summaryrefslogtreecommitdiff
path: root/lib/Target/PowerPC
diff options
context:
space:
mode:
Diffstat (limited to 'lib/Target/PowerPC')
-rw-r--r--lib/Target/PowerPC/MCTargetDesc/PPCELFObjectWriter.cpp2
-rw-r--r--lib/Target/PowerPC/MCTargetDesc/PPCFixupKinds.h4
-rw-r--r--lib/Target/PowerPC/MCTargetDesc/PPCMCTargetDesc.cpp2
-rw-r--r--lib/Target/PowerPC/MCTargetDesc/PPCMCTargetDesc.h2
-rw-r--r--lib/Target/PowerPC/MCTargetDesc/PPCMachObjectWriter.cpp2
-rw-r--r--lib/Target/PowerPC/MCTargetDesc/PPCPredicates.h4
-rw-r--r--lib/Target/PowerPC/PPC.h2
-rw-r--r--lib/Target/PowerPC/PPCBranchSelector.cpp2
-rw-r--r--lib/Target/PowerPC/PPCCallingConv.h2
-rw-r--r--lib/Target/PowerPC/PPCEarlyReturn.cpp2
-rw-r--r--lib/Target/PowerPC/PPCFastISel.cpp6
-rw-r--r--lib/Target/PowerPC/PPCFrameLowering.h2
-rw-r--r--lib/Target/PowerPC/PPCISelDAGToDAG.cpp25
-rw-r--r--lib/Target/PowerPC/PPCISelLowering.cpp107
-rw-r--r--lib/Target/PowerPC/PPCISelLowering.h11
-rw-r--r--lib/Target/PowerPC/PPCInstrAltivec.td50
-rw-r--r--lib/Target/PowerPC/PPCInstrBuilder.h2
-rw-r--r--lib/Target/PowerPC/PPCInstrInfo.cpp23
-rw-r--r--lib/Target/PowerPC/PPCInstrInfo.h2
-rw-r--r--lib/Target/PowerPC/PPCInstrVSX.td24
-rw-r--r--lib/Target/PowerPC/PPCLoopDataPrefetch.cpp2
-rw-r--r--lib/Target/PowerPC/PPCLoopPreIncPrep.cpp4
-rw-r--r--lib/Target/PowerPC/PPCMCInstLower.cpp2
-rw-r--r--lib/Target/PowerPC/PPCSelectionDAGInfo.h2
-rw-r--r--lib/Target/PowerPC/PPCSubtarget.h4
-rw-r--r--lib/Target/PowerPC/PPCTLSDynamicCall.cpp2
-rw-r--r--lib/Target/PowerPC/PPCTOCRegDeps.cpp2
-rw-r--r--lib/Target/PowerPC/PPCTargetStreamer.h2
-rw-r--r--lib/Target/PowerPC/PPCVSXCopy.cpp2
-rw-r--r--lib/Target/PowerPC/PPCVSXFMAMutate.cpp2
-rw-r--r--lib/Target/PowerPC/PPCVSXSwapRemoval.cpp69
31 files changed, 257 insertions, 112 deletions
diff --git a/lib/Target/PowerPC/MCTargetDesc/PPCELFObjectWriter.cpp b/lib/Target/PowerPC/MCTargetDesc/PPCELFObjectWriter.cpp
index 36119d5d7e46..992be5b966c1 100644
--- a/lib/Target/PowerPC/MCTargetDesc/PPCELFObjectWriter.cpp
+++ b/lib/Target/PowerPC/MCTargetDesc/PPCELFObjectWriter.cpp
@@ -31,7 +31,7 @@ namespace {
bool needsRelocateWithSymbol(const MCSymbol &Sym,
unsigned Type) const override;
};
-} // namespace
+}
PPCELFObjectWriter::PPCELFObjectWriter(bool Is64Bit, uint8_t OSABI)
: MCELFObjectTargetWriter(Is64Bit, OSABI,
diff --git a/lib/Target/PowerPC/MCTargetDesc/PPCFixupKinds.h b/lib/Target/PowerPC/MCTargetDesc/PPCFixupKinds.h
index ad614f2ddf35..ae43e59d3cb1 100644
--- a/lib/Target/PowerPC/MCTargetDesc/PPCFixupKinds.h
+++ b/lib/Target/PowerPC/MCTargetDesc/PPCFixupKinds.h
@@ -50,7 +50,7 @@ enum Fixups {
LastTargetFixupKind,
NumTargetFixupKinds = LastTargetFixupKind - FirstTargetFixupKind
};
-} // namespace PPC
-} // namespace llvm
+}
+}
#endif
diff --git a/lib/Target/PowerPC/MCTargetDesc/PPCMCTargetDesc.cpp b/lib/Target/PowerPC/MCTargetDesc/PPCMCTargetDesc.cpp
index 489905b26fcc..5c38fe173d96 100644
--- a/lib/Target/PowerPC/MCTargetDesc/PPCMCTargetDesc.cpp
+++ b/lib/Target/PowerPC/MCTargetDesc/PPCMCTargetDesc.cpp
@@ -219,7 +219,7 @@ public:
llvm_unreachable("Unknown pseudo-op: .localentry");
}
};
-} // namespace
+}
static MCTargetStreamer *createAsmTargetStreamer(MCStreamer &S,
formatted_raw_ostream &OS,
diff --git a/lib/Target/PowerPC/MCTargetDesc/PPCMCTargetDesc.h b/lib/Target/PowerPC/MCTargetDesc/PPCMCTargetDesc.h
index 18818a1c335e..77fe45882289 100644
--- a/lib/Target/PowerPC/MCTargetDesc/PPCMCTargetDesc.h
+++ b/lib/Target/PowerPC/MCTargetDesc/PPCMCTargetDesc.h
@@ -81,7 +81,7 @@ static inline bool isRunOfOnes(unsigned Val, unsigned &MB, unsigned &ME) {
return false;
}
-} // namespace llvm
+} // End llvm namespace
// Generated files will use "namespace PPC". To avoid symbol clash,
// undefine PPC here. PPC may be predefined on some hosts.
