it to something we know about by looking at + // the actual operand type. + if (const char *Repl = LowerXConstraint(OpInfo.ConstraintVT)) { + OpInfo.ConstraintCode = Repl; + OpInfo.ConstraintType = getConstraintType(OpInfo.ConstraintCode); + } + } +} + //===----------------------------------------------------------------------===// // Loop Strength Reduction hooks //===----------------------------------------------------------------------===// @@ -2196,47 +2324,53 @@ static mu magicu64(uint64_t d) /// return a DAG expression to select that will generate the same value by /// multiplying by a magic number. See: /// -SDOperand TargetLowering::BuildSDIV(SDNode *N, SelectionDAG &DAG, - std::vector* Created) const { - MVT::ValueType VT = N->getValueType(0); +SDValue TargetLowering::BuildSDIV(SDNode *N, SelectionDAG &DAG, + std::vector* Created) const { + MVT VT = N->getValueType(0); // Check to see if we can do this. if (!isTypeLegal(VT) || (VT != MVT::i32 && VT != MVT::i64)) - return SDOperand(); // BuildSDIV only operates on i32 or i64 - if (!isOperationLegal(ISD::MULHS, VT)) - return SDOperand(); // Make sure the target supports MULHS. + return SDValue(); // BuildSDIV only operates on i32 or i64 - int64_t d = cast(N->getOperand(1))->getSignExtended(); + int64_t d = cast(N->getOperand(1))->getSExtValue(); ms magics = (VT == MVT::i32) ? magic32(d) : magic64(d); // Multiply the numerator (operand 0) by the magic value - SDOperand Q = DAG.getNode(ISD::MULHS, VT, N->getOperand(0), - DAG.getConstant(magics.m, VT)); + SDValue Q; + if (isOperationLegal(ISD::MULHS, VT)) + Q = DAG.getNode(ISD::MULHS, VT, N->getOperand(0), + DAG.getConstant(magics.m, VT)); + else if (isOperationLegal(ISD::SMUL_LOHI, VT)) + Q = SDValue(DAG.getNode(ISD::SMUL_LOHI, DAG.getVTList(VT, VT), + N->getOperand(0), + DAG.getConstant(magics.m, VT)).getNode(), 1); + else + return SDValue(); // No mulhs or equvialent // If d > 0 and m < 0, add the numerator if (d > 0 && magics.m < 0) { Q = DAG.getNode(ISD::ADD, VT, Q, N->getOperand(0)); if (Created) - Created->push_back(Q.Val); + Created->push_back(Q.getNode()); } // If d < 0 and m > 0, subtract the numerator. if (d < 0 && magics.m > 0) { Q = DAG.getNode(ISD::SUB, VT, Q, N->getOperand(0)); if (Created) - Created->push_back(Q.Val); + Created->push_back(Q.getNode()); } // Shift right algebraic if shift value is nonzero if (magics.s > 0) { Q = DAG.getNode(ISD::SRA, VT, Q, DAG.getConstant(magics.s, getShiftAmountTy())); if (Created) - Created->push_back(Q.Val); + Created->push_back(Q.getNode()); } // Extract the sign bit and add it to the quotient - SDOperand T = - DAG.getNode(ISD::SRL, VT, Q, DAG.getConstant(MVT::getSizeInBits(VT)-1, + SDValue T = + DAG.getNode(ISD::SRL, VT, Q, DAG.getConstant(VT.getSizeInBits()-1, getShiftAmountTy())); if (Created) - Created->push_back(T.Val); + Created->push_back(T.getNode()); return DAG.getNode(ISD::ADD, VT, Q, T); } @@ -2244,39 +2378,45 @@ SDOperand TargetLowering::BuildSDIV(SDNode *N, SelectionDAG &DAG, /// return a DAG expression to select that will generate the same value by /// multiplying by a magic number. See: /// -SDOperand TargetLowering::BuildUDIV(SDNode *N, SelectionDAG &DAG, - std::vector* Created) const { - MVT::ValueType VT = N->getValueType(0); +SDValue TargetLowering::BuildUDIV(SDNode *N, SelectionDAG &DAG, + std::vector* Created) const { + MVT VT = N->getValueType(0); // Check to see if we can do this. if (!isTypeLegal(VT) || (VT != MVT::i32 && VT != MVT::i64)) - return SDOperand(); // BuildUDIV only operates on i32 or i64 - if (!isOperationLegal(ISD::MULHU, VT)) - return SDOperand(); // Make sure the target supports MULHU. + return SDValue(); // BuildUDIV only operates on i32 or i64 - uint64_t d = cast(N->getOperand(1))->getValue(); + uint64_t d = cast(N->getOperand(1))->getZExtValue(); mu magics = (VT == MVT::i32) ? magicu32(d) : magicu64(d); // Multiply the numerator (operand 0) by the magic value - SDOperand Q = DAG.getNode(ISD::MULHU, VT, N->getOperand(0), - DAG.getConstant(magics.m, VT)); + SDValue Q; + if (isOperationLegal(ISD::MULHU, VT)) + Q = DAG.getNode(ISD::MULHU, VT, N->getOperand(0), + DAG.getConstant(magics.m, VT)); + else if (isOperationLegal(ISD::UMUL_LOHI, VT)) + Q = SDValue(DAG.getNode(ISD::UMUL_LOHI, DAG.getVTList(VT, VT), + N->getOperand(0), + DAG.getConstant(magics.m, VT)).getNode(), 1); + else + return SDValue(); // No mulhu or equvialent if (Created) - Created->push_back(Q.Val); + Created->push_back(Q.getNode()); if (magics.a == 0) { return DAG.getNode(ISD::SRL, VT, Q, DAG.getConstant(magics.s, getShiftAmountTy())); } else { - SDOperand NPQ = DAG.getNode(ISD::SUB, VT, N->getOperand(0), Q); + SDValue NPQ = DAG.getNode(ISD::SUB, VT, N->getOperand(0), Q); if (Created) - Created->push_back(NPQ.Val); + Created->push_back(NPQ.getNode()); NPQ = DAG.getNode(ISD::SRL, VT, NPQ, DAG.getConstant(1, getShiftAmountTy())); if (Created) - Created->push_back(NPQ.Val); + Created->push_back(NPQ.getNode()); NPQ = DAG.getNode(ISD::ADD, VT, NPQ, Q); if (Created) - Created->push_back(NPQ.Val); + Created->push_back(NPQ.getNode()); return DAG.getNode(ISD::SRL, VT, NPQ, DAG.getConstant(magics.s-1, getShiftAmountTy())); }