[llvm] r353043 - [AsmPrinter] Remove hidden flag -print-schedule.
Andrea Di Biagio via llvm-commits
llvm-commits at lists.llvm.org
Mon Feb 4 04:51:26 PST 2019
Author: adibiagio
Date: Mon Feb 4 04:51:26 2019
New Revision: 353043
URL: http://llvm.org/viewvc/llvm-project?rev=353043&view=rev
Log:
[AsmPrinter] Remove hidden flag -print-schedule.
This patch removes hidden codegen flag -print-schedule effectively reverting the
logic originally committed as r300311
(https://llvm.org/viewvc/llvm-project?view=revision&revision=300311).
Flag -print-schedule was originally introduced by r300311 to address PR32216
(https://bugs.llvm.org/show_bug.cgi?id=32216). That bug was about adding "Better
testing of schedule model instruction latencies/throughputs".
These days, we can use llvm-mca to test scheduling models. So there is no longer
a need for flag -print-schedule in LLVM. The main use case for PR32216 is
now addressed by llvm-mca.
Flag -print-schedule is mainly used for debugging purposes, and it is only
actually used by x86 specific tests. We already have extensive (latency and
throughput) tests under "test/tools/llvm-mca" for X86 processor models. That
means, most (if not all) existing -print-schedule tests for X86 are redundant.
When flag -print-schedule was first added to LLVM, several files had to be
modified; a few APIs gained new arguments (see for example method
MCAsmStreamer::EmitInstruction), and MCSubtargetInfo/TargetSubtargetInfo gained
a couple of getSchedInfoStr() methods.
Method getSchedInfoStr() had to originally work for both MCInst and
MachineInstr. The original implmentation of getSchedInfoStr() introduced a
subtle layering violation (reported as PR37160 and then fixed/worked-around by
r330615).
In retrospect, that new API could have been designed more optimally. We can
always query MCSchedModel to get the latency and throughput. More importantly,
the "sched-info" string should not have been generated by the subtarget.
Note, r317782 fixed an issue where "print-schedule" didn't work very well in the
presence of inline assembly. That commit is also reverted by this change.
Differential Revision: https://reviews.llvm.org/D57244
Removed:
llvm/trunk/test/CodeGen/X86/3dnow-schedule.ll
llvm/trunk/test/CodeGen/X86/adx-schedule.ll
llvm/trunk/test/CodeGen/X86/aes-schedule.ll
llvm/trunk/test/CodeGen/X86/avx-schedule.ll
llvm/trunk/test/CodeGen/X86/avx2-schedule.ll
llvm/trunk/test/CodeGen/X86/avx512-schedule.ll
llvm/trunk/test/CodeGen/X86/avx512-shuffle-schedule.ll
llvm/trunk/test/CodeGen/X86/avx512vpopcntdq-schedule.ll
llvm/trunk/test/CodeGen/X86/bmi-schedule.ll
llvm/trunk/test/CodeGen/X86/bmi2-schedule.ll
llvm/trunk/test/CodeGen/X86/clflushopt-schedule.ll
llvm/trunk/test/CodeGen/X86/clwb-schedule.ll
llvm/trunk/test/CodeGen/X86/clzero-schedule.ll
llvm/trunk/test/CodeGen/X86/cmov-schedule.ll
llvm/trunk/test/CodeGen/X86/f16c-schedule.ll
llvm/trunk/test/CodeGen/X86/fma-schedule.ll
llvm/trunk/test/CodeGen/X86/fma4-schedule.ll
llvm/trunk/test/CodeGen/X86/fsgsbase-schedule.ll
llvm/trunk/test/CodeGen/X86/lea32-schedule.ll
llvm/trunk/test/CodeGen/X86/lea64-schedule.ll
llvm/trunk/test/CodeGen/X86/lwp-schedule.ll
llvm/trunk/test/CodeGen/X86/lzcnt-schedule.ll
llvm/trunk/test/CodeGen/X86/mmx-schedule.ll
llvm/trunk/test/CodeGen/X86/movbe-schedule.ll
llvm/trunk/test/CodeGen/X86/mwaitx-schedule.ll
llvm/trunk/test/CodeGen/X86/popcnt-schedule.ll
llvm/trunk/test/CodeGen/X86/rdpid-schedule.ll
llvm/trunk/test/CodeGen/X86/rdrand-schedule.ll
llvm/trunk/test/CodeGen/X86/rdseed-schedule.ll
llvm/trunk/test/CodeGen/X86/rtm-schedule.ll
llvm/trunk/test/CodeGen/X86/schedule-x86-64-shld.ll
llvm/trunk/test/CodeGen/X86/schedule-x86_32.ll
llvm/trunk/test/CodeGen/X86/schedule-x86_64.ll
llvm/trunk/test/CodeGen/X86/sha-schedule.ll
llvm/trunk/test/CodeGen/X86/sse-schedule.ll
llvm/trunk/test/CodeGen/X86/sse2-schedule.ll
llvm/trunk/test/CodeGen/X86/sse3-schedule.ll
llvm/trunk/test/CodeGen/X86/sse41-schedule.ll
llvm/trunk/test/CodeGen/X86/sse42-schedule.ll
llvm/trunk/test/CodeGen/X86/sse4a-schedule.ll
llvm/trunk/test/CodeGen/X86/ssse3-schedule.ll
llvm/trunk/test/CodeGen/X86/tbm-schedule.ll
llvm/trunk/test/CodeGen/X86/x87-schedule.ll
llvm/trunk/test/CodeGen/X86/xop-schedule.ll
Modified:
llvm/trunk/include/llvm/CodeGen/AsmPrinter.h
llvm/trunk/include/llvm/CodeGen/TargetSubtargetInfo.h
llvm/trunk/include/llvm/MC/MCObjectStreamer.h
llvm/trunk/include/llvm/MC/MCParser/MCAsmParser.h
llvm/trunk/include/llvm/MC/MCStreamer.h
llvm/trunk/include/llvm/MC/MCSubtargetInfo.h
llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinterInlineAsm.cpp
llvm/trunk/lib/CodeGen/MachineCombiner.cpp
llvm/trunk/lib/CodeGen/TargetSubtargetInfo.cpp
llvm/trunk/lib/MC/MCAsmStreamer.cpp
llvm/trunk/lib/MC/MCObjectStreamer.cpp
llvm/trunk/lib/MC/MCStreamer.cpp
llvm/trunk/lib/Object/RecordStreamer.cpp
llvm/trunk/lib/Object/RecordStreamer.h
llvm/trunk/lib/Target/AArch64/MCTargetDesc/AArch64ELFStreamer.cpp
llvm/trunk/lib/Target/ARM/MCTargetDesc/ARMELFStreamer.cpp
llvm/trunk/lib/Target/Hexagon/MCTargetDesc/HexagonMCELFStreamer.cpp
llvm/trunk/lib/Target/Hexagon/MCTargetDesc/HexagonMCELFStreamer.h
llvm/trunk/lib/Target/Mips/MCTargetDesc/MipsELFStreamer.cpp
llvm/trunk/lib/Target/Mips/MCTargetDesc/MipsELFStreamer.h
llvm/trunk/lib/Target/Mips/MCTargetDesc/MipsNaClELFStreamer.cpp
llvm/trunk/lib/Target/X86/AsmParser/X86AsmInstrumentation.cpp
llvm/trunk/lib/Target/X86/AsmParser/X86AsmInstrumentation.h
llvm/trunk/lib/Target/X86/AsmParser/X86AsmParser.cpp
llvm/trunk/lib/Target/X86/InstPrinter/X86InstComments.cpp
llvm/trunk/lib/Target/X86/MCTargetDesc/X86BaseInfo.h
llvm/trunk/lib/Target/X86/X86MCInstLower.cpp
llvm/trunk/lib/Target/X86/X86Subtarget.h
llvm/trunk/test/CodeGen/X86/mul-constant-i32.ll
llvm/trunk/test/CodeGen/X86/mul-constant-i64.ll
llvm/trunk/test/CodeGen/X86/recip-fastmath.ll
llvm/trunk/test/CodeGen/X86/recip-fastmath2.ll
llvm/trunk/tools/llvm-exegesis/llvm-exegesis.cpp
llvm/trunk/tools/llvm-mca/CodeRegionGenerator.cpp
Modified: llvm/trunk/include/llvm/CodeGen/AsmPrinter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/AsmPrinter.h?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/include/llvm/CodeGen/AsmPrinter.h (original)
+++ llvm/trunk/include/llvm/CodeGen/AsmPrinter.h Mon Feb 4 04:51:26 2019
@@ -121,9 +121,6 @@ public:
using GOTEquivUsePair = std::pair<const GlobalVariable *, unsigned>;
MapVector<const MCSymbol *, GOTEquivUsePair> GlobalGOTEquivs;
- /// Enable print [latency:throughput] in output.
- bool EnablePrintSchedInfo = false;
-
private:
MCSymbol *CurrentFnBegin = nullptr;
MCSymbol *CurrentFnEnd = nullptr;
Modified: llvm/trunk/include/llvm/CodeGen/TargetSubtargetInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/TargetSubtargetInfo.h?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/include/llvm/CodeGen/TargetSubtargetInfo.h (original)
+++ llvm/trunk/include/llvm/CodeGen/TargetSubtargetInfo.h Mon Feb 4 04:51:26 2019
@@ -189,9 +189,6 @@ public:
/// TargetLowering preference). It does not yet disable the postRA scheduler.
virtual bool enableMachineScheduler() const;
- /// Support printing of [latency:throughput] comment in output .S file.
- virtual bool supportPrintSchedInfo() const { return false; }
-
/// True if the machine scheduler should disable the TLI preference
/// for preRA scheduling with the source level scheduler.
virtual bool enableMachineSchedDefaultSched() const { return true; }
@@ -285,10 +282,6 @@ public:
/// possible.
virtual bool enableSubRegLiveness() const { return false; }
- /// Returns string representation of scheduler comment
- std::string getSchedInfoStr(const MachineInstr &MI) const;
- std::string getSchedInfoStr(MCInst const &MCI) const override;
-
/// This is called after a .mir file was loaded.
virtual void mirFileLoaded(MachineFunction &MF) const;
};
Modified: llvm/trunk/include/llvm/MC/MCObjectStreamer.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCObjectStreamer.h?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCObjectStreamer.h (original)
+++ llvm/trunk/include/llvm/MC/MCObjectStreamer.h Mon Feb 4 04:51:26 2019
@@ -115,8 +115,7 @@ public:
void EmitSLEB128Value(const MCExpr *Value) override;
void EmitWeakReference(MCSymbol *Alias, const MCSymbol *Symbol) override;
void ChangeSection(MCSection *Section, const MCExpr *Subsection) override;
- void EmitInstruction(const MCInst &Inst, const MCSubtargetInfo &STI,
- bool = false) override;
+ void EmitInstruction(const MCInst &Inst, const MCSubtargetInfo &STI) override;
/// Emit an instruction to a special fragment, because this instruction
/// can change its size during relaxation.
Modified: llvm/trunk/include/llvm/MC/MCParser/MCAsmParser.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCParser/MCAsmParser.h?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCParser/MCAsmParser.h (original)
+++ llvm/trunk/include/llvm/MC/MCParser/MCAsmParser.h Mon Feb 4 04:51:26 2019
@@ -129,9 +129,6 @@ protected: // Can only create subclasses
/// Flag tracking whether any errors have been encountered.
bool HadError = false;
- /// Enable print [latency:throughput] in output file.
- bool EnablePrintSchedInfo = false;
-
bool ShowParsedOperands = false;
public:
@@ -165,9 +162,6 @@ public:
bool getShowParsedOperands() const { return ShowParsedOperands; }
void setShowParsedOperands(bool Value) { ShowParsedOperands = Value; }
- void setEnablePrintSchedInfo(bool Value) { EnablePrintSchedInfo = Value; }
- bool shouldPrintSchedInfo() const { return EnablePrintSchedInfo; }
-
/// Run the parser on the input source buffer.
virtual bool Run(bool NoInitialTextSection, bool NoFinalize = false) = 0;
Modified: llvm/trunk/include/llvm/MC/MCStreamer.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCStreamer.h?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCStreamer.h (original)
+++ llvm/trunk/include/llvm/MC/MCStreamer.h Mon Feb 4 04:51:26 2019
@@ -952,9 +952,7 @@ public:
virtual void EmitAddrsigSym(const MCSymbol *Sym) {}
/// Emit the given \p Instruction into the current section.
- /// PrintSchedInfo == true then schedul comment should be added to output
- virtual void EmitInstruction(const MCInst &Inst, const MCSubtargetInfo &STI,
- bool PrintSchedInfo = false);
+ virtual void EmitInstruction(const MCInst &Inst, const MCSubtargetInfo &STI);
/// Set the bundle alignment mode from now on in the section.
/// The argument is the power of 2 to which the alignment is set. The
Modified: llvm/trunk/include/llvm/MC/MCSubtargetInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCSubtargetInfo.h?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCSubtargetInfo.h (original)
+++ llvm/trunk/include/llvm/MC/MCSubtargetInfo.h Mon Feb 4 04:51:26 2019
@@ -180,11 +180,6 @@ public:
auto Found = std::lower_bound(ProcDesc.begin(), ProcDesc.end(), CPU);
return Found != ProcDesc.end() && StringRef(Found->Key) == CPU;
}
-
- /// Returns string representation of scheduler comment
- virtual std::string getSchedInfoStr(MCInst const &MCI) const {
- return {};
- }
};
} // end namespace llvm
Modified: llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp Mon Feb 4 04:51:26 2019
@@ -59,7 +59,6 @@
#include "llvm/CodeGen/TargetLowering.h"
#include "llvm/CodeGen/TargetOpcodes.h"
#include "llvm/CodeGen/TargetRegisterInfo.h"
-#include "llvm/CodeGen/TargetSubtargetInfo.h"
#include "llvm/IR/BasicBlock.h"
#include "llvm/IR/Comdat.h"
#include "llvm/IR/Constant.h"
@@ -142,10 +141,6 @@ static const char *const CodeViewLineTab
STATISTIC(EmittedInsts, "Number of machine instrs printed");
-static cl::opt<bool>
- PrintSchedule("print-schedule", cl::Hidden, cl::init(false),
- cl::desc("Print 'sched: [latency:throughput]' in .s output"));
-
char AsmPrinter::ID = 0;
using gcp_map_type = DenseMap<GCStrategy *, std::unique_ptr<GCMetadataPrinter>>;
@@ -746,10 +741,7 @@ void AsmPrinter::EmitFunctionEntryLabel(
}
/// emitComments - Pretty-print comments for instructions.
-/// It returns true iff the sched comment was emitted.
-/// Otherwise it returns false.
-static bool emitComments(const MachineInstr &MI, raw_ostream &CommentOS,
- AsmPrinter *AP) {
+static void emitComments(const MachineInstr &MI, raw_ostream &CommentOS) {
const MachineFunction *MF = MI.getMF();
const TargetInstrInfo *TII = MF->getSubtarget().getInstrInfo();
@@ -757,7 +749,6 @@ static bool emitComments(const MachineIn
int FI;
const MachineFrameInfo &MFI = MF->getFrameInfo();
- bool Commented = false;
auto getSize =
[&MFI](const SmallVectorImpl<const MachineMemOperand *> &Accesses) {
@@ -777,43 +768,24 @@ static bool emitComments(const MachineIn
if (TII->isLoadFromStackSlotPostFE(MI, FI)) {
if (MFI.isSpillSlotObjectIndex(FI)) {
MMO = *MI.memoperands_begin();
- CommentOS << MMO->getSize() << "-byte Reload";
- Commented = true;
+ CommentOS << MMO->getSize() << "-byte Reload\n";
}
} else if (TII->hasLoadFromStackSlot(MI, Accesses)) {
- if (auto Size = getSize(Accesses)) {
- CommentOS << Size << "-byte Folded Reload";
- Commented = true;
- }
+ if (auto Size = getSize(Accesses))
+ CommentOS << Size << "-byte Folded Reload\n";
} else if (TII->isStoreToStackSlotPostFE(MI, FI)) {
if (MFI.isSpillSlotObjectIndex(FI)) {
MMO = *MI.memoperands_begin();
- CommentOS << MMO->getSize() << "-byte Spill";
- Commented = true;
+ CommentOS << MMO->getSize() << "-byte Spill\n";
}
} else if (TII->hasStoreToStackSlot(MI, Accesses)) {
- if (auto Size = getSize(Accesses)) {
- CommentOS << Size << "-byte Folded Spill";
- Commented = true;
- }
+ if (auto Size = getSize(Accesses))
+ CommentOS << Size << "-byte Folded Spill\n";
}
// Check for spill-induced copies
- if (MI.getAsmPrinterFlag(MachineInstr::ReloadReuse)) {
- Commented = true;
- CommentOS << " Reload Reuse";
- }
-
- if (Commented) {
- if (AP->EnablePrintSchedInfo) {
- // If any comment was added above and we need sched info comment then add
- // this new comment just after the above comment w/o "\n" between them.
- CommentOS << " " << MF->getSubtarget().getSchedInfoStr(MI) << "\n";
- return true;
- }
- CommentOS << "\n";
- }
- return false;
+ if (MI.getAsmPrinterFlag(MachineInstr::ReloadReuse))
+ CommentOS << " Reload Reuse\n";
}
/// emitImplicitDef - This method emits the specified machine instruction
@@ -1101,10 +1073,8 @@ void AsmPrinter::EmitFunctionBody() {
}
}
- if (isVerbose() && emitComments(MI, OutStreamer->GetCommentOS(), this)) {
- MachineInstr *MIP = const_cast<MachineInstr *>(&MI);
- MIP->setAsmPrinterFlag(MachineInstr::NoSchedComment);
- }
+ if (isVerbose())
+ emitComments(MI, OutStreamer->GetCommentOS());
switch (MI.getOpcode()) {
case TargetOpcode::CFI_INSTRUCTION:
@@ -1636,11 +1606,6 @@ void AsmPrinter::SetupMachineFunction(Ma
}
ORE = &getAnalysis<MachineOptimizationRemarkEmitterPass>().getORE();
-
- const TargetSubtargetInfo &STI = MF.getSubtarget();
- EnablePrintSchedInfo = PrintSchedule.getNumOccurrences()
- ? PrintSchedule
- : STI.supportPrintSchedInfo();
}
namespace {
Modified: llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinterInlineAsm.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinterInlineAsm.cpp?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinterInlineAsm.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinterInlineAsm.cpp Mon Feb 4 04:51:26 2019
@@ -18,7 +18,6 @@
#include "llvm/CodeGen/MachineModuleInfo.h"
#include "llvm/CodeGen/TargetInstrInfo.h"
#include "llvm/CodeGen/TargetRegisterInfo.h"
-#include "llvm/CodeGen/TargetSubtargetInfo.h"
#include "llvm/IR/Constants.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/InlineAsm.h"
@@ -154,7 +153,6 @@ void AsmPrinter::EmitInlineAsm(StringRef
" we don't have an asm parser for this target\n");
Parser->setAssemblerDialect(Dialect);
Parser->setTargetParser(*TAP.get());
- Parser->setEnablePrintSchedInfo(EnablePrintSchedInfo);
// Enable lexing Masm binary and hex integer literals in intel inline
// assembly.
if (Dialect == InlineAsm::AD_Intel)
Modified: llvm/trunk/lib/CodeGen/MachineCombiner.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineCombiner.cpp?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/MachineCombiner.cpp (original)
+++ llvm/trunk/lib/CodeGen/MachineCombiner.cpp Mon Feb 4 04:51:26 2019
@@ -558,16 +558,13 @@ bool MachineCombiner::combineInstruction
continue;
LLVM_DEBUG(if (dump_intrs) {
- dbgs() << "\tFor the Pattern (" << (int)P << ") these instructions could be removed\n";
- for (auto const *InstrPtr : DelInstrs) {
- dbgs() << "\t\t" << STI->getSchedInfoStr(*InstrPtr) << ": ";
+ dbgs() << "\tFor the Pattern (" << (int)P
+ << ") these instructions could be removed\n";
+ for (auto const *InstrPtr : DelInstrs)
InstrPtr->print(dbgs(), false, false, false, TII);
- }
dbgs() << "\tThese instructions could replace the removed ones\n";
- for (auto const *InstrPtr : InsInstrs) {
- dbgs() << "\t\t" << STI->getSchedInfoStr(*InstrPtr) << ": ";
+ for (auto const *InstrPtr : InsInstrs)
InstrPtr->print(dbgs(), false, false, false, TII);
- }
});
bool SubstituteAlways = false;
Modified: llvm/trunk/lib/CodeGen/TargetSubtargetInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/TargetSubtargetInfo.cpp?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/TargetSubtargetInfo.cpp (original)
+++ llvm/trunk/lib/CodeGen/TargetSubtargetInfo.cpp Mon Feb 4 04:51:26 2019
@@ -11,14 +11,6 @@
//===----------------------------------------------------------------------===//
#include "llvm/CodeGen/TargetSubtargetInfo.h"
-#include "llvm/ADT/Optional.h"
-#include "llvm/CodeGen/MachineInstr.h"
-#include "llvm/CodeGen/TargetInstrInfo.h"
-#include "llvm/CodeGen/TargetSchedule.h"
-#include "llvm/MC/MCInst.h"
-#include "llvm/Support/Format.h"
-#include "llvm/Support/raw_ostream.h"
-#include <string>
using namespace llvm;
@@ -66,64 +58,4 @@ bool TargetSubtargetInfo::useAA() const
return false;
}
-static std::string createSchedInfoStr(unsigned Latency, double RThroughput) {
- static const char *SchedPrefix = " sched: [";
- std::string Comment;
- raw_string_ostream CS(Comment);
- if (RThroughput != 0.0)
- CS << SchedPrefix << Latency << format(":%2.2f", RThroughput)
- << "]";
- else
- CS << SchedPrefix << Latency << ":?]";
- CS.flush();
- return Comment;
-}
-
-/// Returns string representation of scheduler comment
-std::string TargetSubtargetInfo::getSchedInfoStr(const MachineInstr &MI) const {
- if (MI.isPseudo() || MI.isTerminator())
- return std::string();
- // We don't cache TSchedModel because it depends on TargetInstrInfo
- // that could be changed during the compilation
- TargetSchedModel TSchedModel;
- TSchedModel.init(this);
- unsigned Latency = TSchedModel.computeInstrLatency(&MI);
-
- // Add extra latency due to forwarding delays.
- const MCSchedClassDesc &SCDesc = *TSchedModel.resolveSchedClass(&MI);
- Latency +=
- MCSchedModel::getForwardingDelayCycles(getReadAdvanceEntries(SCDesc));
-
- double RThroughput = TSchedModel.computeReciprocalThroughput(&MI);
- return createSchedInfoStr(Latency, RThroughput);
-}
-
-/// Returns string representation of scheduler comment
-std::string TargetSubtargetInfo::getSchedInfoStr(MCInst const &MCI) const {
- // We don't cache TSchedModel because it depends on TargetInstrInfo
- // that could be changed during the compilation
- TargetSchedModel TSchedModel;
- TSchedModel.init(this);
- unsigned Latency;
- if (TSchedModel.hasInstrSchedModel()) {
- Latency = TSchedModel.computeInstrLatency(MCI);
- // Add extra latency due to forwarding delays.
- const MCSchedModel &SM = *TSchedModel.getMCSchedModel();
- unsigned SClassID = getInstrInfo()->get(MCI.getOpcode()).getSchedClass();
- while (SM.getSchedClassDesc(SClassID)->isVariant())
- SClassID = resolveVariantSchedClass(SClassID, &MCI, SM.ProcID);
- const MCSchedClassDesc &SCDesc = *SM.getSchedClassDesc(SClassID);
- Latency +=
- MCSchedModel::getForwardingDelayCycles(getReadAdvanceEntries(SCDesc));
- } else if (TSchedModel.hasInstrItineraries()) {
- auto *ItinData = TSchedModel.getInstrItineraries();
- Latency = ItinData->getStageLatency(
- getInstrInfo()->get(MCI.getOpcode()).getSchedClass());
- } else
- return std::string();
- double RThroughput = TSchedModel.computeReciprocalThroughput(MCI);
- return createSchedInfoStr(Latency, RThroughput);
-}
-
-void TargetSubtargetInfo::mirFileLoaded(MachineFunction &MF) const {
-}
+void TargetSubtargetInfo::mirFileLoaded(MachineFunction &MF) const { }
Modified: llvm/trunk/lib/MC/MCAsmStreamer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCAsmStreamer.cpp?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCAsmStreamer.cpp (original)
+++ llvm/trunk/lib/MC/MCAsmStreamer.cpp Mon Feb 4 04:51:26 2019
@@ -107,10 +107,7 @@ public:
void AddComment(const Twine &T, bool EOL = true) override;
/// Add a comment showing the encoding of an instruction.
- /// If PrintSchedInfo is true, then the comment sched:[x:y] will be added to
- /// the output if supported by the target.
- void AddEncodingComment(const MCInst &Inst, const MCSubtargetInfo &,
- bool PrintSchedInfo);
+ void AddEncodingComment(const MCInst &Inst, const MCSubtargetInfo &);
/// Return a raw_ostream that comments can be written to.
/// Unlike AddComment, you are required to terminate comments with \n if you
@@ -311,8 +308,7 @@ public:
void emitCGProfileEntry(const MCSymbolRefExpr *From,
const MCSymbolRefExpr *To, uint64_t Count) override;
- void EmitInstruction(const MCInst &Inst, const MCSubtargetInfo &STI,
- bool PrintSchedInfo) override;
+ void EmitInstruction(const MCInst &Inst, const MCSubtargetInfo &STI) override;
void EmitBundleAlignMode(unsigned AlignPow2) override;
void EmitBundleLock(bool AlignToEnd) override;
@@ -1739,8 +1735,7 @@ void MCAsmStreamer::emitCGProfileEntry(c
}
void MCAsmStreamer::AddEncodingComment(const MCInst &Inst,
- const MCSubtargetInfo &STI,
- bool PrintSchedInfo) {
+ const MCSubtargetInfo &STI) {
raw_ostream &OS = GetCommentOS();
SmallString<256> Code;
SmallVector<MCFixup, 4> Fixups;
@@ -1819,11 +1814,7 @@ void MCAsmStreamer::AddEncodingComment(c
}
}
}
- OS << "]";
- // If we are not going to add fixup or schedule comments after this point
- // then we have to end the current comment line with "\n".
- if (Fixups.size() || !PrintSchedInfo)
- OS << "\n";
+ OS << "]\n";
for (unsigned i = 0, e = Fixups.size(); i != e; ++i) {
MCFixup &F = Fixups[i];
@@ -1835,18 +1826,15 @@ void MCAsmStreamer::AddEncodingComment(c
}
void MCAsmStreamer::EmitInstruction(const MCInst &Inst,
- const MCSubtargetInfo &STI,
- bool PrintSchedInfo) {
+ const MCSubtargetInfo &STI) {
assert(getCurrentSectionOnly() &&
"Cannot emit contents before setting section!");
// Show the encoding in a comment if we have a code emitter.
- AddEncodingComment(Inst, STI, PrintSchedInfo);
+ AddEncodingComment(Inst, STI);
// Show the MCInst if enabled.
if (ShowInst) {
- if (PrintSchedInfo)
- GetCommentOS() << "\n";
Inst.dump_pretty(GetCommentOS(), InstPrinter.get(), "\n ");
GetCommentOS() << "\n";
}
@@ -1856,12 +1844,6 @@ void MCAsmStreamer::EmitInstruction(cons
else
InstPrinter->printInst(&Inst, OS, "", STI);
- if (PrintSchedInfo) {
- std::string SI = STI.getSchedInfoStr(Inst);
- if (!SI.empty())
- GetCommentOS() << SI;
- }
-
StringRef Comments = CommentToEmit;
if (Comments.size() && Comments.back() != '\n')
GetCommentOS() << "\n";
Modified: llvm/trunk/lib/MC/MCObjectStreamer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCObjectStreamer.cpp?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCObjectStreamer.cpp (original)
+++ llvm/trunk/lib/MC/MCObjectStreamer.cpp Mon Feb 4 04:51:26 2019
@@ -314,7 +314,7 @@ bool MCObjectStreamer::mayHaveInstructio
}
void MCObjectStreamer::EmitInstruction(const MCInst &Inst,
- const MCSubtargetInfo &STI, bool) {
+ const MCSubtargetInfo &STI) {
getAssembler().getBackend().handleCodePaddingInstructionBegin(Inst);
EmitInstructionImpl(Inst, STI);
getAssembler().getBackend().handleCodePaddingInstructionEnd(Inst);
Modified: llvm/trunk/lib/MC/MCStreamer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCStreamer.cpp?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCStreamer.cpp (original)
+++ llvm/trunk/lib/MC/MCStreamer.cpp Mon Feb 4 04:51:26 2019
@@ -952,8 +952,7 @@ void MCStreamer::visitUsedExpr(const MCE
}
}
-void MCStreamer::EmitInstruction(const MCInst &Inst, const MCSubtargetInfo &STI,
- bool) {
+void MCStreamer::EmitInstruction(const MCInst &Inst, const MCSubtargetInfo &) {
// Scan for values.
for (unsigned i = Inst.getNumOperands(); i--;)
if (Inst.getOperand(i).isExpr())
Modified: llvm/trunk/lib/Object/RecordStreamer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Object/RecordStreamer.cpp?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/lib/Object/RecordStreamer.cpp (original)
+++ llvm/trunk/lib/Object/RecordStreamer.cpp Mon Feb 4 04:51:26 2019
@@ -82,7 +82,7 @@ RecordStreamer::const_iterator RecordStr
RecordStreamer::const_iterator RecordStreamer::end() { return Symbols.end(); }
void RecordStreamer::EmitInstruction(const MCInst &Inst,
- const MCSubtargetInfo &STI, bool) {
+ const MCSubtargetInfo &STI) {
MCStreamer::EmitInstruction(Inst, STI);
}
Modified: llvm/trunk/lib/Object/RecordStreamer.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Object/RecordStreamer.h?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/lib/Object/RecordStreamer.h (original)
+++ llvm/trunk/lib/Object/RecordStreamer.h Mon Feb 4 04:51:26 2019
@@ -46,8 +46,7 @@ private:
public:
RecordStreamer(MCContext &Context, const Module &M);
- void EmitInstruction(const MCInst &Inst, const MCSubtargetInfo &STI,
- bool) override;
+ void EmitInstruction(const MCInst &Inst, const MCSubtargetInfo &STI) override;
void EmitLabel(MCSymbol *Symbol, SMLoc Loc = SMLoc()) override;
void EmitAssignment(MCSymbol *Symbol, const MCExpr *Value) override;
bool EmitSymbolAttribute(MCSymbol *Symbol, MCSymbolAttr Attribute) override;
Modified: llvm/trunk/lib/Target/AArch64/MCTargetDesc/AArch64ELFStreamer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/AArch64/MCTargetDesc/AArch64ELFStreamer.cpp?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/lib/Target/AArch64/MCTargetDesc/AArch64ELFStreamer.cpp (original)
+++ llvm/trunk/lib/Target/AArch64/MCTargetDesc/AArch64ELFStreamer.cpp Mon Feb 4 04:51:26 2019
@@ -102,8 +102,8 @@ public:
/// This function is the one used to emit instruction data into the ELF
/// streamer. We override it to add the appropriate mapping symbol if
/// necessary.
- void EmitInstruction(const MCInst &Inst, const MCSubtargetInfo &STI,
- bool) override {
+ void EmitInstruction(const MCInst &Inst,
+ const MCSubtargetInfo &STI) override {
EmitA64MappingSymbol();
MCELFStreamer::EmitInstruction(Inst, STI);
}
Modified: llvm/trunk/lib/Target/ARM/MCTargetDesc/ARMELFStreamer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/MCTargetDesc/ARMELFStreamer.cpp?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/MCTargetDesc/ARMELFStreamer.cpp (original)
+++ llvm/trunk/lib/Target/ARM/MCTargetDesc/ARMELFStreamer.cpp Mon Feb 4 04:51:26 2019
@@ -484,8 +484,8 @@ public:
/// This function is the one used to emit instruction data into the ELF
/// streamer. We override it to add the appropriate mapping symbol if
/// necessary.
- void EmitInstruction(const MCInst &Inst, const MCSubtargetInfo &STI,
- bool) override {
+ void EmitInstruction(const MCInst &Inst,
+ const MCSubtargetInfo &STI) override {
if (IsThumb)
EmitThumbMappingSymbol();
else
Modified: llvm/trunk/lib/Target/Hexagon/MCTargetDesc/HexagonMCELFStreamer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/Hexagon/MCTargetDesc/HexagonMCELFStreamer.cpp?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/lib/Target/Hexagon/MCTargetDesc/HexagonMCELFStreamer.cpp (original)
+++ llvm/trunk/lib/Target/Hexagon/MCTargetDesc/HexagonMCELFStreamer.cpp Mon Feb 4 04:51:26 2019
@@ -59,7 +59,7 @@ HexagonMCELFStreamer::HexagonMCELFStream
MCII(createHexagonMCInstrInfo()) {}
void HexagonMCELFStreamer::EmitInstruction(const MCInst &MCB,
- const MCSubtargetInfo &STI, bool) {
+ const MCSubtargetInfo &STI) {
assert(MCB.getOpcode() == Hexagon::BUNDLE);
assert(HexagonMCInstrInfo::bundleSize(MCB) <= HEXAGON_PACKET_SIZE);
assert(HexagonMCInstrInfo::bundleSize(MCB) > 0);
Modified: llvm/trunk/lib/Target/Hexagon/MCTargetDesc/HexagonMCELFStreamer.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/Hexagon/MCTargetDesc/HexagonMCELFStreamer.h?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/lib/Target/Hexagon/MCTargetDesc/HexagonMCELFStreamer.h (original)
+++ llvm/trunk/lib/Target/Hexagon/MCTargetDesc/HexagonMCELFStreamer.h Mon Feb 4 04:51:26 2019
@@ -30,8 +30,7 @@ public:
std::unique_ptr<MCCodeEmitter> Emitter,
MCAssembler *Assembler);
- void EmitInstruction(const MCInst &Inst, const MCSubtargetInfo &STI,
- bool) override;
+ void EmitInstruction(const MCInst &Inst, const MCSubtargetInfo &STI) override;
void EmitSymbol(const MCInst &Inst);
void HexagonMCEmitLocalCommonSymbol(MCSymbol *Symbol, uint64_t Size,
unsigned ByteAlignment,
Modified: llvm/trunk/lib/Target/Mips/MCTargetDesc/MipsELFStreamer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/Mips/MCTargetDesc/MipsELFStreamer.cpp?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/lib/Target/Mips/MCTargetDesc/MipsELFStreamer.cpp (original)
+++ llvm/trunk/lib/Target/Mips/MCTargetDesc/MipsELFStreamer.cpp Mon Feb 4 04:51:26 2019
@@ -34,7 +34,7 @@ MipsELFStreamer::MipsELFStreamer(MCConte
}
void MipsELFStreamer::EmitInstruction(const MCInst &Inst,
- const MCSubtargetInfo &STI, bool) {
+ const MCSubtargetInfo &STI) {
MCELFStreamer::EmitInstruction(Inst, STI);
MCContext &Context = getContext();
Modified: llvm/trunk/lib/Target/Mips/MCTargetDesc/MipsELFStreamer.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/Mips/MCTargetDesc/MipsELFStreamer.h?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/lib/Target/Mips/MCTargetDesc/MipsELFStreamer.h (original)
+++ llvm/trunk/lib/Target/Mips/MCTargetDesc/MipsELFStreamer.h Mon Feb 4 04:51:26 2019
@@ -41,8 +41,7 @@ public:
/// \p Inst is actually emitted. For example, we can inspect the operands and
/// gather sufficient information that allows us to reason about the register
/// usage for the translation unit.
- void EmitInstruction(const MCInst &Inst, const MCSubtargetInfo &STI,
- bool = false) override;
+ void EmitInstruction(const MCInst &Inst, const MCSubtargetInfo &STI) override;
/// Overriding this function allows us to record all labels that should be
/// marked as microMIPS. Based on this data marking is done in
Modified: llvm/trunk/lib/Target/Mips/MCTargetDesc/MipsNaClELFStreamer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/Mips/MCTargetDesc/MipsNaClELFStreamer.cpp?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/lib/Target/Mips/MCTargetDesc/MipsNaClELFStreamer.cpp (original)
+++ llvm/trunk/lib/Target/Mips/MCTargetDesc/MipsNaClELFStreamer.cpp Mon Feb 4 04:51:26 2019
@@ -143,8 +143,8 @@ private:
public:
/// This function is the one used to emit instruction data into the ELF
/// streamer. We override it to mask dangerous instructions.
- void EmitInstruction(const MCInst &Inst, const MCSubtargetInfo &STI,
- bool) override {
+ void EmitInstruction(const MCInst &Inst,
+ const MCSubtargetInfo &STI) override {
// Sandbox indirect jumps.
if (isIndirectJump(Inst)) {
if (PendingCall)
Modified: llvm/trunk/lib/Target/X86/AsmParser/X86AsmInstrumentation.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/AsmParser/X86AsmInstrumentation.cpp?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/AsmParser/X86AsmInstrumentation.cpp (original)
+++ llvm/trunk/lib/Target/X86/AsmParser/X86AsmInstrumentation.cpp Mon Feb 4 04:51:26 2019
@@ -194,8 +194,7 @@ public:
// X86AsmInstrumentation implementation:
void InstrumentAndEmitInstruction(const MCInst &Inst, OperandVector &Operands,
MCContext &Ctx, const MCInstrInfo &MII,
- MCStreamer &Out,
- /* unused */ bool) override {
+ MCStreamer &Out) override {
InstrumentMOVS(Inst, Operands, Ctx, MII, Out);
if (RepPrefix)
EmitInstruction(Out, MCInstBuilder(X86::REP_PREFIX));
@@ -1043,13 +1042,13 @@ X86AsmInstrumentation::~X86AsmInstrument
void X86AsmInstrumentation::InstrumentAndEmitInstruction(
const MCInst &Inst, OperandVector &Operands, MCContext &Ctx,
- const MCInstrInfo &MII, MCStreamer &Out, bool PrintSchedInfoEnabled) {
- EmitInstruction(Out, Inst, PrintSchedInfoEnabled);
+ const MCInstrInfo &MII, MCStreamer &Out) {
+ EmitInstruction(Out, Inst);
}
-void X86AsmInstrumentation::EmitInstruction(MCStreamer &Out, const MCInst &Inst,
- bool PrintSchedInfoEnabled) {
- Out.EmitInstruction(Inst, *STI, PrintSchedInfoEnabled);
+void X86AsmInstrumentation::EmitInstruction(MCStreamer &Out,
+ const MCInst &Inst) {
+ Out.EmitInstruction(Inst, *STI);
}
unsigned X86AsmInstrumentation::GetFrameRegGeneric(const MCContext &Ctx,
Modified: llvm/trunk/lib/Target/X86/AsmParser/X86AsmInstrumentation.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/AsmParser/X86AsmInstrumentation.h?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/AsmParser/X86AsmInstrumentation.h (original)
+++ llvm/trunk/lib/Target/X86/AsmParser/X86AsmInstrumentation.h Mon Feb 4 04:51:26 2019
@@ -41,8 +41,7 @@ public:
virtual void InstrumentAndEmitInstruction(
const MCInst &Inst,
SmallVectorImpl<std::unique_ptr<MCParsedAsmOperand>> &Operands,
- MCContext &Ctx, const MCInstrInfo &MII, MCStreamer &Out,
- bool PrintSchedInfoEnabled);
+ MCContext &Ctx, const MCInstrInfo &MII, MCStreamer &Out);
protected:
friend X86AsmInstrumentation *
@@ -54,8 +53,7 @@ protected:
unsigned GetFrameRegGeneric(const MCContext &Ctx, MCStreamer &Out);
- void EmitInstruction(MCStreamer &Out, const MCInst &Inst,
- bool PrintSchedInfoEnabled = false);
+ void EmitInstruction(MCStreamer &Out, const MCInst &Inst);
const MCSubtargetInfo *&STI;
Modified: llvm/trunk/lib/Target/X86/AsmParser/X86AsmParser.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/AsmParser/X86AsmParser.cpp?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/AsmParser/X86AsmParser.cpp (original)
+++ llvm/trunk/lib/Target/X86/AsmParser/X86AsmParser.cpp Mon Feb 4 04:51:26 2019
@@ -2864,8 +2864,7 @@ static const char *getSubtargetFeatureNa
void X86AsmParser::EmitInstruction(MCInst &Inst, OperandVector &Operands,
MCStreamer &Out) {
Instrumentation->InstrumentAndEmitInstruction(
- Inst, Operands, getContext(), MII, Out,
- getParser().shouldPrintSchedInfo());
+ Inst, Operands, getContext(), MII, Out);
}
bool X86AsmParser::MatchAndEmitInstruction(SMLoc IDLoc, unsigned &Opcode,
Modified: llvm/trunk/lib/Target/X86/InstPrinter/X86InstComments.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/InstPrinter/X86InstComments.cpp?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/InstPrinter/X86InstComments.cpp (original)
+++ llvm/trunk/lib/Target/X86/InstPrinter/X86InstComments.cpp Mon Feb 4 04:51:26 2019
@@ -1303,6 +1303,7 @@ bool llvm::EmitAnyX86InstComments(const
OS << ']';
--i; // For loop increments element #.