diff --git a/lib/Target/PowerPC/MCTargetDesc/PPCMachObjectWriter.cpp b/lib/Target/PowerPC/MCTargetDesc/PPCMachObjectWriter.cpp
index 9b5491f92491..9d7289658f0f 100644
--- a/lib/Target/PowerPC/MCTargetDesc/PPCMachObjectWriter.cpp
+++ b/lib/Target/PowerPC/MCTargetDesc/PPCMachObjectWriter.cpp
@@ -51,7 +51,7 @@ public:
FixedValue);
}
};
-} // namespace
+}
/// computes the log2 of the size of the relocation,
/// used for relocation_info::r_length.
diff --git a/lib/Target/PowerPC/MCTargetDesc/PPCPredicates.h b/lib/Target/PowerPC/MCTargetDesc/PPCPredicates.h
index ff9b059d906a..6075631a541f 100644
--- a/lib/Target/PowerPC/MCTargetDesc/PPCPredicates.h
+++ b/lib/Target/PowerPC/MCTargetDesc/PPCPredicates.h
@@ -62,7 +62,7 @@ namespace PPC {
/// Assume the condition register is set by MI(a,b), return the predicate if
/// we modify the instructions such that condition register is set by MI(b,a).
Predicate getSwappedPredicate(Predicate Opcode);
-} // namespace PPC
-} // namespace llvm
+}
+}
#endif
diff --git a/lib/Target/PowerPC/PPC.h b/lib/Target/PowerPC/PPC.h
index 49f77b538c1b..ae8d8b4f5dfe 100644
--- a/lib/Target/PowerPC/PPC.h
+++ b/lib/Target/PowerPC/PPC.h
@@ -98,6 +98,6 @@ namespace llvm {
};
} // end namespace PPCII
-} // namespace llvm
+} // end namespace llvm;
#endif
diff --git a/lib/Target/PowerPC/PPCBranchSelector.cpp b/lib/Target/PowerPC/PPCBranchSelector.cpp
index 2b6030aea2b1..940d55ac1f36 100644
--- a/lib/Target/PowerPC/PPCBranchSelector.cpp
+++ b/lib/Target/PowerPC/PPCBranchSelector.cpp
@@ -51,7 +51,7 @@ namespace {
}
};
char PPCBSel::ID = 0;
-} // namespace
+}
INITIALIZE_PASS(PPCBSel, "ppc-branch-select", "PowerPC Branch Selector",
false, false)
diff --git a/lib/Target/PowerPC/PPCCallingConv.h b/lib/Target/PowerPC/PPCCallingConv.h
index 550cac62927e..eb904a858592 100644
--- a/lib/Target/PowerPC/PPCCallingConv.h
+++ b/lib/Target/PowerPC/PPCCallingConv.h
@@ -29,7 +29,7 @@ inline bool CC_PPC_AnyReg_Error(unsigned &, MVT &, MVT &,
return false;
}
-} // namespace llvm
+} // End llvm namespace
#endif
diff --git a/lib/Target/PowerPC/PPCEarlyReturn.cpp b/lib/Target/PowerPC/PPCEarlyReturn.cpp
index 9cd9c2faa51f..fc89753ed94e 100644
--- a/lib/Target/PowerPC/PPCEarlyReturn.cpp
+++ b/lib/Target/PowerPC/PPCEarlyReturn.cpp
@@ -191,7 +191,7 @@ public:
MachineFunctionPass::getAnalysisUsage(AU);
}
};
-} // namespace
+}
INITIALIZE_PASS(PPCEarlyReturn, DEBUG_TYPE,
"PowerPC Early-Return Creation", false, false)
diff --git a/lib/Target/PowerPC/PPCFastISel.cpp b/lib/Target/PowerPC/PPCFastISel.cpp
index 82ff5307d0b7..fafcd76f9d18 100644
--- a/lib/Target/PowerPC/PPCFastISel.cpp
+++ b/lib/Target/PowerPC/PPCFastISel.cpp
@@ -1448,9 +1448,9 @@ bool PPCFastISel::fastLowerCall(CallLoweringInfo &CLI) {
bool IsTailCall = CLI.IsTailCall;
bool IsVarArg = CLI.IsVarArg;
const Value *Callee = CLI.Callee;
- const char *SymName = CLI.SymName;
+ const MCSymbol *Symbol = CLI.Symbol;
- if (!Callee && !SymName)
+ if (!Callee && !Symbol)
return false;
// Allow SelectionDAG isel to handle tail calls.
@@ -2347,4 +2347,4 @@ namespace llvm {
return new PPCFastISel(FuncInfo, LibInfo);
return nullptr;
}
-} // namespace llvm
+}
diff --git a/lib/Target/PowerPC/PPCFrameLowering.h b/lib/Target/PowerPC/PPCFrameLowering.h
index b232863c9614..28d074ecd79d 100644
--- a/lib/Target/PowerPC/PPCFrameLowering.h
+++ b/lib/Target/PowerPC/PPCFrameLowering.h
@@ -93,6 +93,6 @@ public:
const SpillSlot *
getCalleeSavedSpillSlots(unsigned &NumEntries) const override;
};
-} // namespace llvm
+} // End llvm namespace
#endif
diff --git a/lib/Target/PowerPC/PPCISelDAGToDAG.cpp b/lib/Target/PowerPC/PPCISelDAGToDAG.cpp
index 5f9f9f2e341f..c85c2610d2f5 100644
--- a/lib/Target/PowerPC/PPCISelDAGToDAG.cpp
+++ b/lib/Target/PowerPC/PPCISelDAGToDAG.cpp
@@ -234,7 +234,7 @@ private:
SDNode *transferMemOperands(SDNode *N, SDNode *Result);
};
-} // namespace
+}
/// InsertVRSaveCode - Once the entire function has been instruction selected,
/// all virtual registers are created and all machine instructions are built,
@@ -2773,18 +2773,6 @@ SDNode *PPCDAGToDAGISel::Select(SDNode *N) {
else
DM[i] = 1;
- // For little endian, we must swap the input operands and adjust
- // the mask elements (reverse and invert them).
