diff options
Diffstat (limited to 'clang/lib/AST/ExprConstant.cpp')
-rw-r--r-- | clang/lib/AST/ExprConstant.cpp | 258 |
1 files changed, 258 insertions, 0 deletions
diff --git a/clang/lib/AST/ExprConstant.cpp b/clang/lib/AST/ExprConstant.cpp index dfdfef2..149b7d7 100644 --- a/clang/lib/AST/ExprConstant.cpp +++ b/clang/lib/AST/ExprConstant.cpp @@ -11619,6 +11619,44 @@ static bool evalPackBuiltin(const CallExpr *E, EvalInfo &Info, APValue &Result, return true; } +static bool evalPshufbBuiltin(EvalInfo &Info, const CallExpr *Call, + APValue &Out) { + APValue SrcVec, ControlVec; + if (!EvaluateAsRValue(Info, Call->getArg(0), SrcVec)) + return false; + if (!EvaluateAsRValue(Info, Call->getArg(1), ControlVec)) + return false; + + const auto *VT = Call->getType()->getAs<VectorType>(); + if (!VT) + return false; + + QualType ElemT = VT->getElementType(); + unsigned NumElts = VT->getNumElements(); + + SmallVector<APValue, 64> ResultElements; + ResultElements.reserve(NumElts); + + for (unsigned Idx = 0; Idx != NumElts; ++Idx) { + APValue CtlVal = ControlVec.getVectorElt(Idx); + APSInt CtlByte = CtlVal.getInt(); + uint8_t Ctl = static_cast<uint8_t>(CtlByte.getZExtValue()); + + if (Ctl & 0x80) { + APValue Zero(Info.Ctx.MakeIntValue(0, ElemT)); + ResultElements.push_back(Zero); + } else { + unsigned LaneBase = (Idx / 16) * 16; + unsigned SrcOffset = Ctl & 0x0F; + unsigned SrcIdx = LaneBase + SrcOffset; + + ResultElements.push_back(SrcVec.getVectorElt(SrcIdx)); + } + } + Out = APValue(ResultElements.data(), ResultElements.size()); + return true; +} + static bool evalPshufBuiltin(EvalInfo &Info, const CallExpr *Call, bool IsShufHW, APValue &Out) { APValue Vec; @@ -11974,6 +12012,54 @@ bool VectorExprEvaluator::VisitCallExpr(const CallExpr *E) { return Success(APValue(ResultElements.data(), ResultElements.size()), E); } + + case X86::BI__builtin_ia32_vpmadd52luq128: + case X86::BI__builtin_ia32_vpmadd52luq256: + case X86::BI__builtin_ia32_vpmadd52luq512: { + APValue A, B, C; + if (!EvaluateAsRValue(Info, E->getArg(0), A) || + !EvaluateAsRValue(Info, E->getArg(1), B) || + !EvaluateAsRValue(Info, E->getArg(2), C)) + return false; + + unsigned ALen = A.getVectorLength(); + SmallVector<APValue, 4> ResultElements; + ResultElements.reserve(ALen); + + for (unsigned EltNum = 0; EltNum < ALen; EltNum += 1) { + APInt AElt = A.getVectorElt(EltNum).getInt(); + APInt BElt = B.getVectorElt(EltNum).getInt().trunc(52); + APInt CElt = C.getVectorElt(EltNum).getInt().trunc(52); + APSInt ResElt(AElt + (BElt * CElt).zext(64), false); + ResultElements.push_back(APValue(ResElt)); + } + + return Success(APValue(ResultElements.data(), ResultElements.size()), E); + } + case X86::BI__builtin_ia32_vpmadd52huq128: + case X86::BI__builtin_ia32_vpmadd52huq256: + case X86::BI__builtin_ia32_vpmadd52huq512: { + APValue A, B, C; + if (!EvaluateAsRValue(Info, E->getArg(0), A) || + !EvaluateAsRValue(Info, E->getArg(1), B) || + !EvaluateAsRValue(Info, E->getArg(2), C)) + return false; + + unsigned ALen = A.getVectorLength(); + SmallVector<APValue, 4> ResultElements; + ResultElements.reserve(ALen); + + for (unsigned EltNum = 0; EltNum < ALen; EltNum += 1) { + APInt AElt = A.getVectorElt(EltNum).getInt(); + APInt BElt = B.getVectorElt(EltNum).getInt().trunc(52); + APInt CElt = C.getVectorElt(EltNum).getInt().trunc(52); + APSInt ResElt(AElt + llvm::APIntOps::mulhu(BElt, CElt).zext(64), false); + ResultElements.push_back(APValue(ResElt)); + } + + return Success(APValue(ResultElements.data(), ResultElements.size()), E); + } + case