}
+ OS << '\n';
// We successfully added a comment to this instruction.
return true;
Modified: llvm/trunk/lib/Target/X86/MCTargetDesc/X86BaseInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/MCTargetDesc/X86BaseInfo.h?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/MCTargetDesc/X86BaseInfo.h (original)
+++ llvm/trunk/lib/Target/X86/MCTargetDesc/X86BaseInfo.h Mon Feb 4 04:51:26 2019
@@ -59,9 +59,7 @@ namespace X86 {
IP_HAS_REPEAT_NE = 4,
IP_HAS_REPEAT = 8,
IP_HAS_LOCK = 16,
- NO_SCHED_INFO = 32, // Don't add sched comment to the current instr because
- // it was already added
- IP_HAS_NOTRACK = 64
+ IP_HAS_NOTRACK = 32
};
} // end namespace X86;
Modified: llvm/trunk/lib/Target/X86/X86MCInstLower.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86MCInstLower.cpp?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86MCInstLower.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86MCInstLower.cpp Mon Feb 4 04:51:26 2019
@@ -100,9 +100,7 @@ void X86AsmPrinter::StackMapShadowTracke
}
void X86AsmPrinter::EmitAndCountInstruction(MCInst &Inst) {
- OutStreamer->EmitInstruction(Inst, getSubtargetInfo(),
- EnablePrintSchedInfo &&
- !(Inst.getFlags() & X86::NO_SCHED_INFO));
+ OutStreamer->EmitInstruction(Inst, getSubtargetInfo());
SMShadowTracker.count(Inst, getSubtargetInfo(), CodeEmitter.get());
}
@@ -1860,8 +1858,7 @@ void X86AsmPrinter::EmitInstruction(cons
SmallVector<int, 64> Mask;
DecodePSHUFBMask(C, Width, Mask);
if (!Mask.empty())
- OutStreamer->AddComment(getShuffleComment(MI, SrcIdx, SrcIdx, Mask),
- !EnablePrintSchedInfo);
+ OutStreamer->AddComment(getShuffleComment(MI, SrcIdx, SrcIdx, Mask));
}
break;
}
@@ -1933,8 +1930,7 @@ void X86AsmPrinter::EmitInstruction(cons
SmallVector<int, 16> Mask;
DecodeVPERMILPMask(C, ElSize, Width, Mask);
if (!Mask.empty())
- OutStreamer->AddComment(getShuffleComment(MI, SrcIdx, SrcIdx, Mask),
- !EnablePrintSchedInfo);
+ OutStreamer->AddComment(getShuffleComment(MI, SrcIdx, SrcIdx, Mask));
}
break;
}
@@ -1965,8 +1961,7 @@ void X86AsmPrinter::EmitInstruction(cons
SmallVector<int, 16> Mask;
DecodeVPERMIL2PMask(C, (unsigned)CtrlOp.getImm(), ElSize, Width, Mask);
if (!Mask.empty())
- OutStreamer->AddComment(getShuffleComment(MI, 1, 2, Mask),
- !EnablePrintSchedInfo);
+ OutStreamer->AddComment(getShuffleComment(MI, 1, 2, Mask));
}
break;
}
@@ -1983,8 +1978,7 @@ void X86AsmPrinter::EmitInstruction(cons
SmallVector<int, 16> Mask;
DecodeVPPERMMask(C, Width, Mask);
if (!Mask.empty())
- OutStreamer->AddComment(getShuffleComment(MI, 1, 2, Mask),
- !EnablePrintSchedInfo);
+ OutStreamer->AddComment(getShuffleComment(MI, 1, 2, Mask));
}
break;
}
@@ -2001,7 +1995,7 @@ void X86AsmPrinter::EmitInstruction(cons
CS << X86ATTInstPrinter::getRegisterName(DstOp.getReg()) << " = ";
if (auto *CF = dyn_cast<ConstantFP>(C)) {
CS << "0x" << CF->getValueAPF().bitcastToAPInt().toString(16, false);
- OutStreamer->AddComment(CS.str(), !EnablePrintSchedInfo);
+ OutStreamer->AddComment(CS.str());
}
}
break;
@@ -2098,7 +2092,7 @@ void X86AsmPrinter::EmitInstruction(cons
}
}
CS << "]";
- OutStreamer->AddComment(CS.str(), !EnablePrintSchedInfo);
+ OutStreamer->AddComment(CS.str());
} else if (auto *CV = dyn_cast<ConstantVector>(C)) {
CS << "<";
for (int l = 0; l != NumLanes; ++l) {
@@ -2110,7 +2104,7 @@ void X86AsmPrinter::EmitInstruction(cons
}
}
CS << ">";
- OutStreamer->AddComment(CS.str(), !EnablePrintSchedInfo);
+ OutStreamer->AddComment(CS.str());
}
}
break;
@@ -2197,14 +2191,12 @@ void X86AsmPrinter::EmitInstruction(cons
printConstant(C, CS);
}
CS << "]";
- OutStreamer->AddComment(CS.str(), !EnablePrintSchedInfo);
+ OutStreamer->AddComment(CS.str());
}
}
MCInst TmpInst;
MCInstLowering.Lower(MI, TmpInst);
- if (MI->getAsmPrinterFlag(MachineInstr::NoSchedComment))
- TmpInst.setFlags(TmpInst.getFlags() | X86::NO_SCHED_INFO);
// Stackmap shadows cannot include branch targets, so we can count the bytes
// in a call towards the shadow, but must ensure that the no thread returns
Modified: llvm/trunk/lib/Target/X86/X86Subtarget.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86Subtarget.h?rev=353043&r1=353042&r2=353043&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86Subtarget.h (original)
+++ llvm/trunk/lib/Target/X86/X86Subtarget.h Mon Feb 4 04:51:26 2019
@@ -833,9 +833,6 @@ public:
/// Enable the MachineScheduler pass for all X86 subtargets.
bool enableMachineScheduler() const override { return true; }
- // TODO: Update the regression tests and return true.
- bool supportPrintSchedInfo() const override { return false; }
-
bool enableEarlyIfConversion() const override;
AntiDepBreakMode getAntiDepBreakMode() const override {
Removed: llvm/trunk/test/CodeGen/X86/3dnow-schedule.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/3dnow-schedule.ll?rev=353042&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/3dnow-schedule.ll (original)
+++ llvm/trunk/test/CodeGen/X86/3dnow-schedule.ll (removed)
@@ -1,394 +0,0 @@
-; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=x86-64 -mattr=+3dnowa | FileCheck %s --check-prefix=CHECK --check-prefix=GENERIC
-
-define void @test_femms() optsize {
-; CHECK-LABEL: test_femms:
-; CHECK: # %bb.0:
-; CHECK-NEXT: femms # sched: [31:10.33]
-; CHECK-NEXT: retq # sched: [1:1.00]
- call void @llvm.x86.mmx.femms()
- ret void
-}
-declare void @llvm.x86.mmx.femms() nounwind readnone
-
-define i64 @test_pavgusb(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
-; CHECK-LABEL: test_pavgusb:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pavgusb %mm1, %mm0 # sched: [3:1.00]
-; CHECK-NEXT: pavgusb (%rdi), %mm0 # sched: [8:1.00]
-; CHECK-NEXT: movq %mm0, %rax # sched: [2:1.00]
-; CHECK-NEXT: retq # sched: [1:1.00]
- %1 = call x86_mmx @llvm.x86.3dnow.pavgusb(x86_mmx %a0, x86_mmx %a1)
- %2 = load x86_mmx, x86_mmx *%a2, align 8
- %3 = call x86_mmx @llvm.x86.3dnow.pavgusb(x86_mmx %1, x86_mmx %2)
- %4 = bitcast x86_mmx %3 to i64
- ret i64 %4
-}
-declare x86_mmx @llvm.x86.3dnow.pavgusb(x86_mmx, x86_mmx) nounwind readnone
-
-define i64 @test_pf2id(x86_mmx* %a0) optsize {
-; CHECK-LABEL: test_pf2id:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pf2id (%rdi), %mm0 # sched: [9:1.00]
-; CHECK-NEXT: pf2id %mm0, %mm0 # sched: [3:1.00]
-; CHECK-NEXT: movq %mm0, %rax # sched: [2:1.00]
-; CHECK-NEXT: retq # sched: [1:1.00]
- %1 = load x86_mmx, x86_mmx *%a0, align 8
- %2 = call x86_mmx @llvm.x86.3dnow.pf2id(x86_mmx %1)
- %3 = call x86_mmx @llvm.x86.3dnow.pf2id(x86_mmx %2)
- %4 = bitcast x86_mmx %3 to i64
- ret i64 %4
-}
-declare x86_mmx @llvm.x86.3dnow.pf2id(x86_mmx) nounwind readnone
-
-define i64 @test_pf2iw(x86_mmx* %a0) optsize {
-; CHECK-LABEL: test_pf2iw:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pf2iw (%rdi), %mm0 # sched: [9:1.00]
-; CHECK-NEXT: pf2iw %mm0, %mm0 # sched: [3:1.00]
-; CHECK-NEXT: movq %mm0, %rax # sched: [2:1.00]
-; CHECK-NEXT: retq # sched: [1:1.00]
- %1 = load x86_mmx, x86_mmx *%a0, align 8
- %2 = call x86_mmx @llvm.x86.3dnowa.pf2iw(x86_mmx %1)
- %3 = call x86_mmx @llvm.x86.3dnowa.pf2iw(x86_mmx %2)
- %4 = bitcast x86_mmx %3 to i64
- ret i64 %4
-}
-declare x86_mmx @llvm.x86.3dnowa.pf2iw(x86_mmx) nounwind readnone
-
-define i64 @test_pfacc(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
-; CHECK-LABEL: test_pfacc:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pfacc %mm1, %mm0 # sched: [3:1.00]
-; CHECK-NEXT: pfacc (%rdi), %mm0 # sched: [9:1.00]
-; CHECK-NEXT: movq %mm0, %rax # sched: [2:1.00]
-; CHECK-NEXT: retq # sched: [1:1.00]
- %1 = call x86_mmx @llvm.x86.3dnow.pfacc(x86_mmx %a0, x86_mmx %a1)
- %2 = load x86_mmx, x86_mmx *%a2, align 8
- %3 = call x86_mmx @llvm.x86.3dnow.pfacc(x86_mmx %1, x86_mmx %2)
- %4 = bitcast x86_mmx %3 to i64
- ret i64 %4
-}
-declare x86_mmx @llvm.x86.3dnow.pfacc(x86_mmx, x86_mmx) nounwind readnone
-
-define i64 @test_pfadd(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
-; CHECK-LABEL: test_pfadd:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pfadd %mm1, %mm0 # sched: [3:1.00]
-; CHECK-NEXT: pfadd (%rdi), %mm0 # sched: [9:1.00]
-; CHECK-NEXT: movq %mm0, %rax # sched: [2:1.00]
-; CHECK-NEXT: retq # sched: [1:1.00]
- %1 = call x86_mmx @llvm.x86.3dnow.pfadd(x86_mmx %a0, x86_mmx %a1)
- %2 = load x86_mmx, x86_mmx *%a2, align 8
- %3 = call x86_mmx @llvm.x86.3dnow.pfadd(x86_mmx %1, x86_mmx %2)
- %4 = bitcast x86_mmx %3 to i64
- ret i64 %4
-}
-declare x86_mmx @llvm.x86.3dnow.pfadd(x86_mmx, x86_mmx) nounwind readnone
-
-define i64 @test_pfcmpeq(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
-; CHECK-LABEL: test_pfcmpeq:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pfcmpeq %mm1, %mm0 # sched: [3:1.00]
-; CHECK-NEXT: pfcmpeq (%rdi), %mm0 # sched: [9:1.00]
-; CHECK-NEXT: movq %mm0, %rax # sched: [2:1.00]
-; CHECK-NEXT: retq # sched: [1:1.00]
- %1 = call x86_mmx @llvm.x86.3dnow.pfcmpeq(x86_mmx %a0, x86_mmx %a1)
- %2 = load x86_mmx, x86_mmx *%a2, align 8
- %3 = call x86_mmx @llvm.x86.3dnow.pfcmpeq(x86_mmx %1, x86_mmx %2)
- %4 = bitcast x86_mmx %3 to i64
- ret i64 %4
-}
-declare x86_mmx @llvm.x86.3dnow.pfcmpeq(x86_mmx, x86_mmx) nounwind readnone
-
-define i64 @test_pfcmpge(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
-; CHECK-LABEL: test_pfcmpge:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pfcmpge %mm1, %mm0 # sched: [3:1.00]
-; CHECK-NEXT: pfcmpge (%rdi), %mm0 # sched: [9:1.00]
-; CHECK-NEXT: movq %mm0, %rax # sched: [2:1.00]
-; CHECK-NEXT: retq # sched: [1:1.00]
- %1 = call x86_mmx @llvm.x86.3dnow.pfcmpge(x86_mmx %a0, x86_mmx %a1)
- %2 = load x86_mmx, x86_mmx *%a2, align 8
- %3 = call x86_mmx @llvm.x86.3dnow.pfcmpge(x86_mmx %1, x86_mmx %2)
- %4 = bitcast x86_mmx %3 to i64
- ret i64 %4
-}
-declare x86_mmx @llvm.x86.3dnow.pfcmpge(x86_mmx, x86_mmx) nounwind readnone
-
-define i64 @test_pfcmpgt(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
-; CHECK-LABEL: test_pfcmpgt:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pfcmpgt %mm1, %mm0 # sched: [3:1.00]
-; CHECK-NEXT: pfcmpgt (%rdi), %mm0 # sched: [9:1.00]
-; CHECK-NEXT: movq %mm0, %rax # sched: [2:1.00]
-; CHECK-NEXT: retq # sched: [1:1.00]
- %1 = call x86_mmx @llvm.x86.3dnow.pfcmpgt(x86_mmx %a0, x86_mmx %a1)
- %2 = load x86_mmx, x86_mmx *%a2, align 8
- %3 = call x86_mmx @llvm.x86.3dnow.pfcmpgt(x86_mmx %1, x86_mmx %2)
- %4 = bitcast x86_mmx %3 to i64
- ret i64 %4
-}
-declare x86_mmx @llvm.x86.3dnow.pfcmpgt(x86_mmx, x86_mmx) nounwind readnone
-
-define i64 @test_pfmax(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
-; CHECK-LABEL: test_pfmax:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pfmax %mm1, %mm0 # sched: [3:1.00]
-; CHECK-NEXT: pfmax (%rdi), %mm0 # sched: [9:1.00]
-; CHECK-NEXT: movq %mm0, %rax # sched: [2:1.00]
-; CHECK-NEXT: retq # sched: [1:1.00]
- %1 = call x86_mmx @llvm.x86.3dnow.pfmax(x86_mmx %a0, x86_mmx %a1)
- %2 = load x86_mmx, x86_mmx *%a2, align 8
- %3 = call x86_mmx @llvm.x86.3dnow.pfmax(x86_mmx %1, x86_mmx %2)
- %4 = bitcast x86_mmx %3 to i64
- ret i64 %4
-}
-declare x86_mmx @llvm.x86.3dnow.pfmax(x86_mmx, x86_mmx) nounwind readnone
-
-define i64 @test_pfmin(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
-; CHECK-LABEL: test_pfmin:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pfmin %mm1, %mm0 # sched: [3:1.00]
-; CHECK-NEXT: pfmin (%rdi), %mm0 # sched: [9:1.00]
-; CHECK-NEXT: movq %mm0, %rax # sched: [2:1.00]
-; CHECK-NEXT: retq # sched: [1:1.00]
- %1 = call x86_mmx @llvm.x86.3dnow.pfmin(x86_mmx %a0, x86_mmx %a1)
- %2 = load x86_mmx, x86_mmx *%a2, align 8
- %3 = call x86_mmx @llvm.x86.3dnow.pfmin(x86_mmx %1, x86_mmx %2)
- %4 = bitcast x86_mmx %3 to i64
- ret i64 %4
-}
-declare x86_mmx @llvm.x86.3dnow.pfmin(x86_mmx, x86_mmx) nounwind readnone
-
-define i64 @test_pfmul(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
-; CHECK-LABEL: test_pfmul:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pfmul %mm1, %mm0 # sched: [3:1.00]
-; CHECK-NEXT: pfmul (%rdi), %mm0 # sched: [9:1.00]
-; CHECK-NEXT: movq %mm0, %rax # sched: [2:1.00]
-; CHECK-NEXT: retq # sched: [1:1.00]
- %1 = call x86_mmx @llvm.x86.3dnow.pfmul(x86_mmx %a0, x86_mmx %a1)
- %2 = load x86_mmx, x86_mmx *%a2, align 8
- %3 = call x86_mmx @llvm.x86.3dnow.pfmul(x86_mmx %1, x86_mmx %2)
- %4 = bitcast x86_mmx %3 to i64
- ret i64 %4
-}
-declare x86_mmx @llvm.x86.3dnow.pfmul(x86_mmx, x86_mmx) nounwind readnone
-
-define i64 @test_pfnacc(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
-; CHECK-LABEL: test_pfnacc:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pfnacc %mm1, %mm0 # sched: [3:1.00]
-; CHECK-NEXT: pfnacc (%rdi), %mm0 # sched: [9:1.00]
-; CHECK-NEXT: movq %mm0, %rax # sched: [2:1.00]
-; CHECK-NEXT: retq # sched: [1:1.00]
- %1 = call x86_mmx @llvm.x86.3dnowa.pfnacc(x86_mmx %a0, x86_mmx %a1)
- %2 = load x86_mmx, x86_mmx *%a2, align 8
- %3 = call x86_mmx @llvm.x86.3dnowa.pfnacc(x86_mmx %1, x86_mmx %2)
- %4 = bitcast x86_mmx %3 to i64
- ret i64 %4
-}
-declare x86_mmx @llvm.x86.3dnowa.pfnacc(x86_mmx, x86_mmx) nounwind readnone
-
-define i64 @test_pfpnacc(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
-; CHECK-LABEL: test_pfpnacc:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pfpnacc %mm1, %mm0 # sched: [3:1.00]
-; CHECK-NEXT: pfpnacc (%rdi), %mm0 # sched: [9:1.00]
-; CHECK-NEXT: movq %mm0, %rax # sched: [2:1.00]
-; CHECK-NEXT: retq # sched: [1:1.00]
- %1 = call x86_mmx @llvm.x86.3dnowa.pfpnacc(x86_mmx %a0, x86_mmx %a1)
- %2 = load x86_mmx, x86_mmx *%a2, align 8
- %3 = call x86_mmx @llvm.x86.3dnowa.pfpnacc(x86_mmx %1, x86_mmx %2)
- %4 = bitcast x86_mmx %3 to i64
- ret i64 %4
-}
-declare x86_mmx @llvm.x86.3dnowa.pfpnacc(x86_mmx, x86_mmx) nounwind readnone
-
-define i64 @test_pfrcp(x86_mmx* %a0) optsize {
-; CHECK-LABEL: test_pfrcp:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pfrcp (%rdi), %mm0 # sched: [9:1.00]
-; CHECK-NEXT: pfrcp %mm0, %mm0 # sched: [3:1.00]
-; CHECK-NEXT: movq %mm0, %rax # sched: [2:1.00]
-; CHECK-NEXT: retq # sched: [1:1.00]
- %1 = load x86_mmx, x86_mmx *%a0, align 8
- %2 = call x86_mmx @llvm.x86.3dnow.pfrcp(x86_mmx %1)
- %3 = call x86_mmx @llvm.x86.3dnow.pfrcp(x86_mmx %2)
- %4 = bitcast x86_mmx %3 to i64
- ret i64 %4
-}
-declare x86_mmx @llvm.x86.3dnow.pfrcp(x86_mmx) nounwind readnone
-
-define i64 @test_pfrcpit1(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
-; CHECK-LABEL: test_pfrcpit1:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pfrcpit1 %mm1, %mm0 # sched: [3:1.00]
-; CHECK-NEXT: pfrcpit1 (%rdi), %mm0 # sched: [9:1.00]
-; CHECK-NEXT: movq %mm0, %rax # sched: [2:1.00]
-; CHECK-NEXT: retq # sched: [1:1.00]
- %1 = call x86_mmx @llvm.x86.3dnow.pfrcpit1(x86_mmx %a0, x86_mmx %a1)
- %2 = load x86_mmx, x86_mmx *%a2, align 8
- %3 = call x86_mmx @llvm.x86.3dnow.pfrcpit1(x86_mmx %1, x86_mmx %2)
- %4 = bitcast x86_mmx %3 to i64
- ret i64 %4
-}
-declare x86_mmx @llvm.x86.3dnow.pfrcpit1(x86_mmx, x86_mmx) nounwind readnone
-
-define i64 @test_pfrcpit2(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
-; CHECK-LABEL: test_pfrcpit2:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pfrcpit2 %mm1, %mm0 # sched: [3:1.00]
-; CHECK-NEXT: pfrcpit2 (%rdi), %mm0 # sched: [9:1.00]
-; CHECK-NEXT: movq %mm0, %rax # sched: [2:1.00]
-; CHECK-NEXT: retq # sched: [1:1.00]
- %1 = call x86_mmx @llvm.x86.3dnow.pfrcpit2(x86_mmx %a0, x86_mmx %a1)
- %2 = load x86_mmx, x86_mmx *%a2, align 8
- %3 = call x86_mmx @llvm.x86.3dnow.pfrcpit2(x86_mmx %1, x86_mmx %2)
- %4 = bitcast x86_mmx %3 to i64
- ret i64 %4
-}
-declare x86_mmx @llvm.x86.3dnow.pfrcpit2(x86_mmx, x86_mmx) nounwind readnone
-
-define i64 @test_pfrsqit1(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
-; CHECK-LABEL: test_pfrsqit1:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pfrsqit1 %mm1, %mm0 # sched: [3:1.00]
-; CHECK-NEXT: pfrsqit1 (%rdi), %mm0 # sched: [9:1.00]
-; CHECK-NEXT: movq %mm0, %rax # sched: [2:1.00]
-; CHECK-NEXT: retq # sched: [1:1.00]
- %1 = call x86_mmx @llvm.x86.3dnow.pfrsqit1(x86_mmx %a0, x86_mmx %a1)
- %2 = load x86_mmx, x86_mmx *%a2, align 8
- %3 = call x86_mmx @llvm.x86.3dnow.pfrsqit1(x86_mmx %1, x86_mmx %2)
- %4 = bitcast x86_mmx %3 to i64
- ret i64 %4
-}
-declare x86_mmx @llvm.x86.3dnow.pfrsqit1(x86_mmx, x86_mmx) nounwind readnone
-
-define i64 @test_pfrsqrt(x86_mmx* %a0) optsize {
-; CHECK-LABEL: test_pfrsqrt:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pfrsqrt (%rdi), %mm0 # sched: [9:1.00]
-; CHECK-NEXT: pfrsqrt %mm0, %mm0 # sched: [3:1.00]
-; CHECK-NEXT: movq %mm0, %rax # sched: [2:1.00]
-; CHECK-NEXT: retq # sched: [1:1.00]
- %1 = load x86_mmx, x86_mmx *%a0, align 8
- %2 = call x86_mmx @llvm.x86.3dnow.pfrsqrt(x86_mmx %1)
- %3 = call x86_mmx @llvm.x86.3dnow.pfrsqrt(x86_mmx %2)
- %4 = bitcast x86_mmx %3 to i64
- ret i64 %4
-}
-declare x86_mmx @llvm.x86.3dnow.pfrsqrt(x86_mmx) nounwind readnone
-
-define i64 @test_pfsub(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
-; CHECK-LABEL: test_pfsub:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pfsub %mm1, %mm0 # sched: [3:1.00]
-; CHECK-NEXT: pfsub (%rdi), %mm0 # sched: [9:1.00]
-; CHECK-NEXT: movq %mm0, %rax # sched: [2:1.00]
-; CHECK-NEXT: retq # sched: [1:1.00]
- %1 = call x86_mmx @llvm.x86.3dnow.pfsub(x86_mmx %a0, x86_mmx %a1)
- %2 = load x86_mmx, x86_mmx *%a2, align 8
- %3 = call x86_mmx @llvm.x86.3dnow.pfsub(x86_mmx %1, x86_mmx %2)
- %4 = bitcast x86_mmx %3 to i64
- ret i64 %4
-}
-declare x86_mmx @llvm.x86.3dnow.pfsub(x86_mmx, x86_mmx) nounwind readnone
-
-define i64 @test_pfsubr(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
-; CHECK-LABEL: test_pfsubr:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pfsubr %mm1, %mm0 # sched: [3:1.00]
-; CHECK-NEXT: pfsubr (%rdi), %mm0 # sched: [9:1.00]
-; CHECK-NEXT: movq %mm0, %rax # sched: [2:1.00]
-; CHECK-NEXT: retq # sched: [1:1.00]
- %1 = call x86_mmx @llvm.x86.3dnow.pfsubr(x86_mmx %a0, x86_mmx %a1)
- %2 = load x86_mmx, x86_mmx *%a2, align 8
- %3 = call x86_mmx @llvm.x86.3dnow.pfsubr(x86_mmx %1, x86_mmx %2)
- %4 = bitcast x86_mmx %3 to i64
- ret i64 %4
-}
-declare x86_mmx @llvm.x86.3dnow.pfsubr(x86_mmx, x86_mmx) nounwind readnone
-
-define i64 @test_pi2fd(x86_mmx* %a0) optsize {
-; CHECK-LABEL: test_pi2fd:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pi2fd (%rdi), %mm0 # sched: [9:1.00]
-; CHECK-NEXT: pi2fd %mm0, %mm0 # sched: [3:1.00]
-; CHECK-NEXT: movq %mm0, %rax # sched: [2:1.00]
-; CHECK-NEXT: retq # sched: [1:1.00]
- %1 = load x86_mmx, x86_mmx *%a0, align 8
- %2 = call x86_mmx @llvm.x86.3dnow.pi2fd(x86_mmx %1)
- %3 = call x86_mmx @llvm.x86.3dnow.pi2fd(x86_mmx %2)
- %4 = bitcast x86_mmx %3 to i64
- ret i64 %4
-}
-declare x86_mmx @llvm.x86.3dnow.pi2fd(x86_mmx) nounwind readnone
-
-define i64 @test_pi2fw(x86_mmx* %a0) optsize {
-; CHECK-LABEL: test_pi2fw:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pi2fw (%rdi), %mm0 # sched: [9:1.00]
-; CHECK-NEXT: pi2fw %mm0, %mm0 # sched: [3:1.00]
-; CHECK-NEXT: movq %mm0, %rax # sched: [2:1.00]
-; CHECK-NEXT: retq # sched: [1:1.00]
- %1 = load x86_mmx, x86_mmx *%a0, align 8
- %2 = call x86_mmx @llvm.x86.3dnowa.pi2fw(x86_mmx %1)
- %3 = call x86_mmx @llvm.x86.3dnowa.pi2fw(x86_mmx %2)
- %4 = bitcast x86_mmx %3 to i64
- ret i64 %4
-}
-declare x86_mmx @llvm.x86.3dnowa.pi2fw(x86_mmx) nounwind readnone
-
-define i64 @test_pmulhrw(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
-; CHECK-LABEL: test_pmulhrw:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pmulhrw %mm1, %mm0 # sched: [5:1.00]
-; CHECK-NEXT: pmulhrw (%rdi), %mm0 # sched: [10:1.00]
-; CHECK-NEXT: movq %mm0, %rax # sched: [2:1.00]
-; CHECK-NEXT: retq # sched: [1:1.00]
- %1 = call x86_mmx @llvm.x86.3dnow.pmulhrw(x86_mmx %a0, x86_mmx %a1)
- %2 = load x86_mmx, x86_mmx *%a2, align 8
- %3 = call x86_mmx @llvm.x86.3dnow.pmulhrw(x86_mmx %1, x86_mmx %2)
- %4 = bitcast x86_mmx %3 to i64
- ret i64 %4
-}
-declare x86_mmx @llvm.x86.3dnow.pmulhrw(x86_mmx, x86_mmx) nounwind readnone
-
-define void @test_prefetch(i8* %a0) optsize {
-; CHECK-LABEL: test_prefetch:
-; CHECK: # %bb.0:
-; CHECK-NEXT: #APP
-; CHECK-NEXT: prefetch (%rdi) # sched: [5:0.50]
-; CHECK-NEXT: #NO_APP
-; CHECK-NEXT: retq # sched: [1:1.00]
- tail call void asm sideeffect "prefetch $0", "*m"(i8 *%a0) nounwind
- ret void
-}
-
-define void @test_prefetchw(i8* %a0) optsize {
-; CHECK-LABEL: test_prefetchw:
-; CHECK: # %bb.0:
-; CHECK-NEXT: #APP
-; CHECK-NEXT: prefetchw (%rdi) # sched: [5:0.50]
-; CHECK-NEXT: #NO_APP
-; CHECK-NEXT: retq # sched: [1:1.00]
- tail call void asm sideeffect "prefetchw $0", "*m"(i8 *%a0) nounwind
- ret void
-}
-
-define i64 @test_pswapd(x86_mmx* %a0) optsize {
-; CHECK-LABEL: test_pswapd:
-; CHECK: # %bb.0:
-; CHECK-NEXT: pswapd (%rdi), %mm0 # mm0 = mem[1,0] sched: [6:1.00]
-; CHECK-NEXT: pswapd %mm0, %mm0 # mm0 = mm0[1,0] sched: [1:1.00]
-; CHECK-NEXT: movq %mm0, %rax # sched: [2:1.00]
-; CHECK-NEXT: retq # sched: [1:1.00]
- %1 = load x86_mmx, x86_mmx *%a0, align 8
- %2 = call x86_mmx @llvm.x86.3dnowa.pswapd(x86_mmx %1)
- %3 = call x86_mmx @llvm.x86.3dnowa.pswapd(x86_mmx %2)
- %4 = bitcast x86_mmx %3 to i64
- ret i64 %4
-}
-declare x86_mmx @llvm.x86.3dnowa.pswapd(x86_mmx) nounwind readnone
Removed: llvm/trunk/test/CodeGen/X86/adx-schedule.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/adx-schedule.ll?rev=353042&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/adx-schedule.ll (original)
+++ llvm/trunk/test/CodeGen/X86/adx-schedule.ll (removed)
@@ -1,114 +0,0 @@
-; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=x86-64 -mattr=+adx | FileCheck %s --check-prefix=CHECK --check-prefix=GENERIC
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=broadwell | FileCheck %s --check-prefix=CHECK --check-prefix=BROADWELL
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=skylake | FileCheck %s --check-prefix=CHECK --check-prefix=SKYLAKE --check-prefix=SKL
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=skx | FileCheck %s --check-prefix=CHECK --check-prefix=SKYLAKE --check-prefix=SKX
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=knl | FileCheck %s --check-prefix=CHECK --check-prefix=KNL
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=znver1 | FileCheck %s --check-prefix=CHECK --check-prefix=ZNVER1
-
-define void @test_adcx(i32 %a0, i32* %a1, i64 %a2, i64* %a3) optsize {
-; GENERIC-LABEL: test_adcx:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: #APP
-; GENERIC-NEXT: adcxl %edi, %edi # sched: [2:0.67]
-; GENERIC-NEXT: adcxq %rdx, %rdx # sched: [2:0.67]
-; GENERIC-NEXT: adcxl (%rsi), %edi # sched: [7:0.67]
-; GENERIC-NEXT: adcxq (%rcx), %rdx # sched: [7:0.67]
-; GENERIC-NEXT: #NO_APP
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; BROADWELL-LABEL: test_adcx:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: #APP
-; BROADWELL-NEXT: adcxl %edi, %edi # sched: [1:0.50]
-; BROADWELL-NEXT: adcxq %rdx, %rdx # sched: [1:0.50]
-; BROADWELL-NEXT: adcxl (%rsi), %edi # sched: [6:0.50]
-; BROADWELL-NEXT: adcxq (%rcx), %rdx # sched: [6:0.50]
-; BROADWELL-NEXT: #NO_APP
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_adcx:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: #APP
-; SKYLAKE-NEXT: adcxl %edi, %edi # sched: [1:0.50]
-; SKYLAKE-NEXT: adcxq %rdx, %rdx # sched: [1:0.50]
-; SKYLAKE-NEXT: adcxl (%rsi), %edi # sched: [6:0.50]
-; SKYLAKE-NEXT: adcxq (%rcx), %rdx # sched: [6:0.50]
-; SKYLAKE-NEXT: #NO_APP
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; KNL-LABEL: test_adcx:
-; KNL: # %bb.0:
-; KNL-NEXT: #APP
-; KNL-NEXT: adcxl %edi, %edi # sched: [2:0.50]
-; KNL-NEXT: adcxq %rdx, %rdx # sched: [2:0.50]
-; KNL-NEXT: adcxl (%rsi), %edi # sched: [7:0.50]
-; KNL-NEXT: adcxq (%rcx), %rdx # sched: [7:0.50]
-; KNL-NEXT: #NO_APP
-; KNL-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_adcx:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: #APP
-; ZNVER1-NEXT: adcxl %edi, %edi # sched: [1:0.25]
-; ZNVER1-NEXT: adcxq %rdx, %rdx # sched: [1:0.25]
-; ZNVER1-NEXT: adcxl (%rsi), %edi # sched: [5:0.50]
-; ZNVER1-NEXT: adcxq (%rcx), %rdx # sched: [5:0.50]
-; ZNVER1-NEXT: #NO_APP
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- tail call void asm "adcx $0, $0 \0A\09 adcx $2, $2 \0A\09 adcx $1, $0 \0A\09 adcx $3, $2", "r,*m,r,*m"(i32 %a0, i32* %a1, i64 %a2, i64* %a3) nounwind
- ret void
-}
-define void @test_adox(i32 %a0, i32* %a1, i64 %a2, i64* %a3) optsize {
-; GENERIC-LABEL: test_adox:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: #APP
-; GENERIC-NEXT: adoxl %edi, %edi # sched: [2:0.67]
-; GENERIC-NEXT: adoxq %rdx, %rdx # sched: [2:0.67]
-; GENERIC-NEXT: adoxl (%rsi), %edi # sched: [7:0.67]
-; GENERIC-NEXT: adoxq (%rcx), %rdx # sched: [7:0.67]
-; GENERIC-NEXT: #NO_APP
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; BROADWELL-LABEL: test_adox:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: #APP
-; BROADWELL-NEXT: adoxl %edi, %edi # sched: [1:0.50]
-; BROADWELL-NEXT: adoxq %rdx, %rdx # sched: [1:0.50]
-; BROADWELL-NEXT: adoxl (%rsi), %edi # sched: [6:0.50]
-; BROADWELL-NEXT: adoxq (%rcx), %rdx # sched: [6:0.50]
-; BROADWELL-NEXT: #NO_APP
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_adox:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: #APP
-; SKYLAKE-NEXT: adoxl %edi, %edi # sched: [1:0.50]
-; SKYLAKE-NEXT: adoxq %rdx, %rdx # sched: [1:0.50]
-; SKYLAKE-NEXT: adoxl (%rsi), %edi # sched: [6:0.50]
-; SKYLAKE-NEXT: adoxq (%rcx), %rdx # sched: [6:0.50]
-; SKYLAKE-NEXT: #NO_APP
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; KNL-LABEL: test_adox:
-; KNL: # %bb.0:
-; KNL-NEXT: #APP
-; KNL-NEXT: adoxl %edi, %edi # sched: [2:0.50]
-; KNL-NEXT: adoxq %rdx, %rdx # sched: [2:0.50]
-; KNL-NEXT: adoxl (%rsi), %edi # sched: [7:0.50]
-; KNL-NEXT: adoxq (%rcx), %rdx # sched: [7:0.50]
-; KNL-NEXT: #NO_APP
-; KNL-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_adox:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: #APP
-; ZNVER1-NEXT: adoxl %edi, %edi # sched: [1:0.25]
-; ZNVER1-NEXT: adoxq %rdx, %rdx # sched: [1:0.25]
-; ZNVER1-NEXT: adoxl (%rsi), %edi # sched: [5:0.50]
-; ZNVER1-NEXT: adoxq (%rcx), %rdx # sched: [5:0.50]
-; ZNVER1-NEXT: #NO_APP
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- tail call void asm "adox $0, $0 \0A\09 adox $2, $2 \0A\09 adox $1, $0 \0A\09 adox $3, $2", "r,*m,r,*m"(i32 %a0, i32* %a1, i64 %a2, i64* %a3) nounwind
- ret void
-}
Removed: llvm/trunk/test/CodeGen/X86/aes-schedule.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/aes-schedule.ll?rev=353042&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/aes-schedule.ll (original)
+++ llvm/trunk/test/CodeGen/X86/aes-schedule.ll (removed)
@@ -1,751 +0,0 @@
-; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=x86-64 -mattr=+aes | FileCheck %s --check-prefix=CHECK --check-prefix=GENERIC
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=slm -mattr=+aes | FileCheck %s --check-prefix=CHECK --check-prefix=SLM
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=goldmont -mattr=+aes | FileCheck %s --check-prefix=CHECK --check-prefix=SLM
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=sandybridge -mattr=+aes,-avx | FileCheck %s --check-prefixes=CHECK,SANDY-SSE
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=sandybridge -mattr=+aes,-avx2 | FileCheck %s --check-prefixes=CHECK,SANDY
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=ivybridge -mattr=+aes,-avx | FileCheck %s --check-prefixes=CHECK,SANDY-SSE
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=ivybridge -mattr=+aes,-avx2 | FileCheck %s --check-prefixes=CHECK,SANDY
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=haswell -mattr=+aes,-avx | FileCheck %s --check-prefixes=CHECK,HASWELL-SSE
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=haswell -mattr=+aes,-avx2 | FileCheck %s --check-prefixes=CHECK,HASWELL
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=broadwell -mattr=+aes,-avx | FileCheck %s --check-prefixes=CHECK,BROADWELL-SSE
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=broadwell -mattr=+aes,-avx2 | FileCheck %s --check-prefixes=CHECK,BROADWELL
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=skylake -mattr=+aes,-avx | FileCheck %s --check-prefixes=CHECK,SKYLAKE-SSE
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=skylake -mattr=+aes,-avx2 | FileCheck %s --check-prefixes=CHECK,SKYLAKE
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=skx -mattr=+aes,-avx | FileCheck %s --check-prefixes=CHECK,SKX-SSE
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=skx -mattr=+aes,-avx2 | FileCheck %s --check-prefixes=CHECK,SKX
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=bdver2 -mattr=+aes,-avx | FileCheck %s --check-prefixes=CHECK,BDVER2-SSE
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=bdver2 -mattr=+aes,-avx2 | FileCheck %s --check-prefixes=CHECK,BDVER2
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=btver2 -mattr=+aes,-avx | FileCheck %s --check-prefixes=CHECK,BTVER2-SSE
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=btver2 -mattr=+aes,-avx2 | FileCheck %s --check-prefixes=CHECK,BTVER2
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=znver1 -mattr=+aes,-avx | FileCheck %s --check-prefixes=CHECK,ZNVER1-SSE
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=znver1 -mattr=+aes,-avx2 | FileCheck %s --check-prefixes=CHECK,ZNVER1
-
-define <2 x i64> @test_aesdec(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) {
-; GENERIC-LABEL: test_aesdec:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: aesdec %xmm1, %xmm0 # sched: [7:1.00]
-; GENERIC-NEXT: aesdec (%rdi), %xmm0 # sched: [13:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SLM-LABEL: test_aesdec:
-; SLM: # %bb.0:
-; SLM-NEXT: aesdec %xmm1, %xmm0 # sched: [8:5.00]
-; SLM-NEXT: aesdec (%rdi), %xmm0 # sched: [8:5.00]
-; SLM-NEXT: retq # sched: [4:1.00]
-;
-; SANDY-SSE-LABEL: test_aesdec:
-; SANDY-SSE: # %bb.0:
-; SANDY-SSE-NEXT: aesdec %xmm1, %xmm0 # sched: [7:1.00]
-; SANDY-SSE-NEXT: aesdec (%rdi), %xmm0 # sched: [13:1.00]
-; SANDY-SSE-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_aesdec:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vaesdec %xmm1, %xmm0, %xmm0 # sched: [7:1.00]
-; SANDY-NEXT: vaesdec (%rdi), %xmm0, %xmm0 # sched: [13:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-SSE-LABEL: test_aesdec:
-; HASWELL-SSE: # %bb.0:
-; HASWELL-SSE-NEXT: aesdec %xmm1, %xmm0 # sched: [7:1.00]
-; HASWELL-SSE-NEXT: aesdec (%rdi), %xmm0 # sched: [13:1.00]
-; HASWELL-SSE-NEXT: retq # sched: [7:1.00]
-;
-; HASWELL-LABEL: test_aesdec:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vaesdec %xmm1, %xmm0, %xmm0 # sched: [7:1.00]
-; HASWELL-NEXT: vaesdec (%rdi), %xmm0, %xmm0 # sched: [13:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-SSE-LABEL: test_aesdec:
-; BROADWELL-SSE: # %bb.0:
-; BROADWELL-SSE-NEXT: aesdec %xmm1, %xmm0 # sched: [7:1.00]
-; BROADWELL-SSE-NEXT: aesdec (%rdi), %xmm0 # sched: [12:1.00]
-; BROADWELL-SSE-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_aesdec:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vaesdec %xmm1, %xmm0, %xmm0 # sched: [7:1.00]
-; BROADWELL-NEXT: vaesdec (%rdi), %xmm0, %xmm0 # sched: [12:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-SSE-LABEL: test_aesdec:
-; SKYLAKE-SSE: # %bb.0:
-; SKYLAKE-SSE-NEXT: aesdec %xmm1, %xmm0 # sched: [4:1.00]
-; SKYLAKE-SSE-NEXT: aesdec (%rdi), %xmm0 # sched: [10:1.00]
-; SKYLAKE-SSE-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_aesdec:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vaesdec %xmm1, %xmm0, %xmm0 # sched: [4:1.00]
-; SKYLAKE-NEXT: vaesdec (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-SSE-LABEL: test_aesdec:
-; SKX-SSE: # %bb.0:
-; SKX-SSE-NEXT: aesdec %xmm1, %xmm0 # sched: [4:1.00]
-; SKX-SSE-NEXT: aesdec (%rdi), %xmm0 # sched: [10:1.00]
-; SKX-SSE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_aesdec:
-; SKX: # %bb.0:
-; SKX-NEXT: vaesdec %xmm1, %xmm0, %xmm0 # sched: [4:1.00]
-; SKX-NEXT: vaesdec (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-SSE-LABEL: test_aesdec:
-; BDVER2-SSE: # %bb.0:
-; BDVER2-SSE-NEXT: aesdec %xmm1, %xmm0 # sched: [9:1.00]
-; BDVER2-SSE-NEXT: aesdec (%rdi), %xmm0 # sched: [14:1.00]
-; BDVER2-SSE-NEXT: retq # sched: [5:1.00]
-;
-; BDVER2-LABEL: test_aesdec:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vaesdec %xmm1, %xmm0, %xmm0 # sched: [9:1.00]
-; BDVER2-NEXT: vaesdec (%rdi), %xmm0, %xmm0 # sched: [14:1.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-SSE-LABEL: test_aesdec:
-; BTVER2-SSE: # %bb.0:
-; BTVER2-SSE-NEXT: aesdec %xmm1, %xmm0 # sched: [3:1.00]
-; BTVER2-SSE-NEXT: aesdec (%rdi), %xmm0 # sched: [8:1.00]
-; BTVER2-SSE-NEXT: retq # sched: [4:1.00]
-;
-; BTVER2-LABEL: test_aesdec:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vaesdec %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; BTVER2-NEXT: vaesdec (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-SSE-LABEL: test_aesdec:
-; ZNVER1-SSE: # %bb.0:
-; ZNVER1-SSE-NEXT: aesdec %xmm1, %xmm0 # sched: [4:0.50]
-; ZNVER1-SSE-NEXT: aesdec (%rdi), %xmm0 # sched: [11:0.50]
-; ZNVER1-SSE-NEXT: retq # sched: [1:0.50]
-;
-; ZNVER1-LABEL: test_aesdec:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vaesdec %xmm1, %xmm0, %xmm0 # sched: [4:0.50]
-; ZNVER1-NEXT: vaesdec (%rdi), %xmm0, %xmm0 # sched: [11:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = load <2 x i64>, <2 x i64> *%a2, align 16
- %2 = call <2 x i64> @llvm.x86.aesni.aesdec(<2 x i64> %a0, <2 x i64> %a1)
- %3 = call <2 x i64> @llvm.x86.aesni.aesdec(<2 x i64> %2, <2 x i64> %1)
- ret <2 x i64> %3
-}
-declare <2 x i64> @llvm.x86.aesni.aesdec(<2 x i64>, <2 x i64>)
-
-define <2 x i64> @test_aesdeclast(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) {
-; GENERIC-LABEL: test_aesdeclast:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: aesdeclast %xmm1, %xmm0 # sched: [7:1.00]
-; GENERIC-NEXT: aesdeclast (%rdi), %xmm0 # sched: [13:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SLM-LABEL: test_aesdeclast:
-; SLM: # %bb.0:
-; SLM-NEXT: aesdeclast %xmm1, %xmm0 # sched: [8:5.00]
-; SLM-NEXT: aesdeclast (%rdi), %xmm0 # sched: [8:5.00]
-; SLM-NEXT: retq # sched: [4:1.00]
-;
-; SANDY-SSE-LABEL: test_aesdeclast:
-; SANDY-SSE: # %bb.0:
-; SANDY-SSE-NEXT: aesdeclast %xmm1, %xmm0 # sched: [7:1.00]
-; SANDY-SSE-NEXT: aesdeclast (%rdi), %xmm0 # sched: [13:1.00]
-; SANDY-SSE-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_aesdeclast:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vaesdeclast %xmm1, %xmm0, %xmm0 # sched: [7:1.00]
-; SANDY-NEXT: vaesdeclast (%rdi), %xmm0, %xmm0 # sched: [13:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-SSE-LABEL: test_aesdeclast:
-; HASWELL-SSE: # %bb.0:
-; HASWELL-SSE-NEXT: aesdeclast %xmm1, %xmm0 # sched: [7:1.00]
-; HASWELL-SSE-NEXT: aesdeclast (%rdi), %xmm0 # sched: [13:1.00]
-; HASWELL-SSE-NEXT: retq # sched: [7:1.00]
-;
-; HASWELL-LABEL: test_aesdeclast:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vaesdeclast %xmm1, %xmm0, %xmm0 # sched: [7:1.00]
-; HASWELL-NEXT: vaesdeclast (%rdi), %xmm0, %xmm0 # sched: [13:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-SSE-LABEL: test_aesdeclast:
-; BROADWELL-SSE: # %bb.0:
-; BROADWELL-SSE-NEXT: aesdeclast %xmm1, %xmm0 # sched: [7:1.00]
-; BROADWELL-SSE-NEXT: aesdeclast (%rdi), %xmm0 # sched: [12:1.00]
-; BROADWELL-SSE-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_aesdeclast:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vaesdeclast %xmm1, %xmm0, %xmm0 # sched: [7:1.00]
-; BROADWELL-NEXT: vaesdeclast (%rdi), %xmm0, %xmm0 # sched: [12:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-SSE-LABEL: test_aesdeclast:
-; SKYLAKE-SSE: # %bb.0:
-; SKYLAKE-SSE-NEXT: aesdeclast %xmm1, %xmm0 # sched: [4:1.00]
-; SKYLAKE-SSE-NEXT: aesdeclast (%rdi), %xmm0 # sched: [10:1.00]
-; SKYLAKE-SSE-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_aesdeclast:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vaesdeclast %xmm1, %xmm0, %xmm0 # sched: [4:1.00]
-; SKYLAKE-NEXT: vaesdeclast (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-SSE-LABEL: test_aesdeclast:
-; SKX-SSE: # %bb.0:
-; SKX-SSE-NEXT: aesdeclast %xmm1, %xmm0 # sched: [4:1.00]
-; SKX-SSE-NEXT: aesdeclast (%rdi), %xmm0 # sched: [10:1.00]
-; SKX-SSE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_aesdeclast:
-; SKX: # %bb.0:
-; SKX-NEXT: vaesdeclast %xmm1, %xmm0, %xmm0 # sched: [4:1.00]
-; SKX-NEXT: vaesdeclast (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-SSE-LABEL: test_aesdeclast:
-; BDVER2-SSE: # %bb.0:
-; BDVER2-SSE-NEXT: aesdeclast %xmm1, %xmm0 # sched: [9:1.00]
-; BDVER2-SSE-NEXT: aesdeclast (%rdi), %xmm0 # sched: [14:1.00]
-; BDVER2-SSE-NEXT: retq # sched: [5:1.00]
-;
-; BDVER2-LABEL: test_aesdeclast:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vaesdeclast %xmm1, %xmm0, %xmm0 # sched: [9:1.00]
-; BDVER2-NEXT: vaesdeclast (%rdi), %xmm0, %xmm0 # sched: [14:1.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-SSE-LABEL: test_aesdeclast:
-; BTVER2-SSE: # %bb.0:
-; BTVER2-SSE-NEXT: aesdeclast %xmm1, %xmm0 # sched: [3:1.00]
-; BTVER2-SSE-NEXT: aesdeclast (%rdi), %xmm0 # sched: [8:1.00]
-; BTVER2-SSE-NEXT: retq # sched: [4:1.00]
-;
-; BTVER2-LABEL: test_aesdeclast:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vaesdeclast %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; BTVER2-NEXT: vaesdeclast (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-SSE-LABEL: test_aesdeclast:
-; ZNVER1-SSE: # %bb.0:
-; ZNVER1-SSE-NEXT: aesdeclast %xmm1, %xmm0 # sched: [4:0.50]
-; ZNVER1-SSE-NEXT: aesdeclast (%rdi), %xmm0 # sched: [11:0.50]
-; ZNVER1-SSE-NEXT: retq # sched: [1:0.50]
-;
-; ZNVER1-LABEL: test_aesdeclast:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vaesdeclast %xmm1, %xmm0, %xmm0 # sched: [4:0.50]
-; ZNVER1-NEXT: vaesdeclast (%rdi), %xmm0, %xmm0 # sched: [11:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = load <2 x i64>, <2 x i64> *%a2, align 16
- %2 = call <2 x i64> @llvm.x86.aesni.aesdeclast(<2 x i64> %a0, <2 x i64> %a1)
- %3 = call <2 x i64> @llvm.x86.aesni.aesdeclast(<2 x i64> %2, <2 x i64> %1)
- ret <2 x i64> %3
-}
-declare <2 x i64> @llvm.x86.aesni.aesdeclast(<2 x i64>, <2 x i64>)
-
-define <2 x i64> @test_aesenc(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) {
-; GENERIC-LABEL: test_aesenc:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: aesenc %xmm1, %xmm0 # sched: [7:1.00]
-; GENERIC-NEXT: aesenc (%rdi), %xmm0 # sched: [13:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SLM-LABEL: test_aesenc:
-; SLM: # %bb.0:
-; SLM-NEXT: aesenc %xmm1, %xmm0 # sched: [8:5.00]
-; SLM-NEXT: aesenc (%rdi), %xmm0 # sched: [8:5.00]
-; SLM-NEXT: retq # sched: [4:1.00]
-;
-; SANDY-SSE-LABEL: test_aesenc:
-; SANDY-SSE: # %bb.0:
-; SANDY-SSE-NEXT: aesenc %xmm1, %xmm0 # sched: [7:1.00]
-; SANDY-SSE-NEXT: aesenc (%rdi), %xmm0 # sched: [13:1.00]
-; SANDY-SSE-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_aesenc:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vaesenc %xmm1, %xmm0, %xmm0 # sched: [7:1.00]
-; SANDY-NEXT: vaesenc (%rdi), %xmm0, %xmm0 # sched: [13:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-SSE-LABEL: test_aesenc:
-; HASWELL-SSE: # %bb.0:
-; HASWELL-SSE-NEXT: aesenc %xmm1, %xmm0 # sched: [7:1.00]
-; HASWELL-SSE-NEXT: aesenc (%rdi), %xmm0 # sched: [13:1.00]
-; HASWELL-SSE-NEXT: retq # sched: [7:1.00]
-;
-; HASWELL-LABEL: test_aesenc:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vaesenc %xmm1, %xmm0, %xmm0 # sched: [7:1.00]
-; HASWELL-NEXT: vaesenc (%rdi), %xmm0, %xmm0 # sched: [13:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-SSE-LABEL: test_aesenc:
-; BROADWELL-SSE: # %bb.0:
-; BROADWELL-SSE-NEXT: aesenc %xmm1, %xmm0 # sched: [7:1.00]
-; BROADWELL-SSE-NEXT: aesenc (%rdi), %xmm0 # sched: [12:1.00]
-; BROADWELL-SSE-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_aesenc:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vaesenc %xmm1, %xmm0, %xmm0 # sched: [7:1.00]
-; BROADWELL-NEXT: vaesenc (%rdi), %xmm0, %xmm0 # sched: [12:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-SSE-LABEL: test_aesenc:
-; SKYLAKE-SSE: # %bb.0:
-; SKYLAKE-SSE-NEXT: aesenc %xmm1, %xmm0 # sched: [4:1.00]
-; SKYLAKE-SSE-NEXT: aesenc (%rdi), %xmm0 # sched: [10:1.00]
-; SKYLAKE-SSE-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_aesenc:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vaesenc %xmm1, %xmm0, %xmm0 # sched: [4:1.00]
-; SKYLAKE-NEXT: vaesenc (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-SSE-LABEL: test_aesenc:
-; SKX-SSE: # %bb.0:
-; SKX-SSE-NEXT: aesenc %xmm1, %xmm0 # sched: [4:1.00]
-; SKX-SSE-NEXT: aesenc (%rdi), %xmm0 # sched: [10:1.00]
-; SKX-SSE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_aesenc:
-; SKX: # %bb.0:
-; SKX-NEXT: vaesenc %xmm1, %xmm0, %xmm0 # sched: [4:1.00]
-; SKX-NEXT: vaesenc (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-SSE-LABEL: test_aesenc:
-; BDVER2-SSE: # %bb.0:
-; BDVER2-SSE-NEXT: aesenc %xmm1, %xmm0 # sched: [9:1.00]
-; BDVER2-SSE-NEXT: aesenc (%rdi), %xmm0 # sched: [14:1.00]
-; BDVER2-SSE-NEXT: retq # sched: [5:1.00]
-;
-; BDVER2-LABEL: test_aesenc:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vaesenc %xmm1, %xmm0, %xmm0 # sched: [9:1.00]
-; BDVER2-NEXT: vaesenc (%rdi), %xmm0, %xmm0 # sched: [14:1.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-SSE-LABEL: test_aesenc:
-; BTVER2-SSE: # %bb.0:
-; BTVER2-SSE-NEXT: aesenc %xmm1, %xmm0 # sched: [3:1.00]
-; BTVER2-SSE-NEXT: aesenc (%rdi), %xmm0 # sched: [8:1.00]
-; BTVER2-SSE-NEXT: retq # sched: [4:1.00]
-;
-; BTVER2-LABEL: test_aesenc:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vaesenc %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; BTVER2-NEXT: vaesenc (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-SSE-LABEL: test_aesenc:
-; ZNVER1-SSE: # %bb.0:
-; ZNVER1-SSE-NEXT: aesenc %xmm1, %xmm0 # sched: [4:0.50]
-; ZNVER1-SSE-NEXT: aesenc (%rdi), %xmm0 # sched: [11:0.50]
-; ZNVER1-SSE-NEXT: retq # sched: [1:0.50]
-;
-; ZNVER1-LABEL: test_aesenc:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vaesenc %xmm1, %xmm0, %xmm0 # sched: [4:0.50]
-; ZNVER1-NEXT: vaesenc (%rdi), %xmm0, %xmm0 # sched: [11:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = load <2 x i64>, <2 x i64> *%a2, align 16
- %2 = call <2 x i64> @llvm.x86.aesni.aesenc(<2 x i64> %a0, <2 x i64> %a1)
- %3 = call <2 x i64> @llvm.x86.aesni.aesenc(<2 x i64> %2, <2 x i64> %1)
- ret <2 x i64> %3
-}
-declare <2 x i64> @llvm.x86.aesni.aesenc(<2 x i64>, <2 x i64>)
-
-define <2 x i64> @test_aesenclast(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) {
-; GENERIC-LABEL: test_aesenclast:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: aesenclast %xmm1, %xmm0 # sched: [7:1.00]
-; GENERIC-NEXT: aesenclast (%rdi), %xmm0 # sched: [13:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SLM-LABEL: test_aesenclast:
-; SLM: # %bb.0:
-; SLM-NEXT: aesenclast %xmm1, %xmm0 # sched: [8:5.00]
-; SLM-NEXT: aesenclast (%rdi), %xmm0 # sched: [8:5.00]
-; SLM-NEXT: retq # sched: [4:1.00]
-;
-; SANDY-SSE-LABEL: test_aesenclast:
-; SANDY-SSE: # %bb.0:
-; SANDY-SSE-NEXT: aesenclast %xmm1, %xmm0 # sched: [7:1.00]
-; SANDY-SSE-NEXT: aesenclast (%rdi), %xmm0 # sched: [13:1.00]
-; SANDY-SSE-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_aesenclast:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vaesenclast %xmm1, %xmm0, %xmm0 # sched: [7:1.00]
-; SANDY-NEXT: vaesenclast (%rdi), %xmm0, %xmm0 # sched: [13:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-SSE-LABEL: test_aesenclast:
-; HASWELL-SSE: # %bb.0:
-; HASWELL-SSE-NEXT: aesenclast %xmm1, %xmm0 # sched: [7:1.00]
-; HASWELL-SSE-NEXT: aesenclast (%rdi), %xmm0 # sched: [13:1.00]
-; HASWELL-SSE-NEXT: retq # sched: [7:1.00]
-;
-; HASWELL-LABEL: test_aesenclast:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vaesenclast %xmm1, %xmm0, %xmm0 # sched: [7:1.00]
-; HASWELL-NEXT: vaesenclast (%rdi), %xmm0, %xmm0 # sched: [13:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-SSE-LABEL: test_aesenclast:
-; BROADWELL-SSE: # %bb.0:
-; BROADWELL-SSE-NEXT: aesenclast %xmm1, %xmm0 # sched: [7:1.00]
-; BROADWELL-SSE-NEXT: aesenclast (%rdi), %xmm0 # sched: [12:1.00]
-; BROADWELL-SSE-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_aesenclast:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vaesenclast %xmm1, %xmm0, %xmm0 # sched: [7:1.00]
-; BROADWELL-NEXT: vaesenclast (%rdi), %xmm0, %xmm0 # sched: [12:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-SSE-LABEL: test_aesenclast:
-; SKYLAKE-SSE: # %bb.0:
-; SKYLAKE-SSE-NEXT: aesenclast %xmm1, %xmm0 # sched: [4:1.00]
-; SKYLAKE-SSE-NEXT: aesenclast (%rdi), %xmm0 # sched: [10:1.00]