- if (PPCSubTarget->isLittleEndian()) {
- std::swap(Op1, Op2);
- unsigned tmp = DM[0];
- DM[0] = 1 - DM[1];
- DM[1] = 1 - tmp;
- }
-
- SDValue DMV = CurDAG->getTargetConstant(DM[1] | (DM[0] << 1), dl,
- MVT::i32);
-
if (Op1 == Op2 && DM[0] == 0 && DM[1] == 0 &&
Op1.getOpcode() == ISD::SCALAR_TO_VECTOR &&
isa<LoadSDNode>(Op1.getOperand(0))) {
@@ -2800,6 +2788,17 @@ SDNode *PPCDAGToDAGISel::Select(SDNode *N) {
}
}
+ // For little endian, we must swap the input operands and adjust
+ // the mask elements (reverse and invert them).
+ if (PPCSubTarget->isLittleEndian()) {
+ std::swap(Op1, Op2);
+ unsigned tmp = DM[0];
+ DM[0] = 1 - DM[1];
+ DM[1] = 1 - tmp;
+ }
+
+ SDValue DMV = CurDAG->getTargetConstant(DM[1] | (DM[0] << 1), dl,
+ MVT::i32);
SDValue Ops[] = { Op1, Op2, DMV };
return CurDAG->SelectNodeTo(N, PPC::XXPERMDI, N->getValueType(0), Ops);
}
diff --git a/lib/Target/PowerPC/PPCISelLowering.cpp b/lib/Target/PowerPC/PPCISelLowering.cpp
index 1cdfb4178544..594472bbb47b 100644
--- a/lib/Target/PowerPC/PPCISelLowering.cpp
+++ b/lib/Target/PowerPC/PPCISelLowering.cpp
@@ -1279,6 +1279,99 @@ bool PPC::isVMRGHShuffleMask(ShuffleVectorSDNode *N, unsigned UnitSize,
}
}
+/**
+ * \brief Common function used to match vmrgew and vmrgow shuffles
+ *
+ * The indexOffset determines whether to look for even or odd words in
+ * the shuffle mask. This is based on the of the endianness of the target
+ * machine.
+ * - Little Endian:
+ * - Use offset of 0 to check for odd elements
+ * - Use offset of 4 to check for even elements
+ * - Big Endian:
+ * - Use offset of 0 to check for even elements
+ * - Use offset of 4 to check for odd elements
+ * A detailed description of the vector element ordering for little endian and
+ * big endian can be found at
+ * http://www.ibm.com/developerworks/library/l-ibm-xl-c-cpp-compiler/index.html
+ * Targeting your applications - what little endian and big endian IBM XL C/C++
+ * compiler differences mean to you
+ *
+ * The mask to the shuffle vector instruction specifies the indices of the
+ * elements from the two input vectors to place in the result. The elements are
+ * numbered in array-access order, starting with the first vector. These vectors
+ * are always of type v16i8, thus each vector will contain 16 elements of size
+ * 8. More info on the shuffle vector can be found in the
+ * http://llvm.org/docs/LangRef.html#shufflevector-instruction
+ * Language Reference.
+ *
+ * The RHSStartValue indicates whether the same input vectors are used (unary)
+ * or two different input vectors are used, based on the following:
+ * - If the instruction uses the same vector for both inputs, the range of the
+ * indices will be 0 to 15. In this case, the RHSStart value passed should
+ * be 0.
+ * - If the instruction has two different vectors then the range of the
+ * indices will be 0 to 31. In this case, the RHSStart value passed should
+ * be 16 (indices 0-15 specify elements in the first vector while indices 16
+ * to 31 specify elements in the second vector).
+ *
+ * \param[in] N The shuffle vector SD Node to analyze
+ * \param[in] IndexOffset Specifies whether to look for even or odd elements
+ * \param[in] RHSStartValue Specifies the starting index for the righthand input
+ * vector to the shuffle_vector instruction
+ * \return true iff this shuffle vector represents an even or odd word merge
+ */
+static bool isVMerge(ShuffleVectorSDNode *N, unsigned IndexOffset,
+ unsigned RHSStartValue) {
+ if (N->getValueType(0) != MVT::v16i8)
+ return false;
+
+ for (unsigned i = 0; i < 2; ++i)
+ for (unsigned j = 0; j < 4; ++j)
+ if (!isConstantOrUndef(N->getMaskElt(i*4+j),
+ i*RHSStartValue+j+IndexOffset) ||
+ !isConstantOrUndef(N->getMaskElt(i*4+j+8),
+ i*RHSStartValue+j+IndexOffset+8))
+ return false;
+ return true;
+}
+
+/**
+ * \brief Determine if the specified shuffle mask is suitable for the vmrgew or
+ * vmrgow instructions.
+ *
+ * \param[in] N The shuffle vector SD Node to analyze
+ * \param[in] CheckEven Check for an even merge (true) or an odd merge (false)
+ * \param[in] ShuffleKind Identify the type of merge:
+ * - 0 = big-endian merge with two different inputs;
+ * - 1 = either-endian merge with two identical inputs;
+ * - 2 = little-endian merge with two different inputs (inputs are swapped for
+ * little-endian merges).
+ * \param[in] DAG The current SelectionDAG
+ * \return true iff this shuffle mask
+ */
+bool PPC::isVMRGEOShuffleMask(ShuffleVectorSDNode *N, bool CheckEven,
+ unsigned ShuffleKind, SelectionDAG &DAG) {
+ if (DAG.getTarget().getDataLayout()->isLittleEndian()) {
+ unsigned indexOffset = CheckEven ? 4 : 0;
+ if (ShuffleKind == 1) // Unary
+ return isVMerge(N, indexOffset, 0);
+ else if (ShuffleKind == 2) // swapped
+ return isVMerge(N, indexOffset, 16);
+ else
+ return false;
+ }
+ else {
+ unsigned indexOffset = CheckEven ? 0 : 4;
+ if (ShuffleKind == 1) // Unary
+ return isVMerge(N, indexOffset, 0);
+ else if (ShuffleKind == 0) // Normal
+ return isVMerge(N, indexOffset, 16);
+ else
+ return false;
+ }
+ return false;
+}
/// isVSLDOIShuffleMask - If this is a vsldoi shuffle mask, return the shift
/// amount, otherwise return -1.