clang::X86::BI__builtin_ia32_vprotbi: case clang::X86::BI__builtin_ia32_vprotdi: case clang::X86::BI__builtin_ia32_vprotqi: @@ -12193,6 +12279,15 @@ bool VectorExprEvaluator::VisitCallExpr(const CallExpr *E) { return Success(APValue(ResultElements.data(), ResultElements.size()), E); } + case X86::BI__builtin_ia32_pshufb128: + case X86::BI__builtin_ia32_pshufb256: + case X86::BI__builtin_ia32_pshufb512: { + APValue R; + if (!evalPshufbBuiltin(Info, E, R)) + return false; + return Success(R, E); + } + case X86::BI__builtin_ia32_pshuflw: case X86::BI__builtin_ia32_pshuflw256: case X86::BI__builtin_ia32_pshuflw512: { @@ -12381,6 +12476,169 @@ bool VectorExprEvaluator::VisitCallExpr(const CallExpr *E) { return Success(APValue(ResultElements.data(), ResultElements.size()), E); } + case clang::X86::BI__builtin_ia32_phaddw128: + case clang::X86::BI__builtin_ia32_phaddw256: + case clang::X86::BI__builtin_ia32_phaddd128: + case clang::X86::BI__builtin_ia32_phaddd256: + case clang::X86::BI__builtin_ia32_phaddsw128: + case clang::X86::BI__builtin_ia32_phaddsw256: + + case clang::X86::BI__builtin_ia32_phsubw128: + case clang::X86::BI__builtin_ia32_phsubw256: + case clang::X86::BI__builtin_ia32_phsubd128: + case clang::X86::BI__builtin_ia32_phsubd256: + case clang::X86::BI__builtin_ia32_phsubsw128: + case clang::X86::BI__builtin_ia32_phsubsw256: { + APValue SourceLHS, SourceRHS; + if (!EvaluateAsRValue(Info, E->getArg(0), SourceLHS) || + !EvaluateAsRValue(Info, E->getArg(1), SourceRHS)) + return false; + QualType DestEltTy = E->getType()->castAs<VectorType>()->getElementType(); + bool DestUnsigned = DestEltTy->isUnsignedIntegerOrEnumerationType(); + + unsigned NumElts = SourceLHS.getVectorLength(); + unsigned EltBits = Info.Ctx.getIntWidth(DestEltTy); + unsigned EltsPerLane = 128 / EltBits; + SmallVector<APValue, 4> ResultElements; + ResultElements.reserve(NumElts); + + for (unsigned LaneStart = 0; LaneStart != NumElts; + LaneStart += EltsPerLane) { + for (unsigned I = 0; I != EltsPerLane; I += 2) { + APSInt LHSA = SourceLHS.getVectorElt(LaneStart + I).getInt(); + APSInt LHSB = SourceLHS.getVectorElt(LaneStart + I + 1).getInt(); + switch (E->getBuiltinCallee()) { + case clang::X86::BI__builtin_ia32_phaddw128: + case clang::X86::BI__builtin_ia32_phaddw256: + case clang::X86::BI__builtin_ia32_phaddd128: + case clang::X86::BI__builtin_ia32_phaddd256: { + APSInt Res(LHSA + LHSB, DestUnsigned); + ResultElements.push_back(APValue(Res)); + break; + } + case clang::X86::BI__builtin_ia32_phaddsw128: + case clang::X86::BI__builtin_ia32_phaddsw256: { + APSInt Res(LHSA.sadd_sat(LHSB)); + ResultElements.push_back(APValue(Res)); + break; + } + case clang::X86::BI__builtin_ia32_phsubw128: + case clang::X86::BI__builtin_ia32_phsubw256: + case clang::X86::BI__builtin_ia32_phsubd128: + case clang::X86::BI__builtin_ia32_phsubd256: { + APSInt Res(LHSA - LHSB, DestUnsigned); + ResultElements.push_back(APValue(Res)); + break; + } + case clang::X86::BI__builtin_ia32_phsubsw128: + case clang::X86::BI__builtin_ia32_phsubsw256: { + APSInt Res(LHSA.ssub_sat(LHSB)); + ResultElements.push_back(APValue(Res)); + break; + } + } + } + for (unsigned I = 0; I != EltsPerLane; I += 2) { + APSInt RHSA = SourceRHS.getVectorElt(LaneStart + I).getInt(); + APSInt RHSB = SourceRHS.getVectorElt(LaneStart + I + 1).getInt(); + switch (E->getBuiltinCallee()) { + case clang::X86::BI__builtin_ia32_phaddw128: + case clang::X86::BI__builtin_ia32_phaddw256: + case clang::X86::BI__builtin_ia32_phaddd128: + case clang::X86::BI__builtin_ia32_phaddd256: { + APSInt Res(RHSA + RHSB, DestUnsigned); + ResultElements.push_back(APValue(Res)); + break; + } + case clang::X86::BI__builtin_ia32_phaddsw128: + case clang::X86::BI__builtin_ia32_phaddsw256: { + APSInt Res(RHSA.sadd_sat(RHSB)); + ResultElements.push_back(APValue(Res)); + break; + } + case clang::X86::BI__builtin_ia32_phsubw128: + case clang::X86::BI__builtin_ia32_phsubw256: + case clang::X86::BI__builtin_ia32_phsubd128: + case clang::X86::BI__builtin_ia32_phsubd256: { + APSInt Res(RHSA - RHSB, DestUnsigned); + ResultElements.push_back(APValue(Res)); + break; + } + case clang::X86::BI__builtin_ia32_phsubsw128: + case clang::X86::BI__builtin_ia32_phsubsw256: { + APSInt Res(RHSA.ssub_sat(RHSB)); + ResultElements.push_back(APValue(Res)); + break; + } + } + } + } + return Success(APValue(ResultElements.data(), ResultElements.size()), E); + } + case clang::X86::BI__builtin_ia32_haddpd: + case clang::X86::BI__builtin_ia32_haddps: + case clang::X86::BI__builtin_ia32_haddps256: + case clang::X86::BI__builtin_ia32_haddpd256: + case clang::X86::BI__builtin_ia32_hsubpd: + case clang::X86::BI__builtin_ia32_hsubps: + case clang::X86::BI__builtin_ia32_hsubps256: + case clang::X86::BI__builtin_ia32_hsubpd256: { + APValue SourceLHS, SourceRHS; + if (!EvaluateAsRValue(Info, E->getArg(0), SourceLHS) || + !EvaluateAsRValue(Info, E->getArg(1), SourceRHS)) + return false; + unsigned NumElts = SourceLHS.getVectorLength(); + SmallVector<APValue, 4> ResultElements; + ResultElements.reserve(NumElts); + llvm::RoundingMode RM = getActiveRoundingMode(getEvalInfo(), E); + QualType DestEltTy = E->getType()->castAs<VectorType>()->getElementType(); + unsigned EltBits = Info.Ctx.getTypeSize(DestEltTy); + unsigned NumLanes = NumElts * EltBits / 128; + unsigned NumElemsPerLane = NumElts / NumLanes; + unsigned HalfElemsPerLane = NumElemsPerLane / 2; + + for (unsigned L = 0; L != NumElts; L += NumElemsPerLane) { + for (unsigned I = 0; I != HalfElemsPerLane; ++I) { + APFloat LHSA = SourceLHS.getVectorElt(L + (2 * I) + 0).getFloat(); + APFloat LHSB = SourceLHS.getVectorElt(L + (2 * I) + 1).getFloat(); + switch (E->getBuiltinCallee()) { + case clang::X86::BI__builtin_ia32_haddpd: + case clang::X86::BI__builtin_ia32_haddps: + case clang::X86::BI__builtin_ia32_haddps256: + case clang::X86::BI__builtin_ia32_haddpd256: + LHSA.add(LHSB, RM); + break; + case clang::X86::BI__builtin_ia32_hsubpd: + case clang::X86::BI__builtin_ia32_hsubps: + case clang::X86::BI__builtin_ia32_hsubps256: + case clang::X86::BI__builtin_ia32_hsubpd256: + LHSA.subtract(LHSB, RM); + break; + } + ResultElements.push_back(APValue(LHSA)); + } + for (unsigned I = 0; I != HalfElemsPerLane; ++I) { + APFloat RHSA = SourceRHS.getVectorElt(L + (2 * I) + 0).getFloat(); + APFloat RHSB = SourceRHS.getVectorElt(L + (2 * I) + 1).getFloat(); + switch (E->getBuiltinCallee()) { + case clang::X86::BI__builtin_ia32_haddpd: + case clang::X86::BI__builtin_ia32_haddps: + case clang::X86::BI__builtin_ia32_haddps256: + case clang::X86::BI__builtin_ia32_haddpd256: + RHSA.add(RHSB, RM); + break; + case clang::X86::BI__builtin_ia32_hsubpd: + case clang::X86::BI__builtin_ia32_hsubps: + case clang::X86::BI__builtin_ia32_hsubps256: + case clang::X86::BI__builtin_ia32_hsubpd256: + RHSA.subtract(RHSB, RM); + break; + } + ResultElements.push_back(APValue(RHSA)); + } + } + return Success(APValue(ResultElements.data(), ResultElements.size()), E); + } case Builtin::BI__builtin_elementwise_fshl: case Builtin::BI__builtin_elementwise_fshr: { APValue SourceHi, SourceLo, SourceShift; |