-; SKYLAKE-SSE-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_aesenclast:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vaesenclast %xmm1, %xmm0, %xmm0 # sched: [4:1.00]
-; SKYLAKE-NEXT: vaesenclast (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-SSE-LABEL: test_aesenclast:
-; SKX-SSE: # %bb.0:
-; SKX-SSE-NEXT: aesenclast %xmm1, %xmm0 # sched: [4:1.00]
-; SKX-SSE-NEXT: aesenclast (%rdi), %xmm0 # sched: [10:1.00]
-; SKX-SSE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_aesenclast:
-; SKX: # %bb.0:
-; SKX-NEXT: vaesenclast %xmm1, %xmm0, %xmm0 # sched: [4:1.00]
-; SKX-NEXT: vaesenclast (%rdi), %xmm0, %xmm0 # sched: [10:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-SSE-LABEL: test_aesenclast:
-; BDVER2-SSE: # %bb.0:
-; BDVER2-SSE-NEXT: aesenclast %xmm1, %xmm0 # sched: [9:1.00]
-; BDVER2-SSE-NEXT: aesenclast (%rdi), %xmm0 # sched: [14:1.00]
-; BDVER2-SSE-NEXT: retq # sched: [5:1.00]
-;
-; BDVER2-LABEL: test_aesenclast:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vaesenclast %xmm1, %xmm0, %xmm0 # sched: [9:1.00]
-; BDVER2-NEXT: vaesenclast (%rdi), %xmm0, %xmm0 # sched: [14:1.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-SSE-LABEL: test_aesenclast:
-; BTVER2-SSE: # %bb.0:
-; BTVER2-SSE-NEXT: aesenclast %xmm1, %xmm0 # sched: [3:1.00]
-; BTVER2-SSE-NEXT: aesenclast (%rdi), %xmm0 # sched: [8:1.00]
-; BTVER2-SSE-NEXT: retq # sched: [4:1.00]
-;
-; BTVER2-LABEL: test_aesenclast:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vaesenclast %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; BTVER2-NEXT: vaesenclast (%rdi), %xmm0, %xmm0 # sched: [8:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-SSE-LABEL: test_aesenclast:
-; ZNVER1-SSE: # %bb.0:
-; ZNVER1-SSE-NEXT: aesenclast %xmm1, %xmm0 # sched: [4:0.50]
-; ZNVER1-SSE-NEXT: aesenclast (%rdi), %xmm0 # sched: [11:0.50]
-; ZNVER1-SSE-NEXT: retq # sched: [1:0.50]
-;
-; ZNVER1-LABEL: test_aesenclast:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vaesenclast %xmm1, %xmm0, %xmm0 # sched: [4:0.50]
-; ZNVER1-NEXT: vaesenclast (%rdi), %xmm0, %xmm0 # sched: [11:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = load <2 x i64>, <2 x i64> *%a2, align 16
- %2 = call <2 x i64> @llvm.x86.aesni.aesenclast(<2 x i64> %a0, <2 x i64> %a1)
- %3 = call <2 x i64> @llvm.x86.aesni.aesenclast(<2 x i64> %2, <2 x i64> %1)
- ret <2 x i64> %3
-}
-declare <2 x i64> @llvm.x86.aesni.aesenclast(<2 x i64>, <2 x i64>)
-
-define <2 x i64> @test_aesimc(<2 x i64> %a0, <2 x i64> *%a1) {
-; GENERIC-LABEL: test_aesimc:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: aesimc %xmm0, %xmm1 # sched: [12:2.00]
-; GENERIC-NEXT: aesimc (%rdi), %xmm0 # sched: [18:2.00]
-; GENERIC-NEXT: por %xmm1, %xmm0 # sched: [1:0.33]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SLM-LABEL: test_aesimc:
-; SLM: # %bb.0:
-; SLM-NEXT: aesimc %xmm0, %xmm1 # sched: [8:5.00]
-; SLM-NEXT: aesimc (%rdi), %xmm0 # sched: [8:5.00]
-; SLM-NEXT: por %xmm1, %xmm0 # sched: [1:0.50]
-; SLM-NEXT: retq # sched: [4:1.00]
-;
-; SANDY-SSE-LABEL: test_aesimc:
-; SANDY-SSE: # %bb.0:
-; SANDY-SSE-NEXT: aesimc %xmm0, %xmm1 # sched: [12:2.00]
-; SANDY-SSE-NEXT: aesimc (%rdi), %xmm0 # sched: [18:2.00]
-; SANDY-SSE-NEXT: por %xmm1, %xmm0 # sched: [1:0.33]
-; SANDY-SSE-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_aesimc:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vaesimc %xmm0, %xmm0 # sched: [12:2.00]
-; SANDY-NEXT: vaesimc (%rdi), %xmm1 # sched: [18:2.00]
-; SANDY-NEXT: vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-SSE-LABEL: test_aesimc:
-; HASWELL-SSE: # %bb.0:
-; HASWELL-SSE-NEXT: aesimc %xmm0, %xmm1 # sched: [14:2.00]
-; HASWELL-SSE-NEXT: aesimc (%rdi), %xmm0 # sched: [20:2.00]
-; HASWELL-SSE-NEXT: por %xmm1, %xmm0 # sched: [1:0.33]
-; HASWELL-SSE-NEXT: retq # sched: [7:1.00]
-;
-; HASWELL-LABEL: test_aesimc:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vaesimc %xmm0, %xmm0 # sched: [14:2.00]
-; HASWELL-NEXT: vaesimc (%rdi), %xmm1 # sched: [20:2.00]
-; HASWELL-NEXT: vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-SSE-LABEL: test_aesimc:
-; BROADWELL-SSE: # %bb.0:
-; BROADWELL-SSE-NEXT: aesimc (%rdi), %xmm1 # sched: [19:2.00]
-; BROADWELL-SSE-NEXT: aesimc %xmm0, %xmm0 # sched: [14:2.00]
-; BROADWELL-SSE-NEXT: por %xmm0, %xmm1 # sched: [1:0.33]
-; BROADWELL-SSE-NEXT: movdqa %xmm1, %xmm0 # sched: [1:0.33]
-; BROADWELL-SSE-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_aesimc:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vaesimc (%rdi), %xmm1 # sched: [19:2.00]
-; BROADWELL-NEXT: vaesimc %xmm0, %xmm0 # sched: [14:2.00]
-; BROADWELL-NEXT: vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-SSE-LABEL: test_aesimc:
-; SKYLAKE-SSE: # %bb.0:
-; SKYLAKE-SSE-NEXT: aesimc %xmm0, %xmm1 # sched: [8:2.00]
-; SKYLAKE-SSE-NEXT: aesimc (%rdi), %xmm0 # sched: [14:2.00]
-; SKYLAKE-SSE-NEXT: por %xmm1, %xmm0 # sched: [1:0.33]
-; SKYLAKE-SSE-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_aesimc:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vaesimc %xmm0, %xmm0 # sched: [8:2.00]
-; SKYLAKE-NEXT: vaesimc (%rdi), %xmm1 # sched: [14:2.00]
-; SKYLAKE-NEXT: vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-SSE-LABEL: test_aesimc:
-; SKX-SSE: # %bb.0:
-; SKX-SSE-NEXT: aesimc %xmm0, %xmm1 # sched: [8:2.00]
-; SKX-SSE-NEXT: aesimc (%rdi), %xmm0 # sched: [14:2.00]
-; SKX-SSE-NEXT: por %xmm1, %xmm0 # sched: [1:0.33]
-; SKX-SSE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_aesimc:
-; SKX: # %bb.0:
-; SKX-NEXT: vaesimc %xmm0, %xmm0 # sched: [8:2.00]
-; SKX-NEXT: vaesimc (%rdi), %xmm1 # sched: [14:2.00]
-; SKX-NEXT: vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-SSE-LABEL: test_aesimc:
-; BDVER2-SSE: # %bb.0:
-; BDVER2-SSE-NEXT: aesimc %xmm0, %xmm1 # sched: [5:1.00]
-; BDVER2-SSE-NEXT: aesimc (%rdi), %xmm0 # sched: [10:1.00]
-; BDVER2-SSE-NEXT: por %xmm1, %xmm0 # sched: [2:0.50]
-; BDVER2-SSE-NEXT: retq # sched: [5:1.00]
-;
-; BDVER2-LABEL: test_aesimc:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vaesimc (%rdi), %xmm1 # sched: [10:1.00]
-; BDVER2-NEXT: vaesimc %xmm0, %xmm0 # sched: [5:1.00]
-; BDVER2-NEXT: vpor %xmm1, %xmm0, %xmm0 # sched: [2:0.50]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-SSE-LABEL: test_aesimc:
-; BTVER2-SSE: # %bb.0:
-; BTVER2-SSE-NEXT: aesimc %xmm0, %xmm1 # sched: [2:1.00]
-; BTVER2-SSE-NEXT: aesimc (%rdi), %xmm0 # sched: [7:1.00]
-; BTVER2-SSE-NEXT: por %xmm1, %xmm0 # sched: [1:0.50]
-; BTVER2-SSE-NEXT: retq # sched: [4:1.00]
-;
-; BTVER2-LABEL: test_aesimc:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vaesimc (%rdi), %xmm1 # sched: [7:1.00]
-; BTVER2-NEXT: vaesimc %xmm0, %xmm0 # sched: [2:1.00]
-; BTVER2-NEXT: vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-SSE-LABEL: test_aesimc:
-; ZNVER1-SSE: # %bb.0:
-; ZNVER1-SSE-NEXT: aesimc %xmm0, %xmm1 # sched: [4:0.50]
-; ZNVER1-SSE-NEXT: aesimc (%rdi), %xmm0 # sched: [11:0.50]
-; ZNVER1-SSE-NEXT: por %xmm1, %xmm0 # sched: [1:0.25]
-; ZNVER1-SSE-NEXT: retq # sched: [1:0.50]
-;
-; ZNVER1-LABEL: test_aesimc:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vaesimc (%rdi), %xmm1 # sched: [11:0.50]
-; ZNVER1-NEXT: vaesimc %xmm0, %xmm0 # sched: [4:0.50]
-; ZNVER1-NEXT: vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = load <2 x i64>, <2 x i64> *%a1, align 16
- %2 = call <2 x i64> @llvm.x86.aesni.aesimc(<2 x i64> %a0)
- %3 = call <2 x i64> @llvm.x86.aesni.aesimc(<2 x i64> %1)
- %4 = or <2 x i64> %2, %3
- ret <2 x i64> %4
-}
-declare <2 x i64> @llvm.x86.aesni.aesimc(<2 x i64>)
-
-define <2 x i64> @test_aeskeygenassist(<2 x i64> %a0, <2 x i64> *%a1) {
-; GENERIC-LABEL: test_aeskeygenassist:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: aeskeygenassist $7, %xmm0, %xmm1 # sched: [8:3.67]
-; GENERIC-NEXT: aeskeygenassist $7, (%rdi), %xmm0 # sched: [8:3.33]
-; GENERIC-NEXT: por %xmm1, %xmm0 # sched: [1:0.33]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SLM-LABEL: test_aeskeygenassist:
-; SLM: # %bb.0:
-; SLM-NEXT: aeskeygenassist $7, %xmm0, %xmm1 # sched: [8:5.00]
-; SLM-NEXT: aeskeygenassist $7, (%rdi), %xmm0 # sched: [8:5.00]
-; SLM-NEXT: por %xmm1, %xmm0 # sched: [1:0.50]
-; SLM-NEXT: retq # sched: [4:1.00]
-;
-; SANDY-SSE-LABEL: test_aeskeygenassist:
-; SANDY-SSE: # %bb.0:
-; SANDY-SSE-NEXT: aeskeygenassist $7, %xmm0, %xmm1 # sched: [8:3.67]
-; SANDY-SSE-NEXT: aeskeygenassist $7, (%rdi), %xmm0 # sched: [8:3.33]
-; SANDY-SSE-NEXT: por %xmm1, %xmm0 # sched: [1:0.33]
-; SANDY-SSE-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_aeskeygenassist:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vaeskeygenassist $7, %xmm0, %xmm0 # sched: [8:3.67]
-; SANDY-NEXT: vaeskeygenassist $7, (%rdi), %xmm1 # sched: [8:3.33]
-; SANDY-NEXT: vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-SSE-LABEL: test_aeskeygenassist:
-; HASWELL-SSE: # %bb.0:
-; HASWELL-SSE-NEXT: aeskeygenassist $7, %xmm0, %xmm1 # sched: [29:7.00]
-; HASWELL-SSE-NEXT: aeskeygenassist $7, (%rdi), %xmm0 # sched: [34:7.00]
-; HASWELL-SSE-NEXT: por %xmm1, %xmm0 # sched: [1:0.33]
-; HASWELL-SSE-NEXT: retq # sched: [7:1.00]
-;
-; HASWELL-LABEL: test_aeskeygenassist:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vaeskeygenassist $7, %xmm0, %xmm0 # sched: [29:7.00]
-; HASWELL-NEXT: vaeskeygenassist $7, (%rdi), %xmm1 # sched: [34:7.00]
-; HASWELL-NEXT: vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-SSE-LABEL: test_aeskeygenassist:
-; BROADWELL-SSE: # %bb.0:
-; BROADWELL-SSE-NEXT: aeskeygenassist $7, %xmm0, %xmm1 # sched: [29:7.00]
-; BROADWELL-SSE-NEXT: aeskeygenassist $7, (%rdi), %xmm0 # sched: [33:7.00]
-; BROADWELL-SSE-NEXT: por %xmm1, %xmm0 # sched: [1:0.33]
-; BROADWELL-SSE-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_aeskeygenassist:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vaeskeygenassist $7, %xmm0, %xmm0 # sched: [29:7.00]
-; BROADWELL-NEXT: vaeskeygenassist $7, (%rdi), %xmm1 # sched: [33:7.00]
-; BROADWELL-NEXT: vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-SSE-LABEL: test_aeskeygenassist:
-; SKYLAKE-SSE: # %bb.0:
-; SKYLAKE-SSE-NEXT: aeskeygenassist $7, %xmm0, %xmm1 # sched: [20:6.00]
-; SKYLAKE-SSE-NEXT: aeskeygenassist $7, (%rdi), %xmm0 # sched: [25:6.00]
-; SKYLAKE-SSE-NEXT: por %xmm1, %xmm0 # sched: [1:0.33]
-; SKYLAKE-SSE-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_aeskeygenassist:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vaeskeygenassist $7, %xmm0, %xmm0 # sched: [20:6.00]
-; SKYLAKE-NEXT: vaeskeygenassist $7, (%rdi), %xmm1 # sched: [25:6.00]
-; SKYLAKE-NEXT: vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-SSE-LABEL: test_aeskeygenassist:
-; SKX-SSE: # %bb.0:
-; SKX-SSE-NEXT: aeskeygenassist $7, %xmm0, %xmm1 # sched: [20:6.00]
-; SKX-SSE-NEXT: aeskeygenassist $7, (%rdi), %xmm0 # sched: [25:6.00]
-; SKX-SSE-NEXT: por %xmm1, %xmm0 # sched: [1:0.33]
-; SKX-SSE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_aeskeygenassist:
-; SKX: # %bb.0:
-; SKX-NEXT: vaeskeygenassist $7, %xmm0, %xmm0 # sched: [20:6.00]
-; SKX-NEXT: vaeskeygenassist $7, (%rdi), %xmm1 # sched: [25:6.00]
-; SKX-NEXT: vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-SSE-LABEL: test_aeskeygenassist:
-; BDVER2-SSE: # %bb.0:
-; BDVER2-SSE-NEXT: aeskeygenassist $7, %xmm0, %xmm1 # sched: [5:1.00]
-; BDVER2-SSE-NEXT: aeskeygenassist $7, (%rdi), %xmm0 # sched: [10:1.00]
-; BDVER2-SSE-NEXT: por %xmm1, %xmm0 # sched: [2:0.50]
-; BDVER2-SSE-NEXT: retq # sched: [5:1.00]
-;
-; BDVER2-LABEL: test_aeskeygenassist:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vaeskeygenassist $7, (%rdi), %xmm1 # sched: [10:1.00]
-; BDVER2-NEXT: vaeskeygenassist $7, %xmm0, %xmm0 # sched: [5:1.00]
-; BDVER2-NEXT: vpor %xmm1, %xmm0, %xmm0 # sched: [2:0.50]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-SSE-LABEL: test_aeskeygenassist:
-; BTVER2-SSE: # %bb.0:
-; BTVER2-SSE-NEXT: aeskeygenassist $7, %xmm0, %xmm1 # sched: [2:1.00]
-; BTVER2-SSE-NEXT: aeskeygenassist $7, (%rdi), %xmm0 # sched: [7:1.00]
-; BTVER2-SSE-NEXT: por %xmm1, %xmm0 # sched: [1:0.50]
-; BTVER2-SSE-NEXT: retq # sched: [4:1.00]
-;
-; BTVER2-LABEL: test_aeskeygenassist:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vaeskeygenassist $7, (%rdi), %xmm1 # sched: [7:1.00]
-; BTVER2-NEXT: vaeskeygenassist $7, %xmm0, %xmm0 # sched: [2:1.00]
-; BTVER2-NEXT: vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-SSE-LABEL: test_aeskeygenassist:
-; ZNVER1-SSE: # %bb.0:
-; ZNVER1-SSE-NEXT: aeskeygenassist $7, %xmm0, %xmm1 # sched: [4:0.50]
-; ZNVER1-SSE-NEXT: aeskeygenassist $7, (%rdi), %xmm0 # sched: [11:0.50]
-; ZNVER1-SSE-NEXT: por %xmm1, %xmm0 # sched: [1:0.25]
-; ZNVER1-SSE-NEXT: retq # sched: [1:0.50]
-;
-; ZNVER1-LABEL: test_aeskeygenassist:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vaeskeygenassist $7, (%rdi), %xmm1 # sched: [11:0.50]
-; ZNVER1-NEXT: vaeskeygenassist $7, %xmm0, %xmm0 # sched: [4:0.50]
-; ZNVER1-NEXT: vpor %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = load <2 x i64>, <2 x i64> *%a1, align 16
- %2 = call <2 x i64> @llvm.x86.aesni.aeskeygenassist(<2 x i64> %a0, i8 7)
- %3 = call <2 x i64> @llvm.x86.aesni.aeskeygenassist(<2 x i64> %1, i8 7)
- %4 = or <2 x i64> %2, %3
- ret <2 x i64> %4
-}
-declare <2 x i64> @llvm.x86.aesni.aeskeygenassist(<2 x i64>, i8)
Removed: llvm/trunk/test/CodeGen/X86/avx-schedule.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/avx-schedule.ll?rev=353042&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/avx-schedule.ll (original)
+++ llvm/trunk/test/CodeGen/X86/avx-schedule.ll (removed)
@@ -1,6120 +0,0 @@
-; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=x86-64 -mattr=+avx | FileCheck %s --check-prefix=CHECK --check-prefix=GENERIC
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=sandybridge -mattr=-avx2 | FileCheck %s --check-prefix=CHECK --check-prefix=SANDY
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=ivybridge -mattr=-avx2 | FileCheck %s --check-prefix=CHECK --check-prefix=SANDY
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=haswell -mattr=-avx2 | FileCheck %s --check-prefix=CHECK --check-prefix=HASWELL
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=broadwell -mattr=-avx2 | FileCheck %s --check-prefix=CHECK --check-prefix=BROADWELL
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=skylake -mattr=-avx2 | FileCheck %s --check-prefix=CHECK --check-prefix=SKYLAKE
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=skx -mattr=-avx2 | FileCheck %s --check-prefix=CHECK --check-prefix=SKX
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=bdver2 -mattr=-avx2 | FileCheck %s --check-prefix=CHECK --check-prefix=BDVER2
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=btver2 -mattr=-avx2 | FileCheck %s --check-prefix=CHECK --check-prefix=BTVER2
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=znver1 -mattr=-avx2 | FileCheck %s --check-prefix=CHECK --check-prefix=ZNVER1
-
-define <4 x double> @test_addpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
-; GENERIC-LABEL: test_addpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: vaddpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_addpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: vaddpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_addpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vaddpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_addpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vaddpd (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_addpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vaddpd (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_addpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vaddpd (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_addpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: vaddpd (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_addpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: vaddpd (%rdi), %ymm0, %ymm0 # sched: [8:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_addpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: vaddpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = fadd <4 x double> %a0, %a1
- %2 = load <4 x double>, <4 x double> *%a2, align 32
- %3 = fadd <4 x double> %1, %2
- ret <4 x double> %3
-}
-
-define <8 x float> @test_addps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {
-; GENERIC-LABEL: test_addps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: vaddps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_addps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: vaddps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_addps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vaddps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_addps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vaddps (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_addps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vaddps (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_addps:
-; SKX: # %bb.0:
-; SKX-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vaddps (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_addps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: vaddps (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_addps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: vaddps (%rdi), %ymm0, %ymm0 # sched: [8:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_addps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: vaddps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = fadd <8 x float> %a0, %a1
- %2 = load <8 x float>, <8 x float> *%a2, align 32
- %3 = fadd <8 x float> %1, %2
- ret <8 x float> %3
-}
-
-define <4 x double> @test_addsubpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
-; GENERIC-LABEL: test_addsubpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vaddsubpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: vaddsubpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_addsubpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vaddsubpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: vaddsubpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_addsubpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vaddsubpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vaddsubpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_addsubpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vaddsubpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vaddsubpd (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_addsubpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vaddsubpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vaddsubpd (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_addsubpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vaddsubpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vaddsubpd (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_addsubpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vaddsubpd %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: vaddsubpd (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_addsubpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vaddsubpd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: vaddsubpd (%rdi), %ymm0, %ymm0 # sched: [8:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_addsubpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vaddsubpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: vaddsubpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x double> @llvm.x86.avx.addsub.pd.256(<4 x double> %a0, <4 x double> %a1)
- %2 = load <4 x double>, <4 x double> *%a2, align 32
- %3 = call <4 x double> @llvm.x86.avx.addsub.pd.256(<4 x double> %1, <4 x double> %2)
- ret <4 x double> %3
-}
-declare <4 x double> @llvm.x86.avx.addsub.pd.256(<4 x double>, <4 x double>) nounwind readnone
-
-define <8 x float> @test_addsubps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {
-; GENERIC-LABEL: test_addsubps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vaddsubps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: vaddsubps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_addsubps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vaddsubps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: vaddsubps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_addsubps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vaddsubps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vaddsubps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_addsubps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vaddsubps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vaddsubps (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_addsubps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vaddsubps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vaddsubps (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_addsubps:
-; SKX: # %bb.0:
-; SKX-NEXT: vaddsubps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vaddsubps (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_addsubps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vaddsubps %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: vaddsubps (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_addsubps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vaddsubps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: vaddsubps (%rdi), %ymm0, %ymm0 # sched: [8:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_addsubps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vaddsubps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: vaddsubps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x float> @llvm.x86.avx.addsub.ps.256(<8 x float> %a0, <8 x float> %a1)
- %2 = load <8 x float>, <8 x float> *%a2, align 32
- %3 = call <8 x float> @llvm.x86.avx.addsub.ps.256(<8 x float> %1, <8 x float> %2)
- ret <8 x float> %3
-}
-declare <8 x float> @llvm.x86.avx.addsub.ps.256(<8 x float>, <8 x float>) nounwind readnone
-
-define <4 x double> @test_andnotpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
-; GENERIC-LABEL: test_andnotpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vandnpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vandnpd (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; GENERIC-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_andnotpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vandnpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SANDY-NEXT: vandnpd (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; SANDY-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_andnotpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vandnpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: vandnpd (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_andnotpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vandnpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: vandnpd (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_andnotpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vandnpd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: vandnpd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_andnotpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vandnpd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: vandnpd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_andnotpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vandnpd %ymm1, %ymm0, %ymm0 # sched: [2:1.00]
-; BDVER2-NEXT: vandnpd (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BDVER2-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_andnotpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vandnpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BTVER2-NEXT: vandnpd (%rdi), %ymm0, %ymm0 # sched: [6:2.00]
-; BTVER2-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_andnotpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vandnpd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vandnpd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = bitcast <4 x double> %a0 to <4 x i64>
- %2 = bitcast <4 x double> %a1 to <4 x i64>
- %3 = xor <4 x i64> %1, <i64 -1, i64 -1, i64 -1, i64 -1>
- %4 = and <4 x i64> %3, %2
- %5 = load <4 x double>, <4 x double> *%a2, align 32
- %6 = bitcast <4 x double> %5 to <4 x i64>
- %7 = xor <4 x i64> %4, <i64 -1, i64 -1, i64 -1, i64 -1>
- %8 = and <4 x i64> %6, %7
- %9 = bitcast <4 x i64> %8 to <4 x double>
- %10 = fadd <4 x double> %a1, %9
- ret <4 x double> %10
-}
-
-define <8 x float> @test_andnotps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {
-; GENERIC-LABEL: test_andnotps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vandnps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vandnps (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; GENERIC-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_andnotps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vandnps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SANDY-NEXT: vandnps (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; SANDY-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_andnotps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vandnps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: vandnps (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_andnotps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vandnps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: vandnps (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_andnotps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vandnps %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: vandnps (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_andnotps:
-; SKX: # %bb.0:
-; SKX-NEXT: vandnps %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: vandnps (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_andnotps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vandnps %ymm1, %ymm0, %ymm0 # sched: [2:1.00]
-; BDVER2-NEXT: vandnps (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BDVER2-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_andnotps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vandnps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BTVER2-NEXT: vandnps (%rdi), %ymm0, %ymm0 # sched: [6:2.00]
-; BTVER2-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_andnotps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vandnps %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vandnps (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = bitcast <8 x float> %a0 to <4 x i64>
- %2 = bitcast <8 x float> %a1 to <4 x i64>
- %3 = xor <4 x i64> %1, <i64 -1, i64 -1, i64 -1, i64 -1>
- %4 = and <4 x i64> %3, %2
- %5 = load <8 x float>, <8 x float> *%a2, align 32
- %6 = bitcast <8 x float> %5 to <4 x i64>
- %7 = xor <4 x i64> %4, <i64 -1, i64 -1, i64 -1, i64 -1>
- %8 = and <4 x i64> %6, %7
- %9 = bitcast <4 x i64> %8 to <8 x float>
- %10 = fadd <8 x float> %a1, %9
- ret <8 x float> %10
-}
-
-define <4 x double> @test_andpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
-; GENERIC-LABEL: test_andpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vandpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vandpd (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; GENERIC-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_andpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vandpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SANDY-NEXT: vandpd (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; SANDY-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_andpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vandpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: vandpd (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_andpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vandpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: vandpd (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_andpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vandpd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: vandpd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_andpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vandpd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: vandpd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_andpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vandpd %ymm1, %ymm0, %ymm0 # sched: [2:1.00]
-; BDVER2-NEXT: vandpd (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BDVER2-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_andpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vandpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BTVER2-NEXT: vandpd (%rdi), %ymm0, %ymm0 # sched: [6:2.00]
-; BTVER2-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_andpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vandpd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vandpd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = bitcast <4 x double> %a0 to <4 x i64>
- %2 = bitcast <4 x double> %a1 to <4 x i64>
- %3 = and <4 x i64> %1, %2
- %4 = load <4 x double>, <4 x double> *%a2, align 32
- %5 = bitcast <4 x double> %4 to <4 x i64>
- %6 = and <4 x i64> %3, %5
- %7 = bitcast <4 x i64> %6 to <4 x double>
- %8 = fadd <4 x double> %a1, %7
- ret <4 x double> %8
-}
-
-define <8 x float> @test_andps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {
-; GENERIC-LABEL: test_andps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vandps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vandps (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; GENERIC-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_andps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vandps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SANDY-NEXT: vandps (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; SANDY-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_andps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vandps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: vandps (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_andps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vandps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: vandps (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_andps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vandps %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: vandps (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_andps:
-; SKX: # %bb.0:
-; SKX-NEXT: vandps %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: vandps (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_andps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vandps %ymm1, %ymm0, %ymm0 # sched: [2:1.00]
-; BDVER2-NEXT: vandps (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BDVER2-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_andps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vandps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BTVER2-NEXT: vandps (%rdi), %ymm0, %ymm0 # sched: [6:2.00]
-; BTVER2-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_andps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vandps %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vandps (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = bitcast <8 x float> %a0 to <4 x i64>
- %2 = bitcast <8 x float> %a1 to <4 x i64>
- %3 = and <4 x i64> %1, %2
- %4 = load <8 x float>, <8 x float> *%a2, align 32
- %5 = bitcast <8 x float> %4 to <4 x i64>
- %6 = and <4 x i64> %3, %5
- %7 = bitcast <4 x i64> %6 to <8 x float>
- %8 = fadd <8 x float> %a1, %7
- ret <8 x float> %8
-}
-
-define <4 x double> @test_blendpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
-; GENERIC-LABEL: test_blendpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3] sched: [1:0.50]
-; GENERIC-NEXT: vblendpd {{.*#+}} ymm1 = ymm0[0,1],mem[2,3] sched: [8:0.50]
-; GENERIC-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_blendpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3] sched: [1:0.50]
-; SANDY-NEXT: vblendpd {{.*#+}} ymm1 = ymm0[0,1],mem[2,3] sched: [8:0.50]
-; SANDY-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_blendpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3] sched: [1:0.33]
-; HASWELL-NEXT: vblendpd {{.*#+}} ymm1 = ymm0[0,1],mem[2,3] sched: [8:0.50]
-; HASWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_blendpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3] sched: [1:0.33]
-; BROADWELL-NEXT: vblendpd {{.*#+}} ymm1 = ymm0[0,1],mem[2,3] sched: [7:0.50]
-; BROADWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_blendpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3] sched: [1:0.33]
-; SKYLAKE-NEXT: vblendpd {{.*#+}} ymm1 = ymm0[0,1],mem[2,3] sched: [8:0.50]
-; SKYLAKE-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_blendpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3] sched: [1:0.33]
-; SKX-NEXT: vblendpd {{.*#+}} ymm1 = ymm0[0,1],mem[2,3] sched: [8:0.50]
-; SKX-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_blendpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3] sched: [2:1.00]
-; BDVER2-NEXT: vblendpd {{.*#+}} ymm1 = ymm0[0,1],mem[2,3] sched: [7:1.00]
-; BDVER2-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_blendpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3] sched: [1:1.00]
-; BTVER2-NEXT: vblendpd {{.*#+}} ymm1 = ymm0[0,1],mem[2,3] sched: [6:2.00]
-; BTVER2-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_blendpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3] sched: [1:0.50]
-; ZNVER1-NEXT: vblendpd {{.*#+}} ymm1 = ymm0[0,1],mem[2,3] sched: [8:0.50]
-; ZNVER1-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <4 x double> %a0, <4 x double> %a1, <4 x i32> <i32 0, i32 5, i32 6, i32 3>
- %2 = load <4 x double>, <4 x double> *%a2, align 32
- %3 = shufflevector <4 x double> %1, <4 x double> %2, <4 x i32> <i32 0, i32 1, i32 6, i32 7>
- %4 = fadd <4 x double> %1, %3
- ret <4 x double> %4
-}
-
-define <8 x float> @test_blendps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {
-; GENERIC-LABEL: test_blendps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3,4,5,6,7] sched: [1:0.50]
-; GENERIC-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],mem[2],ymm1[3],mem[4,5,6],ymm1[7] sched: [8:0.50]
-; GENERIC-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_blendps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3,4,5,6,7] sched: [1:0.50]
-; SANDY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],mem[2],ymm1[3],mem[4,5,6],ymm1[7] sched: [8:0.50]
-; SANDY-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_blendps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3,4,5,6,7] sched: [1:0.33]
-; HASWELL-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],mem[2],ymm1[3],mem[4,5,6],ymm1[7] sched: [8:0.50]
-; HASWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_blendps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3,4,5,6,7] sched: [1:0.33]
-; BROADWELL-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],mem[2],ymm1[3],mem[4,5,6],ymm1[7] sched: [7:0.50]
-; BROADWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_blendps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3,4,5,6,7] sched: [1:0.33]
-; SKYLAKE-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],mem[2],ymm1[3],mem[4,5,6],ymm1[7] sched: [8:0.50]
-; SKYLAKE-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_blendps:
-; SKX: # %bb.0:
-; SKX-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3,4,5,6,7] sched: [1:0.33]
-; SKX-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],mem[2],ymm1[3],mem[4,5,6],ymm1[7] sched: [8:0.50]
-; SKX-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_blendps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3,4,5,6,7] sched: [2:1.00]
-; BDVER2-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],mem[2],ymm1[3],mem[4,5,6],ymm1[7] sched: [7:1.00]
-; BDVER2-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_blendps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3,4,5,6,7] sched: [1:1.00]
-; BTVER2-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],mem[2],ymm1[3],mem[4,5,6],ymm1[7] sched: [6:2.00]
-; BTVER2-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_blendps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2],ymm0[3,4,5,6,7] sched: [1:0.50]
-; ZNVER1-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],mem[2],ymm1[3],mem[4,5,6],ymm1[7] sched: [8:0.50]
-; ZNVER1-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <8 x float> %a0, <8 x float> %a1, <8 x i32> <i32 0, i32 9, i32 10, i32 3, i32 4, i32 5, i32 6, i32 7>
- %2 = load <8 x float>, <8 x float> *%a2, align 32
- %3 = shufflevector <8 x float> %a1, <8 x float> %2, <8 x i32> <i32 0, i32 1, i32 10, i32 3, i32 12, i32 13, i32 14, i32 7>
- %4 = fadd <8 x float> %1, %3
- ret <8 x float> %4
-}
-
-define <4 x double> @test_blendvpd(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2, <4 x double> *%a3) {
-; GENERIC-LABEL: test_blendvpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vblendvpd %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:1.00]
-; GENERIC-NEXT: vblendvpd %ymm2, (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_blendvpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vblendvpd %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:1.00]
-; SANDY-NEXT: vblendvpd %ymm2, (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_blendvpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vblendvpd %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:2.00]
-; HASWELL-NEXT: vblendvpd %ymm2, (%rdi), %ymm0, %ymm0 # sched: [9:2.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_blendvpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vblendvpd %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:2.00]
-; BROADWELL-NEXT: vblendvpd %ymm2, (%rdi), %ymm0, %ymm0 # sched: [8:2.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_blendvpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vblendvpd %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:0.67]
-; SKYLAKE-NEXT: vblendvpd %ymm2, (%rdi), %ymm0, %ymm0 # sched: [9:0.67]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_blendvpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vblendvpd %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:0.67]
-; SKX-NEXT: vblendvpd %ymm2, (%rdi), %ymm0, %ymm0 # sched: [9:0.67]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_blendvpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vblendvpd %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:3.00]
-; BDVER2-NEXT: vblendvpd %ymm2, (%rdi), %ymm0, %ymm0 # sched: [7:3.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_blendvpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vblendvpd %ymm2, %ymm1, %ymm0, %ymm0 # sched: [3:3.00]
-; BTVER2-NEXT: vblendvpd %ymm2, (%rdi), %ymm0, %ymm0 # sched: [8:3.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_blendvpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vblendvpd %ymm2, %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; ZNVER1-NEXT: vblendvpd %ymm2, (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x double> @llvm.x86.avx.blendv.pd.256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2)
- %2 = load <4 x double>, <4 x double> *%a3, align 32
- %3 = call <4 x double> @llvm.x86.avx.blendv.pd.256(<4 x double> %1, <4 x double> %2, <4 x double> %a2)
- ret <4 x double> %3
-}
-declare <4 x double> @llvm.x86.avx.blendv.pd.256(<4 x double>, <4 x double>, <4 x double>) nounwind readnone
-
-define <8 x float> @test_blendvps(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2, <8 x float> *%a3) {
-; GENERIC-LABEL: test_blendvps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vblendvps %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:1.00]
-; GENERIC-NEXT: vblendvps %ymm2, (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_blendvps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vblendvps %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:1.00]
-; SANDY-NEXT: vblendvps %ymm2, (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_blendvps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vblendvps %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:2.00]
-; HASWELL-NEXT: vblendvps %ymm2, (%rdi), %ymm0, %ymm0 # sched: [9:2.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_blendvps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vblendvps %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:2.00]
-; BROADWELL-NEXT: vblendvps %ymm2, (%rdi), %ymm0, %ymm0 # sched: [8:2.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_blendvps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vblendvps %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:0.67]
-; SKYLAKE-NEXT: vblendvps %ymm2, (%rdi), %ymm0, %ymm0 # sched: [9:0.67]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_blendvps:
-; SKX: # %bb.0:
-; SKX-NEXT: vblendvps %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:0.67]
-; SKX-NEXT: vblendvps %ymm2, (%rdi), %ymm0, %ymm0 # sched: [9:0.67]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_blendvps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vblendvps %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:3.00]
-; BDVER2-NEXT: vblendvps %ymm2, (%rdi), %ymm0, %ymm0 # sched: [7:3.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_blendvps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vblendvps %ymm2, %ymm1, %ymm0, %ymm0 # sched: [3:3.00]
-; BTVER2-NEXT: vblendvps %ymm2, (%rdi), %ymm0, %ymm0 # sched: [8:3.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_blendvps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vblendvps %ymm2, %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; ZNVER1-NEXT: vblendvps %ymm2, (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x float> @llvm.x86.avx.blendv.ps.256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2)
- %2 = load <8 x float>, <8 x float> *%a3, align 32
- %3 = call <8 x float> @llvm.x86.avx.blendv.ps.256(<8 x float> %1, <8 x float> %2, <8 x float> %a2)
- ret <8 x float> %3
-}
-declare <8 x float> @llvm.x86.avx.blendv.ps.256(<8 x float>, <8 x float>, <8 x float>) nounwind readnone
-
-define <8 x float> @test_broadcastf128(<4 x float> *%a0) {
-; GENERIC-LABEL: test_broadcastf128:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1] sched: [7:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_broadcastf128:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1] sched: [7:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_broadcastf128:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1] sched: [7:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_broadcastf128:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1] sched: [6:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_broadcastf128:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1] sched: [7:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_broadcastf128:
-; SKX: # %bb.0:
-; SKX-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1] sched: [7:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_broadcastf128:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1] sched: [7:0.50]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_broadcastf128:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1] sched: [6:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_broadcastf128:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vbroadcastf128 {{.*#+}} ymm0 = mem[0,1,0,1] sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = load <4 x float>, <4 x float> *%a0, align 32
- %2 = shufflevector <4 x float> %1, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3>
- ret <8 x float> %2
-}
-
-define <4 x double> @test_broadcastsd_ymm(double *%a0) {