@@ -3765,7 +3858,7 @@ struct TailCallArgumentInfo {
TailCallArgumentInfo() : FrameIdx(0) {}
};
-} // namespace
+}
/// StoreTailCallArgumentsToStackSlot - Stores arguments to their stack slot.
static void
@@ -7046,7 +7139,9 @@ SDValue PPCTargetLowering::LowerVECTOR_SHUFFLE(SDValue Op,
PPC::isVMRGLShuffleMask(SVOp, 4, 1, DAG) ||
PPC::isVMRGHShuffleMask(SVOp, 1, 1, DAG) ||
PPC::isVMRGHShuffleMask(SVOp, 2, 1, DAG) ||
- PPC::isVMRGHShuffleMask(SVOp, 4, 1, DAG)) {
+ PPC::isVMRGHShuffleMask(SVOp, 4, 1, DAG) ||
+ PPC::isVMRGEOShuffleMask(SVOp, true, 1, DAG) ||
+ PPC::isVMRGEOShuffleMask(SVOp, false, 1, DAG)) {
return Op;
}
}
@@ -7064,7 +7159,9 @@ SDValue PPCTargetLowering::LowerVECTOR_SHUFFLE(SDValue Op,
PPC::isVMRGLShuffleMask(SVOp, 4, ShuffleKind, DAG) ||
PPC::isVMRGHShuffleMask(SVOp, 1, ShuffleKind, DAG) ||
PPC::isVMRGHShuffleMask(SVOp, 2, ShuffleKind, DAG) ||
- PPC::isVMRGHShuffleMask(SVOp, 4, ShuffleKind, DAG))
+ PPC::isVMRGHShuffleMask(SVOp, 4, ShuffleKind, DAG) ||
+ PPC::isVMRGEOShuffleMask(SVOp, true, ShuffleKind, DAG) ||
+ PPC::isVMRGEOShuffleMask(SVOp, false, ShuffleKind, DAG))
return Op;
// Check to see if this is a shuffle of 4-byte values. If so, we can use our
@@ -9863,7 +9960,9 @@ SDValue PPCTargetLowering::expandVSXLoadForLE(SDNode *N,
case ISD::INTRINSIC_W_CHAIN: {
MemIntrinsicSDNode *Intrin = cast<MemIntrinsicSDNode>(N);
Chain = Intrin->getChain();
- Base = Intrin->getBasePtr();
+ // Similarly to the store case below, Intrin->getBasePtr() doesn't get
+ // us what we want. Get operand 2 instead.
+ Base = Intrin->getOperand(2);
MMO = Intrin->getMemOperand();
break;
}
diff --git a/lib/Target/PowerPC/PPCISelLowering.h b/lib/Target/PowerPC/PPCISelLowering.h
index c33d60565b79..02242b512a4f 100644
--- a/lib/Target/PowerPC/PPCISelLowering.h
+++ b/lib/Target/PowerPC/PPCISelLowering.h
@@ -353,7 +353,7 @@ namespace llvm {
/// the last operand.
TOC_ENTRY
};
- } // namespace PPCISD
+ }
/// Define some predicates that are used for node matching.
namespace PPC {
@@ -382,6 +382,11 @@ namespace llvm {
bool isVMRGHShuffleMask(ShuffleVectorSDNode *N, unsigned UnitSize,
unsigned ShuffleKind, SelectionDAG &DAG);
+ /// isVMRGEOShuffleMask - Return true if this is a shuffle mask suitable for
+ /// a VMRGEW or VMRGOW instruction
+ bool isVMRGEOShuffleMask(ShuffleVectorSDNode *N, bool CheckEven,
+ unsigned ShuffleKind, SelectionDAG &DAG);
+
/// isVSLDOIShuffleMask - If this is a vsldoi shuffle mask, return the
/// shift amount, otherwise return -1.
int isVSLDOIShuffleMask(SDNode *N, unsigned ShuffleKind,
@@ -405,7 +410,7 @@ namespace llvm {
/// If this is a qvaligni shuffle mask, return the shift
/// amount, otherwise return -1.
int isQVALIGNIShuffleMask(SDNode *N);
- } // namespace PPC
+ }
class PPCTargetLowering : public TargetLowering {
const PPCSubtarget &Subtarget;
@@ -871,6 +876,6 @@ namespace llvm {
CCValAssign::LocInfo &LocInfo,
ISD::ArgFlagsTy &ArgFlags,
CCState &State);
-} // namespace llvm
+}
#endif // LLVM_TARGET_POWERPC_PPC32ISELLOWERING_H
diff --git a/lib/Target/PowerPC/PPCInstrAltivec.td b/lib/Target/PowerPC/PPCInstrAltivec.td
index 9ff604bbee9d..cb0271fe8d0c 100644
--- a/lib/Target/PowerPC/PPCInstrAltivec.td
+++ b/lib/Target/PowerPC/PPCInstrAltivec.td
@@ -155,6 +155,33 @@ def vmrghw_swapped_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
}]>;
+def vmrgew_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
+ (vector_shuffle node:$lhs, node:$rhs), [{
+ return PPC::isVMRGEOShuffleMask(cast<ShuffleVectorSDNode>(N), true, 0, *CurDAG);
+}]>;
+def vmrgow_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
+ (vector_shuffle node:$lhs, node:$rhs), [{
+ return PPC::isVMRGEOShuffleMask(cast<ShuffleVectorSDNode>(N), false, 0, *CurDAG);
+}]>;
+def vmrgew_unary_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
+ (vector_shuffle node:$lhs, node:$rhs), [{
+ return PPC::isVMRGEOShuffleMask(cast<ShuffleVectorSDNode>(N), true, 1, *CurDAG);
+}]>;
+def vmrgow_unary_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
+ (vector_shuffle node:$lhs, node:$rhs), [{
+ return PPC::isVMRGEOShuffleMask(cast<ShuffleVectorSDNode>(N), false, 1, *CurDAG);
+}]>;
+def vmrgew_swapped_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
+ (vector_shuffle node:$lhs, node:$rhs), [{
+ return PPC::isVMRGEOShuffleMask(cast<ShuffleVectorSDNode>(N), true, 2, *CurDAG);
+}]>;
+def vmrgow_swapped_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
+ (vector_shuffle node:$lhs, node:$rhs), [{
+ return PPC::isVMRGEOShuffleMask(cast<ShuffleVectorSDNode>(N), false, 2, *CurDAG);
+}]>;
+
+
+
def VSLDOI_get_imm : SDNodeXForm<vector_shuffle, [{
return getI32Imm(PPC::isVSLDOIShuffleMask(N, 0, *CurDAG), SDLoc(N));
}]>;
@@ -1008,6 +1035,29 @@ def VMINSD : VX1_Int_Ty<962, "vminsd", int_ppc_altivec_vminsd, v2i64>;
def VMINUD : VX1_Int_Ty<706, "vminud", int_ppc_altivec_vminud, v2i64>;
} // isCommutable
+// Vector merge
+def VMRGEW : VXForm_1<1932, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB),
+ "vmrgew $vD, $vA, $vB", IIC_VecFP,
+ [(set v16i8:$vD, (vmrgew_shuffle v16i8:$vA, v16i8:$vB))]>;
+def VMRGOW : VXForm_1<1676, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB),
+ "vmrgow $vD, $vA, $vB", IIC_VecFP,
+ [(set v16i8:$vD, (vmrgow_shuffle v16i8:$vA, v16i8:$vB))]>;
+
+// Match vmrgew(x,x) and vmrgow(x,x)
+def:Pat<(vmrgew_unary_shuffle v16i8:$vA, undef),
+ (VMRGEW $vA, $vA)>;
+def:Pat<(vmrgow_unary_shuffle v16i8:$vA, undef),
+ (VMRGOW $vA, $vA)>;
+
+// Match vmrgew(y,x) and vmrgow(y,x), i.e., swapped operands. These fragments
+// are matched for little-endian, where the inputs must be swapped for correct
+// semantics.w
+def:Pat<(vmrgew_swapped_shuffle v16i8:$vA, v16i8:$vB),
+ (VMRGEW $vB, $vA)>;
+def:Pat<(vmrgow_swapped_shuffle v16i8:$vA, v16i8:$vB),
+ (VMRGOW $vB, $vA)>;
+
+
// Vector shifts
def VRLD : VX1_Int_Ty<196, "vrld", int_ppc_altivec_vrld, v2i64>;
def VSLD : VXForm_1<1476, (outs vrrc:$vD), (ins vrrc:$vA, vrrc:$vB),
diff --git a/lib/Target/PowerPC/PPCInstrBuilder.h b/lib/Target/PowerPC/PPCInstrBuilder.h
index ec94fa5580ff..cf71b1c59869 100644
--- a/lib/Target/PowerPC/PPCInstrBuilder.h
+++ b/lib/Target/PowerPC/PPCInstrBuilder.h
@@ -38,6 +38,6 @@ addFrameReference(const MachineInstrBuilder &MIB, int FI, int Offset = 0,
return MIB.addFrameIndex(FI).addImm(Offset);
}
-} // namespace llvm
+} // End llvm namespace
#endif
diff --git a/lib/Target/PowerPC/PPCInstrInfo.cpp b/lib/Target/PowerPC/PPCInstrInfo.cpp
index d3bb7a63c622..696a83860e53 100644
--- a/lib/Target/PowerPC/PPCInstrInfo.cpp
+++ b/lib/Target/PowerPC/PPCInstrInfo.cpp
@@ -352,15 +352,10 @@ bool PPCInstrInfo::AnalyzeBranch(MachineBasicBlock &MBB,MachineBasicBlock *&TBB,
bool isPPC64 = Subtarget.isPPC64();
// If the block has no terminators, it just falls into the block after it.
- MachineBasicBlock::iterator I = MBB.end();
- if (I == MBB.begin())
+ MachineBasicBlock::iterator I = MBB.getLastNonDebugInstr();
+ if (I == MBB.end())
return false;
- --I;
- while (I->isDebugValue()) {
- if (I == MBB.begin())
- return false;
- --I;
- }
+
if (!isUnpredicatedTerminator(I))
return false;
@@ -513,14 +508,10 @@ bool PPCInstrInfo::AnalyzeBranch(MachineBasicBlock &MBB,MachineBasicBlock *&TBB,
}
unsigned PPCInstrInfo::RemoveBranch(MachineBasicBlock &MBB) const {
- MachineBasicBlock::iterator I = MBB.end();
- if (I == MBB.begin()) return 0;
- --I;
- while (I->isDebugValue()) {
- if (I == MBB.begin())
- return 0;
- --I;
- }
+ MachineBasicBlock::iterator I = MBB.getLastNonDebugInstr();
+ if (I == MBB.end())
+ return 0;
+
if (I->getOpcode() != PPC::B && I->getOpcode() != PPC::BCC &&
I->getOpcode() != PPC::BC && I->getOpcode() != PPC::BCn &&
I->getOpcode() != PPC::BDNZ8 && I->getOpcode() != PPC::BDNZ &&
diff --git a/lib/Target/PowerPC/PPCInstrInfo.h b/lib/Target/PowerPC/PPCInstrInfo.h
index 39bf4547733c..e2d6346aa532 100644
--- a/lib/Target/PowerPC/PPCInstrInfo.h
+++ b/lib/Target/PowerPC/PPCInstrInfo.h
@@ -237,6 +237,6 @@ public:
void getNoopForMachoTarget(MCInst &NopInst) const override;
};
-} // namespace llvm
+}
#endif
diff --git a/lib/Target/PowerPC/PPCInstrVSX.td b/lib/Target/PowerPC/PPCInstrVSX.td