-; GENERIC-LABEL: test_broadcastsd_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vbroadcastsd (%rdi), %ymm0 # sched: [7:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_broadcastsd_ymm:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vbroadcastsd (%rdi), %ymm0 # sched: [7:0.50]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_broadcastsd_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vbroadcastsd (%rdi), %ymm0 # sched: [7:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_broadcastsd_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vbroadcastsd (%rdi), %ymm0 # sched: [6:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_broadcastsd_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vbroadcastsd (%rdi), %ymm0 # sched: [7:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_broadcastsd_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vbroadcastsd (%rdi), %ymm0 # sched: [7:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_broadcastsd_ymm:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vbroadcastsd (%rdi), %ymm0 # sched: [6:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_broadcastsd_ymm:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vbroadcastsd (%rdi), %ymm0 # sched: [6:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_broadcastsd_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vbroadcastsd (%rdi), %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = load double, double *%a0, align 8
- %2 = insertelement <4 x double> undef, double %1, i32 0
- %3 = shufflevector <4 x double> %2, <4 x double> undef, <4 x i32> zeroinitializer
- ret <4 x double> %3
-}
-
-define <4 x float> @test_broadcastss(float *%a0) {
-; GENERIC-LABEL: test_broadcastss:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vbroadcastss (%rdi), %xmm0 # sched: [6:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_broadcastss:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vbroadcastss (%rdi), %xmm0 # sched: [6:0.50]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_broadcastss:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vbroadcastss (%rdi), %xmm0 # sched: [6:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_broadcastss:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vbroadcastss (%rdi), %xmm0 # sched: [5:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_broadcastss:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vbroadcastss (%rdi), %xmm0 # sched: [6:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_broadcastss:
-; SKX: # %bb.0:
-; SKX-NEXT: vbroadcastss (%rdi), %xmm0 # sched: [6:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_broadcastss:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vbroadcastss (%rdi), %xmm0 # sched: [7:0.50]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_broadcastss:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vbroadcastss (%rdi), %xmm0 # sched: [6:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_broadcastss:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vbroadcastss (%rdi), %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = load float, float *%a0, align 4
- %2 = insertelement <4 x float> undef, float %1, i32 0
- %3 = shufflevector <4 x float> %2, <4 x float> undef, <4 x i32> zeroinitializer
- ret <4 x float> %3
-}
-
-define <8 x float> @test_broadcastss_ymm(float *%a0) {
-; GENERIC-LABEL: test_broadcastss_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vbroadcastss (%rdi), %ymm0 # sched: [7:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_broadcastss_ymm:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vbroadcastss (%rdi), %ymm0 # sched: [7:0.50]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_broadcastss_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vbroadcastss (%rdi), %ymm0 # sched: [7:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_broadcastss_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vbroadcastss (%rdi), %ymm0 # sched: [6:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_broadcastss_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vbroadcastss (%rdi), %ymm0 # sched: [7:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_broadcastss_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vbroadcastss (%rdi), %ymm0 # sched: [7:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_broadcastss_ymm:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vbroadcastss (%rdi), %ymm0 # sched: [6:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_broadcastss_ymm:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vbroadcastss (%rdi), %ymm0 # sched: [6:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_broadcastss_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vbroadcastss (%rdi), %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = load float, float *%a0, align 4
- %2 = insertelement <8 x float> undef, float %1, i32 0
- %3 = shufflevector <8 x float> %2, <8 x float> undef, <8 x i32> zeroinitializer
- ret <8 x float> %3
-}
-
-define <4 x double> @test_cmppd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
-; GENERIC-LABEL: test_cmppd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vcmpeqpd %ymm1, %ymm0, %ymm1 # sched: [3:1.00]
-; GENERIC-NEXT: vcmpeqpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; GENERIC-NEXT: vorpd %ymm0, %ymm1, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_cmppd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vcmpeqpd %ymm1, %ymm0, %ymm1 # sched: [3:1.00]
-; SANDY-NEXT: vcmpeqpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; SANDY-NEXT: vorpd %ymm0, %ymm1, %ymm0 # sched: [1:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_cmppd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vcmpeqpd %ymm1, %ymm0, %ymm1 # sched: [3:1.00]
-; HASWELL-NEXT: vcmpeqpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; HASWELL-NEXT: vorpd %ymm0, %ymm1, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_cmppd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vcmpeqpd %ymm1, %ymm0, %ymm1 # sched: [3:1.00]
-; BROADWELL-NEXT: vcmpeqpd (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; BROADWELL-NEXT: vorpd %ymm0, %ymm1, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_cmppd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vcmpeqpd %ymm1, %ymm0, %ymm1 # sched: [4:0.50]
-; SKYLAKE-NEXT: vcmpeqpd (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKYLAKE-NEXT: vorpd %ymm0, %ymm1, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_cmppd:
-; SKX: # %bb.0:
-; SKX-NEXT: vcmpeqpd %ymm1, %ymm0, %ymm1 # sched: [4:0.50]
-; SKX-NEXT: vcmpeqpd (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKX-NEXT: vorpd %ymm0, %ymm1, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_cmppd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vcmpeqpd %ymm1, %ymm0, %ymm1 # sched: [2:2.00]
-; BDVER2-NEXT: vcmpeqpd (%rdi), %ymm0, %ymm0 # sched: [7:2.00]
-; BDVER2-NEXT: vorpd %ymm0, %ymm1, %ymm0 # sched: [2:1.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_cmppd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vcmpeqpd %ymm1, %ymm0, %ymm1 # sched: [2:2.00]
-; BTVER2-NEXT: vcmpeqpd (%rdi), %ymm0, %ymm0 # sched: [7:2.00]
-; BTVER2-NEXT: vorpd %ymm0, %ymm1, %ymm0 # sched: [1:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_cmppd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vcmpeqpd %ymm1, %ymm0, %ymm1 # sched: [3:1.00]
-; ZNVER1-NEXT: vcmpeqpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT: vorpd %ymm0, %ymm1, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = fcmp oeq <4 x double> %a0, %a1
- %2 = load <4 x double>, <4 x double> *%a2, align 32
- %3 = fcmp oeq <4 x double> %a0, %2
- %4 = sext <4 x i1> %1 to <4 x i64>
- %5 = sext <4 x i1> %3 to <4 x i64>
- %6 = or <4 x i64> %4, %5
- %7 = bitcast <4 x i64> %6 to <4 x double>
- ret <4 x double> %7
-}
-
-define <8 x float> @test_cmpps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {
-; GENERIC-LABEL: test_cmpps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vcmpeqps %ymm1, %ymm0, %ymm1 # sched: [3:1.00]
-; GENERIC-NEXT: vcmpeqps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; GENERIC-NEXT: vorps %ymm0, %ymm1, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_cmpps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vcmpeqps %ymm1, %ymm0, %ymm1 # sched: [3:1.00]
-; SANDY-NEXT: vcmpeqps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; SANDY-NEXT: vorps %ymm0, %ymm1, %ymm0 # sched: [1:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_cmpps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vcmpeqps %ymm1, %ymm0, %ymm1 # sched: [3:1.00]
-; HASWELL-NEXT: vcmpeqps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; HASWELL-NEXT: vorps %ymm0, %ymm1, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_cmpps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vcmpeqps %ymm1, %ymm0, %ymm1 # sched: [3:1.00]
-; BROADWELL-NEXT: vcmpeqps (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; BROADWELL-NEXT: vorps %ymm0, %ymm1, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_cmpps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vcmpeqps %ymm1, %ymm0, %ymm1 # sched: [4:0.50]
-; SKYLAKE-NEXT: vcmpeqps (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKYLAKE-NEXT: vorps %ymm0, %ymm1, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_cmpps:
-; SKX: # %bb.0:
-; SKX-NEXT: vcmpeqps %ymm1, %ymm0, %ymm1 # sched: [4:0.50]
-; SKX-NEXT: vcmpeqps (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKX-NEXT: vorps %ymm0, %ymm1, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_cmpps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vcmpeqps %ymm1, %ymm0, %ymm1 # sched: [2:2.00]
-; BDVER2-NEXT: vcmpeqps (%rdi), %ymm0, %ymm0 # sched: [7:2.00]
-; BDVER2-NEXT: vorps %ymm0, %ymm1, %ymm0 # sched: [2:1.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_cmpps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vcmpeqps %ymm1, %ymm0, %ymm1 # sched: [2:2.00]
-; BTVER2-NEXT: vcmpeqps (%rdi), %ymm0, %ymm0 # sched: [7:2.00]
-; BTVER2-NEXT: vorps %ymm0, %ymm1, %ymm0 # sched: [1:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_cmpps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vcmpeqps %ymm1, %ymm0, %ymm1 # sched: [3:1.00]
-; ZNVER1-NEXT: vcmpeqps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT: vorps %ymm0, %ymm1, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = fcmp oeq <8 x float> %a0, %a1
- %2 = load <8 x float>, <8 x float> *%a2, align 32
- %3 = fcmp oeq <8 x float> %a0, %2
- %4 = sext <8 x i1> %1 to <8 x i32>
- %5 = sext <8 x i1> %3 to <8 x i32>
- %6 = or <8 x i32> %4, %5
- %7 = bitcast <8 x i32> %6 to <8 x float>
- ret <8 x float> %7
-}
-
-define <4 x double> @test_cvtdq2pd(<4 x i32> %a0, <4 x i32> *%a1) {
-; GENERIC-LABEL: test_cvtdq2pd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vcvtdq2pd %xmm0, %ymm0 # sched: [4:1.00]
-; GENERIC-NEXT: vcvtdq2pd (%rdi), %ymm1 # sched: [10:1.00]
-; GENERIC-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_cvtdq2pd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vcvtdq2pd %xmm0, %ymm0 # sched: [4:1.00]
-; SANDY-NEXT: vcvtdq2pd (%rdi), %ymm1 # sched: [10:1.00]
-; SANDY-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_cvtdq2pd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vcvtdq2pd %xmm0, %ymm0 # sched: [6:1.00]
-; HASWELL-NEXT: vcvtdq2pd (%rdi), %ymm1 # sched: [12:1.00]
-; HASWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_cvtdq2pd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vcvtdq2pd (%rdi), %ymm1 # sched: [11:1.00]
-; BROADWELL-NEXT: vcvtdq2pd %xmm0, %ymm0 # sched: [6:1.00]
-; BROADWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_cvtdq2pd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vcvtdq2pd %xmm0, %ymm0 # sched: [7:1.00]
-; SKYLAKE-NEXT: vcvtdq2pd (%rdi), %ymm1 # sched: [13:1.00]
-; SKYLAKE-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_cvtdq2pd:
-; SKX: # %bb.0:
-; SKX-NEXT: vcvtdq2pd %xmm0, %ymm0 # sched: [7:1.00]
-; SKX-NEXT: vcvtdq2pd (%rdi), %ymm1 # sched: [13:1.00]
-; SKX-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_cvtdq2pd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vcvtdq2pd (%rdi), %ymm1 # sched: [13:2.00]
-; BDVER2-NEXT: vcvtdq2pd %xmm0, %ymm0 # sched: [8:2.00]
-; BDVER2-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_cvtdq2pd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vcvtdq2pd (%rdi), %ymm1 # sched: [8:2.00]
-; BTVER2-NEXT: vcvtdq2pd %xmm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_cvtdq2pd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vcvtdq2pd (%rdi), %ymm1 # sched: [12:1.00]
-; ZNVER1-NEXT: vcvtdq2pd %xmm0, %ymm0 # sched: [5:1.00]
-; ZNVER1-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = sitofp <4 x i32> %a0 to <4 x double>
- %2 = load <4 x i32>, <4 x i32> *%a1, align 16
- %3 = sitofp <4 x i32> %2 to <4 x double>
- %4 = fadd <4 x double> %1, %3
- ret <4 x double> %4
-}
-
-define <8 x float> @test_cvtdq2ps(<8 x i32> %a0, <8 x i32> *%a1) {
-; GENERIC-LABEL: test_cvtdq2ps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vcvtdq2ps %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: vcvtdq2ps (%rdi), %ymm1 # sched: [10:1.00]
-; GENERIC-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_cvtdq2ps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vcvtdq2ps %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: vmovaps (%rdi), %xmm1 # sched: [6:0.50]
-; SANDY-NEXT: vinsertf128 $1, 16(%rdi), %ymm1, %ymm1 # sched: [7:0.50]
-; SANDY-NEXT: vcvtdq2ps %ymm1, %ymm1 # sched: [3:1.00]
-; SANDY-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_cvtdq2ps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vcvtdq2ps %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vcvtdq2ps (%rdi), %ymm1 # sched: [10:1.00]
-; HASWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_cvtdq2ps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vcvtdq2ps %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vcvtdq2ps (%rdi), %ymm1 # sched: [9:1.00]
-; BROADWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_cvtdq2ps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vcvtdq2ps %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vcvtdq2ps (%rdi), %ymm1 # sched: [11:0.50]
-; SKYLAKE-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_cvtdq2ps:
-; SKX: # %bb.0:
-; SKX-NEXT: vcvtdq2ps %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vcvtdq2ps (%rdi), %ymm1 # sched: [11:0.50]
-; SKX-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_cvtdq2ps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vcvtdq2ps (%rdi), %ymm1 # sched: [9:2.00]
-; BDVER2-NEXT: vcvtdq2ps %ymm0, %ymm0 # sched: [4:2.00]
-; BDVER2-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_cvtdq2ps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vcvtdq2ps (%rdi), %ymm1 # sched: [8:2.00]
-; BTVER2-NEXT: vcvtdq2ps %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_cvtdq2ps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vcvtdq2ps (%rdi), %ymm1 # sched: [12:1.00]
-; ZNVER1-NEXT: vcvtdq2ps %ymm0, %ymm0 # sched: [5:1.00]
-; ZNVER1-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = sitofp <8 x i32> %a0 to <8 x float>
- %2 = load <8 x i32>, <8 x i32> *%a1, align 16
- %3 = sitofp <8 x i32> %2 to <8 x float>
- %4 = fadd <8 x float> %1, %3
- ret <8 x float> %4
-}
-
-define <8 x i32> @test_cvtpd2dq(<4 x double> %a0, <4 x double> *%a1) {
-; GENERIC-LABEL: test_cvtpd2dq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vcvtpd2dq %ymm0, %xmm0 # sched: [4:1.00]
-; GENERIC-NEXT: vcvtpd2dqy (%rdi), %xmm1 # sched: [11:1.00]
-; GENERIC-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_cvtpd2dq:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vcvtpd2dq %ymm0, %xmm0 # sched: [4:1.00]
-; SANDY-NEXT: vcvtpd2dqy (%rdi), %xmm1 # sched: [11:1.00]
-; SANDY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_cvtpd2dq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vcvtpd2dq %ymm0, %xmm0 # sched: [6:1.00]
-; HASWELL-NEXT: vcvtpd2dqy (%rdi), %xmm1 # sched: [8:1.00]
-; HASWELL-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_cvtpd2dq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vcvtpd2dq %ymm0, %xmm0 # sched: [6:1.00]
-; BROADWELL-NEXT: vcvtpd2dqy (%rdi), %xmm1 # sched: [8:1.00]
-; BROADWELL-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_cvtpd2dq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vcvtpd2dq %ymm0, %xmm0 # sched: [7:1.00]
-; SKYLAKE-NEXT: vcvtpd2dqy (%rdi), %xmm1 # sched: [8:1.00]
-; SKYLAKE-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_cvtpd2dq:
-; SKX: # %bb.0:
-; SKX-NEXT: vcvtpd2dq %ymm0, %xmm0 # sched: [7:1.00]
-; SKX-NEXT: vcvtpd2dqy (%rdi), %xmm1 # sched: [8:0.50]
-; SKX-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_cvtpd2dq:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vcvtpd2dqy (%rdi), %xmm1 # sched: [13:2.00]
-; BDVER2-NEXT: vcvtpd2dq %ymm0, %xmm0 # sched: [8:2.00]
-; BDVER2-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [2:0.50]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_cvtpd2dq:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vcvtpd2dqy (%rdi), %xmm1 # sched: [11:2.00]
-; BTVER2-NEXT: vcvtpd2dq %ymm0, %xmm0 # sched: [6:2.00]
-; BTVER2-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_cvtpd2dq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vcvtpd2dqy (%rdi), %xmm1 # sched: [12:1.00]
-; ZNVER1-NEXT: vcvtpd2dq %ymm0, %xmm0 # sched: [5:1.00]
-; ZNVER1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [2:0.67]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x i32> @llvm.x86.avx.cvt.pd2dq.256(<4 x double> %a0)
- %2 = load <4 x double>, <4 x double> *%a1, align 32
- %3 = call <4 x i32> @llvm.x86.avx.cvt.pd2dq.256(<4 x double> %2)
- %4 = shufflevector <4 x i32> %1, <4 x i32> %3, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
- ret <8 x i32> %4
-}
-declare <4 x i32> @llvm.x86.avx.cvt.pd2dq.256(<4 x double>) nounwind readnone
-
-define <8 x i32> @test_cvttpd2dq(<4 x double> %a0, <4 x double> *%a1) {
-; GENERIC-LABEL: test_cvttpd2dq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vcvttpd2dq %ymm0, %xmm0 # sched: [4:1.00]
-; GENERIC-NEXT: vcvttpd2dqy (%rdi), %xmm1 # sched: [11:1.00]
-; GENERIC-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_cvttpd2dq:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vcvttpd2dq %ymm0, %xmm0 # sched: [4:1.00]
-; SANDY-NEXT: vcvttpd2dqy (%rdi), %xmm1 # sched: [11:1.00]
-; SANDY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_cvttpd2dq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vcvttpd2dq %ymm0, %xmm0 # sched: [6:1.00]
-; HASWELL-NEXT: vcvttpd2dqy (%rdi), %xmm1 # sched: [8:1.00]
-; HASWELL-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_cvttpd2dq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vcvttpd2dq %ymm0, %xmm0 # sched: [6:1.00]
-; BROADWELL-NEXT: vcvttpd2dqy (%rdi), %xmm1 # sched: [8:1.00]
-; BROADWELL-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_cvttpd2dq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vcvttpd2dq %ymm0, %xmm0 # sched: [7:1.00]
-; SKYLAKE-NEXT: vcvttpd2dqy (%rdi), %xmm1 # sched: [8:1.00]
-; SKYLAKE-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_cvttpd2dq:
-; SKX: # %bb.0:
-; SKX-NEXT: vcvttpd2dq %ymm0, %xmm0 # sched: [7:1.00]
-; SKX-NEXT: vcvttpd2dqy (%rdi), %xmm1 # sched: [8:0.50]
-; SKX-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_cvttpd2dq:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vcvttpd2dqy (%rdi), %xmm1 # sched: [13:2.00]
-; BDVER2-NEXT: vcvttpd2dq %ymm0, %xmm0 # sched: [8:2.00]
-; BDVER2-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [2:0.50]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_cvttpd2dq:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vcvttpd2dqy (%rdi), %xmm1 # sched: [11:2.00]
-; BTVER2-NEXT: vcvttpd2dq %ymm0, %xmm0 # sched: [6:2.00]
-; BTVER2-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_cvttpd2dq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vcvttpd2dqy (%rdi), %xmm1 # sched: [12:1.00]
-; ZNVER1-NEXT: vcvttpd2dq %ymm0, %xmm0 # sched: [5:1.00]
-; ZNVER1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [2:0.67]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = fptosi <4 x double> %a0 to <4 x i32>
- %2 = load <4 x double>, <4 x double> *%a1, align 32
- %3 = fptosi <4 x double> %2 to <4 x i32>
- %4 = shufflevector <4 x i32> %1, <4 x i32> %3, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
- ret <8 x i32> %4
-}
-
-define <8 x float> @test_cvtpd2ps(<4 x double> %a0, <4 x double> *%a1) {
-; GENERIC-LABEL: test_cvtpd2ps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vcvtpd2ps %ymm0, %xmm0 # sched: [4:1.00]
-; GENERIC-NEXT: vcvtpd2psy (%rdi), %xmm1 # sched: [11:1.00]
-; GENERIC-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_cvtpd2ps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vcvtpd2ps %ymm0, %xmm0 # sched: [4:1.00]
-; SANDY-NEXT: vcvtpd2psy (%rdi), %xmm1 # sched: [11:1.00]
-; SANDY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_cvtpd2ps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vcvtpd2ps %ymm0, %xmm0 # sched: [6:1.00]
-; HASWELL-NEXT: vcvtpd2psy (%rdi), %xmm1 # sched: [8:1.00]
-; HASWELL-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_cvtpd2ps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vcvtpd2ps %ymm0, %xmm0 # sched: [6:1.00]
-; BROADWELL-NEXT: vcvtpd2psy (%rdi), %xmm1 # sched: [8:1.00]
-; BROADWELL-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_cvtpd2ps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vcvtpd2ps %ymm0, %xmm0 # sched: [7:1.00]
-; SKYLAKE-NEXT: vcvtpd2psy (%rdi), %xmm1 # sched: [8:1.00]
-; SKYLAKE-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_cvtpd2ps:
-; SKX: # %bb.0:
-; SKX-NEXT: vcvtpd2ps %ymm0, %xmm0 # sched: [7:1.00]
-; SKX-NEXT: vcvtpd2psy (%rdi), %xmm1 # sched: [8:1.00]
-; SKX-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_cvtpd2ps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vcvtpd2psy (%rdi), %xmm1 # sched: [13:2.00]
-; BDVER2-NEXT: vcvtpd2ps %ymm0, %xmm0 # sched: [8:2.00]
-; BDVER2-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [2:0.50]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_cvtpd2ps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vcvtpd2psy (%rdi), %xmm1 # sched: [11:2.00]
-; BTVER2-NEXT: vcvtpd2ps %ymm0, %xmm0 # sched: [6:2.00]
-; BTVER2-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_cvtpd2ps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vcvtpd2psy (%rdi), %xmm1 # sched: [11:1.00]
-; ZNVER1-NEXT: vcvtpd2ps %ymm0, %xmm0 # sched: [5:1.00]
-; ZNVER1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 # sched: [2:0.67]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = fptrunc <4 x double> %a0 to <4 x float>
- %2 = load <4 x double>, <4 x double> *%a1, align 32
- %3 = fptrunc <4 x double> %2 to <4 x float>
- %4 = shufflevector <4 x float> %1, <4 x float> %3, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
- ret <8 x float> %4
-}
-
-define <8 x i32> @test_cvtps2dq(<8 x float> %a0, <8 x float> *%a1) {
-; GENERIC-LABEL: test_cvtps2dq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vcvtps2dq %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: vcvtps2dq (%rdi), %ymm1 # sched: [10:1.00]
-; GENERIC-NEXT: vorpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_cvtps2dq:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vcvtps2dq %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: vcvtps2dq (%rdi), %ymm1 # sched: [10:1.00]
-; SANDY-NEXT: vorpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_cvtps2dq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vcvtps2dq %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vcvtps2dq (%rdi), %ymm1 # sched: [10:1.00]
-; HASWELL-NEXT: vorpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_cvtps2dq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vcvtps2dq %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vcvtps2dq (%rdi), %ymm1 # sched: [9:1.00]
-; BROADWELL-NEXT: vorpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_cvtps2dq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vcvtps2dq %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vcvtps2dq (%rdi), %ymm1 # sched: [11:0.50]
-; SKYLAKE-NEXT: vorpd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_cvtps2dq:
-; SKX: # %bb.0:
-; SKX-NEXT: vcvtps2dq %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vcvtps2dq (%rdi), %ymm1 # sched: [11:0.50]
-; SKX-NEXT: vorpd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_cvtps2dq:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vcvtps2dq (%rdi), %ymm1 # sched: [9:2.00]
-; BDVER2-NEXT: vcvtps2dq %ymm0, %ymm0 # sched: [4:2.00]
-; BDVER2-NEXT: vorpd %ymm1, %ymm0, %ymm0 # sched: [2:1.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_cvtps2dq:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vcvtps2dq (%rdi), %ymm1 # sched: [8:2.00]
-; BTVER2-NEXT: vcvtps2dq %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: vorpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_cvtps2dq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vcvtps2dq (%rdi), %ymm1 # sched: [12:1.00]
-; ZNVER1-NEXT: vcvtps2dq %ymm0, %ymm0 # sched: [5:1.00]
-; ZNVER1-NEXT: vorpd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x i32> @llvm.x86.avx.cvt.ps2dq.256(<8 x float> %a0)
- %2 = load <8 x float>, <8 x float> *%a1, align 32
- %3 = call <8 x i32> @llvm.x86.avx.cvt.ps2dq.256(<8 x float> %2)
- %4 = or <8 x i32> %1, %3
- ret <8 x i32> %4
-}
-declare <8 x i32> @llvm.x86.avx.cvt.ps2dq.256(<8 x float>) nounwind readnone
-
-define <8 x i32> @test_cvttps2dq(<8 x float> %a0, <8 x float> *%a1) {
-; GENERIC-LABEL: test_cvttps2dq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vcvttps2dq %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: vcvttps2dq (%rdi), %ymm1 # sched: [10:1.00]
-; GENERIC-NEXT: vorps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_cvttps2dq:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vcvttps2dq %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: vcvttps2dq (%rdi), %ymm1 # sched: [10:1.00]
-; SANDY-NEXT: vorps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_cvttps2dq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vcvttps2dq %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vcvttps2dq (%rdi), %ymm1 # sched: [10:1.00]
-; HASWELL-NEXT: vorps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_cvttps2dq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vcvttps2dq %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vcvttps2dq (%rdi), %ymm1 # sched: [9:1.00]
-; BROADWELL-NEXT: vorps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_cvttps2dq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vcvttps2dq %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vcvttps2dq (%rdi), %ymm1 # sched: [11:0.50]
-; SKYLAKE-NEXT: vorps %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_cvttps2dq:
-; SKX: # %bb.0:
-; SKX-NEXT: vcvttps2dq %ymm0, %ymm0 # sched: [3:0.50]
-; SKX-NEXT: vcvttps2dq (%rdi), %ymm1 # sched: [11:0.50]
-; SKX-NEXT: vorps %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_cvttps2dq:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vcvttps2dq (%rdi), %ymm1 # sched: [9:2.00]
-; BDVER2-NEXT: vcvttps2dq %ymm0, %ymm0 # sched: [4:2.00]
-; BDVER2-NEXT: vorps %ymm1, %ymm0, %ymm0 # sched: [2:1.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_cvttps2dq:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vcvttps2dq (%rdi), %ymm1 # sched: [8:2.00]
-; BTVER2-NEXT: vcvttps2dq %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: vorps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_cvttps2dq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vcvttps2dq (%rdi), %ymm1 # sched: [12:1.00]
-; ZNVER1-NEXT: vcvttps2dq %ymm0, %ymm0 # sched: [5:1.00]
-; ZNVER1-NEXT: vorps %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = fptosi <8 x float> %a0 to <8 x i32>
- %2 = load <8 x float>, <8 x float> *%a1, align 32
- %3 = fptosi <8 x float> %2 to <8 x i32>
- %4 = or <8 x i32> %1, %3
- ret <8 x i32> %4
-}
-
-define <4 x double> @test_divpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
-; GENERIC-LABEL: test_divpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vdivpd %ymm1, %ymm0, %ymm0 # sched: [45:44.00]
-; GENERIC-NEXT: vdivpd (%rdi), %ymm0, %ymm0 # sched: [52:44.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_divpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vdivpd %ymm1, %ymm0, %ymm0 # sched: [45:44.00]
-; SANDY-NEXT: vdivpd (%rdi), %ymm0, %ymm0 # sched: [52:44.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_divpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vdivpd %ymm1, %ymm0, %ymm0 # sched: [35:28.00]
-; HASWELL-NEXT: vdivpd (%rdi), %ymm0, %ymm0 # sched: [42:28.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_divpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vdivpd %ymm1, %ymm0, %ymm0 # sched: [23:16.00]
-; BROADWELL-NEXT: vdivpd (%rdi), %ymm0, %ymm0 # sched: [29:16.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_divpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vdivpd %ymm1, %ymm0, %ymm0 # sched: [14:5.00]
-; SKYLAKE-NEXT: vdivpd (%rdi), %ymm0, %ymm0 # sched: [21:8.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_divpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vdivpd %ymm1, %ymm0, %ymm0 # sched: [14:5.00]
-; SKX-NEXT: vdivpd (%rdi), %ymm0, %ymm0 # sched: [21:8.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_divpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vdivpd %ymm1, %ymm0, %ymm0 # sched: [9:19.00]
-; BDVER2-NEXT: vdivpd (%rdi), %ymm0, %ymm0 # sched: [14:19.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_divpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vdivpd %ymm1, %ymm0, %ymm0 # sched: [38:38.00]
-; BTVER2-NEXT: vdivpd (%rdi), %ymm0, %ymm0 # sched: [43:38.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_divpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vdivpd %ymm1, %ymm0, %ymm0 # sched: [15:15.00]
-; ZNVER1-NEXT: vdivpd (%rdi), %ymm0, %ymm0 # sched: [22:22.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = fdiv <4 x double> %a0, %a1
- %2 = load <4 x double>, <4 x double> *%a2, align 32
- %3 = fdiv <4 x double> %1, %2
- ret <4 x double> %3
-}
-
-define <8 x float> @test_divps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {
-; GENERIC-LABEL: test_divps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vdivps %ymm1, %ymm0, %ymm0 # sched: [29:28.00]
-; GENERIC-NEXT: vdivps (%rdi), %ymm0, %ymm0 # sched: [36:28.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_divps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vdivps %ymm1, %ymm0, %ymm0 # sched: [29:28.00]
-; SANDY-NEXT: vdivps (%rdi), %ymm0, %ymm0 # sched: [36:28.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_divps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vdivps %ymm1, %ymm0, %ymm0 # sched: [21:14.00]
-; HASWELL-NEXT: vdivps (%rdi), %ymm0, %ymm0 # sched: [28:14.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_divps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vdivps %ymm1, %ymm0, %ymm0 # sched: [17:10.00]
-; BROADWELL-NEXT: vdivps (%rdi), %ymm0, %ymm0 # sched: [23:10.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_divps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vdivps %ymm1, %ymm0, %ymm0 # sched: [11:5.00]
-; SKYLAKE-NEXT: vdivps (%rdi), %ymm0, %ymm0 # sched: [18:5.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_divps:
-; SKX: # %bb.0:
-; SKX-NEXT: vdivps %ymm1, %ymm0, %ymm0 # sched: [11:5.00]
-; SKX-NEXT: vdivps (%rdi), %ymm0, %ymm0 # sched: [18:5.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_divps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vdivps %ymm1, %ymm0, %ymm0 # sched: [9:19.00]
-; BDVER2-NEXT: vdivps (%rdi), %ymm0, %ymm0 # sched: [14:19.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_divps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vdivps %ymm1, %ymm0, %ymm0 # sched: [38:38.00]
-; BTVER2-NEXT: vdivps (%rdi), %ymm0, %ymm0 # sched: [43:38.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_divps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vdivps %ymm1, %ymm0, %ymm0 # sched: [12:12.00]
-; ZNVER1-NEXT: vdivps (%rdi), %ymm0, %ymm0 # sched: [19:19.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = fdiv <8 x float> %a0, %a1
- %2 = load <8 x float>, <8 x float> *%a2, align 32
- %3 = fdiv <8 x float> %1, %2
- ret <8 x float> %3
-}
-
-define <8 x float> @test_dpps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {
-; GENERIC-LABEL: test_dpps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vdpps $7, %ymm1, %ymm0, %ymm0 # sched: [12:2.00]
-; GENERIC-NEXT: vdpps $7, (%rdi), %ymm0, %ymm0 # sched: [19:2.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_dpps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vdpps $7, %ymm1, %ymm0, %ymm0 # sched: [12:2.00]
-; SANDY-NEXT: vdpps $7, (%rdi), %ymm0, %ymm0 # sched: [19:2.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_dpps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vdpps $7, %ymm1, %ymm0, %ymm0 # sched: [14:2.00]
-; HASWELL-NEXT: vdpps $7, (%rdi), %ymm0, %ymm0 # sched: [21:2.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_dpps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vdpps $7, %ymm1, %ymm0, %ymm0 # sched: [14:2.00]
-; BROADWELL-NEXT: vdpps $7, (%rdi), %ymm0, %ymm0 # sched: [20:2.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_dpps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vdpps $7, %ymm1, %ymm0, %ymm0 # sched: [13:1.50]
-; SKYLAKE-NEXT: vdpps $7, (%rdi), %ymm0, %ymm0 # sched: [20:1.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_dpps:
-; SKX: # %bb.0:
-; SKX-NEXT: vdpps $7, %ymm1, %ymm0, %ymm0 # sched: [13:1.33]
-; SKX-NEXT: vdpps $7, (%rdi), %ymm0, %ymm0 # sched: [20:1.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_dpps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vdpps $7, %ymm1, %ymm0, %ymm0 # sched: [27:3.00]
-; BDVER2-NEXT: vdpps $7, (%rdi), %ymm0, %ymm0 # sched: [32:3.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_dpps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vdpps $7, %ymm1, %ymm0, %ymm0 # sched: [12:6.00]
-; BTVER2-NEXT: vdpps $7, (%rdi), %ymm0, %ymm0 # sched: [17:6.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_dpps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vdpps $7, %ymm1, %ymm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: vdpps $7, (%rdi), %ymm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x float> @llvm.x86.avx.dp.ps.256(<8 x float> %a0, <8 x float> %a1, i8 7)
- %2 = load <8 x float>, <8 x float> *%a2, align 32
- %3 = call <8 x float> @llvm.x86.avx.dp.ps.256(<8 x float> %1, <8 x float> %2, i8 7)
- ret <8 x float> %3
-}
-declare <8 x float> @llvm.x86.avx.dp.ps.256(<8 x float>, <8 x float>, i8) nounwind readnone
-
-define <4 x float> @test_extractf128(<8 x float> %a0, <8 x float> %a1, <4 x float> *%a2) {
-; GENERIC-LABEL: test_extractf128:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vextractf128 $1, %ymm0, %xmm0 # sched: [1:1.00]
-; GENERIC-NEXT: vextractf128 $1, %ymm1, (%rdi) # sched: [1:1.00]
-; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_extractf128:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vextractf128 $1, %ymm0, %xmm0 # sched: [1:1.00]
-; SANDY-NEXT: vextractf128 $1, %ymm1, (%rdi) # sched: [1:1.00]
-; SANDY-NEXT: vzeroupper # sched: [1:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_extractf128:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vextractf128 $1, %ymm0, %xmm0 # sched: [3:1.00]
-; HASWELL-NEXT: vextractf128 $1, %ymm1, (%rdi) # sched: [1:1.00]
-; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_extractf128:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vextractf128 $1, %ymm0, %xmm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vextractf128 $1, %ymm1, (%rdi) # sched: [1:1.00]
-; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_extractf128:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vextractf128 $1, %ymm0, %xmm0 # sched: [3:1.00]
-; SKYLAKE-NEXT: vextractf128 $1, %ymm1, (%rdi) # sched: [1:1.00]
-; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_extractf128:
-; SKX: # %bb.0:
-; SKX-NEXT: vextractf128 $1, %ymm0, %xmm0 # sched: [3:1.00]
-; SKX-NEXT: vextractf128 $1, %ymm1, (%rdi) # sched: [1:1.00]
-; SKX-NEXT: vzeroupper # sched: [0:0.67]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_extractf128:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vextractf128 $1, %ymm0, %xmm0 # sched: [2:0.50]
-; BDVER2-NEXT: vextractf128 $1, %ymm1, (%rdi) # sched: [7:0.50]
-; BDVER2-NEXT: vzeroupper # sched: [46:4.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_extractf128:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vextractf128 $1, %ymm0, %xmm0 # sched: [1:0.50]
-; BTVER2-NEXT: vextractf128 $1, %ymm1, (%rdi) # sched: [1:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_extractf128:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vextractf128 $1, %ymm0, %xmm0 # sched: [1:0.33]
-; ZNVER1-NEXT: vextractf128 $1, %ymm1, (%rdi) # sched: [8:0.50]
-; ZNVER1-NEXT: vzeroupper # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <8 x float> %a0, <8 x float> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
- %2 = shufflevector <8 x float> %a1, <8 x float> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
- store <4 x float> %2, <4 x float> *%a2
- ret <4 x float> %1
-}
-
-define <4 x double> @test_haddpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
-; GENERIC-LABEL: test_haddpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vhaddpd %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; GENERIC-NEXT: vhaddpd (%rdi), %ymm0, %ymm0 # sched: [12:2.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_haddpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vhaddpd %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; SANDY-NEXT: vhaddpd (%rdi), %ymm0, %ymm0 # sched: [12:2.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_haddpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vhaddpd %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; HASWELL-NEXT: vhaddpd (%rdi), %ymm0, %ymm0 # sched: [12:2.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_haddpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vhaddpd %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BROADWELL-NEXT: vhaddpd (%rdi), %ymm0, %ymm0 # sched: [11:2.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_haddpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vhaddpd %ymm1, %ymm0, %ymm0 # sched: [6:2.00]
-; SKYLAKE-NEXT: vhaddpd (%rdi), %ymm0, %ymm0 # sched: [13:2.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_haddpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vhaddpd %ymm1, %ymm0, %ymm0 # sched: [6:2.00]
-; SKX-NEXT: vhaddpd (%rdi), %ymm0, %ymm0 # sched: [13:2.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_haddpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vhaddpd %ymm1, %ymm0, %ymm0 # sched: [11:2.00]
-; BDVER2-NEXT: vhaddpd (%rdi), %ymm0, %ymm0 # sched: [16:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_haddpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vhaddpd %ymm1, %ymm0, %ymm0 # sched: [4:2.00]
-; BTVER2-NEXT: vhaddpd (%rdi), %ymm0, %ymm0 # sched: [9:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_haddpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vhaddpd %ymm1, %ymm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: vhaddpd (%rdi), %ymm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x double> @llvm.x86.avx.hadd.pd.256(<4 x double> %a0, <4 x double> %a1)
- %2 = load <4 x double>, <4 x double> *%a2, align 32
- %3 = call <4 x double> @llvm.x86.avx.hadd.pd.256(<4 x double> %1, <4 x double> %2)
- ret <4 x double> %3
-}
-declare <4 x double> @llvm.x86.avx.hadd.pd.256(<4 x double>, <4 x double>) nounwind readnone
-
-define <8 x float> @test_haddps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {
-; GENERIC-LABEL: test_haddps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vhaddps %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; GENERIC-NEXT: vhaddps (%rdi), %ymm0, %ymm0 # sched: [12:2.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_haddps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vhaddps %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; SANDY-NEXT: vhaddps (%rdi), %ymm0, %ymm0 # sched: [12:2.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_haddps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vhaddps %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; HASWELL-NEXT: vhaddps (%rdi), %ymm0, %ymm0 # sched: [12:2.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_haddps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vhaddps %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BROADWELL-NEXT: vhaddps (%rdi), %ymm0, %ymm0 # sched: [11:2.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_haddps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vhaddps %ymm1, %ymm0, %ymm0 # sched: [6:2.00]
-; SKYLAKE-NEXT: vhaddps (%rdi), %ymm0, %ymm0 # sched: [13:2.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_haddps:
-; SKX: # %bb.0:
-; SKX-NEXT: vhaddps %ymm1, %ymm0, %ymm0 # sched: [6:2.00]
-; SKX-NEXT: vhaddps (%rdi), %ymm0, %ymm0 # sched: [13:2.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_haddps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vhaddps %ymm1, %ymm0, %ymm0 # sched: [11:2.00]
-; BDVER2-NEXT: vhaddps (%rdi), %ymm0, %ymm0 # sched: [16:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_haddps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vhaddps %ymm1, %ymm0, %ymm0 # sched: [4:2.00]
-; BTVER2-NEXT: vhaddps (%rdi), %ymm0, %ymm0 # sched: [9:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_haddps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vhaddps %ymm1, %ymm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: vhaddps (%rdi), %ymm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x float> @llvm.x86.avx.hadd.ps.256(<8 x float> %a0, <8 x float> %a1)
- %2 = load <8 x float>, <8 x float> *%a2, align 32
- %3 = call <8 x float> @llvm.x86.avx.hadd.ps.256(<8 x float> %1, <8 x float> %2)
- ret <8 x float> %3
-}
-declare <8 x float> @llvm.x86.avx.hadd.ps.256(<8 x float>, <8 x float>) nounwind readnone
-
-define <4 x double> @test_hsubpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
-; GENERIC-LABEL: test_hsubpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vhsubpd %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; GENERIC-NEXT: vhsubpd (%rdi), %ymm0, %ymm0 # sched: [12:2.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_hsubpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vhsubpd %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; SANDY-NEXT: vhsubpd (%rdi), %ymm0, %ymm0 # sched: [12:2.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_hsubpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vhsubpd %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; HASWELL-NEXT: vhsubpd (%rdi), %ymm0, %ymm0 # sched: [12:2.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_hsubpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vhsubpd %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BROADWELL-NEXT: vhsubpd (%rdi), %ymm0, %ymm0 # sched: [11:2.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_hsubpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vhsubpd %ymm1, %ymm0, %ymm0 # sched: [6:2.00]
-; SKYLAKE-NEXT: vhsubpd (%rdi), %ymm0, %ymm0 # sched: [13:2.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_hsubpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vhsubpd %ymm1, %ymm0, %ymm0 # sched: [6:2.00]
-; SKX-NEXT: vhsubpd (%rdi), %ymm0, %ymm0 # sched: [13:2.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_hsubpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vhsubpd %ymm1, %ymm0, %ymm0 # sched: [11:2.00]
-; BDVER2-NEXT: vhsubpd (%rdi), %ymm0, %ymm0 # sched: [16:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_hsubpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vhsubpd %ymm1, %ymm0, %ymm0 # sched: [4:2.00]
-; BTVER2-NEXT: vhsubpd (%rdi), %ymm0, %ymm0 # sched: [9:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_hsubpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vhsubpd %ymm1, %ymm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: vhsubpd (%rdi), %ymm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x double> @llvm.x86.avx.hsub.pd.256(<4 x double> %a0, <4 x double> %a1)
- %2 = load <4 x double>, <4 x double> *%a2, align 32
- %3 = call <4 x double> @llvm.x86.avx.hsub.pd.256(<4 x double> %1, <4 x double> %2)
- ret <4 x double> %3
-}
-declare <4 x double> @llvm.x86.avx.hsub.pd.256(<4 x double>, <4 x double>) nounwind readnone
-
-define <8 x float> @test_hsubps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {
-; GENERIC-LABEL: test_hsubps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vhsubps %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; GENERIC-NEXT: vhsubps (%rdi), %ymm0, %ymm0 # sched: [12:2.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_hsubps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vhsubps %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; SANDY-NEXT: vhsubps (%rdi), %ymm0, %ymm0 # sched: [12:2.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_hsubps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vhsubps %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; HASWELL-NEXT: vhsubps (%rdi), %ymm0, %ymm0 # sched: [12:2.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_hsubps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vhsubps %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BROADWELL-NEXT: vhsubps (%rdi), %ymm0, %ymm0 # sched: [11:2.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_hsubps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vhsubps %ymm1, %ymm0, %ymm0 # sched: [6:2.00]
-; SKYLAKE-NEXT: vhsubps (%rdi), %ymm0, %ymm0 # sched: [13:2.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_hsubps:
-; SKX: # %bb.0:
-; SKX-NEXT: vhsubps %ymm1, %ymm0, %ymm0 # sched: [6:2.00]
-; SKX-NEXT: vhsubps (%rdi), %ymm0, %ymm0 # sched: [13:2.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_hsubps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vhsubps %ymm1, %ymm0, %ymm0 # sched: [11:2.00]
-; BDVER2-NEXT: vhsubps (%rdi), %ymm0, %ymm0 # sched: [16:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_hsubps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vhsubps %ymm1, %ymm0, %ymm0 # sched: [4:2.00]
-; BTVER2-NEXT: vhsubps (%rdi), %ymm0, %ymm0 # sched: [9:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_hsubps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vhsubps %ymm1, %ymm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: vhsubps (%rdi), %ymm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x float> @llvm.x86.avx.hsub.ps.256(<8 x float> %a0, <8 x float> %a1)
- %2 = load <8 x float>, <8 x float> *%a2, align 32
- %3 = call <8 x float> @llvm.x86.avx.hsub.ps.256(<8 x float> %1, <8 x float> %2)
- ret <8 x float> %3
-}
-declare <8 x float> @llvm.x86.avx.hsub.ps.256(<8 x float>, <8 x float>) nounwind readnone
-
-define <8 x float> @test_insertf128(<8 x float> %a0, <4 x float> %a1, <4 x float> *%a2) {
-; GENERIC-LABEL: test_insertf128:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 # sched: [1:1.00]
-; GENERIC-NEXT: vinsertf128 $1, (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; GENERIC-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_insertf128:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 # sched: [1:1.00]
-; SANDY-NEXT: vinsertf128 $1, (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; SANDY-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_insertf128:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 # sched: [3:1.00]
-; HASWELL-NEXT: vinsertf128 $1, (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; HASWELL-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_insertf128:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 # sched: [3:1.00]
-; BROADWELL-NEXT: vinsertf128 $1, (%rdi), %ymm0, %ymm0 # sched: [6:0.50]
-; BROADWELL-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_insertf128:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 # sched: [3:1.00]
-; SKYLAKE-NEXT: vinsertf128 $1, (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; SKYLAKE-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_insertf128:
-; SKX: # %bb.0:
-; SKX-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 # sched: [3:1.00]
-; SKX-NEXT: vinsertf128 $1, (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; SKX-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_insertf128:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 # sched: [2:0.50]
-; BDVER2-NEXT: vinsertf128 $1, (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BDVER2-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_insertf128:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 # sched: [1:1.00]
-; BTVER2-NEXT: vinsertf128 $1, (%rdi), %ymm0, %ymm0 # sched: [6:1.00]
-; BTVER2-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_insertf128:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 # sched: [2:0.67]
-; ZNVER1-NEXT: vinsertf128 $1, (%rdi), %ymm0, %ymm0 # sched: [9:0.67]
-; ZNVER1-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <4 x float> %a1, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
- %2 = shufflevector <8 x float> %a0, <8 x float> %1, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
- %3 = load <4 x float>, <4 x float> *%a2, align 16
- %4 = shufflevector <4 x float> %3, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
- %5 = shufflevector <8 x float> %a0, <8 x float> %4, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
- %6 = fadd <8 x float> %2, %5
- ret <8 x float> %6
-}
-
-define <32 x i8> @test_lddqu(i8* %a0) {
-; GENERIC-LABEL: test_lddqu:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vlddqu (%rdi), %ymm0 # sched: [7:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_lddqu:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vlddqu (%rdi), %ymm0 # sched: [7:0.50]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_lddqu:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vlddqu (%rdi), %ymm0 # sched: [7:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_lddqu:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vlddqu (%rdi), %ymm0 # sched: [6:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_lddqu:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vlddqu (%rdi), %ymm0 # sched: [7:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_lddqu:
-; SKX: # %bb.0:
-; SKX-NEXT: vlddqu (%rdi), %ymm0 # sched: [7:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_lddqu:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vlddqu (%rdi), %ymm0 # sched: [5:0.50]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_lddqu:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vlddqu (%rdi), %ymm0 # sched: [5:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_lddqu:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vlddqu (%rdi), %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <32 x i8> @llvm.x86.avx.ldu.dq.256(i8* %a0)
- ret <32 x i8> %1
-}
-declare <32 x i8> @llvm.x86.avx.ldu.dq.256(i8*) nounwind readonly
-
-define <2 x double> @test_maskmovpd(i8* %a0, <2 x i64> %a1, <2 x double> %a2) {
-; GENERIC-LABEL: test_maskmovpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vmaskmovpd (%rdi), %xmm0, %xmm2 # sched: [8:1.00]
-; GENERIC-NEXT: vmaskmovpd %xmm1, %xmm0, (%rdi) # sched: [5:1.00]
-; GENERIC-NEXT: vmovapd %xmm2, %xmm0 # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_maskmovpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vmaskmovpd (%rdi), %xmm0, %xmm2 # sched: [8:1.00]
-; SANDY-NEXT: vmaskmovpd %xmm1, %xmm0, (%rdi) # sched: [5:1.00]
-; SANDY-NEXT: vmovapd %xmm2, %xmm0 # sched: [1:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_maskmovpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vmaskmovpd (%rdi), %xmm0, %xmm2 # sched: [8:2.00]
-; HASWELL-NEXT: vmaskmovpd %xmm1, %xmm0, (%rdi) # sched: [5:1.00]
-; HASWELL-NEXT: vmovapd %xmm2, %xmm0 # sched: [1:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_maskmovpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vmaskmovpd (%rdi), %xmm0, %xmm2 # sched: [7:2.00]
-; BROADWELL-NEXT: vmaskmovpd %xmm1, %xmm0, (%rdi) # sched: [5:1.00]
-; BROADWELL-NEXT: vmovapd %xmm2, %xmm0 # sched: [1:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_maskmovpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vmaskmovpd (%rdi), %xmm0, %xmm2 # sched: [7:0.50]
-; SKYLAKE-NEXT: vmaskmovpd %xmm1, %xmm0, (%rdi) # sched: [2:1.00]
-; SKYLAKE-NEXT: vmovapd %xmm2, %xmm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_maskmovpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vmaskmovpd (%rdi), %xmm0, %xmm2 # sched: [7:0.50]
-; SKX-NEXT: vmaskmovpd %xmm1, %xmm0, (%rdi) # sched: [2:1.00]
-; SKX-NEXT: vmovapd %xmm2, %xmm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_maskmovpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vmaskmovpd (%rdi), %xmm0, %xmm2 # sched: [6:1.00]
-; BDVER2-NEXT: vmaskmovpd %xmm1, %xmm0, (%rdi) # sched: [6:2.00]
-; BDVER2-NEXT: vmovapd %xmm2, %xmm0 # sched: [1:0.50]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_maskmovpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vmaskmovpd (%rdi), %xmm0, %xmm2 # sched: [6:1.00]
-; BTVER2-NEXT: vmaskmovpd %xmm1, %xmm0, (%rdi) # sched: [6:2.00]
-; BTVER2-NEXT: vmovapd %xmm2, %xmm0 # sched: [1:0.50]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_maskmovpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vmaskmovpd (%rdi), %xmm0, %xmm2 # sched: [8:0.50]
-; ZNVER1-NEXT: vmaskmovpd %xmm1, %xmm0, (%rdi) # sched: [4:0.50]
-; ZNVER1-NEXT: vmovapd %xmm2, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <2 x double> @llvm.x86.avx.maskload.pd(i8* %a0, <2 x i64> %a1)
- call void @llvm.x86.avx.maskstore.pd(i8* %a0, <2 x i64> %a1, <2 x double> %a2)
- ret <2 x double> %1
-}
-declare <2 x double> @llvm.x86.avx.maskload.pd(i8*, <2 x i64>) nounwind readonly
-declare void @llvm.x86.avx.maskstore.pd(i8*, <2 x i64>, <2 x double>) nounwind
-
-define <4 x double> @test_maskmovpd_ymm(i8* %a0, <4 x i64> %a1, <4 x double> %a2) {
-; GENERIC-LABEL: test_maskmovpd_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vmaskmovpd (%rdi), %ymm0, %ymm2 # sched: [9:1.00]
-; GENERIC-NEXT: vmaskmovpd %ymm1, %ymm0, (%rdi) # sched: [5:1.00]
-; GENERIC-NEXT: vmovapd %ymm2, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_maskmovpd_ymm:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vmaskmovpd (%rdi), %ymm0, %ymm2 # sched: [9:1.00]
-; SANDY-NEXT: vmaskmovpd %ymm1, %ymm0, (%rdi) # sched: [5:1.00]
-; SANDY-NEXT: vmovapd %ymm2, %ymm0 # sched: [1:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_maskmovpd_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vmaskmovpd (%rdi), %ymm0, %ymm2 # sched: [9:2.00]
-; HASWELL-NEXT: vmaskmovpd %ymm1, %ymm0, (%rdi) # sched: [5:1.00]
-; HASWELL-NEXT: vmovapd %ymm2, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_maskmovpd_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vmaskmovpd (%rdi), %ymm0, %ymm2 # sched: [8:2.00]
-; BROADWELL-NEXT: vmaskmovpd %ymm1, %ymm0, (%rdi) # sched: [5:1.00]
-; BROADWELL-NEXT: vmovapd %ymm2, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_maskmovpd_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vmaskmovpd (%rdi), %ymm0, %ymm2 # sched: [8:0.50]
-; SKYLAKE-NEXT: vmaskmovpd %ymm1, %ymm0, (%rdi) # sched: [2:1.00]
-; SKYLAKE-NEXT: vmovapd %ymm2, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_maskmovpd_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vmaskmovpd (%rdi), %ymm0, %ymm2 # sched: [8:0.50]
-; SKX-NEXT: vmaskmovpd %ymm1, %ymm0, (%rdi) # sched: [2:1.00]
-; SKX-NEXT: vmovapd %ymm2, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_maskmovpd_ymm:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vmaskmovpd (%rdi), %ymm0, %ymm2 # sched: [6:2.00]
-; BDVER2-NEXT: vmaskmovpd %ymm1, %ymm0, (%rdi) # sched: [6:2.00]
-; BDVER2-NEXT: vmovapd %ymm2, %ymm0 # sched: [2:1.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_maskmovpd_ymm:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vmaskmovpd (%rdi), %ymm0, %ymm2 # sched: [6:2.00]
-; BTVER2-NEXT: vmaskmovpd %ymm1, %ymm0, (%rdi) # sched: [6:2.00]
-; BTVER2-NEXT: vmovapd %ymm2, %ymm0 # sched: [1:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_maskmovpd_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vmaskmovpd (%rdi), %ymm0, %ymm2 # sched: [8:1.00]
-; ZNVER1-NEXT: vmaskmovpd %ymm1, %ymm0, (%rdi) # sched: [5:1.00]
-; ZNVER1-NEXT: vmovapd %ymm2, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x double> @llvm.x86.avx.maskload.pd.256(i8* %a0, <4 x i64> %a1)
- call void @llvm.x86.avx.maskstore.pd.256(i8* %a0, <4 x i64> %a1, <4 x double> %a2)
- ret <4 x double> %1
-}
-declare <4 x double> @llvm.x86.avx.maskload.pd.256(i8*, <4 x i64>) nounwind readonly
-declare void @llvm.x86.avx.maskstore.pd.256(i8*, <4 x i64>, <4 x double>) nounwind
-
-define <4 x float> @test_maskmovps(i8* %a0, <4 x i32> %a1, <4 x float> %a2) {
-; GENERIC-LABEL: test_maskmovps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vmaskmovps (%rdi), %xmm0, %xmm2 # sched: [8:1.00]
-; GENERIC-NEXT: vmaskmovps %xmm1, %xmm0, (%rdi) # sched: [5:1.00]
-; GENERIC-NEXT: vmovaps %xmm2, %xmm0 # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_maskmovps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vmaskmovps (%rdi), %xmm0, %xmm2 # sched: [8:1.00]
-; SANDY-NEXT: vmaskmovps %xmm1, %xmm0, (%rdi) # sched: [5:1.00]
-; SANDY-NEXT: vmovaps %xmm2, %xmm0 # sched: [1:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_maskmovps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vmaskmovps (%rdi), %xmm0, %xmm2 # sched: [8:2.00]
-; HASWELL-NEXT: vmaskmovps %xmm1, %xmm0, (%rdi) # sched: [5:1.00]
-; HASWELL-NEXT: vmovaps %xmm2, %xmm0 # sched: [1:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_maskmovps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vmaskmovps (%rdi), %xmm0, %xmm2 # sched: [7:2.00]
-; BROADWELL-NEXT: vmaskmovps %xmm1, %xmm0, (%rdi) # sched: [5:1.00]
-; BROADWELL-NEXT: vmovaps %xmm2, %xmm0 # sched: [1:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_maskmovps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vmaskmovps (%rdi), %xmm0, %xmm2 # sched: [7:0.50]
-; SKYLAKE-NEXT: vmaskmovps %xmm1, %xmm0, (%rdi) # sched: [2:1.00]
-; SKYLAKE-NEXT: vmovaps %xmm2, %xmm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_maskmovps:
-; SKX: # %bb.0:
-; SKX-NEXT: vmaskmovps (%rdi), %xmm0, %xmm2 # sched: [7:0.50]
-; SKX-NEXT: vmaskmovps %xmm1, %xmm0, (%rdi) # sched: [2:1.00]
-; SKX-NEXT: vmovaps %xmm2, %xmm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_maskmovps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vmaskmovps (%rdi), %xmm0, %xmm2 # sched: [6:1.00]
-; BDVER2-NEXT: vmaskmovps %xmm1, %xmm0, (%rdi) # sched: [6:2.00]
-; BDVER2-NEXT: vmovaps %xmm2, %xmm0 # sched: [1:0.50]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_maskmovps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vmaskmovps (%rdi), %xmm0, %xmm2 # sched: [6:1.00]
-; BTVER2-NEXT: vmaskmovps %xmm1, %xmm0, (%rdi) # sched: [6:2.00]
-; BTVER2-NEXT: vmovaps %xmm2, %xmm0 # sched: [1:0.50]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_maskmovps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vmaskmovps (%rdi), %xmm0, %xmm2 # sched: [8:0.50]
-; ZNVER1-NEXT: vmaskmovps %xmm1, %xmm0, (%rdi) # sched: [4:0.50]
-; ZNVER1-NEXT: vmovaps %xmm2, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x float> @llvm.x86.avx.maskload.ps(i8* %a0, <4 x i32> %a1)
- call void @llvm.x86.avx.maskstore.ps(i8* %a0, <4 x i32> %a1, <4 x float> %a2)
- ret <4 x float> %1
-}
-declare <4 x float> @llvm.x86.avx.maskload.ps(i8*, <4 x i32>) nounwind readonly
-declare void @llvm.x86.avx.maskstore.ps(i8*, <4 x i32>, <4 x float>) nounwind
-
-define <8 x float> @test_maskmovps_ymm(i8* %a0, <8 x i32> %a1, <8 x float> %a2) {
-; GENERIC-LABEL: test_maskmovps_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vmaskmovps (%rdi), %ymm0, %ymm2 # sched: [9:1.00]
-; GENERIC-NEXT: vmaskmovps %ymm1, %ymm0, (%rdi) # sched: [5:1.00]
-; GENERIC-NEXT: vmovaps %ymm2, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_maskmovps_ymm:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vmaskmovps (%rdi), %ymm0, %ymm2 # sched: [9:1.00]
-; SANDY-NEXT: vmaskmovps %ymm1, %ymm0, (%rdi) # sched: [5:1.00]
-; SANDY-NEXT: vmovaps %ymm2, %ymm0 # sched: [1:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_maskmovps_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vmaskmovps (%rdi), %ymm0, %ymm2 # sched: [9:2.00]
-; HASWELL-NEXT: vmaskmovps %ymm1, %ymm0, (%rdi) # sched: [5:1.00]
-; HASWELL-NEXT: vmovaps %ymm2, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_maskmovps_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vmaskmovps (%rdi), %ymm0, %ymm2 # sched: [8:2.00]
-; BROADWELL-NEXT: vmaskmovps %ymm1, %ymm0, (%rdi) # sched: [5:1.00]
-; BROADWELL-NEXT: vmovaps %ymm2, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_maskmovps_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vmaskmovps (%rdi), %ymm0, %ymm2 # sched: [8:0.50]
-; SKYLAKE-NEXT: vmaskmovps %ymm1, %ymm0, (%rdi) # sched: [2:1.00]
-; SKYLAKE-NEXT: vmovaps %ymm2, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_maskmovps_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vmaskmovps (%rdi), %ymm0, %ymm2 # sched: [8:0.50]
-; SKX-NEXT: vmaskmovps %ymm1, %ymm0, (%rdi) # sched: [2:1.00]
-; SKX-NEXT: vmovaps %ymm2, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_maskmovps_ymm:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vmaskmovps (%rdi), %ymm0, %ymm2 # sched: [6:2.00]
-; BDVER2-NEXT: vmaskmovps %ymm1, %ymm0, (%rdi) # sched: [6:2.00]
-; BDVER2-NEXT: vmovaps %ymm2, %ymm0 # sched: [2:1.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_maskmovps_ymm:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vmaskmovps (%rdi), %ymm0, %ymm2 # sched: [6:2.00]
-; BTVER2-NEXT: vmaskmovps %ymm1, %ymm0, (%rdi) # sched: [6:2.00]
-; BTVER2-NEXT: vmovaps %ymm2, %ymm0 # sched: [1:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_maskmovps_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vmaskmovps (%rdi), %ymm0, %ymm2 # sched: [8:1.00]
-; ZNVER1-NEXT: vmaskmovps %ymm1, %ymm0, (%rdi) # sched: [5:1.00]
-; ZNVER1-NEXT: vmovaps %ymm2, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x float> @llvm.x86.avx.maskload.ps.256(i8* %a0, <8 x i32> %a1)
- call void @llvm.x86.avx.maskstore.ps.256(i8* %a0, <8 x i32> %a1, <8 x float> %a2)
- ret <8 x float> %1
-}
-declare <8 x float> @llvm.x86.avx.maskload.ps.256(i8*, <8 x i32>) nounwind readonly
-declare void @llvm.x86.avx.maskstore.ps.256(i8*, <8 x i32>, <8 x float>) nounwind
-
-define <4 x double> @test_maxpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
-; GENERIC-LABEL: test_maxpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vmaxpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: vmaxpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_maxpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vmaxpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: vmaxpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_maxpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vmaxpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vmaxpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_maxpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vmaxpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vmaxpd (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_maxpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vmaxpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vmaxpd (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_maxpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vmaxpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vmaxpd (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_maxpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vmaxpd %ymm1, %ymm0, %ymm0 # sched: [2:2.00]
-; BDVER2-NEXT: vmaxpd (%rdi), %ymm0, %ymm0 # sched: [7:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_maxpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vmaxpd %ymm1, %ymm0, %ymm0 # sched: [2:2.00]
-; BTVER2-NEXT: vmaxpd (%rdi), %ymm0, %ymm0 # sched: [7:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_maxpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vmaxpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: vmaxpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x double> @llvm.x86.avx.max.pd.256(<4 x double> %a0, <4 x double> %a1)
- %2 = load <4 x double>, <4 x double> *%a2, align 32
- %3 = call <4 x double> @llvm.x86.avx.max.pd.256(<4 x double> %1, <4 x double> %2)
- ret <4 x double> %3
-}
-declare <4 x double> @llvm.x86.avx.max.pd.256(<4 x double>, <4 x double>) nounwind readnone
-
-define <8 x float> @test_maxps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {
-; GENERIC-LABEL: test_maxps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vmaxps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: vmaxps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_maxps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vmaxps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: vmaxps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_maxps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vmaxps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vmaxps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_maxps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vmaxps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vmaxps (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_maxps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vmaxps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vmaxps (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_maxps:
-; SKX: # %bb.0:
-; SKX-NEXT: vmaxps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vmaxps (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_maxps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vmaxps %ymm1, %ymm0, %ymm0 # sched: [2:2.00]
-; BDVER2-NEXT: vmaxps (%rdi), %ymm0, %ymm0 # sched: [7:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_maxps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vmaxps %ymm1, %ymm0, %ymm0 # sched: [2:2.00]
-; BTVER2-NEXT: vmaxps (%rdi), %ymm0, %ymm0 # sched: [7:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_maxps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vmaxps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: vmaxps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x float> @llvm.x86.avx.max.ps.256(<8 x float> %a0, <8 x float> %a1)
- %2 = load <8 x float>, <8 x float> *%a2, align 32
- %3 = call <8 x float> @llvm.x86.avx.max.ps.256(<8 x float> %1, <8 x float> %2)
- ret <8 x float> %3
-}
-declare <8 x float> @llvm.x86.avx.max.ps.256(<8 x float>, <8 x float>) nounwind readnone
-
-define <4 x double> @test_minpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
-; GENERIC-LABEL: test_minpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vminpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: vminpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_minpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vminpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: vminpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_minpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vminpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vminpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_minpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vminpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vminpd (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_minpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vminpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vminpd (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_minpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vminpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vminpd (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_minpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vminpd %ymm1, %ymm0, %ymm0 # sched: [2:2.00]
-; BDVER2-NEXT: vminpd (%rdi), %ymm0, %ymm0 # sched: [7:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_minpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vminpd %ymm1, %ymm0, %ymm0 # sched: [2:2.00]
-; BTVER2-NEXT: vminpd (%rdi), %ymm0, %ymm0 # sched: [7:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_minpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vminpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: vminpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x double> @llvm.x86.avx.min.pd.256(<4 x double> %a0, <4 x double> %a1)
- %2 = load <4 x double>, <4 x double> *%a2, align 32
- %3 = call <4 x double> @llvm.x86.avx.min.pd.256(<4 x double> %1, <4 x double> %2)
- ret <4 x double> %3
-}
-declare <4 x double> @llvm.x86.avx.min.pd.256(<4 x double>, <4 x double>) nounwind readnone
-
-define <8 x float> @test_minps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {
-; GENERIC-LABEL: test_minps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vminps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: vminps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_minps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vminps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: vminps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_minps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vminps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vminps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_minps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vminps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vminps (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_minps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vminps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vminps (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_minps:
-; SKX: # %bb.0:
-; SKX-NEXT: vminps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vminps (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_minps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vminps %ymm1, %ymm0, %ymm0 # sched: [2:2.00]
-; BDVER2-NEXT: vminps (%rdi), %ymm0, %ymm0 # sched: [7:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_minps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vminps %ymm1, %ymm0, %ymm0 # sched: [2:2.00]
-; BTVER2-NEXT: vminps (%rdi), %ymm0, %ymm0 # sched: [7:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_minps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vminps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: vminps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x float> @llvm.x86.avx.min.ps.256(<8 x float> %a0, <8 x float> %a1)
- %2 = load <8 x float>, <8 x float> *%a2, align 32
- %3 = call <8 x float> @llvm.x86.avx.min.ps.256(<8 x float> %1, <8 x float> %2)
- ret <8 x float> %3
-}
-declare <8 x float> @llvm.x86.avx.min.ps.256(<8 x float>, <8 x float>) nounwind readnone
-
-define <4 x double> @test_movapd(<4 x double> *%a0, <4 x double> *%a1) {
-; GENERIC-LABEL: test_movapd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vmovapd (%rdi), %ymm0 # sched: [7:0.50]
-; GENERIC-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: vmovapd %ymm0, (%rsi) # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_movapd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vmovapd (%rdi), %ymm0 # sched: [7:0.50]
-; SANDY-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: vmovapd %ymm0, (%rsi) # sched: [1:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_movapd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vmovapd (%rdi), %ymm0 # sched: [7:0.50]
-; HASWELL-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vmovapd %ymm0, (%rsi) # sched: [1:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_movapd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vmovapd (%rdi), %ymm0 # sched: [6:0.50]
-; BROADWELL-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vmovapd %ymm0, (%rsi) # sched: [1:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_movapd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vmovapd (%rdi), %ymm0 # sched: [7:0.50]
-; SKYLAKE-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vmovapd %ymm0, (%rsi) # sched: [1:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_movapd:
-; SKX: # %bb.0:
-; SKX-NEXT: vmovapd (%rdi), %ymm0 # sched: [7:0.50]
-; SKX-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vmovapd %ymm0, (%rsi) # sched: [1:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_movapd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vmovapd (%rdi), %ymm0 # sched: [5:0.50]
-; BDVER2-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: vmovapd %ymm0, (%rsi) # sched: [1:1.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_movapd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vmovapd (%rdi), %ymm0 # sched: [5:1.00]
-; BTVER2-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: vmovapd %ymm0, (%rsi) # sched: [1:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_movapd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vmovapd (%rdi), %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: vmovapd %ymm0, (%rsi) # sched: [1:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = load <4 x double>, <4 x double> *%a0, align 32
- %2 = fadd <4 x double> %1, %1
- store <4 x double> %2, <4 x double> *%a1, align 32
- ret <4 x double> %2
-}
-
-define <8 x float> @test_movaps(<8 x float> *%a0, <8 x float> *%a1) {
-; GENERIC-LABEL: test_movaps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vmovaps (%rdi), %ymm0 # sched: [7:0.50]
-; GENERIC-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: vmovaps %ymm0, (%rsi) # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_movaps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vmovaps (%rdi), %ymm0 # sched: [7:0.50]
-; SANDY-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: vmovaps %ymm0, (%rsi) # sched: [1:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_movaps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vmovaps (%rdi), %ymm0 # sched: [7:0.50]
-; HASWELL-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vmovaps %ymm0, (%rsi) # sched: [1:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_movaps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vmovaps (%rdi), %ymm0 # sched: [6:0.50]
-; BROADWELL-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vmovaps %ymm0, (%rsi) # sched: [1:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_movaps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vmovaps (%rdi), %ymm0 # sched: [7:0.50]
-; SKYLAKE-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vmovaps %ymm0, (%rsi) # sched: [1:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_movaps:
-; SKX: # %bb.0:
-; SKX-NEXT: vmovaps (%rdi), %ymm0 # sched: [7:0.50]
-; SKX-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vmovaps %ymm0, (%rsi) # sched: [1:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_movaps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vmovaps (%rdi), %ymm0 # sched: [5:0.50]
-; BDVER2-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: vmovaps %ymm0, (%rsi) # sched: [1:1.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_movaps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vmovaps (%rdi), %ymm0 # sched: [5:1.00]
-; BTVER2-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: vmovaps %ymm0, (%rsi) # sched: [1:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_movaps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vmovaps (%rdi), %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: vmovaps %ymm0, (%rsi) # sched: [1:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = load <8 x float>, <8 x float> *%a0, align 32
- %2 = fadd <8 x float> %1, %1
- store <8 x float> %2, <8 x float> *%a1, align 32
- ret <8 x float> %2
-}
-
-define <4 x double> @test_movddup(<4 x double> %a0, <4 x double> *%a1) {
-; GENERIC-LABEL: test_movddup:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2] sched: [1:1.00]
-; GENERIC-NEXT: vmovddup {{.*#+}} ymm1 = mem[0,0,2,2] sched: [7:0.50]
-; GENERIC-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_movddup:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2] sched: [1:1.00]
-; SANDY-NEXT: vmovddup {{.*#+}} ymm1 = mem[0,0,2,2] sched: [7:0.50]
-; SANDY-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_movddup:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2] sched: [1:1.00]
-; HASWELL-NEXT: vmovddup {{.*#+}} ymm1 = mem[0,0,2,2] sched: [7:0.50]
-; HASWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_movddup:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2] sched: [1:1.00]
-; BROADWELL-NEXT: vmovddup {{.*#+}} ymm1 = mem[0,0,2,2] sched: [6:0.50]
-; BROADWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_movddup:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2] sched: [1:1.00]
-; SKYLAKE-NEXT: vmovddup {{.*#+}} ymm1 = mem[0,0,2,2] sched: [7:0.50]
-; SKYLAKE-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_movddup:
-; SKX: # %bb.0:
-; SKX-NEXT: vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2] sched: [1:1.00]
-; SKX-NEXT: vmovddup {{.*#+}} ymm1 = mem[0,0,2,2] sched: [7:0.50]
-; SKX-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_movddup:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vmovddup {{.*#+}} ymm1 = mem[0,0,2,2] sched: [7:1.00]
-; BDVER2-NEXT: vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2] sched: [2:1.00]
-; BDVER2-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_movddup:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vmovddup {{.*#+}} ymm1 = mem[0,0,2,2] sched: [6:2.00]
-; BTVER2-NEXT: vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2] sched: [1:1.00]
-; BTVER2-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_movddup:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vmovddup {{.*#+}} ymm1 = mem[0,0,2,2] sched: [8:0.50]
-; ZNVER1-NEXT: vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2] sched: [1:0.50]
-; ZNVER1-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <4 x double> %a0, <4 x double> undef, <4 x i32> <i32 0, i32 0, i32 2, i32 2>
- %2 = load <4 x double>, <4 x double> *%a1, align 32
- %3 = shufflevector <4 x double> %2, <4 x double> undef, <4 x i32> <i32 0, i32 0, i32 2, i32 2>
- %4 = fadd <4 x double> %1, %3
- ret <4 x double> %4
-}
-
-define i32 @test_movmskpd(<4 x double> %a0) {
-; GENERIC-LABEL: test_movmskpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vmovmskpd %ymm0, %eax # sched: [2:1.00]
-; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_movmskpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vmovmskpd %ymm0, %eax # sched: [2:1.00]
-; SANDY-NEXT: vzeroupper # sched: [1:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_movmskpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vmovmskpd %ymm0, %eax # sched: [3:1.00]
-; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_movmskpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vmovmskpd %ymm0, %eax # sched: [3:1.00]
-; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_movmskpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vmovmskpd %ymm0, %eax # sched: [2:1.00]
-; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_movmskpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vmovmskpd %ymm0, %eax # sched: [2:1.00]
-; SKX-NEXT: vzeroupper # sched: [0:0.67]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_movmskpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vmovmskpd %ymm0, %eax # sched: [10:1.00]
-; BDVER2-NEXT: vzeroupper # sched: [46:4.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_movmskpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vmovmskpd %ymm0, %eax # sched: [3:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_movmskpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vmovmskpd %ymm0, %eax # sched: [1:1.00]
-; ZNVER1-NEXT: vzeroupper # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call i32 @llvm.x86.avx.movmsk.pd.256(<4 x double> %a0)
- ret i32 %1
-}
-declare i32 @llvm.x86.avx.movmsk.pd.256(<4 x double>) nounwind readnone
-
-define i32 @test_movmskps(<8 x float> %a0) {
-; GENERIC-LABEL: test_movmskps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vmovmskps %ymm0, %eax # sched: [2:1.00]
-; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_movmskps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vmovmskps %ymm0, %eax # sched: [2:1.00]
-; SANDY-NEXT: vzeroupper # sched: [1:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_movmskps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vmovmskps %ymm0, %eax # sched: [3:1.00]
-; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_movmskps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vmovmskps %ymm0, %eax # sched: [3:1.00]
-; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_movmskps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vmovmskps %ymm0, %eax # sched: [2:1.00]
-; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_movmskps:
-; SKX: # %bb.0:
-; SKX-NEXT: vmovmskps %ymm0, %eax # sched: [2:1.00]
-; SKX-NEXT: vzeroupper # sched: [0:0.67]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_movmskps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vmovmskps %ymm0, %eax # sched: [10:1.00]
-; BDVER2-NEXT: vzeroupper # sched: [46:4.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_movmskps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vmovmskps %ymm0, %eax # sched: [3:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_movmskps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vmovmskps %ymm0, %eax # sched: [1:1.00]
-; ZNVER1-NEXT: vzeroupper # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call i32 @llvm.x86.avx.movmsk.ps.256(<8 x float> %a0)
- ret i32 %1
-}
-declare i32 @llvm.x86.avx.movmsk.ps.256(<8 x float>) nounwind readnone
-
-define void @test_movntdq(<4 x i64> %a0, <4 x i64> *%a1) {
-; GENERIC-LABEL: test_movntdq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: #APP
-; GENERIC-NEXT: vmovntdq %ymm0, (%rdi) # sched: [1:1.00]
-; GENERIC-NEXT: #NO_APP
-; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_movntdq:
-; SANDY: # %bb.0:
-; SANDY-NEXT: #APP
-; SANDY-NEXT: vmovntdq %ymm0, (%rdi) # sched: [1:1.00]
-; SANDY-NEXT: #NO_APP
-; SANDY-NEXT: vzeroupper # sched: [1:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_movntdq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: #APP
-; HASWELL-NEXT: vmovntdq %ymm0, (%rdi) # sched: [1:1.00]
-; HASWELL-NEXT: #NO_APP
-; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_movntdq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: #APP
-; BROADWELL-NEXT: vmovntdq %ymm0, (%rdi) # sched: [1:1.00]
-; BROADWELL-NEXT: #NO_APP
-; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_movntdq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: #APP
-; SKYLAKE-NEXT: vmovntdq %ymm0, (%rdi) # sched: [1:1.00]
-; SKYLAKE-NEXT: #NO_APP
-; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_movntdq:
-; SKX: # %bb.0:
-; SKX-NEXT: #APP
-; SKX-NEXT: vmovntdq %ymm0, (%rdi) # sched: [1:1.00]
-; SKX-NEXT: #NO_APP
-; SKX-NEXT: vzeroupper # sched: [0:0.67]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_movntdq:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: #APP
-; BDVER2-NEXT: vmovntdq %ymm0, (%rdi) # sched: [2:2.00]
-; BDVER2-NEXT: #NO_APP
-; BDVER2-NEXT: vzeroupper # sched: [46:4.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_movntdq:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: #APP
-; BTVER2-NEXT: vmovntdq %ymm0, (%rdi) # sched: [2:2.00]
-; BTVER2-NEXT: #NO_APP
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_movntdq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: #APP
-; ZNVER1-NEXT: vmovntdq %ymm0, (%rdi) # sched: [1:0.50]
-; ZNVER1-NEXT: #NO_APP
-; ZNVER1-NEXT: vzeroupper # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- call void asm sideeffect "vmovntdq $0, $1", "x,*m"(<4 x i64> %a0, <4 x i64> *%a1)
- ret void
-}
-
-define <4 x double> @test_movntpd(<4 x double> %a0, <4 x double> *%a1) {
-; GENERIC-LABEL: test_movntpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: vmovntpd %ymm0, (%rdi) # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_movntpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: vmovntpd %ymm0, (%rdi) # sched: [1:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_movntpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vmovntpd %ymm0, (%rdi) # sched: [1:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_movntpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vmovntpd %ymm0, (%rdi) # sched: [1:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_movntpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vmovntpd %ymm0, (%rdi) # sched: [1:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_movntpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vmovntpd %ymm0, (%rdi) # sched: [1:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_movntpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: vmovntpd %ymm0, (%rdi) # sched: [3:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_movntpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: vmovntpd %ymm0, (%rdi) # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_movntpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: vmovntpd %ymm0, (%rdi) # sched: [1:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = fadd <4 x double> %a0, %a0
- store <4 x double> %1, <4 x double> *%a1, align 32, !nontemporal !0
- ret <4 x double> %1
-}
-
-define <8 x float> @test_movntps(<8 x float> %a0, <8 x float> *%a1) {
-; GENERIC-LABEL: test_movntps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: vmovntps %ymm0, (%rdi) # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_movntps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: vmovntps %ymm0, (%rdi) # sched: [1:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_movntps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vmovntps %ymm0, (%rdi) # sched: [1:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_movntps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vmovntps %ymm0, (%rdi) # sched: [1:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_movntps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vmovntps %ymm0, (%rdi) # sched: [1:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_movntps:
-; SKX: # %bb.0:
-; SKX-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vmovntps %ymm0, (%rdi) # sched: [1:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_movntps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: vmovntps %ymm0, (%rdi) # sched: [3:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_movntps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: vmovntps %ymm0, (%rdi) # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_movntps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: vmovntps %ymm0, (%rdi) # sched: [1:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = fadd <8 x float> %a0, %a0
- store <8 x float> %1, <8 x float> *%a1, align 32, !nontemporal !0
- ret <8 x float> %1
-}
-
-define <8 x float> @test_movshdup(<8 x float> %a0, <8 x float> *%a1) {
-; GENERIC-LABEL: test_movshdup:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vmovshdup {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7] sched: [1:1.00]
-; GENERIC-NEXT: vmovshdup {{.*#+}} ymm1 = mem[1,1,3,3,5,5,7,7] sched: [7:0.50]
-; GENERIC-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_movshdup:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vmovshdup {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7] sched: [1:1.00]
-; SANDY-NEXT: vmovshdup {{.*#+}} ymm1 = mem[1,1,3,3,5,5,7,7] sched: [7:0.50]
-; SANDY-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_movshdup:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vmovshdup {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7] sched: [1:1.00]
-; HASWELL-NEXT: vmovshdup {{.*#+}} ymm1 = mem[1,1,3,3,5,5,7,7] sched: [7:0.50]
-; HASWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_movshdup:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vmovshdup {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7] sched: [1:1.00]
-; BROADWELL-NEXT: vmovshdup {{.*#+}} ymm1 = mem[1,1,3,3,5,5,7,7] sched: [6:0.50]
-; BROADWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_movshdup:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vmovshdup {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7] sched: [1:1.00]
-; SKYLAKE-NEXT: vmovshdup {{.*#+}} ymm1 = mem[1,1,3,3,5,5,7,7] sched: [7:0.50]
-; SKYLAKE-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_movshdup:
-; SKX: # %bb.0:
-; SKX-NEXT: vmovshdup {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7] sched: [1:1.00]
-; SKX-NEXT: vmovshdup {{.*#+}} ymm1 = mem[1,1,3,3,5,5,7,7] sched: [7:0.50]
-; SKX-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_movshdup:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vmovshdup {{.*#+}} ymm1 = mem[1,1,3,3,5,5,7,7] sched: [7:1.00]
-; BDVER2-NEXT: vmovshdup {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7] sched: [2:1.00]