index d08b80871f3e..43ba4994fde6 100644
--- a/lib/Target/PowerPC/PPCInstrVSX.td
+++ b/lib/Target/PowerPC/PPCInstrVSX.td
@@ -457,22 +457,34 @@ let Uses = [RM] in {
defm XVCMPEQDP : XX3Form_Rcr<60, 99,
(outs vsrc:$XT), (ins vsrc:$XA, vsrc:$XB),
- "xvcmpeqdp", "$XT, $XA, $XB", IIC_VecFPCompare, []>;
+ "xvcmpeqdp", "$XT, $XA, $XB", IIC_VecFPCompare,
+ [(set v2i64:$XT,
+ (int_ppc_vsx_xvcmpeqdp v2f64:$XA, v2f64:$XB))]>;
defm XVCMPEQSP : XX3Form_Rcr<60, 67,
(outs vsrc:$XT), (ins vsrc:$XA, vsrc:$XB),
- "xvcmpeqsp", "$XT, $XA, $XB", IIC_VecFPCompare, []>;
+ "xvcmpeqsp", "$XT, $XA, $XB", IIC_VecFPCompare,
+ [(set v4i32:$XT,
+ (int_ppc_vsx_xvcmpeqsp v4f32:$XA, v4f32:$XB))]>;
defm XVCMPGEDP : XX3Form_Rcr<60, 115,
(outs vsrc:$XT), (ins vsrc:$XA, vsrc:$XB),
- "xvcmpgedp", "$XT, $XA, $XB", IIC_VecFPCompare, []>;
+ "xvcmpgedp", "$XT, $XA, $XB", IIC_VecFPCompare,
+ [(set v2i64:$XT,
+ (int_ppc_vsx_xvcmpgedp v2f64:$XA, v2f64:$XB))]>;
defm XVCMPGESP : XX3Form_Rcr<60, 83,
(outs vsrc:$XT), (ins vsrc:$XA, vsrc:$XB),
- "xvcmpgesp", "$XT, $XA, $XB", IIC_VecFPCompare, []>;
+ "xvcmpgesp", "$XT, $XA, $XB", IIC_VecFPCompare,
+ [(set v4i32:$XT,
+ (int_ppc_vsx_xvcmpgesp v4f32:$XA, v4f32:$XB))]>;
defm XVCMPGTDP : XX3Form_Rcr<60, 107,
(outs vsrc:$XT), (ins vsrc:$XA, vsrc:$XB),
- "xvcmpgtdp", "$XT, $XA, $XB", IIC_VecFPCompare, []>;
+ "xvcmpgtdp", "$XT, $XA, $XB", IIC_VecFPCompare,
+ [(set v2i64:$XT,
+ (int_ppc_vsx_xvcmpgtdp v2f64:$XA, v2f64:$XB))]>;
defm XVCMPGTSP : XX3Form_Rcr<60, 75,
(outs vsrc:$XT), (ins vsrc:$XA, vsrc:$XB),
- "xvcmpgtsp", "$XT, $XA, $XB", IIC_VecFPCompare, []>;
+ "xvcmpgtsp", "$XT, $XA, $XB", IIC_VecFPCompare,
+ [(set v4i32:$XT,
+ (int_ppc_vsx_xvcmpgtsp v4f32:$XA, v4f32:$XB))]>;
// Move Instructions
def XSABSDP : XX2Form<60, 345,
diff --git a/lib/Target/PowerPC/PPCLoopDataPrefetch.cpp b/lib/Target/PowerPC/PPCLoopDataPrefetch.cpp
index e783b5e65333..b4e1c099f190 100644
--- a/lib/Target/PowerPC/PPCLoopDataPrefetch.cpp
+++ b/lib/Target/PowerPC/PPCLoopDataPrefetch.cpp
@@ -88,7 +88,7 @@ namespace {
const TargetTransformInfo *TTI;
const DataLayout *DL;
};
-} // namespace
+}
char PPCLoopDataPrefetch::ID = 0;
INITIALIZE_PASS_BEGIN(PPCLoopDataPrefetch, "ppc-loop-data-prefetch",
diff --git a/lib/Target/PowerPC/PPCLoopPreIncPrep.cpp b/lib/Target/PowerPC/PPCLoopPreIncPrep.cpp
index 1891b6315c51..b6e7799402e1 100644
--- a/lib/Target/PowerPC/PPCLoopPreIncPrep.cpp
+++ b/lib/Target/PowerPC/PPCLoopPreIncPrep.cpp
@@ -87,7 +87,7 @@ namespace {
LoopInfo *LI;
ScalarEvolution *SE;
};
-} // namespace
+}
char PPCLoopPreIncPrep::ID = 0;
static const char *name = "Prepare loop for pre-inc. addressing modes";
@@ -113,7 +113,7 @@ namespace {
protected:
ScalarEvolution *SE;
};
-} // namespace
+}
static bool IsPtrInBounds(Value *BasePtr) {
Value *StrippedBasePtr = BasePtr;
diff --git a/lib/Target/PowerPC/PPCMCInstLower.cpp b/lib/Target/PowerPC/PPCMCInstLower.cpp
index c44d5d70f8dc..76837ecb32de 100644
--- a/lib/Target/PowerPC/PPCMCInstLower.cpp
+++ b/lib/Target/PowerPC/PPCMCInstLower.cpp
@@ -57,7 +57,7 @@ static MCSymbol *GetSymbolFromOperand(const MachineOperand &MO, AsmPrinter &AP){
if (!MO.isGlobal()) {
assert(MO.isSymbol() && "Isn't a symbol reference");
- Mang->getNameWithPrefix(Name, MO.getSymbolName());
+ Mangler::getNameWithPrefix(Name, MO.getSymbolName(), *DL);
} else {
const GlobalValue *GV = MO.getGlobal();
TM.getNameWithPrefix(Name, GV, *Mang);
diff --git a/lib/Target/PowerPC/PPCSelectionDAGInfo.h b/lib/Target/PowerPC/PPCSelectionDAGInfo.h
index d2eaeb42dbc4..2c1378d5670d 100644
--- a/lib/Target/PowerPC/PPCSelectionDAGInfo.h
+++ b/lib/Target/PowerPC/PPCSelectionDAGInfo.h
@@ -26,6 +26,6 @@ public:
~PPCSelectionDAGInfo();
};
-} // namespace llvm
+}
#endif
diff --git a/lib/Target/PowerPC/PPCSubtarget.h b/lib/Target/PowerPC/PPCSubtarget.h
index ea17e1c189b8..e9cc3d4bd5bc 100644
--- a/lib/Target/PowerPC/PPCSubtarget.h
+++ b/lib/Target/PowerPC/PPCSubtarget.h
@@ -58,7 +58,7 @@ namespace PPC {
DIR_PWR8,
DIR_64
};
-} // namespace PPC
+}
class GlobalValue;
class TargetMachine;
@@ -286,6 +286,6 @@ public:
bool enableSubRegLiveness() const override;
};
-} // namespace llvm
+} // End llvm namespace
#endif
diff --git a/lib/Target/PowerPC/PPCTLSDynamicCall.cpp b/lib/Target/PowerPC/PPCTLSDynamicCall.cpp
index 7a9db0fabb07..2dc0d825c80d 100644
--- a/lib/Target/PowerPC/PPCTLSDynamicCall.cpp
+++ b/lib/Target/PowerPC/PPCTLSDynamicCall.cpp
@@ -156,7 +156,7 @@ public:
MachineFunctionPass::getAnalysisUsage(AU);
}
};
-} // namespace
+}
INITIALIZE_PASS_BEGIN(PPCTLSDynamicCall, DEBUG_TYPE,
"PowerPC TLS Dynamic Call Fixup", false, false)
diff --git a/lib/Target/PowerPC/PPCTOCRegDeps.cpp b/lib/Target/PowerPC/PPCTOCRegDeps.cpp
index 61b963fe6da5..bf165c9edc6e 100644
--- a/lib/Target/PowerPC/PPCTOCRegDeps.cpp
+++ b/lib/Target/PowerPC/PPCTOCRegDeps.cpp
@@ -145,7 +145,7 @@ public:
MachineFunctionPass::getAnalysisUsage(AU);
}
};
-} // namespace
+}
INITIALIZE_PASS(PPCTOCRegDeps, DEBUG_TYPE,
"PowerPC TOC Register Dependencies", false, false)
diff --git a/lib/Target/PowerPC/PPCTargetStreamer.h b/lib/Target/PowerPC/PPCTargetStreamer.h
index a5c4c23c7901..dbe7617d3542 100644
--- a/lib/Target/PowerPC/PPCTargetStreamer.h
+++ b/lib/Target/PowerPC/PPCTargetStreamer.h
@@ -22,6 +22,6 @@ public:
virtual void emitAbiVersion(int AbiVersion) = 0;
virtual void emitLocalEntry(MCSymbolELF *S, const MCExpr *LocalOffset) = 0;
};
-} // namespace llvm
+}
#endif
diff --git a/lib/Target/PowerPC/PPCVSXCopy.cpp b/lib/Target/PowerPC/PPCVSXCopy.cpp
index 537db656fd60..5e3ae2a4471b 100644
--- a/lib/Target/PowerPC/PPCVSXCopy.cpp
+++ b/lib/Target/PowerPC/PPCVSXCopy.cpp
@@ -165,7 +165,7 @@ public:
MachineFunctionPass::getAnalysisUsage(AU);
}
};
-} // namespace
+}
INITIALIZE_PASS(PPCVSXCopy, DEBUG_TYPE,
"PowerPC VSX Copy Legalization", false, false)
diff --git a/lib/Target/PowerPC/PPCVSXFMAMutate.cpp b/lib/Target/PowerPC/PPCVSXFMAMutate.cpp
index a029ddf0bc08..f352fa647ace 100644
--- a/lib/Target/PowerPC/PPCVSXFMAMutate.cpp
+++ b/lib/Target/PowerPC/PPCVSXFMAMutate.cpp
@@ -317,7 +317,7 @@ public:
MachineFunctionPass::getAnalysisUsage(AU);
}
};
-} // namespace
+}
INITIALIZE_PASS_BEGIN(PPCVSXFMAMutate, DEBUG_TYPE,
"PowerPC VSX FMA Mutation", false, false)
diff --git a/lib/Target/PowerPC/PPCVSXSwapRemoval.cpp b/lib/Target/PowerPC/PPCVSXSwapRemoval.cpp
index 939293a5638e..e7ab71ac2106 100644
--- a/lib/Target/PowerPC/PPCVSXSwapRemoval.cpp
+++ b/lib/Target/PowerPC/PPCVSXSwapRemoval.cpp
@@ -79,7 +79,6 @@ struct PPCVSXSwapEntry {
unsigned int IsStore : 1;
unsigned int IsSwap : 1;
unsigned int MentionsPhysVR : 1;
- unsigned int HasImplicitSubreg : 1;
unsigned int IsSwappable : 1;
unsigned int SpecialHandling : 3;
unsigned int WebRejected : 1;
@@ -224,7 +223,6 @@ bool PPCVSXSwapRemoval::gatherVectorInstructions() {
for (MachineInstr &MI : MBB) {
bool RelevantInstr = false;
- bool ImplicitSubreg = false;
for (const MachineOperand &MO : MI.operands()) {
if (!MO.isReg())
@@ -232,8 +230,6 @@ bool PPCVSXSwapRemoval::gatherVectorInstructions() {
unsigned Reg = MO.getReg();
if (isVecReg(Reg)) {
RelevantInstr = true;
- if (MO.getSubReg() != 0)
- ImplicitSubreg = true;
break;
}
}
@@ -249,9 +245,6 @@ bool PPCVSXSwapRemoval::gatherVectorInstructions() {
PPCVSXSwapEntry SwapEntry{};
int VecIdx = addSwapEntry(&MI, SwapEntry);
- if (ImplicitSubreg)
- SwapVector[VecIdx].HasImplicitSubreg = 1;
-
switch(MI.getOpcode()) {
default:
// Unless noted otherwise, an instruction is considered
@@ -260,7 +253,7 @@ bool PPCVSXSwapRemoval::gatherVectorInstructions() {
// select, compare, etc.).
SwapVector[VecIdx].IsSwappable = 1;
break;
- case PPC::XXPERMDI:
+ case PPC::XXPERMDI: {
// This is a swap if it is of the form XXPERMDI t, s, s, 2.
// Unfortunately, MachineCSE ignores COPY and SUBREG_TO_REG, so we
// can also see XXPERMDI t, SUBREG_TO_REG(s), SUBREG_TO_REG(s), 2,
@@ -268,9 +261,8 @@ bool PPCVSXSwapRemoval::gatherVectorInstructions() {
// SUBREG_TO_REG to find the real source value for comparison.
// If the real source value is a physical register, then mark the
// XXPERMDI as mentioning a physical register.