-; BDVER2-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_movshdup:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vmovshdup {{.*#+}} ymm1 = mem[1,1,3,3,5,5,7,7] sched: [6:2.00]
-; BTVER2-NEXT: vmovshdup {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7] sched: [1:1.00]
-; BTVER2-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_movshdup:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vmovshdup {{.*#+}} ymm1 = mem[1,1,3,3,5,5,7,7] sched: [8:0.50]
-; ZNVER1-NEXT: vmovshdup {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7] sched: [1:0.50]
-; ZNVER1-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <8 x float> %a0, <8 x float> undef, <8 x i32> <i32 1, i32 1, i32 3, i32 3, i32 5, i32 5, i32 7, i32 7>
- %2 = load <8 x float>, <8 x float> *%a1, align 32
- %3 = shufflevector <8 x float> %2, <8 x float> undef, <8 x i32> <i32 1, i32 1, i32 3, i32 3, i32 5, i32 5, i32 7, i32 7>
- %4 = fadd <8 x float> %1, %3
- ret <8 x float> %4
-}
-
-define <8 x float> @test_movsldup(<8 x float> %a0, <8 x float> *%a1) {
-; GENERIC-LABEL: test_movsldup:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vmovsldup {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6] sched: [1:1.00]
-; GENERIC-NEXT: vmovsldup {{.*#+}} ymm1 = mem[0,0,2,2,4,4,6,6] sched: [7:0.50]
-; GENERIC-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_movsldup:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vmovsldup {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6] sched: [1:1.00]
-; SANDY-NEXT: vmovsldup {{.*#+}} ymm1 = mem[0,0,2,2,4,4,6,6] sched: [7:0.50]
-; SANDY-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_movsldup:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vmovsldup {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6] sched: [1:1.00]
-; HASWELL-NEXT: vmovsldup {{.*#+}} ymm1 = mem[0,0,2,2,4,4,6,6] sched: [7:0.50]
-; HASWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_movsldup:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vmovsldup {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6] sched: [1:1.00]
-; BROADWELL-NEXT: vmovsldup {{.*#+}} ymm1 = mem[0,0,2,2,4,4,6,6] sched: [6:0.50]
-; BROADWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_movsldup:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vmovsldup {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6] sched: [1:1.00]
-; SKYLAKE-NEXT: vmovsldup {{.*#+}} ymm1 = mem[0,0,2,2,4,4,6,6] sched: [7:0.50]
-; SKYLAKE-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_movsldup:
-; SKX: # %bb.0:
-; SKX-NEXT: vmovsldup {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6] sched: [1:1.00]
-; SKX-NEXT: vmovsldup {{.*#+}} ymm1 = mem[0,0,2,2,4,4,6,6] sched: [7:0.50]
-; SKX-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_movsldup:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vmovsldup {{.*#+}} ymm1 = mem[0,0,2,2,4,4,6,6] sched: [7:1.00]
-; BDVER2-NEXT: vmovsldup {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6] sched: [2:1.00]
-; BDVER2-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_movsldup:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vmovsldup {{.*#+}} ymm1 = mem[0,0,2,2,4,4,6,6] sched: [6:2.00]
-; BTVER2-NEXT: vmovsldup {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6] sched: [1:1.00]
-; BTVER2-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_movsldup:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vmovsldup {{.*#+}} ymm1 = mem[0,0,2,2,4,4,6,6] sched: [8:0.50]
-; ZNVER1-NEXT: vmovsldup {{.*#+}} ymm0 = ymm0[0,0,2,2,4,4,6,6] sched: [1:0.50]
-; ZNVER1-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <8 x float> %a0, <8 x float> undef, <8 x i32> <i32 0, i32 0, i32 2, i32 2, i32 4, i32 4, i32 6, i32 6>
- %2 = load <8 x float>, <8 x float> *%a1, align 32
- %3 = shufflevector <8 x float> %2, <8 x float> undef, <8 x i32> <i32 0, i32 0, i32 2, i32 2, i32 4, i32 4, i32 6, i32 6>
- %4 = fadd <8 x float> %1, %3
- ret <8 x float> %4
-}
-
-define <4 x double> @test_movupd(<4 x double> *%a0, <4 x double> *%a1) {
-; GENERIC-LABEL: test_movupd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vmovupd (%rdi), %ymm0 # sched: [7:0.50]
-; GENERIC-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: vmovupd %ymm0, (%rsi) # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_movupd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vmovups (%rdi), %xmm0 # sched: [6:0.50]
-; SANDY-NEXT: vinsertf128 $1, 16(%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; SANDY-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: vextractf128 $1, %ymm0, 16(%rsi) # sched: [1:1.00]
-; SANDY-NEXT: vmovupd %xmm0, (%rsi) # sched: [1:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_movupd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vmovupd (%rdi), %ymm0 # sched: [7:0.50]
-; HASWELL-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vmovupd %ymm0, (%rsi) # sched: [1:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_movupd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vmovupd (%rdi), %ymm0 # sched: [6:0.50]
-; BROADWELL-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vmovupd %ymm0, (%rsi) # sched: [1:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_movupd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vmovupd (%rdi), %ymm0 # sched: [7:0.50]
-; SKYLAKE-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vmovupd %ymm0, (%rsi) # sched: [1:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_movupd:
-; SKX: # %bb.0:
-; SKX-NEXT: vmovupd (%rdi), %ymm0 # sched: [7:0.50]
-; SKX-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vmovupd %ymm0, (%rsi) # sched: [1:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_movupd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vmovupd (%rdi), %ymm0 # sched: [5:0.50]
-; BDVER2-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: vmovupd %ymm0, (%rsi) # sched: [1:1.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_movupd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vmovupd (%rdi), %ymm0 # sched: [5:1.00]
-; BTVER2-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: vmovupd %ymm0, (%rsi) # sched: [1:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_movupd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vmovupd (%rdi), %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: vmovupd %ymm0, (%rsi) # sched: [1:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = load <4 x double>, <4 x double> *%a0, align 1
- %2 = fadd <4 x double> %1, %1
- store <4 x double> %2, <4 x double> *%a1, align 1
- ret <4 x double> %2
-}
-
-define <8 x float> @test_movups(<8 x float> *%a0, <8 x float> *%a1) {
-; GENERIC-LABEL: test_movups:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vmovups (%rdi), %ymm0 # sched: [7:0.50]
-; GENERIC-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: vmovups %ymm0, (%rsi) # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_movups:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vmovups (%rdi), %xmm0 # sched: [6:0.50]
-; SANDY-NEXT: vinsertf128 $1, 16(%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; SANDY-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: vextractf128 $1, %ymm0, 16(%rsi) # sched: [1:1.00]
-; SANDY-NEXT: vmovups %xmm0, (%rsi) # sched: [1:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_movups:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vmovups (%rdi), %ymm0 # sched: [7:0.50]
-; HASWELL-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vmovups %ymm0, (%rsi) # sched: [1:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_movups:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vmovups (%rdi), %ymm0 # sched: [6:0.50]
-; BROADWELL-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vmovups %ymm0, (%rsi) # sched: [1:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_movups:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vmovups (%rdi), %ymm0 # sched: [7:0.50]
-; SKYLAKE-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vmovups %ymm0, (%rsi) # sched: [1:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_movups:
-; SKX: # %bb.0:
-; SKX-NEXT: vmovups (%rdi), %ymm0 # sched: [7:0.50]
-; SKX-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vmovups %ymm0, (%rsi) # sched: [1:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_movups:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vmovups (%rdi), %ymm0 # sched: [5:0.50]
-; BDVER2-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: vmovups %ymm0, (%rsi) # sched: [1:1.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_movups:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vmovups (%rdi), %ymm0 # sched: [5:1.00]
-; BTVER2-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: vmovups %ymm0, (%rsi) # sched: [1:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_movups:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vmovups (%rdi), %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: vmovups %ymm0, (%rsi) # sched: [1:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = load <8 x float>, <8 x float> *%a0, align 1
- %2 = fadd <8 x float> %1, %1
- store <8 x float> %2, <8 x float> *%a1, align 1
- ret <8 x float> %2
-}
-
-define <4 x double> @test_mulpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
-; GENERIC-LABEL: test_mulpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vmulpd %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; GENERIC-NEXT: vmulpd (%rdi), %ymm0, %ymm0 # sched: [12:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_mulpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vmulpd %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; SANDY-NEXT: vmulpd (%rdi), %ymm0, %ymm0 # sched: [12:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_mulpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vmulpd %ymm1, %ymm0, %ymm0 # sched: [5:0.50]
-; HASWELL-NEXT: vmulpd (%rdi), %ymm0, %ymm0 # sched: [12:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_mulpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vmulpd %ymm1, %ymm0, %ymm0 # sched: [3:0.50]
-; BROADWELL-NEXT: vmulpd (%rdi), %ymm0, %ymm0 # sched: [9:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_mulpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vmulpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vmulpd (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_mulpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vmulpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vmulpd (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_mulpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vmulpd %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: vmulpd (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_mulpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vmulpd %ymm1, %ymm0, %ymm0 # sched: [4:4.00]
-; BTVER2-NEXT: vmulpd (%rdi), %ymm0, %ymm0 # sched: [9:4.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_mulpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vmulpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; ZNVER1-NEXT: vmulpd (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = fmul <4 x double> %a0, %a1
- %2 = load <4 x double>, <4 x double> *%a2, align 32
- %3 = fmul <4 x double> %1, %2
- ret <4 x double> %3
-}
-
-define <8 x float> @test_mulps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {
-; GENERIC-LABEL: test_mulps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vmulps %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; GENERIC-NEXT: vmulps (%rdi), %ymm0, %ymm0 # sched: [12:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_mulps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vmulps %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; SANDY-NEXT: vmulps (%rdi), %ymm0, %ymm0 # sched: [12:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_mulps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vmulps %ymm1, %ymm0, %ymm0 # sched: [5:0.50]
-; HASWELL-NEXT: vmulps (%rdi), %ymm0, %ymm0 # sched: [12:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_mulps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vmulps %ymm1, %ymm0, %ymm0 # sched: [3:0.50]
-; BROADWELL-NEXT: vmulps (%rdi), %ymm0, %ymm0 # sched: [9:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_mulps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vmulps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vmulps (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_mulps:
-; SKX: # %bb.0:
-; SKX-NEXT: vmulps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vmulps (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_mulps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vmulps %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: vmulps (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_mulps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vmulps %ymm1, %ymm0, %ymm0 # sched: [2:2.00]
-; BTVER2-NEXT: vmulps (%rdi), %ymm0, %ymm0 # sched: [7:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_mulps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vmulps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; ZNVER1-NEXT: vmulps (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = fmul <8 x float> %a0, %a1
- %2 = load <8 x float>, <8 x float> *%a2, align 32
- %3 = fmul <8 x float> %1, %2
- ret <8 x float> %3
-}
-
-define <4 x double> @orpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
-; GENERIC-LABEL: orpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vorpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vorpd (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; GENERIC-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: orpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vorpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SANDY-NEXT: vorpd (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; SANDY-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: orpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vorpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: vorpd (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: orpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vorpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: vorpd (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: orpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vorpd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: vorpd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: orpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vorpd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: vorpd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: orpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vorpd %ymm1, %ymm0, %ymm0 # sched: [2:1.00]
-; BDVER2-NEXT: vorpd (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BDVER2-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: orpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vorpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BTVER2-NEXT: vorpd (%rdi), %ymm0, %ymm0 # sched: [6:2.00]
-; BTVER2-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: orpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vorpd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vorpd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = bitcast <4 x double> %a0 to <4 x i64>
- %2 = bitcast <4 x double> %a1 to <4 x i64>
- %3 = or <4 x i64> %1, %2
- %4 = load <4 x double>, <4 x double> *%a2, align 32
- %5 = bitcast <4 x double> %4 to <4 x i64>
- %6 = or <4 x i64> %3, %5
- %7 = bitcast <4 x i64> %6 to <4 x double>
- %8 = fadd <4 x double> %a1, %7
- ret <4 x double> %8
-}
-
-define <8 x float> @test_orps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {
-; GENERIC-LABEL: test_orps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vorps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vorps (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; GENERIC-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_orps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vorps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SANDY-NEXT: vorps (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; SANDY-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_orps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vorps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: vorps (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_orps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vorps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: vorps (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_orps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vorps %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: vorps (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_orps:
-; SKX: # %bb.0:
-; SKX-NEXT: vorps %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: vorps (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_orps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vorps %ymm1, %ymm0, %ymm0 # sched: [2:1.00]
-; BDVER2-NEXT: vorps (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BDVER2-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_orps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vorps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BTVER2-NEXT: vorps (%rdi), %ymm0, %ymm0 # sched: [6:2.00]
-; BTVER2-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_orps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vorps %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vorps (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = bitcast <8 x float> %a0 to <4 x i64>
- %2 = bitcast <8 x float> %a1 to <4 x i64>
- %3 = or <4 x i64> %1, %2
- %4 = load <8 x float>, <8 x float> *%a2, align 32
- %5 = bitcast <8 x float> %4 to <4 x i64>
- %6 = or <4 x i64> %3, %5
- %7 = bitcast <4 x i64> %6 to <8 x float>
- %8 = fadd <8 x float> %a1, %7
- ret <8 x float> %8
-}
-
-define <4 x double> @test_perm2f128(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
-; GENERIC-LABEL: test_perm2f128:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1] sched: [1:1.00]
-; GENERIC-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],mem[0,1] sched: [8:1.00]
-; GENERIC-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_perm2f128:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1] sched: [1:1.00]
-; SANDY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],mem[0,1] sched: [8:1.00]
-; SANDY-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_perm2f128:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1] sched: [3:1.00]
-; HASWELL-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],mem[0,1] sched: [10:1.00]
-; HASWELL-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_perm2f128:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1] sched: [3:1.00]
-; BROADWELL-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],mem[0,1] sched: [9:1.00]
-; BROADWELL-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_perm2f128:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1] sched: [3:1.00]
-; SKYLAKE-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],mem[0,1] sched: [10:1.00]
-; SKYLAKE-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_perm2f128:
-; SKX: # %bb.0:
-; SKX-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1] sched: [3:1.00]
-; SKX-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],mem[0,1] sched: [10:1.00]
-; SKX-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_perm2f128:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1] sched: [4:0.50]
-; BDVER2-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],mem[0,1] sched: [8:0.50]
-; BDVER2-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_perm2f128:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1] sched: [1:1.00]
-; BTVER2-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],mem[0,1] sched: [6:1.00]
-; BTVER2-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_perm2f128:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1] sched: [100:0.25]
-; ZNVER1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],mem[0,1] sched: [100:0.25]
-; ZNVER1-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <4 x double> %a0, <4 x double> %a1, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
- %2 = load <4 x double>, <4 x double> *%a2, align 32
- %3 = shufflevector <4 x double> %a0, <4 x double> %2, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
- %4 = fadd <4 x double> %1, %3
- ret <4 x double> %4
-}
-
-define <2 x double> @test_permilpd(<2 x double> %a0, <2 x double> *%a1) {
-; GENERIC-LABEL: test_permilpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0] sched: [1:1.00]
-; GENERIC-NEXT: vpermilpd {{.*#+}} xmm1 = mem[1,0] sched: [7:1.00]
-; GENERIC-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_permilpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0] sched: [1:1.00]
-; SANDY-NEXT: vpermilpd {{.*#+}} xmm1 = mem[1,0] sched: [7:1.00]
-; SANDY-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_permilpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0] sched: [1:1.00]
-; HASWELL-NEXT: vpermilpd {{.*#+}} xmm1 = mem[1,0] sched: [7:1.00]
-; HASWELL-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_permilpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0] sched: [1:1.00]
-; BROADWELL-NEXT: vpermilpd {{.*#+}} xmm1 = mem[1,0] sched: [6:1.00]
-; BROADWELL-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_permilpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0] sched: [1:1.00]
-; SKYLAKE-NEXT: vpermilpd {{.*#+}} xmm1 = mem[1,0] sched: [7:1.00]
-; SKYLAKE-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_permilpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0] sched: [1:1.00]
-; SKX-NEXT: vpermilpd {{.*#+}} xmm1 = mem[1,0] sched: [7:1.00]
-; SKX-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_permilpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vpermilpd {{.*#+}} xmm1 = mem[1,0] sched: [7:0.50]
-; BDVER2-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0] sched: [2:0.50]
-; BDVER2-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_permilpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vpermilpd {{.*#+}} xmm1 = mem[1,0] sched: [6:1.00]
-; BTVER2-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0] sched: [1:0.50]
-; BTVER2-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_permilpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpermilpd {{.*#+}} xmm1 = mem[1,0] sched: [8:0.50]
-; ZNVER1-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0] sched: [1:0.50]
-; ZNVER1-NEXT: vaddpd %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <2 x double> %a0, <2 x double> undef, <2 x i32> <i32 1, i32 0>
- %2 = load <2 x double>, <2 x double> *%a1, align 16
- %3 = shufflevector <2 x double> %2, <2 x double> undef, <2 x i32> <i32 1, i32 0>
- %4 = fadd <2 x double> %1, %3
- ret <2 x double> %4
-}
-
-define <4 x double> @test_permilpd_ymm(<4 x double> %a0, <4 x double> *%a1) {
-; GENERIC-LABEL: test_permilpd_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,3] sched: [1:1.00]
-; GENERIC-NEXT: vpermilpd {{.*#+}} ymm1 = mem[1,0,2,3] sched: [8:1.00]
-; GENERIC-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_permilpd_ymm:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,3] sched: [1:1.00]
-; SANDY-NEXT: vpermilpd {{.*#+}} ymm1 = mem[1,0,2,3] sched: [8:1.00]
-; SANDY-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_permilpd_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,3] sched: [1:1.00]
-; HASWELL-NEXT: vpermilpd {{.*#+}} ymm1 = mem[1,0,2,3] sched: [8:1.00]
-; HASWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_permilpd_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,3] sched: [1:1.00]
-; BROADWELL-NEXT: vpermilpd {{.*#+}} ymm1 = mem[1,0,2,3] sched: [7:1.00]
-; BROADWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_permilpd_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,3] sched: [1:1.00]
-; SKYLAKE-NEXT: vpermilpd {{.*#+}} ymm1 = mem[1,0,2,3] sched: [8:1.00]
-; SKYLAKE-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_permilpd_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,3] sched: [1:1.00]
-; SKX-NEXT: vpermilpd {{.*#+}} ymm1 = mem[1,0,2,3] sched: [8:1.00]
-; SKX-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_permilpd_ymm:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vpermilpd {{.*#+}} ymm1 = mem[1,0,2,3] sched: [7:1.00]
-; BDVER2-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,3] sched: [2:1.00]
-; BDVER2-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_permilpd_ymm:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vpermilpd {{.*#+}} ymm1 = mem[1,0,2,3] sched: [6:2.00]
-; BTVER2-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,3] sched: [1:1.00]
-; BTVER2-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_permilpd_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpermilpd {{.*#+}} ymm1 = mem[1,0,2,3] sched: [8:0.50]
-; ZNVER1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,3] sched: [1:0.50]
-; ZNVER1-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <4 x double> %a0, <4 x double> undef, <4 x i32> <i32 1, i32 0, i32 2, i32 3>
- %2 = load <4 x double>, <4 x double> *%a1, align 32
- %3 = shufflevector <4 x double> %2, <4 x double> undef, <4 x i32> <i32 1, i32 0, i32 2, i32 3>
- %4 = fadd <4 x double> %1, %3
- ret <4 x double> %4
-}
-
-define <4 x float> @test_permilps(<4 x float> %a0, <4 x float> *%a1) {
-; GENERIC-LABEL: test_permilps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0] sched: [1:1.00]
-; GENERIC-NEXT: vpermilps {{.*#+}} xmm1 = mem[3,2,1,0] sched: [7:1.00]
-; GENERIC-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_permilps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0] sched: [1:1.00]
-; SANDY-NEXT: vpermilps {{.*#+}} xmm1 = mem[3,2,1,0] sched: [7:1.00]
-; SANDY-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_permilps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0] sched: [1:1.00]
-; HASWELL-NEXT: vpermilps {{.*#+}} xmm1 = mem[3,2,1,0] sched: [7:1.00]
-; HASWELL-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_permilps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0] sched: [1:1.00]
-; BROADWELL-NEXT: vpermilps {{.*#+}} xmm1 = mem[3,2,1,0] sched: [6:1.00]
-; BROADWELL-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_permilps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0] sched: [1:1.00]
-; SKYLAKE-NEXT: vpermilps {{.*#+}} xmm1 = mem[3,2,1,0] sched: [7:1.00]
-; SKYLAKE-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_permilps:
-; SKX: # %bb.0:
-; SKX-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0] sched: [1:1.00]
-; SKX-NEXT: vpermilps {{.*#+}} xmm1 = mem[3,2,1,0] sched: [7:1.00]
-; SKX-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_permilps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vpermilps {{.*#+}} xmm1 = mem[3,2,1,0] sched: [7:0.50]
-; BDVER2-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0] sched: [2:0.50]
-; BDVER2-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [5:1.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_permilps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vpermilps {{.*#+}} xmm1 = mem[3,2,1,0] sched: [6:1.00]
-; BTVER2-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0] sched: [1:0.50]
-; BTVER2-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_permilps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpermilps {{.*#+}} xmm1 = mem[3,2,1,0] sched: [8:0.50]
-; ZNVER1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[3,2,1,0] sched: [1:0.50]
-; ZNVER1-NEXT: vaddps %xmm1, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <4 x float> %a0, <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
- %2 = load <4 x float>, <4 x float> *%a1, align 16
- %3 = shufflevector <4 x float> %2, <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
- %4 = fadd <4 x float> %1, %3
- ret <4 x float> %4
-}
-
-define <8 x float> @test_permilps_ymm(<8 x float> %a0, <8 x float> *%a1) {
-; GENERIC-LABEL: test_permilps_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4] sched: [1:1.00]
-; GENERIC-NEXT: vpermilps {{.*#+}} ymm1 = mem[3,2,1,0,7,6,5,4] sched: [8:1.00]
-; GENERIC-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_permilps_ymm:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4] sched: [1:1.00]
-; SANDY-NEXT: vpermilps {{.*#+}} ymm1 = mem[3,2,1,0,7,6,5,4] sched: [8:1.00]
-; SANDY-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_permilps_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4] sched: [1:1.00]
-; HASWELL-NEXT: vpermilps {{.*#+}} ymm1 = mem[3,2,1,0,7,6,5,4] sched: [8:1.00]
-; HASWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_permilps_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4] sched: [1:1.00]
-; BROADWELL-NEXT: vpermilps {{.*#+}} ymm1 = mem[3,2,1,0,7,6,5,4] sched: [7:1.00]
-; BROADWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_permilps_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4] sched: [1:1.00]
-; SKYLAKE-NEXT: vpermilps {{.*#+}} ymm1 = mem[3,2,1,0,7,6,5,4] sched: [8:1.00]
-; SKYLAKE-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_permilps_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4] sched: [1:1.00]
-; SKX-NEXT: vpermilps {{.*#+}} ymm1 = mem[3,2,1,0,7,6,5,4] sched: [8:1.00]
-; SKX-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_permilps_ymm:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vpermilps {{.*#+}} ymm1 = mem[3,2,1,0,7,6,5,4] sched: [7:1.00]
-; BDVER2-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4] sched: [2:1.00]
-; BDVER2-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_permilps_ymm:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vpermilps {{.*#+}} ymm1 = mem[3,2,1,0,7,6,5,4] sched: [6:2.00]
-; BTVER2-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4] sched: [1:1.00]
-; BTVER2-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_permilps_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpermilps {{.*#+}} ymm1 = mem[3,2,1,0,7,6,5,4] sched: [8:0.50]
-; ZNVER1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4] sched: [1:0.50]
-; ZNVER1-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <8 x float> %a0, <8 x float> undef, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
- %2 = load <8 x float>, <8 x float> *%a1, align 32
- %3 = shufflevector <8 x float> %2, <8 x float> undef, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
- %4 = fadd <8 x float> %1, %3
- ret <8 x float> %4
-}
-
-define <2 x double> @test_permilvarpd(<2 x double> %a0, <2 x i64> %a1, <2 x i64> *%a2) {
-; GENERIC-LABEL: test_permilvarpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpermilpd %xmm1, %xmm0, %xmm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpermilpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_permilvarpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vpermilpd %xmm1, %xmm0, %xmm0 # sched: [1:1.00]
-; SANDY-NEXT: vpermilpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_permilvarpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpermilpd %xmm1, %xmm0, %xmm0 # sched: [1:1.00]
-; HASWELL-NEXT: vpermilpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_permilvarpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpermilpd %xmm1, %xmm0, %xmm0 # sched: [1:1.00]
-; BROADWELL-NEXT: vpermilpd (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_permilvarpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpermilpd %xmm1, %xmm0, %xmm0 # sched: [1:1.00]
-; SKYLAKE-NEXT: vpermilpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_permilvarpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpermilpd %xmm1, %xmm0, %xmm0 # sched: [1:1.00]
-; SKX-NEXT: vpermilpd (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_permilvarpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vpermilpd %xmm1, %xmm0, %xmm0 # sched: [3:2.00]
-; BDVER2-NEXT: vpermilpd (%rdi), %xmm0, %xmm0 # sched: [8:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_permilvarpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vpermilpd %xmm1, %xmm0, %xmm0 # sched: [3:2.00]
-; BTVER2-NEXT: vpermilpd (%rdi), %xmm0, %xmm0 # sched: [8:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_permilvarpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpermilpd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; ZNVER1-NEXT: vpermilpd (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <2 x double> @llvm.x86.avx.vpermilvar.pd(<2 x double> %a0, <2 x i64> %a1)
- %2 = load <2 x i64>, <2 x i64> *%a2, align 16
- %3 = call <2 x double> @llvm.x86.avx.vpermilvar.pd(<2 x double> %1, <2 x i64> %2)
- ret <2 x double> %3
-}
-declare <2 x double> @llvm.x86.avx.vpermilvar.pd(<2 x double>, <2 x i64>) nounwind readnone
-
-define <4 x double> @test_permilvarpd_ymm(<4 x double> %a0, <4 x i64> %a1, <4 x i64> *%a2) {
-; GENERIC-LABEL: test_permilvarpd_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpermilpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpermilpd (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_permilvarpd_ymm:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vpermilpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SANDY-NEXT: vpermilpd (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_permilvarpd_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpermilpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: vpermilpd (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_permilvarpd_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpermilpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: vpermilpd (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_permilvarpd_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpermilpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SKYLAKE-NEXT: vpermilpd (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_permilvarpd_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vpermilpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SKX-NEXT: vpermilpd (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_permilvarpd_ymm:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vpermilpd %ymm1, %ymm0, %ymm0 # sched: [3:3.00]
-; BDVER2-NEXT: vpermilpd (%rdi), %ymm0, %ymm0 # sched: [8:3.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_permilvarpd_ymm:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vpermilpd %ymm1, %ymm0, %ymm0 # sched: [4:3.00]
-; BTVER2-NEXT: vpermilpd (%rdi), %ymm0, %ymm0 # sched: [9:3.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_permilvarpd_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpermilpd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; ZNVER1-NEXT: vpermilpd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x double> @llvm.x86.avx.vpermilvar.pd.256(<4 x double> %a0, <4 x i64> %a1)
- %2 = load <4 x i64>, <4 x i64> *%a2, align 32
- %3 = call <4 x double> @llvm.x86.avx.vpermilvar.pd.256(<4 x double> %1, <4 x i64> %2)
- ret <4 x double> %3
-}
-declare <4 x double> @llvm.x86.avx.vpermilvar.pd.256(<4 x double>, <4 x i64>) nounwind readnone
-
-define <4 x float> @test_permilvarps(<4 x float> %a0, <4 x i32> %a1, <4 x i32> *%a2) {
-; GENERIC-LABEL: test_permilvarps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpermilps %xmm1, %xmm0, %xmm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpermilps (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_permilvarps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vpermilps %xmm1, %xmm0, %xmm0 # sched: [1:1.00]
-; SANDY-NEXT: vpermilps (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_permilvarps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpermilps %xmm1, %xmm0, %xmm0 # sched: [1:1.00]
-; HASWELL-NEXT: vpermilps (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_permilvarps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpermilps %xmm1, %xmm0, %xmm0 # sched: [1:1.00]
-; BROADWELL-NEXT: vpermilps (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_permilvarps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpermilps %xmm1, %xmm0, %xmm0 # sched: [1:1.00]
-; SKYLAKE-NEXT: vpermilps (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_permilvarps:
-; SKX: # %bb.0:
-; SKX-NEXT: vpermilps %xmm1, %xmm0, %xmm0 # sched: [1:1.00]
-; SKX-NEXT: vpermilps (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_permilvarps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vpermilps %xmm1, %xmm0, %xmm0 # sched: [3:2.00]
-; BDVER2-NEXT: vpermilps (%rdi), %xmm0, %xmm0 # sched: [8:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_permilvarps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vpermilps %xmm1, %xmm0, %xmm0 # sched: [3:2.00]
-; BTVER2-NEXT: vpermilps (%rdi), %xmm0, %xmm0 # sched: [8:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_permilvarps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpermilps %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; ZNVER1-NEXT: vpermilps (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x float> @llvm.x86.avx.vpermilvar.ps(<4 x float> %a0, <4 x i32> %a1)
- %2 = load <4 x i32>, <4 x i32> *%a2, align 16
- %3 = call <4 x float> @llvm.x86.avx.vpermilvar.ps(<4 x float> %1, <4 x i32> %2)
- ret <4 x float> %3
-}
-declare <4 x float> @llvm.x86.avx.vpermilvar.ps(<4 x float>, <4 x i32>) nounwind readnone
-
-define <8 x float> @test_permilvarps_ymm(<8 x float> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
-; GENERIC-LABEL: test_permilvarps_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpermilps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpermilps (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_permilvarps_ymm:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vpermilps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SANDY-NEXT: vpermilps (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_permilvarps_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpermilps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: vpermilps (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_permilvarps_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpermilps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: vpermilps (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_permilvarps_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpermilps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SKYLAKE-NEXT: vpermilps (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_permilvarps_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vpermilps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SKX-NEXT: vpermilps (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_permilvarps_ymm:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vpermilps %ymm1, %ymm0, %ymm0 # sched: [3:3.00]
-; BDVER2-NEXT: vpermilps (%rdi), %ymm0, %ymm0 # sched: [8:3.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_permilvarps_ymm:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vpermilps %ymm1, %ymm0, %ymm0 # sched: [4:3.00]
-; BTVER2-NEXT: vpermilps (%rdi), %ymm0, %ymm0 # sched: [9:3.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_permilvarps_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpermilps %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; ZNVER1-NEXT: vpermilps (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x float> @llvm.x86.avx.vpermilvar.ps.256(<8 x float> %a0, <8 x i32> %a1)
- %2 = load <8 x i32>, <8 x i32> *%a2, align 32
- %3 = call <8 x float> @llvm.x86.avx.vpermilvar.ps.256(<8 x float> %1, <8 x i32> %2)
- ret <8 x float> %3
-}
-declare <8 x float> @llvm.x86.avx.vpermilvar.ps.256(<8 x float>, <8 x i32>) nounwind readnone
-
-define <8 x float> @test_rcpps(<8 x float> %a0, <8 x float> *%a1) {
-; GENERIC-LABEL: test_rcpps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vrcpps (%rdi), %ymm1 # sched: [14:2.00]
-; GENERIC-NEXT: vrcpps %ymm0, %ymm0 # sched: [7:2.00]
-; GENERIC-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_rcpps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vrcpps (%rdi), %ymm1 # sched: [14:2.00]
-; SANDY-NEXT: vrcpps %ymm0, %ymm0 # sched: [7:2.00]
-; SANDY-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_rcpps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vrcpps (%rdi), %ymm1 # sched: [18:2.00]
-; HASWELL-NEXT: vrcpps %ymm0, %ymm0 # sched: [11:2.00]
-; HASWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_rcpps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vrcpps %ymm0, %ymm0 # sched: [11:2.00]
-; BROADWELL-NEXT: vrcpps (%rdi), %ymm1 # sched: [17:2.00]
-; BROADWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_rcpps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vrcpps %ymm0, %ymm0 # sched: [4:1.00]
-; SKYLAKE-NEXT: vrcpps (%rdi), %ymm1 # sched: [11:1.00]
-; SKYLAKE-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_rcpps:
-; SKX: # %bb.0:
-; SKX-NEXT: vrcpps %ymm0, %ymm0 # sched: [4:1.00]
-; SKX-NEXT: vrcpps (%rdi), %ymm1 # sched: [11:1.00]
-; SKX-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_rcpps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vrcpps (%rdi), %ymm1 # sched: [10:2.00]
-; BDVER2-NEXT: vrcpps %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_rcpps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vrcpps (%rdi), %ymm1 # sched: [7:2.00]
-; BTVER2-NEXT: vrcpps %ymm0, %ymm0 # sched: [2:2.00]
-; BTVER2-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_rcpps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vrcpps (%rdi), %ymm1 # sched: [12:0.50]
-; ZNVER1-NEXT: vrcpps %ymm0, %ymm0 # sched: [5:0.50]
-; ZNVER1-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x float> @llvm.x86.avx.rcp.ps.256(<8 x float> %a0)
- %2 = load <8 x float>, <8 x float> *%a1, align 32
- %3 = call <8 x float> @llvm.x86.avx.rcp.ps.256(<8 x float> %2)
- %4 = fadd <8 x float> %1, %3
- ret <8 x float> %4
-}
-declare <8 x float> @llvm.x86.avx.rcp.ps.256(<8 x float>) nounwind readnone
-
-define <4 x double> @test_roundpd(<4 x double> %a0, <4 x double> *%a1) {
-; GENERIC-LABEL: test_roundpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vroundpd $7, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: vroundpd $7, (%rdi), %ymm1 # sched: [10:1.00]
-; GENERIC-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_roundpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vroundpd $7, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: vroundpd $7, (%rdi), %ymm1 # sched: [10:1.00]
-; SANDY-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_roundpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vroundpd $7, %ymm0, %ymm0 # sched: [6:0.50]
-; HASWELL-NEXT: vroundpd $7, (%rdi), %ymm1 # sched: [13:2.00]
-; HASWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_roundpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vroundpd $7, (%rdi), %ymm1 # sched: [12:2.00]
-; BROADWELL-NEXT: vroundpd $7, %ymm0, %ymm0 # sched: [6:0.50]
-; BROADWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_roundpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vroundpd $7, %ymm0, %ymm0 # sched: [8:1.00]
-; SKYLAKE-NEXT: vroundpd $7, (%rdi), %ymm1 # sched: [15:1.00]
-; SKYLAKE-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_roundpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vroundpd $7, %ymm0, %ymm0 # sched: [8:1.00]
-; SKX-NEXT: vroundpd $7, (%rdi), %ymm1 # sched: [15:1.00]
-; SKX-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_roundpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vroundpd $7, (%rdi), %ymm1 # sched: [9:2.00]
-; BDVER2-NEXT: vroundpd $7, %ymm0, %ymm0 # sched: [4:2.00]
-; BDVER2-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_roundpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vroundpd $7, (%rdi), %ymm1 # sched: [8:2.00]
-; BTVER2-NEXT: vroundpd $7, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_roundpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vroundpd $7, (%rdi), %ymm1 # sched: [11:1.00]
-; ZNVER1-NEXT: vroundpd $7, %ymm0, %ymm0 # sched: [4:1.00]
-; ZNVER1-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x double> @llvm.x86.avx.round.pd.256(<4 x double> %a0, i32 7)
- %2 = load <4 x double>, <4 x double> *%a1, align 32
- %3 = call <4 x double> @llvm.x86.avx.round.pd.256(<4 x double> %2, i32 7)
- %4 = fadd <4 x double> %1, %3
- ret <4 x double> %4
-}
-declare <4 x double> @llvm.x86.avx.round.pd.256(<4 x double>, i32) nounwind readnone
-
-define <8 x float> @test_roundps(<8 x float> %a0, <8 x float> *%a1) {