- // Any other form of XXPERMDI is lane-sensitive and unsafe
- // for the optimization.
- if (MI.getOperand(3).getImm() == 2) {
+ int immed = MI.getOperand(3).getImm();
+ if (immed == 2) {
unsigned trueReg1 = lookThruCopyLike(MI.getOperand(1).getReg(),
VecIdx);
unsigned trueReg2 = lookThruCopyLike(MI.getOperand(2).getReg(),
@@ -278,7 +270,26 @@ bool PPCVSXSwapRemoval::gatherVectorInstructions() {
if (trueReg1 == trueReg2)
SwapVector[VecIdx].IsSwap = 1;
}
+ // This is a doubleword splat if it is of the form
+ // XXPERMDI t, s, s, 0 or XXPERMDI t, s, s, 3. As above we
+ // must look through chains of copy-likes to find the source
+ // register. We turn off the marking for mention of a physical
+ // register, because splatting it is safe; the optimization
+ // will not swap the value in the physical register.
+ else if (immed == 0 || immed == 3) {
+ unsigned trueReg1 = lookThruCopyLike(MI.getOperand(1).getReg(),
+ VecIdx);
+ unsigned trueReg2 = lookThruCopyLike(MI.getOperand(2).getReg(),
+ VecIdx);
+ if (trueReg1 == trueReg2) {
+ SwapVector[VecIdx].IsSwappable = 1;
+ SwapVector[VecIdx].MentionsPhysVR = 0;
+ }
+ }
+ // Any other form of XXPERMDI is lane-sensitive and unsafe
+ // for the optimization.
break;
+ }
case PPC::LVX:
// Non-permuting loads are currently unsafe. We can use special
// handling for this in the future. By not marking these as
@@ -307,14 +318,6 @@ bool PPCVSXSwapRemoval::gatherVectorInstructions() {
SwapVector[VecIdx].IsStore = 1;
SwapVector[VecIdx].IsSwap = 1;
break;
- case PPC::SUBREG_TO_REG:
- // These are fine provided they are moving between full vector
- // register classes. For example, the VRs are a subset of the
- // VSRs, but each VR and each VSR is a full 128-bit register.
- if (isVecReg(MI.getOperand(0).getReg()) &&
- isVecReg(MI.getOperand(2).getReg()))
- SwapVector[VecIdx].IsSwappable = 1;
- break;
case PPC::COPY:
// These are fine provided they are moving between full vector
// register classes.
@@ -349,7 +352,6 @@ bool PPCVSXSwapRemoval::gatherVectorInstructions() {
case PPC::LVSL:
case PPC::LVSR:
case PPC::LVXL:
- case PPC::LXVDSX:
case PPC::STVEBX:
case PPC::STVEHX:
case PPC::STVEWX:
@@ -457,23 +459,19 @@ int PPCVSXSwapRemoval::addSwapEntry(MachineInstr *MI,
// such operations to the ultimate source register. If a
// physical register is encountered, we stop the search and
// flag the swap entry indicated by VecIdx (the original
-// XXPERMDI) as mentioning a physical register. Similarly
-// for implicit subregister mentions (which should never
-// happen).
+// XXPERMDI) as mentioning a physical register.
unsigned PPCVSXSwapRemoval::lookThruCopyLike(unsigned SrcReg,
unsigned VecIdx) {
MachineInstr *MI = MRI->getVRegDef(SrcReg);
if (!MI->isCopyLike())
return SrcReg;
- unsigned CopySrcReg, CopySrcSubreg;
- if (MI->isCopy()) {
+ unsigned CopySrcReg;
+ if (MI->isCopy())
CopySrcReg = MI->getOperand(1).getReg();
- CopySrcSubreg = MI->getOperand(1).getSubReg();
- } else {
+ else {
assert(MI->isSubregToReg() && "bad opcode for lookThruCopyLike");
CopySrcReg = MI->getOperand(2).getReg();
- CopySrcSubreg = MI->getOperand(2).getSubReg();
}
if (!TargetRegisterInfo::isVirtualRegister(CopySrcReg)) {
@@ -481,11 +479,6 @@ unsigned PPCVSXSwapRemoval::lookThruCopyLike(unsigned SrcReg,
return CopySrcReg;
}
- if (CopySrcSubreg != 0) {
- SwapVector[VecIdx].HasImplicitSubreg = 1;
- return CopySrcReg;
- }
-
return lookThruCopyLike(CopySrcReg, VecIdx);
}
@@ -552,11 +545,9 @@ void PPCVSXSwapRemoval::recordUnoptimizableWebs() {
for (unsigned EntryIdx = 0; EntryIdx < SwapVector.size(); ++EntryIdx) {
int Repr = EC->getLeaderValue(SwapVector[EntryIdx].VSEId);
- // Reject webs containing mentions of physical registers or implicit
- // subregs, or containing operations that we don't know how to handle
- // in a lane-permuted region.
+ // Reject webs containing mentions of physical registers, or containing
+ // operations that we don't know how to handle in a lane-permuted region.
if (SwapVector[EntryIdx].MentionsPhysVR ||
- SwapVector[EntryIdx].HasImplicitSubreg ||
!(SwapVector[EntryIdx].IsSwappable || SwapVector[EntryIdx].IsSwap)) {
SwapVector[Repr].WebRejected = 1;
@@ -765,8 +756,6 @@ void PPCVSXSwapRemoval::dumpSwapVector() {
DEBUG(dbgs() << "swap ");
if (SwapVector[EntryIdx].MentionsPhysVR)
DEBUG(dbgs() << "physreg ");
- if (SwapVector[EntryIdx].HasImplicitSubreg)
- DEBUG(dbgs() << "implsubreg ");
if (SwapVector[EntryIdx].IsSwappable) {
DEBUG(dbgs() << "swappable ");
@@ -809,7 +798,7 @@ void PPCVSXSwapRemoval::dumpSwapVector() {
DEBUG(dbgs() << "\n");
}
-} // namespace
+} // end default namespace
INITIALIZE_PASS_BEGIN(PPCVSXSwapRemoval, DEBUG_TYPE,
"PowerPC VSX Swap Removal", false, false)