-; GENERIC-LABEL: test_roundps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vroundps $7, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: vroundps $7, (%rdi), %ymm1 # sched: [10:1.00]
-; GENERIC-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_roundps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vroundps $7, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: vroundps $7, (%rdi), %ymm1 # sched: [10:1.00]
-; SANDY-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_roundps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vroundps $7, %ymm0, %ymm0 # sched: [6:0.50]
-; HASWELL-NEXT: vroundps $7, (%rdi), %ymm1 # sched: [13:2.00]
-; HASWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_roundps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vroundps $7, (%rdi), %ymm1 # sched: [12:2.00]
-; BROADWELL-NEXT: vroundps $7, %ymm0, %ymm0 # sched: [6:0.50]
-; BROADWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_roundps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vroundps $7, %ymm0, %ymm0 # sched: [8:1.00]
-; SKYLAKE-NEXT: vroundps $7, (%rdi), %ymm1 # sched: [15:1.00]
-; SKYLAKE-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_roundps:
-; SKX: # %bb.0:
-; SKX-NEXT: vroundps $7, %ymm0, %ymm0 # sched: [8:1.00]
-; SKX-NEXT: vroundps $7, (%rdi), %ymm1 # sched: [15:1.00]
-; SKX-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_roundps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vroundps $7, (%rdi), %ymm1 # sched: [9:2.00]
-; BDVER2-NEXT: vroundps $7, %ymm0, %ymm0 # sched: [4:2.00]
-; BDVER2-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_roundps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vroundps $7, (%rdi), %ymm1 # sched: [8:2.00]
-; BTVER2-NEXT: vroundps $7, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_roundps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vroundps $7, (%rdi), %ymm1 # sched: [11:1.00]
-; ZNVER1-NEXT: vroundps $7, %ymm0, %ymm0 # sched: [4:1.00]
-; ZNVER1-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x float> @llvm.x86.avx.round.ps.256(<8 x float> %a0, i32 7)
- %2 = load <8 x float>, <8 x float> *%a1, align 32
- %3 = call <8 x float> @llvm.x86.avx.round.ps.256(<8 x float> %2, i32 7)
- %4 = fadd <8 x float> %1, %3
- ret <8 x float> %4
-}
-declare <8 x float> @llvm.x86.avx.round.ps.256(<8 x float>, i32) nounwind readnone
-
-define <8 x float> @test_rsqrtps(<8 x float> %a0, <8 x float> *%a1) {
-; GENERIC-LABEL: test_rsqrtps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vrsqrtps (%rdi), %ymm1 # sched: [14:2.00]
-; GENERIC-NEXT: vrsqrtps %ymm0, %ymm0 # sched: [7:2.00]
-; GENERIC-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_rsqrtps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vrsqrtps (%rdi), %ymm1 # sched: [14:2.00]
-; SANDY-NEXT: vrsqrtps %ymm0, %ymm0 # sched: [7:2.00]
-; SANDY-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_rsqrtps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vrsqrtps (%rdi), %ymm1 # sched: [18:2.00]
-; HASWELL-NEXT: vrsqrtps %ymm0, %ymm0 # sched: [11:2.00]
-; HASWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_rsqrtps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vrsqrtps %ymm0, %ymm0 # sched: [11:2.00]
-; BROADWELL-NEXT: vrsqrtps (%rdi), %ymm1 # sched: [17:2.00]
-; BROADWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_rsqrtps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vrsqrtps %ymm0, %ymm0 # sched: [4:1.00]
-; SKYLAKE-NEXT: vrsqrtps (%rdi), %ymm1 # sched: [11:1.00]
-; SKYLAKE-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_rsqrtps:
-; SKX: # %bb.0:
-; SKX-NEXT: vrsqrtps %ymm0, %ymm0 # sched: [4:1.00]
-; SKX-NEXT: vrsqrtps (%rdi), %ymm1 # sched: [11:1.00]
-; SKX-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_rsqrtps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vrsqrtps (%rdi), %ymm1 # sched: [10:2.00]
-; BDVER2-NEXT: vrsqrtps %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_rsqrtps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vrsqrtps (%rdi), %ymm1 # sched: [7:2.00]
-; BTVER2-NEXT: vrsqrtps %ymm0, %ymm0 # sched: [2:2.00]
-; BTVER2-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_rsqrtps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vrsqrtps (%rdi), %ymm1 # sched: [12:0.50]
-; ZNVER1-NEXT: vrsqrtps %ymm0, %ymm0 # sched: [5:1.00]
-; ZNVER1-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x float> @llvm.x86.avx.rsqrt.ps.256(<8 x float> %a0)
- %2 = load <8 x float>, <8 x float> *%a1, align 32
- %3 = call <8 x float> @llvm.x86.avx.rsqrt.ps.256(<8 x float> %2)
- %4 = fadd <8 x float> %1, %3
- ret <8 x float> %4
-}
-declare <8 x float> @llvm.x86.avx.rsqrt.ps.256(<8 x float>) nounwind readnone
-
-define <4 x double> @test_shufpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
-; GENERIC-LABEL: test_shufpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[0],ymm0[2],ymm1[3] sched: [1:1.00]
-; GENERIC-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],mem[0],ymm1[2],mem[3] sched: [8:1.00]
-; GENERIC-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_shufpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[0],ymm0[2],ymm1[3] sched: [1:1.00]
-; SANDY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],mem[0],ymm1[2],mem[3] sched: [8:1.00]
-; SANDY-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_shufpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[0],ymm0[2],ymm1[3] sched: [1:1.00]
-; HASWELL-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],mem[0],ymm1[2],mem[3] sched: [8:1.00]
-; HASWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_shufpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[0],ymm0[2],ymm1[3] sched: [1:1.00]
-; BROADWELL-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],mem[0],ymm1[2],mem[3] sched: [7:1.00]
-; BROADWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_shufpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[0],ymm0[2],ymm1[3] sched: [1:1.00]
-; SKYLAKE-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],mem[0],ymm1[2],mem[3] sched: [8:1.00]
-; SKYLAKE-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_shufpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[0],ymm0[2],ymm1[3] sched: [1:1.00]
-; SKX-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],mem[0],ymm1[2],mem[3] sched: [8:1.00]
-; SKX-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_shufpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[0],ymm0[2],ymm1[3] sched: [2:1.00]
-; BDVER2-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],mem[0],ymm1[2],mem[3] sched: [7:1.00]
-; BDVER2-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_shufpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[0],ymm0[2],ymm1[3] sched: [1:1.00]
-; BTVER2-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],mem[0],ymm1[2],mem[3] sched: [6:2.00]
-; BTVER2-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_shufpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1],ymm1[0],ymm0[2],ymm1[3] sched: [1:0.50]
-; ZNVER1-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1],mem[0],ymm1[2],mem[3] sched: [8:0.50]
-; ZNVER1-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <4 x double> %a0, <4 x double> %a1, <4 x i32> <i32 1, i32 4, i32 2, i32 7>
- %2 = load <4 x double>, <4 x double> *%a2, align 32
- %3 = shufflevector <4 x double> %a1, <4 x double> %2, <4 x i32> <i32 1, i32 4, i32 2, i32 7>
- %4 = fadd <4 x double> %1, %3
- ret <4 x double> %4
-}
-
-define <8 x float> @test_shufps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) nounwind {
-; GENERIC-LABEL: test_shufps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4] sched: [1:1.00]
-; GENERIC-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,3],mem[0,0],ymm1[4,7],mem[4,4] sched: [8:1.00]
-; GENERIC-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_shufps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4] sched: [1:1.00]
-; SANDY-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,3],mem[0,0],ymm1[4,7],mem[4,4] sched: [8:1.00]
-; SANDY-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_shufps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4] sched: [1:1.00]
-; HASWELL-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,3],mem[0,0],ymm1[4,7],mem[4,4] sched: [8:1.00]
-; HASWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_shufps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4] sched: [1:1.00]
-; BROADWELL-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,3],mem[0,0],ymm1[4,7],mem[4,4] sched: [7:1.00]
-; BROADWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_shufps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4] sched: [1:1.00]
-; SKYLAKE-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,3],mem[0,0],ymm1[4,7],mem[4,4] sched: [8:1.00]
-; SKYLAKE-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_shufps:
-; SKX: # %bb.0:
-; SKX-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4] sched: [1:1.00]
-; SKX-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,3],mem[0,0],ymm1[4,7],mem[4,4] sched: [8:1.00]
-; SKX-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_shufps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4] sched: [2:1.00]
-; BDVER2-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,3],mem[0,0],ymm1[4,7],mem[4,4] sched: [7:1.00]
-; BDVER2-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_shufps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4] sched: [1:1.00]
-; BTVER2-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,3],mem[0,0],ymm1[4,7],mem[4,4] sched: [6:2.00]
-; BTVER2-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_shufps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vshufps {{.*#+}} ymm0 = ymm0[0,0],ymm1[0,0],ymm0[4,4],ymm1[4,4] sched: [1:0.50]
-; ZNVER1-NEXT: vshufps {{.*#+}} ymm1 = ymm1[0,3],mem[0,0],ymm1[4,7],mem[4,4] sched: [8:0.50]
-; ZNVER1-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <8 x float> %a0, <8 x float> %a1, <8 x i32> <i32 0, i32 0, i32 8, i32 8, i32 4, i32 4, i32 12, i32 12>
- %2 = load <8 x float>, <8 x float> *%a2, align 32
- %3 = shufflevector <8 x float> %a1, <8 x float> %2, <8 x i32> <i32 0, i32 3, i32 8, i32 8, i32 4, i32 7, i32 12, i32 12>
- %4 = fadd <8 x float> %1, %3
- ret <8 x float> %4
-}
-
-define <4 x double> @test_sqrtpd(<4 x double> %a0, <4 x double> *%a1) {
-; GENERIC-LABEL: test_sqrtpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vsqrtpd (%rdi), %ymm1 # sched: [52:44.00]
-; GENERIC-NEXT: vsqrtpd %ymm0, %ymm0 # sched: [45:44.00]
-; GENERIC-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_sqrtpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vsqrtpd (%rdi), %ymm1 # sched: [52:44.00]
-; SANDY-NEXT: vsqrtpd %ymm0, %ymm0 # sched: [45:44.00]
-; SANDY-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_sqrtpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vsqrtpd (%rdi), %ymm1 # sched: [42:28.00]
-; HASWELL-NEXT: vsqrtpd %ymm0, %ymm0 # sched: [35:28.00]
-; HASWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_sqrtpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vsqrtpd %ymm0, %ymm0 # sched: [29:28.00]
-; BROADWELL-NEXT: vsqrtpd (%rdi), %ymm1 # sched: [35:28.00]
-; BROADWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_sqrtpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vsqrtpd %ymm0, %ymm0 # sched: [18:12.00]
-; SKYLAKE-NEXT: vsqrtpd (%rdi), %ymm1 # sched: [25:12.00]
-; SKYLAKE-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_sqrtpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vsqrtpd %ymm0, %ymm0 # sched: [18:12.00]
-; SKX-NEXT: vsqrtpd (%rdi), %ymm1 # sched: [25:12.00]
-; SKX-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_sqrtpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vsqrtpd (%rdi), %ymm1 # sched: [14:27.00]
-; BDVER2-NEXT: vsqrtpd %ymm0, %ymm0 # sched: [9:27.00]
-; BDVER2-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_sqrtpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vsqrtpd (%rdi), %ymm1 # sched: [59:54.00]
-; BTVER2-NEXT: vsqrtpd %ymm0, %ymm0 # sched: [54:54.00]
-; BTVER2-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_sqrtpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vsqrtpd (%rdi), %ymm1 # sched: [47:40.00]
-; ZNVER1-NEXT: vsqrtpd %ymm0, %ymm0 # sched: [40:40.00]
-; ZNVER1-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x double> @llvm.x86.avx.sqrt.pd.256(<4 x double> %a0)
- %2 = load <4 x double>, <4 x double> *%a1, align 32
- %3 = call <4 x double> @llvm.x86.avx.sqrt.pd.256(<4 x double> %2)
- %4 = fadd <4 x double> %1, %3
- ret <4 x double> %4
-}
-declare <4 x double> @llvm.x86.avx.sqrt.pd.256(<4 x double>) nounwind readnone
-
-define <8 x float> @test_sqrtps(<8 x float> %a0, <8 x float> *%a1) {
-; GENERIC-LABEL: test_sqrtps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vsqrtps (%rdi), %ymm1 # sched: [36:28.00]
-; GENERIC-NEXT: vsqrtps %ymm0, %ymm0 # sched: [29:28.00]
-; GENERIC-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_sqrtps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vsqrtps (%rdi), %ymm1 # sched: [36:28.00]
-; SANDY-NEXT: vsqrtps %ymm0, %ymm0 # sched: [29:28.00]
-; SANDY-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_sqrtps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vsqrtps (%rdi), %ymm1 # sched: [28:14.00]
-; HASWELL-NEXT: vsqrtps %ymm0, %ymm0 # sched: [21:14.00]
-; HASWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_sqrtps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vsqrtps %ymm0, %ymm0 # sched: [21:14.00]
-; BROADWELL-NEXT: vsqrtps (%rdi), %ymm1 # sched: [27:14.00]
-; BROADWELL-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_sqrtps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vsqrtps %ymm0, %ymm0 # sched: [12:6.00]
-; SKYLAKE-NEXT: vsqrtps (%rdi), %ymm1 # sched: [19:6.00]
-; SKYLAKE-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_sqrtps:
-; SKX: # %bb.0:
-; SKX-NEXT: vsqrtps %ymm0, %ymm0 # sched: [12:6.00]
-; SKX-NEXT: vsqrtps (%rdi), %ymm1 # sched: [19:6.00]
-; SKX-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_sqrtps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vsqrtps (%rdi), %ymm1 # sched: [14:21.00]
-; BDVER2-NEXT: vsqrtps %ymm0, %ymm0 # sched: [9:21.00]
-; BDVER2-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_sqrtps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vsqrtps (%rdi), %ymm1 # sched: [47:42.00]
-; BTVER2-NEXT: vsqrtps %ymm0, %ymm0 # sched: [42:42.00]
-; BTVER2-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_sqrtps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vsqrtps (%rdi), %ymm1 # sched: [35:28.00]
-; ZNVER1-NEXT: vsqrtps %ymm0, %ymm0 # sched: [28:28.00]
-; ZNVER1-NEXT: vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x float> @llvm.x86.avx.sqrt.ps.256(<8 x float> %a0)
- %2 = load <8 x float>, <8 x float> *%a1, align 32
- %3 = call <8 x float> @llvm.x86.avx.sqrt.ps.256(<8 x float> %2)
- %4 = fadd <8 x float> %1, %3
- ret <8 x float> %4
-}
-declare <8 x float> @llvm.x86.avx.sqrt.ps.256(<8 x float>) nounwind readnone
-
-define <4 x double> @test_subpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
-; GENERIC-LABEL: test_subpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vsubpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: vsubpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_subpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vsubpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: vsubpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_subpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vsubpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vsubpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_subpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vsubpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vsubpd (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_subpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vsubpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vsubpd (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_subpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vsubpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vsubpd (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_subpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vsubpd %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: vsubpd (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_subpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vsubpd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: vsubpd (%rdi), %ymm0, %ymm0 # sched: [8:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_subpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vsubpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: vsubpd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = fsub <4 x double> %a0, %a1
- %2 = load <4 x double>, <4 x double> *%a2, align 32
- %3 = fsub <4 x double> %1, %2
- ret <4 x double> %3
-}
-
-define <8 x float> @test_subps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {
-; GENERIC-LABEL: test_subps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vsubps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: vsubps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_subps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vsubps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: vsubps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_subps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vsubps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vsubps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_subps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vsubps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vsubps (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_subps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vsubps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vsubps (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_subps:
-; SKX: # %bb.0:
-; SKX-NEXT: vsubps %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vsubps (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_subps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vsubps %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: vsubps (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_subps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vsubps %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: vsubps (%rdi), %ymm0, %ymm0 # sched: [8:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_subps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vsubps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: vsubps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = fsub <8 x float> %a0, %a1
- %2 = load <8 x float>, <8 x float> *%a2, align 32
- %3 = fsub <8 x float> %1, %2
- ret <8 x float> %3
-}
-
-define i32 @test_testpd(<2 x double> %a0, <2 x double> %a1, <2 x double> *%a2) {
-; GENERIC-LABEL: test_testpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: xorl %eax, %eax # sched: [0:0.25]
-; GENERIC-NEXT: vtestpd %xmm1, %xmm0 # sched: [1:1.00]
-; GENERIC-NEXT: setb %al # sched: [1:0.50]
-; GENERIC-NEXT: vtestpd (%rdi), %xmm0 # sched: [7:1.00]
-; GENERIC-NEXT: adcl $0, %eax # sched: [2:0.67]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_testpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: xorl %eax, %eax # sched: [0:0.25]
-; SANDY-NEXT: vtestpd %xmm1, %xmm0 # sched: [1:1.00]
-; SANDY-NEXT: setb %al # sched: [1:0.50]
-; SANDY-NEXT: vtestpd (%rdi), %xmm0 # sched: [7:1.00]
-; SANDY-NEXT: adcl $0, %eax # sched: [2:0.67]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_testpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: xorl %eax, %eax # sched: [1:0.25]
-; HASWELL-NEXT: vtestpd %xmm1, %xmm0 # sched: [1:1.00]
-; HASWELL-NEXT: setb %al # sched: [1:0.50]
-; HASWELL-NEXT: vtestpd (%rdi), %xmm0 # sched: [7:1.00]
-; HASWELL-NEXT: adcl $0, %eax # sched: [2:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_testpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: xorl %eax, %eax # sched: [1:0.25]
-; BROADWELL-NEXT: vtestpd %xmm1, %xmm0 # sched: [1:1.00]
-; BROADWELL-NEXT: setb %al # sched: [1:0.50]
-; BROADWELL-NEXT: vtestpd (%rdi), %xmm0 # sched: [6:1.00]
-; BROADWELL-NEXT: adcl $0, %eax # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_testpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: xorl %eax, %eax # sched: [1:0.25]
-; SKYLAKE-NEXT: vtestpd %xmm1, %xmm0 # sched: [2:1.00]
-; SKYLAKE-NEXT: setb %al # sched: [1:0.50]
-; SKYLAKE-NEXT: vtestpd (%rdi), %xmm0 # sched: [8:1.00]
-; SKYLAKE-NEXT: adcl $0, %eax # sched: [1:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_testpd:
-; SKX: # %bb.0:
-; SKX-NEXT: xorl %eax, %eax # sched: [1:0.25]
-; SKX-NEXT: vtestpd %xmm1, %xmm0 # sched: [2:1.00]
-; SKX-NEXT: setb %al # sched: [1:0.50]
-; SKX-NEXT: vtestpd (%rdi), %xmm0 # sched: [8:1.00]
-; SKX-NEXT: adcl $0, %eax # sched: [1:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_testpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: xorl %eax, %eax # sched: [0:0.25]
-; BDVER2-NEXT: vtestpd %xmm1, %xmm0 # sched: [1:1.00]
-; BDVER2-NEXT: setb %al # sched: [1:0.50]
-; BDVER2-NEXT: vtestpd (%rdi), %xmm0 # sched: [6:1.00]
-; BDVER2-NEXT: adcl $0, %eax # sched: [1:1.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_testpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: xorl %eax, %eax # sched: [0:0.50]
-; BTVER2-NEXT: vtestpd %xmm1, %xmm0 # sched: [3:1.00]
-; BTVER2-NEXT: setb %al # sched: [1:0.50]
-; BTVER2-NEXT: vtestpd (%rdi), %xmm0 # sched: [8:1.00]
-; BTVER2-NEXT: adcl $0, %eax # sched: [1:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_testpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: xorl %eax, %eax # sched: [1:0.25]
-; ZNVER1-NEXT: vtestpd %xmm1, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT: setb %al # sched: [1:0.25]
-; ZNVER1-NEXT: vtestpd (%rdi), %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT: adcl $0, %eax # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call i32 @llvm.x86.avx.vtestc.pd(<2 x double> %a0, <2 x double> %a1)
- %2 = load <2 x double>, <2 x double> *%a2, align 16
- %3 = call i32 @llvm.x86.avx.vtestc.pd(<2 x double> %a0, <2 x double> %2)
- %4 = add i32 %1, %3
- ret i32 %4
-}
-declare i32 @llvm.x86.avx.vtestc.pd(<2 x double>, <2 x double>) nounwind readnone
-
-define i32 @test_testpd_ymm(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
-; GENERIC-LABEL: test_testpd_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: xorl %eax, %eax # sched: [0:0.25]
-; GENERIC-NEXT: vtestpd %ymm1, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: setb %al # sched: [1:0.50]
-; GENERIC-NEXT: vtestpd (%rdi), %ymm0 # sched: [8:1.00]
-; GENERIC-NEXT: adcl $0, %eax # sched: [2:0.67]
-; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_testpd_ymm:
-; SANDY: # %bb.0:
-; SANDY-NEXT: xorl %eax, %eax # sched: [0:0.25]
-; SANDY-NEXT: vtestpd %ymm1, %ymm0 # sched: [1:1.00]
-; SANDY-NEXT: setb %al # sched: [1:0.50]
-; SANDY-NEXT: vtestpd (%rdi), %ymm0 # sched: [8:1.00]
-; SANDY-NEXT: adcl $0, %eax # sched: [2:0.67]
-; SANDY-NEXT: vzeroupper # sched: [1:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_testpd_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: xorl %eax, %eax # sched: [1:0.25]
-; HASWELL-NEXT: vtestpd %ymm1, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: setb %al # sched: [1:0.50]
-; HASWELL-NEXT: vtestpd (%rdi), %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: adcl $0, %eax # sched: [2:0.50]
-; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_testpd_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: xorl %eax, %eax # sched: [1:0.25]
-; BROADWELL-NEXT: vtestpd %ymm1, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: setb %al # sched: [1:0.50]
-; BROADWELL-NEXT: vtestpd (%rdi), %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: adcl $0, %eax # sched: [1:0.50]
-; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_testpd_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: xorl %eax, %eax # sched: [1:0.25]
-; SKYLAKE-NEXT: vtestpd %ymm1, %ymm0 # sched: [2:1.00]
-; SKYLAKE-NEXT: setb %al # sched: [1:0.50]
-; SKYLAKE-NEXT: vtestpd (%rdi), %ymm0 # sched: [9:1.00]
-; SKYLAKE-NEXT: adcl $0, %eax # sched: [1:0.50]
-; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_testpd_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: xorl %eax, %eax # sched: [1:0.25]
-; SKX-NEXT: vtestpd %ymm1, %ymm0 # sched: [2:1.00]
-; SKX-NEXT: setb %al # sched: [1:0.50]
-; SKX-NEXT: vtestpd (%rdi), %ymm0 # sched: [9:1.00]
-; SKX-NEXT: adcl $0, %eax # sched: [1:0.50]
-; SKX-NEXT: vzeroupper # sched: [0:0.67]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_testpd_ymm:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: xorl %eax, %eax # sched: [0:0.25]
-; BDVER2-NEXT: vtestpd %ymm1, %ymm0 # sched: [1:1.00]
-; BDVER2-NEXT: setb %al # sched: [1:0.50]
-; BDVER2-NEXT: vtestpd (%rdi), %ymm0 # sched: [6:1.00]
-; BDVER2-NEXT: adcl $0, %eax # sched: [1:1.00]
-; BDVER2-NEXT: vzeroupper # sched: [46:4.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_testpd_ymm:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: xorl %eax, %eax # sched: [0:0.50]
-; BTVER2-NEXT: vtestpd %ymm1, %ymm0 # sched: [4:2.00]
-; BTVER2-NEXT: setb %al # sched: [1:0.50]
-; BTVER2-NEXT: vtestpd (%rdi), %ymm0 # sched: [9:2.00]
-; BTVER2-NEXT: adcl $0, %eax # sched: [1:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_testpd_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: xorl %eax, %eax # sched: [1:0.25]
-; ZNVER1-NEXT: vtestpd %ymm1, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: setb %al # sched: [1:0.25]
-; ZNVER1-NEXT: vtestpd (%rdi), %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: adcl $0, %eax # sched: [1:0.25]
-; ZNVER1-NEXT: vzeroupper # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call i32 @llvm.x86.avx.vtestc.pd.256(<4 x double> %a0, <4 x double> %a1)
- %2 = load <4 x double>, <4 x double> *%a2, align 32
- %3 = call i32 @llvm.x86.avx.vtestc.pd.256(<4 x double> %a0, <4 x double> %2)
- %4 = add i32 %1, %3
- ret i32 %4
-}
-declare i32 @llvm.x86.avx.vtestc.pd.256(<4 x double>, <4 x double>) nounwind readnone
-
-define i32 @test_testps(<4 x float> %a0, <4 x float> %a1, <4 x float> *%a2) {
-; GENERIC-LABEL: test_testps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: xorl %eax, %eax # sched: [0:0.25]
-; GENERIC-NEXT: vtestps %xmm1, %xmm0 # sched: [1:1.00]
-; GENERIC-NEXT: setb %al # sched: [1:0.50]
-; GENERIC-NEXT: vtestps (%rdi), %xmm0 # sched: [7:1.00]
-; GENERIC-NEXT: adcl $0, %eax # sched: [2:0.67]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_testps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: xorl %eax, %eax # sched: [0:0.25]
-; SANDY-NEXT: vtestps %xmm1, %xmm0 # sched: [1:1.00]
-; SANDY-NEXT: setb %al # sched: [1:0.50]
-; SANDY-NEXT: vtestps (%rdi), %xmm0 # sched: [7:1.00]
-; SANDY-NEXT: adcl $0, %eax # sched: [2:0.67]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_testps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: xorl %eax, %eax # sched: [1:0.25]
-; HASWELL-NEXT: vtestps %xmm1, %xmm0 # sched: [1:1.00]
-; HASWELL-NEXT: setb %al # sched: [1:0.50]
-; HASWELL-NEXT: vtestps (%rdi), %xmm0 # sched: [7:1.00]
-; HASWELL-NEXT: adcl $0, %eax # sched: [2:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_testps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: xorl %eax, %eax # sched: [1:0.25]
-; BROADWELL-NEXT: vtestps %xmm1, %xmm0 # sched: [1:1.00]
-; BROADWELL-NEXT: setb %al # sched: [1:0.50]
-; BROADWELL-NEXT: vtestps (%rdi), %xmm0 # sched: [6:1.00]
-; BROADWELL-NEXT: adcl $0, %eax # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_testps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: xorl %eax, %eax # sched: [1:0.25]
-; SKYLAKE-NEXT: vtestps %xmm1, %xmm0 # sched: [2:1.00]
-; SKYLAKE-NEXT: setb %al # sched: [1:0.50]
-; SKYLAKE-NEXT: vtestps (%rdi), %xmm0 # sched: [8:1.00]
-; SKYLAKE-NEXT: adcl $0, %eax # sched: [1:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_testps:
-; SKX: # %bb.0:
-; SKX-NEXT: xorl %eax, %eax # sched: [1:0.25]
-; SKX-NEXT: vtestps %xmm1, %xmm0 # sched: [2:1.00]
-; SKX-NEXT: setb %al # sched: [1:0.50]
-; SKX-NEXT: vtestps (%rdi), %xmm0 # sched: [8:1.00]
-; SKX-NEXT: adcl $0, %eax # sched: [1:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_testps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: xorl %eax, %eax # sched: [0:0.25]
-; BDVER2-NEXT: vtestps %xmm1, %xmm0 # sched: [1:1.00]
-; BDVER2-NEXT: setb %al # sched: [1:0.50]
-; BDVER2-NEXT: vtestps (%rdi), %xmm0 # sched: [6:1.00]
-; BDVER2-NEXT: adcl $0, %eax # sched: [1:1.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_testps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: xorl %eax, %eax # sched: [0:0.50]
-; BTVER2-NEXT: vtestps %xmm1, %xmm0 # sched: [3:1.00]
-; BTVER2-NEXT: setb %al # sched: [1:0.50]
-; BTVER2-NEXT: vtestps (%rdi), %xmm0 # sched: [8:1.00]
-; BTVER2-NEXT: adcl $0, %eax # sched: [1:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_testps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: xorl %eax, %eax # sched: [1:0.25]
-; ZNVER1-NEXT: vtestps %xmm1, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT: setb %al # sched: [1:0.25]
-; ZNVER1-NEXT: vtestps (%rdi), %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT: adcl $0, %eax # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call i32 @llvm.x86.avx.vtestc.ps(<4 x float> %a0, <4 x float> %a1)
- %2 = load <4 x float>, <4 x float> *%a2, align 16
- %3 = call i32 @llvm.x86.avx.vtestc.ps(<4 x float> %a0, <4 x float> %2)
- %4 = add i32 %1, %3
- ret i32 %4
-}
-declare i32 @llvm.x86.avx.vtestc.ps(<4 x float>, <4 x float>) nounwind readnone
-
-define i32 @test_testps_ymm(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {
-; GENERIC-LABEL: test_testps_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: xorl %eax, %eax # sched: [0:0.25]
-; GENERIC-NEXT: vtestps %ymm1, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: setb %al # sched: [1:0.50]
-; GENERIC-NEXT: vtestps (%rdi), %ymm0 # sched: [8:1.00]
-; GENERIC-NEXT: adcl $0, %eax # sched: [2:0.67]
-; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_testps_ymm:
-; SANDY: # %bb.0:
-; SANDY-NEXT: xorl %eax, %eax # sched: [0:0.25]
-; SANDY-NEXT: vtestps %ymm1, %ymm0 # sched: [1:1.00]
-; SANDY-NEXT: setb %al # sched: [1:0.50]
-; SANDY-NEXT: vtestps (%rdi), %ymm0 # sched: [8:1.00]
-; SANDY-NEXT: adcl $0, %eax # sched: [2:0.67]
-; SANDY-NEXT: vzeroupper # sched: [1:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_testps_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: xorl %eax, %eax # sched: [1:0.25]
-; HASWELL-NEXT: vtestps %ymm1, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: setb %al # sched: [1:0.50]
-; HASWELL-NEXT: vtestps (%rdi), %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: adcl $0, %eax # sched: [2:0.50]
-; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_testps_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: xorl %eax, %eax # sched: [1:0.25]
-; BROADWELL-NEXT: vtestps %ymm1, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: setb %al # sched: [1:0.50]
-; BROADWELL-NEXT: vtestps (%rdi), %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: adcl $0, %eax # sched: [1:0.50]
-; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_testps_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: xorl %eax, %eax # sched: [1:0.25]
-; SKYLAKE-NEXT: vtestps %ymm1, %ymm0 # sched: [2:1.00]
-; SKYLAKE-NEXT: setb %al # sched: [1:0.50]
-; SKYLAKE-NEXT: vtestps (%rdi), %ymm0 # sched: [9:1.00]
-; SKYLAKE-NEXT: adcl $0, %eax # sched: [1:0.50]
-; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_testps_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: xorl %eax, %eax # sched: [1:0.25]
-; SKX-NEXT: vtestps %ymm1, %ymm0 # sched: [2:1.00]
-; SKX-NEXT: setb %al # sched: [1:0.50]
-; SKX-NEXT: vtestps (%rdi), %ymm0 # sched: [9:1.00]
-; SKX-NEXT: adcl $0, %eax # sched: [1:0.50]
-; SKX-NEXT: vzeroupper # sched: [0:0.67]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_testps_ymm:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: xorl %eax, %eax # sched: [0:0.25]
-; BDVER2-NEXT: vtestps %ymm1, %ymm0 # sched: [1:1.00]
-; BDVER2-NEXT: setb %al # sched: [1:0.50]
-; BDVER2-NEXT: vtestps (%rdi), %ymm0 # sched: [6:1.00]
-; BDVER2-NEXT: adcl $0, %eax # sched: [1:1.00]
-; BDVER2-NEXT: vzeroupper # sched: [46:4.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_testps_ymm:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: xorl %eax, %eax # sched: [0:0.50]
-; BTVER2-NEXT: vtestps %ymm1, %ymm0 # sched: [4:2.00]
-; BTVER2-NEXT: setb %al # sched: [1:0.50]
-; BTVER2-NEXT: vtestps (%rdi), %ymm0 # sched: [9:2.00]
-; BTVER2-NEXT: adcl $0, %eax # sched: [1:1.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_testps_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: xorl %eax, %eax # sched: [1:0.25]
-; ZNVER1-NEXT: vtestps %ymm1, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: setb %al # sched: [1:0.25]
-; ZNVER1-NEXT: vtestps (%rdi), %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: adcl $0, %eax # sched: [1:0.25]
-; ZNVER1-NEXT: vzeroupper # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call i32 @llvm.x86.avx.vtestc.ps.256(<8 x float> %a0, <8 x float> %a1)
- %2 = load <8 x float>, <8 x float> *%a2, align 32
- %3 = call i32 @llvm.x86.avx.vtestc.ps.256(<8 x float> %a0, <8 x float> %2)
- %4 = add i32 %1, %3
- ret i32 %4
-}
-declare i32 @llvm.x86.avx.vtestc.ps.256(<8 x float>, <8 x float>) nounwind readnone
-
-define <4 x double> @test_unpckhpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
-; GENERIC-LABEL: test_unpckhpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3] sched: [1:1.00]
-; GENERIC-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],mem[1],ymm1[3],mem[3] sched: [8:1.00]
-; GENERIC-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_unpckhpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3] sched: [1:1.00]
-; SANDY-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],mem[1],ymm1[3],mem[3] sched: [8:1.00]
-; SANDY-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_unpckhpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3] sched: [1:1.00]
-; HASWELL-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],mem[1],ymm1[3],mem[3] sched: [8:1.00]
-; HASWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_unpckhpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3] sched: [1:1.00]
-; BROADWELL-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],mem[1],ymm1[3],mem[3] sched: [7:1.00]
-; BROADWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_unpckhpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3] sched: [1:1.00]
-; SKYLAKE-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],mem[1],ymm1[3],mem[3] sched: [8:1.00]
-; SKYLAKE-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_unpckhpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3] sched: [1:1.00]
-; SKX-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],mem[1],ymm1[3],mem[3] sched: [8:1.00]
-; SKX-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_unpckhpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3] sched: [2:1.00]
-; BDVER2-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],mem[1],ymm1[3],mem[3] sched: [7:1.00]
-; BDVER2-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_unpckhpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3] sched: [1:1.00]
-; BTVER2-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],mem[1],ymm1[3],mem[3] sched: [6:2.00]
-; BTVER2-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_unpckhpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vunpckhpd {{.*#+}} ymm0 = ymm0[1],ymm1[1],ymm0[3],ymm1[3] sched: [1:0.50]
-; ZNVER1-NEXT: vunpckhpd {{.*#+}} ymm1 = ymm1[1],mem[1],ymm1[3],mem[3] sched: [8:0.50]
-; ZNVER1-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <4 x double> %a0, <4 x double> %a1, <4 x i32> <i32 1, i32 5, i32 3, i32 7>
- %2 = load <4 x double>, <4 x double> *%a2, align 32
- %3 = shufflevector <4 x double> %a1, <4 x double> %2, <4 x i32> <i32 1, i32 5, i32 3, i32 7>
- %4 = fadd <4 x double> %1, %3
- ret <4 x double> %4
-}
-
-define <8 x float> @test_unpckhps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) nounwind {
-; GENERIC-LABEL: test_unpckhps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] sched: [1:1.00]
-; GENERIC-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7] sched: [8:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_unpckhps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] sched: [1:1.00]
-; SANDY-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7] sched: [8:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_unpckhps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] sched: [1:1.00]
-; HASWELL-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7] sched: [8:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_unpckhps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] sched: [1:1.00]
-; BROADWELL-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7] sched: [7:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_unpckhps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] sched: [1:1.00]
-; SKYLAKE-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7] sched: [8:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_unpckhps:
-; SKX: # %bb.0:
-; SKX-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] sched: [1:1.00]
-; SKX-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7] sched: [8:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_unpckhps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] sched: [2:1.00]
-; BDVER2-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7] sched: [7:1.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_unpckhps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] sched: [1:1.00]
-; BTVER2-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7] sched: [6:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_unpckhps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] sched: [1:0.50]
-; ZNVER1-NEXT: vunpckhps {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7] sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <8 x float> %a0, <8 x float> %a1, <8 x i32> <i32 2, i32 10, i32 3, i32 11, i32 6, i32 14, i32 7, i32 15>
- %2 = load <8 x float>, <8 x float> *%a2, align 32
- %3 = shufflevector <8 x float> %1, <8 x float> %2, <8 x i32> <i32 2, i32 10, i32 3, i32 11, i32 6, i32 14, i32 7, i32 15>
- ret <8 x float> %3
-}
-
-define <4 x double> @test_unpcklpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
-; GENERIC-LABEL: test_unpcklpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2] sched: [1:1.00]
-; GENERIC-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[2],mem[2] sched: [8:1.00]
-; GENERIC-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_unpcklpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2] sched: [1:1.00]
-; SANDY-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[2],mem[2] sched: [8:1.00]
-; SANDY-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_unpcklpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2] sched: [1:1.00]
-; HASWELL-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[2],mem[2] sched: [8:1.00]
-; HASWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_unpcklpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2] sched: [1:1.00]
-; BROADWELL-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[2],mem[2] sched: [7:1.00]
-; BROADWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_unpcklpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2] sched: [1:1.00]
-; SKYLAKE-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[2],mem[2] sched: [8:1.00]
-; SKYLAKE-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_unpcklpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2] sched: [1:1.00]
-; SKX-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[2],mem[2] sched: [8:1.00]
-; SKX-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_unpcklpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2] sched: [2:1.00]
-; BDVER2-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[2],mem[2] sched: [7:1.00]
-; BDVER2-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_unpcklpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2] sched: [1:1.00]
-; BTVER2-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[2],mem[2] sched: [6:2.00]
-; BTVER2-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_unpcklpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2] sched: [1:0.50]
-; ZNVER1-NEXT: vunpcklpd {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[2],mem[2] sched: [8:0.50]
-; ZNVER1-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <4 x double> %a0, <4 x double> %a1, <4 x i32> <i32 0, i32 4, i32 2, i32 6>
- %2 = load <4 x double>, <4 x double> *%a2, align 32
- %3 = shufflevector <4 x double> %a1, <4 x double> %2, <4 x i32> <i32 0, i32 4, i32 2, i32 6>
- %4 = fadd <4 x double> %1, %3
- ret <4 x double> %4
-}
-
-define <8 x float> @test_unpcklps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) nounwind {
-; GENERIC-LABEL: test_unpcklps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] sched: [1:1.00]
-; GENERIC-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [8:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_unpcklps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] sched: [1:1.00]
-; SANDY-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [8:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_unpcklps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] sched: [1:1.00]
-; HASWELL-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [8:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_unpcklps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] sched: [1:1.00]
-; BROADWELL-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [7:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_unpcklps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] sched: [1:1.00]
-; SKYLAKE-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [8:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_unpcklps:
-; SKX: # %bb.0:
-; SKX-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] sched: [1:1.00]
-; SKX-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [8:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_unpcklps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] sched: [2:1.00]
-; BDVER2-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [7:1.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_unpcklps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] sched: [1:1.00]
-; BTVER2-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [6:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_unpcklps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] sched: [1:0.50]
-; ZNVER1-NEXT: vunpcklps {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <8 x float> %a0, <8 x float> %a1, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 4, i32 12, i32 5, i32 13>
- %2 = load <8 x float>, <8 x float> *%a2, align 32
- %3 = shufflevector <8 x float> %1, <8 x float> %2, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 4, i32 12, i32 5, i32 13>
- ret <8 x float> %3
-}
-
-define <4 x double> @test_xorpd(<4 x double> %a0, <4 x double> %a1, <4 x double> *%a2) {
-; GENERIC-LABEL: test_xorpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vxorpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vxorpd (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; GENERIC-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_xorpd:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vxorpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SANDY-NEXT: vxorpd (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; SANDY-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_xorpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vxorpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: vxorpd (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_xorpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vxorpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: vxorpd (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_xorpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vxorpd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: vxorpd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_xorpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vxorpd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: vxorpd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_xorpd:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vxorpd %ymm1, %ymm0, %ymm0 # sched: [2:1.00]
-; BDVER2-NEXT: vxorpd (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BDVER2-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_xorpd:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vxorpd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BTVER2-NEXT: vxorpd (%rdi), %ymm0, %ymm0 # sched: [6:2.00]
-; BTVER2-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_xorpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vxorpd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vxorpd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: vaddpd %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = bitcast <4 x double> %a0 to <4 x i64>
- %2 = bitcast <4 x double> %a1 to <4 x i64>
- %3 = xor <4 x i64> %1, %2
- %4 = load <4 x double>, <4 x double> *%a2, align 32
- %5 = bitcast <4 x double> %4 to <4 x i64>
- %6 = xor <4 x i64> %3, %5
- %7 = bitcast <4 x i64> %6 to <4 x double>
- %8 = fadd <4 x double> %a1, %7
- ret <4 x double> %8
-}
-
-define <8 x float> @test_xorps(<8 x float> %a0, <8 x float> %a1, <8 x float> *%a2) {
-; GENERIC-LABEL: test_xorps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vxorps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vxorps (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; GENERIC-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_xorps:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vxorps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SANDY-NEXT: vxorps (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; SANDY-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_xorps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vxorps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: vxorps (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_xorps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vxorps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: vxorps (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_xorps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vxorps %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: vxorps (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_xorps:
-; SKX: # %bb.0:
-; SKX-NEXT: vxorps %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: vxorps (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_xorps:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vxorps %ymm1, %ymm0, %ymm0 # sched: [2:1.00]
-; BDVER2-NEXT: vxorps (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BDVER2-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [5:2.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_xorps:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vxorps %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BTVER2-NEXT: vxorps (%rdi), %ymm0, %ymm0 # sched: [6:2.00]
-; BTVER2-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:2.00]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_xorps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vxorps %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vxorps (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = bitcast <8 x float> %a0 to <4 x i64>
- %2 = bitcast <8 x float> %a1 to <4 x i64>
- %3 = xor <4 x i64> %1, %2
- %4 = load <8 x float>, <8 x float> *%a2, align 32
- %5 = bitcast <8 x float> %4 to <4 x i64>
- %6 = xor <4 x i64> %3, %5
- %7 = bitcast <4 x i64> %6 to <8 x float>
- %8 = fadd <8 x float> %a1, %7
- ret <8 x float> %8
-}
-
-define void @test_zeroall() {
-; GENERIC-LABEL: test_zeroall:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vzeroall # sched: [9:2.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_zeroall:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vzeroall # sched: [9:2.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_zeroall:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vzeroall # sched: [8:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_zeroall:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vzeroall # sched: [8:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_zeroall:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vzeroall # sched: [16:4.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_zeroall:
-; SKX: # %bb.0:
-; SKX-NEXT: vzeroall # sched: [12:5.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_zeroall:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vzeroall # sched: [90:8.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_zeroall:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vzeroall # sched: [90:36.50]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_zeroall:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vzeroall # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- call void @llvm.x86.avx.vzeroall()
- ret void
-}
-declare void @llvm.x86.avx.vzeroall() nounwind
-
-define void @test_zeroupper() {
-; GENERIC-LABEL: test_zeroupper:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_zeroupper:
-; SANDY: # %bb.0:
-; SANDY-NEXT: vzeroupper # sched: [1:1.00]
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_zeroupper:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_zeroupper:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_zeroupper:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_zeroupper:
-; SKX: # %bb.0:
-; SKX-NEXT: vzeroupper # sched: [0:0.67]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_zeroupper:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: vzeroupper # sched: [46:4.00]
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_zeroupper:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: vzeroupper # sched: [46:18.50]
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_zeroupper:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vzeroupper # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- call void @llvm.x86.avx.vzeroupper()
- ret void
-}
-
-define void @test_avx256_zero_idioms() {
-; GENERIC-LABEL: test_avx256_zero_idioms:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: #APP
-; GENERIC-NEXT: vxorps %ymm0, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vxorpd %ymm1, %ymm1, %ymm1 # sched: [1:1.00]
-; GENERIC-NEXT: vandnps %ymm2, %ymm2, %ymm2 # sched: [1:1.00]
-; GENERIC-NEXT: vandnpd %ymm3, %ymm3, %ymm3 # sched: [1:1.00]
-; GENERIC-NEXT: #NO_APP
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; SANDY-LABEL: test_avx256_zero_idioms:
-; SANDY: # %bb.0:
-; SANDY-NEXT: #APP
-; SANDY-NEXT: vxorps %ymm0, %ymm0, %ymm0 # sched: [1:1.00]
-; SANDY-NEXT: vxorpd %ymm1, %ymm1, %ymm1 # sched: [1:1.00]
-; SANDY-NEXT: vandnps %ymm2, %ymm2, %ymm2 # sched: [1:1.00]
-; SANDY-NEXT: vandnpd %ymm3, %ymm3, %ymm3 # sched: [1:1.00]
-; SANDY-NEXT: #NO_APP
-; SANDY-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_avx256_zero_idioms:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: #APP
-; HASWELL-NEXT: vxorps %ymm0, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: vxorpd %ymm1, %ymm1, %ymm1 # sched: [1:1.00]
-; HASWELL-NEXT: vandnps %ymm2, %ymm2, %ymm2 # sched: [1:1.00]
-; HASWELL-NEXT: vandnpd %ymm3, %ymm3, %ymm3 # sched: [1:1.00]
-; HASWELL-NEXT: #NO_APP
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_avx256_zero_idioms:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: #APP
-; BROADWELL-NEXT: vxorps %ymm0, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: vxorpd %ymm1, %ymm1, %ymm1 # sched: [1:1.00]
-; BROADWELL-NEXT: vandnps %ymm2, %ymm2, %ymm2 # sched: [1:1.00]
-; BROADWELL-NEXT: vandnpd %ymm3, %ymm3, %ymm3 # sched: [1:1.00]
-; BROADWELL-NEXT: #NO_APP
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_avx256_zero_idioms:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: #APP
-; SKYLAKE-NEXT: vxorps %ymm0, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: vxorpd %ymm1, %ymm1, %ymm1 # sched: [1:0.33]
-; SKYLAKE-NEXT: vandnps %ymm2, %ymm2, %ymm2 # sched: [1:0.33]
-; SKYLAKE-NEXT: vandnpd %ymm3, %ymm3, %ymm3 # sched: [1:0.33]
-; SKYLAKE-NEXT: #NO_APP
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_avx256_zero_idioms:
-; SKX: # %bb.0:
-; SKX-NEXT: #APP
-; SKX-NEXT: vxorps %ymm0, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: vxorpd %ymm1, %ymm1, %ymm1 # sched: [1:0.33]
-; SKX-NEXT: vandnps %ymm2, %ymm2, %ymm2 # sched: [1:0.33]
-; SKX-NEXT: vandnpd %ymm3, %ymm3, %ymm3 # sched: [1:0.33]
-; SKX-NEXT: #NO_APP
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; BDVER2-LABEL: test_avx256_zero_idioms:
-; BDVER2: # %bb.0:
-; BDVER2-NEXT: #APP
-; BDVER2-NEXT: vxorps %ymm0, %ymm0, %ymm0 # sched: [2:1.00]
-; BDVER2-NEXT: vxorpd %ymm1, %ymm1, %ymm1 # sched: [2:1.00]
-; BDVER2-NEXT: vandnps %ymm2, %ymm2, %ymm2 # sched: [2:1.00]
-; BDVER2-NEXT: vandnpd %ymm3, %ymm3, %ymm3 # sched: [2:1.00]
-; BDVER2-NEXT: #NO_APP
-; BDVER2-NEXT: retq # sched: [5:1.00]
-;
-; BTVER2-LABEL: test_avx256_zero_idioms:
-; BTVER2: # %bb.0:
-; BTVER2-NEXT: #APP
-; BTVER2-NEXT: vxorps %ymm0, %ymm0, %ymm0 # sched: [1:0.50]
-; BTVER2-NEXT: vxorpd %ymm1, %ymm1, %ymm1 # sched: [1:0.50]
-; BTVER2-NEXT: vandnps %ymm2, %ymm2, %ymm2 # sched: [1:0.50]
-; BTVER2-NEXT: vandnpd %ymm3, %ymm3, %ymm3 # sched: [1:0.50]
-; BTVER2-NEXT: #NO_APP
-; BTVER2-NEXT: retq # sched: [4:1.00]
-;
-; ZNVER1-LABEL: test_avx256_zero_idioms:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: #APP
-; ZNVER1-NEXT: vxorps %ymm0, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vxorpd %ymm1, %ymm1, %ymm1 # sched: [1:0.25]
-; ZNVER1-NEXT: vandnps %ymm2, %ymm2, %ymm2 # sched: [1:0.25]
-; ZNVER1-NEXT: vandnpd %ymm3, %ymm3, %ymm3 # sched: [1:0.25]
-; ZNVER1-NEXT: #NO_APP
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- call void asm sideeffect "vxorps %ymm0, %ymm0, %ymm0\0Avxorpd %ymm1, %ymm1, %ymm1\0Avandnps %ymm2, %ymm2, %ymm2\0Avandnpd %ymm3, %ymm3, %ymm3", ""()
- ret void
-}
-declare void @llvm.x86.avx.vzeroupper() nounwind
-
-!0 = !{i32 1}
Removed: llvm/trunk/test/CodeGen/X86/avx2-schedule.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/avx2-schedule.ll?rev=353042&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/avx2-schedule.ll (original)
+++ llvm/trunk/test/CodeGen/X86/avx2-schedule.ll (removed)
@@ -1,7111 +0,0 @@
-; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=x86-64 -mattr=+avx2 | FileCheck %s --check-prefix=CHECK --check-prefix=GENERIC
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=haswell -mattr=-avx512f | FileCheck %s --check-prefix=CHECK --check-prefix=HASWELL
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=broadwell -mattr=-avx512f | FileCheck %s --check-prefix=CHECK --check-prefix=BROADWELL
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=skylake -mattr=-avx512f | FileCheck %s --check-prefix=CHECK --check-prefix=SKYLAKE
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=skx -mattr=-avx512f | FileCheck %s --check-prefix=CHECK --check-prefix=SKX
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=znver1 -mattr=-avx512f | FileCheck %s --check-prefix=CHECK --check-prefix=ZNVER1
-
-define <8 x i32> @test_broadcasti128(<8 x i32> %a0, <4 x i32> *%a1) {
-; GENERIC-LABEL: test_broadcasti128:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vbroadcasti128 {{.*#+}} ymm1 = mem[0,1,0,1] sched: [6:1.00]
-; GENERIC-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_broadcasti128:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vbroadcasti128 {{.*#+}} ymm1 = mem[0,1,0,1] sched: [7:0.50]
-; HASWELL-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_broadcasti128:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vbroadcasti128 {{.*#+}} ymm1 = mem[0,1,0,1] sched: [6:0.50]
-; BROADWELL-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_broadcasti128:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vbroadcasti128 {{.*#+}} ymm1 = mem[0,1,0,1] sched: [7:0.50]
-; SKYLAKE-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_broadcasti128:
-; SKX: # %bb.0:
-; SKX-NEXT: vbroadcasti128 {{.*#+}} ymm1 = mem[0,1,0,1] sched: [7:0.50]
-; SKX-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_broadcasti128:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vbroadcasti128 {{.*#+}} ymm1 = mem[0,1,0,1] sched: [8:0.50]
-; ZNVER1-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = load <4 x i32>, <4 x i32> *%a1, align 16
- %2 = shufflevector <4 x i32> %1, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3>
- %3 = add <8 x i32> %2, %a0
- ret <8 x i32> %3
-}
-
-define <4 x double> @test_broadcastsd_ymm(<2 x double> %a0) {
-; GENERIC-LABEL: test_broadcastsd_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vbroadcastsd %xmm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_broadcastsd_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vbroadcastsd %xmm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_broadcastsd_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vbroadcastsd %xmm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_broadcastsd_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vbroadcastsd %xmm0, %ymm0 # sched: [3:1.00]
-; SKYLAKE-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_broadcastsd_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vbroadcastsd %xmm0, %ymm0 # sched: [3:1.00]
-; SKX-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_broadcastsd_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vbroadcastsd %xmm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <2 x double> %a0, <2 x double> undef, <4 x i32> zeroinitializer
- %2 = fadd <4 x double> %1, %1
- ret <4 x double> %2
-}
-
-define <4 x float> @test_broadcastss(<4 x float> %a0) {
-; GENERIC-LABEL: test_broadcastss:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vbroadcastss %xmm0, %xmm0 # sched: [1:1.00]
-; GENERIC-NEXT: vaddps %xmm0, %xmm0, %xmm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_broadcastss:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vbroadcastss %xmm0, %xmm0 # sched: [1:1.00]
-; HASWELL-NEXT: vaddps %xmm0, %xmm0, %xmm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_broadcastss:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vbroadcastss %xmm0, %xmm0 # sched: [1:1.00]
-; BROADWELL-NEXT: vaddps %xmm0, %xmm0, %xmm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_broadcastss:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vbroadcastss %xmm0, %xmm0 # sched: [1:1.00]
-; SKYLAKE-NEXT: vaddps %xmm0, %xmm0, %xmm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_broadcastss:
-; SKX: # %bb.0:
-; SKX-NEXT: vbroadcastss %xmm0, %xmm0 # sched: [1:1.00]
-; SKX-NEXT: vaddps %xmm0, %xmm0, %xmm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_broadcastss:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vbroadcastss %xmm0, %xmm0 # sched: [1:0.50]
-; ZNVER1-NEXT: vaddps %xmm0, %xmm0, %xmm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <4 x float> %a0, <4 x float> undef, <4 x i32> zeroinitializer
- %2 = fadd <4 x float> %1, %1
- ret <4 x float> %2
-}
-
-define <8 x float> @test_broadcastss_ymm(<4 x float> %a0) {
-; GENERIC-LABEL: test_broadcastss_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vbroadcastss %xmm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_broadcastss_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vbroadcastss %xmm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_broadcastss_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vbroadcastss %xmm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_broadcastss_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vbroadcastss %xmm0, %ymm0 # sched: [3:1.00]
-; SKYLAKE-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_broadcastss_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vbroadcastss %xmm0, %ymm0 # sched: [3:1.00]
-; SKX-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_broadcastss_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vbroadcastss %xmm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <4 x float> %a0, <4 x float> undef, <8 x i32> zeroinitializer
- %2 = fadd <8 x float> %1, %1
- ret <8 x float> %2
-}
-
-define <4 x i32> @test_extracti128(<8 x i16> %a0, <4 x i32> *%a1) {
-; GENERIC-LABEL: test_extracti128:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmovzxwd {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero sched: [1:1.00]
-; GENERIC-NEXT: vextracti128 $1, %ymm1, %xmm0 # sched: [1:1.00]
-; GENERIC-NEXT: vextracti128 $1, %ymm1, (%rdi) # sched: [1:1.00]
-; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_extracti128:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmovzxwd {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero sched: [3:1.00]
-; HASWELL-NEXT: vextracti128 $1, %ymm1, %xmm0 # sched: [3:1.00]
-; HASWELL-NEXT: vextracti128 $1, %ymm1, (%rdi) # sched: [1:1.00]
-; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_extracti128:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmovzxwd {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero sched: [3:1.00]
-; BROADWELL-NEXT: vextracti128 $1, %ymm1, %xmm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vextracti128 $1, %ymm1, (%rdi) # sched: [1:1.00]
-; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_extracti128:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmovzxwd {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero sched: [3:1.00]
-; SKYLAKE-NEXT: vextracti128 $1, %ymm1, %xmm0 # sched: [3:1.00]
-; SKYLAKE-NEXT: vextracti128 $1, %ymm1, (%rdi) # sched: [1:1.00]
-; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_extracti128:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmovzxwd {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero sched: [3:1.00]
-; SKX-NEXT: vextracti128 $1, %ymm1, %xmm0 # sched: [3:1.00]
-; SKX-NEXT: vextracti128 $1, %ymm1, (%rdi) # sched: [1:1.00]
-; SKX-NEXT: vzeroupper # sched: [0:0.67]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_extracti128:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmovzxwd {{.*#+}} ymm1 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero sched: [1:0.50]
-; ZNVER1-NEXT: vextracti128 $1, %ymm1, %xmm0 # sched: [2:0.25]
-; ZNVER1-NEXT: vextracti128 $1, %ymm1, (%rdi) # sched: [1:0.50]
-; ZNVER1-NEXT: vzeroupper # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %z = zext <8 x i16> %a0 to <8 x i32>
- %ext = shufflevector <8 x i32> %z, <8 x i32> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
- store <4 x i32> %ext, <4 x i32> *%a1
- ret <4 x i32> %ext
-}
-
-define <2 x double> @test_gatherdpd(<2 x double> %a0, i8* %a1, <4 x i32> %a2, <2 x double> %a3) {
-; GENERIC-LABEL: test_gatherdpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vgatherdpd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [5:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_gatherdpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vgatherdpd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [26:2.67]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_gatherdpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vgatherdpd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [25:3.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_gatherdpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vgatherdpd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_gatherdpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vgatherdpd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_gatherdpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vgatherdpd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <2 x double> @llvm.x86.avx2.gather.d.pd(<2 x double> %a0, i8* %a1, <4 x i32> %a2, <2 x double> %a3, i8 2)
- ret <2 x double> %1
-}
-declare <2 x double> @llvm.x86.avx2.gather.d.pd(<2 x double>, i8*, <4 x i32>, <2 x double>, i8) nounwind readonly
-
-define <4 x double> @test_gatherdpd_ymm(<4 x double> %a0, i8* %a1, <4 x i32> %a2, <4 x double> %a3) {
-; GENERIC-LABEL: test_gatherdpd_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vgatherdpd %ymm2, (%rdi,%xmm1,8), %ymm0 # sched: [5:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_gatherdpd_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vgatherdpd %ymm2, (%rdi,%xmm1,8), %ymm0 # sched: [27:4.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_gatherdpd_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vgatherdpd %ymm2, (%rdi,%xmm1,8), %ymm0 # sched: [26:5.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_gatherdpd_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vgatherdpd %ymm2, (%rdi,%xmm1,8), %ymm0 # sched: [25:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_gatherdpd_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vgatherdpd %ymm2, (%rdi,%xmm1,8), %ymm0 # sched: [25:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_gatherdpd_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vgatherdpd %ymm2, (%rdi,%xmm1,8), %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x double> @llvm.x86.avx2.gather.d.pd.256(<4 x double> %a0, i8* %a1, <4 x i32> %a2, <4 x double> %a3, i8 8)
- ret <4 x double> %1
-}
-declare <4 x double> @llvm.x86.avx2.gather.d.pd.256(<4 x double>, i8*, <4 x i32>, <4 x double>, i8) nounwind readonly
-
-define <4 x float> @test_gatherdps(<4 x float> %a0, i8* %a1, <4 x i32> %a2, <4 x float> %a3) {
-; GENERIC-LABEL: test_gatherdps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vgatherdps %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [5:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_gatherdps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vgatherdps %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [25:3.67]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_gatherdps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vgatherdps %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [25:3.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_gatherdps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vgatherdps %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_gatherdps:
-; SKX: # %bb.0:
-; SKX-NEXT: vgatherdps %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_gatherdps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vgatherdps %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x float> @llvm.x86.avx2.gather.d.ps(<4 x float> %a0, i8* %a1, <4 x i32> %a2, <4 x float> %a3, i8 2)
- ret <4 x float> %1
-}
-declare <4 x float> @llvm.x86.avx2.gather.d.ps(<4 x float>, i8*, <4 x i32>, <4 x float>, i8) nounwind readonly
-
-define <8 x float> @test_gatherdps_ymm(<8 x float> %a0, i8* %a1, <8 x i32> %a2, <8 x float> %a3) {
-; GENERIC-LABEL: test_gatherdps_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vgatherdps %ymm2, (%rdi,%ymm1,4), %ymm0 # sched: [5:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_gatherdps_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vgatherdps %ymm2, (%rdi,%ymm1,4), %ymm0 # sched: [27:6.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_gatherdps_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vgatherdps %ymm2, (%rdi,%ymm1,4), %ymm0 # sched: [26:4.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_gatherdps_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vgatherdps %ymm2, (%rdi,%ymm1,4), %ymm0 # sched: [25:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_gatherdps_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vgatherdps %ymm2, (%rdi,%ymm1,4), %ymm0 # sched: [25:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_gatherdps_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vgatherdps %ymm2, (%rdi,%ymm1,4), %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x float> @llvm.x86.avx2.gather.d.ps.256(<8 x float> %a0, i8* %a1, <8 x i32> %a2, <8 x float> %a3, i8 4)
- ret <8 x float> %1
-}
-declare <8 x float> @llvm.x86.avx2.gather.d.ps.256(<8 x float>, i8*, <8 x i32>, <8 x float>, i8) nounwind readonly
-
-define <2 x double> @test_gatherqpd(<2 x double> %a0, i8* %a1, <2 x i64> %a2, <2 x double> %a3) {
-; GENERIC-LABEL: test_gatherqpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vgatherqpd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [5:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_gatherqpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vgatherqpd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [23:3.33]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_gatherqpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vgatherqpd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:3.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_gatherqpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vgatherqpd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_gatherqpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vgatherqpd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_gatherqpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vgatherqpd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <2 x double> @llvm.x86.avx2.gather.q.pd(<2 x double> %a0, i8* %a1, <2 x i64> %a2, <2 x double> %a3, i8 2)
- ret <2 x double> %1
-}
-declare <2 x double> @llvm.x86.avx2.gather.q.pd(<2 x double>, i8*, <2 x i64>, <2 x double>, i8) nounwind readonly
-
-define <4 x double> @test_gatherqpd_ymm(<4 x double> %a0, i8* %a1, <4 x i64> %a2, <4 x double> %a3) {
-; GENERIC-LABEL: test_gatherqpd_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vgatherqpd %ymm2, (%rdi,%ymm1,8), %ymm0 # sched: [5:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_gatherqpd_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vgatherqpd %ymm2, (%rdi,%ymm1,8), %ymm0 # sched: [24:5.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_gatherqpd_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vgatherqpd %ymm2, (%rdi,%ymm1,8), %ymm0 # sched: [23:3.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_gatherqpd_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vgatherqpd %ymm2, (%rdi,%ymm1,8), %ymm0 # sched: [25:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_gatherqpd_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vgatherqpd %ymm2, (%rdi,%ymm1,8), %ymm0 # sched: [25:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_gatherqpd_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vgatherqpd %ymm2, (%rdi,%ymm1,8), %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x double> @llvm.x86.avx2.gather.q.pd.256(<4 x double> %a0, i8* %a1, <4 x i64> %a2, <4 x double> %a3, i8 8)
- ret <4 x double> %1
-}
-declare <4 x double> @llvm.x86.avx2.gather.q.pd.256(<4 x double>, i8*, <4 x i64>, <4 x double>, i8) nounwind readonly
-
-define <4 x float> @test_gatherqps(<4 x float> %a0, i8* %a1, <2 x i64> %a2, <4 x float> %a3) {
-; GENERIC-LABEL: test_gatherqps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vgatherqps %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [5:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_gatherqps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vgatherqps %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [25:3.67]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_gatherqps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vgatherqps %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [27:5.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_gatherqps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vgatherqps %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_gatherqps:
-; SKX: # %bb.0:
-; SKX-NEXT: vgatherqps %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_gatherqps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vgatherqps %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x float> @llvm.x86.avx2.gather.q.ps(<4 x float> %a0, i8* %a1, <2 x i64> %a2, <4 x float> %a3, i8 2)
- ret <4 x float> %1
-}
-declare <4 x float> @llvm.x86.avx2.gather.q.ps(<4 x float>, i8*, <2 x i64>, <4 x float>, i8) nounwind readonly
-
-define <4 x float> @test_gatherqps_ymm(<4 x float> %a0, i8* %a1, <4 x i64> %a2, <4 x float> %a3) {
-; GENERIC-LABEL: test_gatherqps_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vgatherqps %xmm2, (%rdi,%ymm1,4), %xmm0 # sched: [5:0.50]
-; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_gatherqps_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vgatherqps %xmm2, (%rdi,%ymm1,4), %xmm0 # sched: [28:3.67]
-; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_gatherqps_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vgatherqps %xmm2, (%rdi,%ymm1,4), %xmm0 # sched: [24:5.00]
-; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_gatherqps_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vgatherqps %xmm2, (%rdi,%ymm1,4), %xmm0 # sched: [25:1.00]
-; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_gatherqps_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vgatherqps %xmm2, (%rdi,%ymm1,4), %xmm0 # sched: [25:1.00]
-; SKX-NEXT: vzeroupper # sched: [0:0.67]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_gatherqps_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vgatherqps %xmm2, (%rdi,%ymm1,4), %xmm0 # sched: [100:0.25]
-; ZNVER1-NEXT: vzeroupper # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x float> @llvm.x86.avx2.gather.q.ps.256(<4 x float> %a0, i8* %a1, <4 x i64> %a2, <4 x float> %a3, i8 4)
- ret <4 x float> %1
-}
-declare <4 x float> @llvm.x86.avx2.gather.q.ps.256(<4 x float>, i8*, <4 x i64>, <4 x float>, i8) nounwind readonly
-
-define <8 x i32> @test_inserti128(<8 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {
-; GENERIC-LABEL: test_inserti128:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 # sched: [1:1.00]
-; GENERIC-NEXT: vinserti128 $1, (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; GENERIC-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_inserti128:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 # sched: [3:1.00]
-; HASWELL-NEXT: vinserti128 $1, (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; HASWELL-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_inserti128:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 # sched: [3:1.00]
-; BROADWELL-NEXT: vinserti128 $1, (%rdi), %ymm0, %ymm0 # sched: [6:0.50]
-; BROADWELL-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_inserti128:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 # sched: [3:1.00]
-; SKYLAKE-NEXT: vinserti128 $1, (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; SKYLAKE-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_inserti128:
-; SKX: # %bb.0:
-; SKX-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 # sched: [3:1.00]
-; SKX-NEXT: vinserti128 $1, (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; SKX-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_inserti128:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 # sched: [2:0.25]
-; ZNVER1-NEXT: vinserti128 $1, (%rdi), %ymm0, %ymm0 # sched: [9:0.50]
-; ZNVER1-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <4 x i32> %a1, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
- %2 = shufflevector <8 x i32> %a0, <8 x i32> %1, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
- %3 = load <4 x i32>, <4 x i32> *%a2, align 16
- %4 = shufflevector <4 x i32> %3, <4 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
- %5 = shufflevector <8 x i32> %a0, <8 x i32> %4, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
- %6 = add <8 x i32> %2, %5
- ret <8 x i32> %6
-}
-
-define <4 x i64> @test_movntdqa(i8* %a0) {
-; GENERIC-LABEL: test_movntdqa:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vmovntdqa (%rdi), %ymm0 # sched: [7:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_movntdqa:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vmovntdqa (%rdi), %ymm0 # sched: [7:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_movntdqa:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vmovntdqa (%rdi), %ymm0 # sched: [6:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_movntdqa:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vmovntdqa (%rdi), %ymm0 # sched: [7:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_movntdqa:
-; SKX: # %bb.0:
-; SKX-NEXT: vmovntdqa (%rdi), %ymm0 # sched: [7:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_movntdqa:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vmovntdqa (%rdi), %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x i64> @llvm.x86.avx2.movntdqa(i8* %a0)
- ret <4 x i64> %1
-}
-declare <4 x i64> @llvm.x86.avx2.movntdqa(i8*) nounwind readonly
-
-define <16 x i16> @test_mpsadbw(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) {
-; GENERIC-LABEL: test_mpsadbw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vmpsadbw $7, %ymm1, %ymm0, %ymm0 # sched: [7:1.00]
-; GENERIC-NEXT: vmpsadbw $7, (%rdi), %ymm0, %ymm0 # sched: [14:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_mpsadbw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vmpsadbw $7, %ymm1, %ymm0, %ymm0 # sched: [7:2.00]
-; HASWELL-NEXT: vmpsadbw $7, (%rdi), %ymm0, %ymm0 # sched: [14:2.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_mpsadbw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vmpsadbw $7, %ymm1, %ymm0, %ymm0 # sched: [7:2.00]
-; BROADWELL-NEXT: vmpsadbw $7, (%rdi), %ymm0, %ymm0 # sched: [13:2.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_mpsadbw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vmpsadbw $7, %ymm1, %ymm0, %ymm0 # sched: [4:2.00]
-; SKYLAKE-NEXT: vmpsadbw $7, (%rdi), %ymm0, %ymm0 # sched: [11:2.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_mpsadbw:
-; SKX: # %bb.0:
-; SKX-NEXT: vmpsadbw $7, %ymm1, %ymm0, %ymm0 # sched: [4:2.00]
-; SKX-NEXT: vmpsadbw $7, (%rdi), %ymm0, %ymm0 # sched: [11:2.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_mpsadbw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vmpsadbw $7, %ymm1, %ymm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: vmpsadbw $7, (%rdi), %ymm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <16 x i16> @llvm.x86.avx2.mpsadbw(<32 x i8> %a0, <32 x i8> %a1, i8 7)
- %2 = bitcast <16 x i16> %1 to <32 x i8>
- %3 = load <32 x i8>, <32 x i8> *%a2, align 32
- %4 = call <16 x i16> @llvm.x86.avx2.mpsadbw(<32 x i8> %2, <32 x i8> %3, i8 7)
- ret <16 x i16> %4
-}
-declare <16 x i16> @llvm.x86.avx2.mpsadbw(<32 x i8>, <32 x i8>, i8) nounwind readnone
-
-define <32 x i8> @test_pabsb(<32 x i8> %a0, <32 x i8> *%a1) {
-; GENERIC-LABEL: test_pabsb:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpabsb %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpabsb (%rdi), %ymm1 # sched: [8:0.50]
-; GENERIC-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pabsb:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpabsb %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpabsb (%rdi), %ymm1 # sched: [8:0.50]
-; HASWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pabsb:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpabsb %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpabsb (%rdi), %ymm1 # sched: [7:0.50]
-; BROADWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pabsb:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpabsb %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpabsb (%rdi), %ymm1 # sched: [8:0.50]
-; SKYLAKE-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pabsb:
-; SKX: # %bb.0:
-; SKX-NEXT: vpabsb %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpabsb (%rdi), %ymm1 # sched: [8:0.50]
-; SKX-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pabsb:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpabsb (%rdi), %ymm1 # sched: [8:0.50]
-; ZNVER1-NEXT: vpabsb %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <32 x i8> @llvm.x86.avx2.pabs.b(<32 x i8> %a0)
- %2 = load <32 x i8>, <32 x i8> *%a1, align 32
- %3 = call <32 x i8> @llvm.x86.avx2.pabs.b(<32 x i8> %2)
- %4 = or <32 x i8> %1, %3
- ret <32 x i8> %4
-}
-declare <32 x i8> @llvm.x86.avx2.pabs.b(<32 x i8>) nounwind readnone
-
-define <8 x i32> @test_pabsd(<8 x i32> %a0, <8 x i32> *%a1) {
-; GENERIC-LABEL: test_pabsd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpabsd %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpabsd (%rdi), %ymm1 # sched: [8:0.50]
-; GENERIC-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pabsd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpabsd %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpabsd (%rdi), %ymm1 # sched: [8:0.50]
-; HASWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pabsd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpabsd %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpabsd (%rdi), %ymm1 # sched: [7:0.50]
-; BROADWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pabsd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpabsd %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpabsd (%rdi), %ymm1 # sched: [8:0.50]
-; SKYLAKE-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pabsd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpabsd %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpabsd (%rdi), %ymm1 # sched: [8:0.50]
-; SKX-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pabsd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpabsd (%rdi), %ymm1 # sched: [8:0.50]
-; ZNVER1-NEXT: vpabsd %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x i32> @llvm.x86.avx2.pabs.d(<8 x i32> %a0)
- %2 = load <8 x i32>, <8 x i32> *%a1, align 32
- %3 = call <8 x i32> @llvm.x86.avx2.pabs.d(<8 x i32> %2)
- %4 = or <8 x i32> %1, %3
- ret <8 x i32> %4
-}
-declare <8 x i32> @llvm.x86.avx2.pabs.d(<8 x i32>) nounwind readnone
-
-define <16 x i16> @test_pabsw(<16 x i16> %a0, <16 x i16> *%a1) {
-; GENERIC-LABEL: test_pabsw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpabsw %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpabsw (%rdi), %ymm1 # sched: [8:0.50]
-; GENERIC-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pabsw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpabsw %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpabsw (%rdi), %ymm1 # sched: [8:0.50]
-; HASWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pabsw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpabsw %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpabsw (%rdi), %ymm1 # sched: [7:0.50]
-; BROADWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pabsw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpabsw %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpabsw (%rdi), %ymm1 # sched: [8:0.50]
-; SKYLAKE-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pabsw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpabsw %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpabsw (%rdi), %ymm1 # sched: [8:0.50]
-; SKX-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pabsw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpabsw (%rdi), %ymm1 # sched: [8:0.50]
-; ZNVER1-NEXT: vpabsw %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <16 x i16> @llvm.x86.avx2.pabs.w(<16 x i16> %a0)
- %2 = load <16 x i16>, <16 x i16> *%a1, align 32
- %3 = call <16 x i16> @llvm.x86.avx2.pabs.w(<16 x i16> %2)
- %4 = or <16 x i16> %1, %3
- ret <16 x i16> %4
-}
-declare <16 x i16> @llvm.x86.avx2.pabs.w(<16 x i16>) nounwind readnone
-
-define <16 x i16> @test_packssdw(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
-; GENERIC-LABEL: test_packssdw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpackssdw %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpackssdw (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_packssdw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpackssdw %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: vpackssdw (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_packssdw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpackssdw %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: vpackssdw (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_packssdw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpackssdw %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SKYLAKE-NEXT: vpackssdw (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_packssdw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpackssdw %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SKX-NEXT: vpackssdw (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_packssdw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpackssdw %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpackssdw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> %a0, <8 x i32> %a1)
- %2 = bitcast <16 x i16> %1 to <8 x i32>
- %3 = load <8 x i32>, <8 x i32> *%a2, align 32
- %4 = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> %2, <8 x i32> %3)
- ret <16 x i16> %4
-}
-declare <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32>, <8 x i32>) nounwind readnone
-
-define <32 x i8> @test_packsswb(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_packsswb:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpacksswb %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpacksswb (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_packsswb:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpacksswb %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: vpacksswb (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_packsswb:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpacksswb %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: vpacksswb (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_packsswb:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpacksswb %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SKYLAKE-NEXT: vpacksswb (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_packsswb:
-; SKX: # %bb.0:
-; SKX-NEXT: vpacksswb %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SKX-NEXT: vpacksswb (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_packsswb:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpacksswb %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpacksswb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <32 x i8> @llvm.x86.avx2.packsswb(<16 x i16> %a0, <16 x i16> %a1)
- %2 = bitcast <32 x i8> %1 to <16 x i16>
- %3 = load <16 x i16>, <16 x i16> *%a2, align 32
- %4 = call <32 x i8> @llvm.x86.avx2.packsswb(<16 x i16> %2, <16 x i16> %3)
- ret <32 x i8> %4
-}
-declare <32 x i8> @llvm.x86.avx2.packsswb(<16 x i16>, <16 x i16>) nounwind readnone
-
-define <16 x i16> @test_packusdw(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
-; GENERIC-LABEL: test_packusdw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpackusdw (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_packusdw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: vpackusdw (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_packusdw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: vpackusdw (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_packusdw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SKYLAKE-NEXT: vpackusdw (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_packusdw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SKX-NEXT: vpackusdw (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_packusdw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpackusdw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32> %a0, <8 x i32> %a1)
- %2 = bitcast <16 x i16> %1 to <8 x i32>
- %3 = load <8 x i32>, <8 x i32> *%a2, align 32
- %4 = call <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32> %2, <8 x i32> %3)
- ret <16 x i16> %4
-}
-declare <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32>, <8 x i32>) nounwind readnone
-
-define <32 x i8> @test_packuswb(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_packuswb:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpackuswb %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpackuswb (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_packuswb:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpackuswb %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: vpackuswb (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_packuswb:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpackuswb %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: vpackuswb (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_packuswb:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpackuswb %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SKYLAKE-NEXT: vpackuswb (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_packuswb:
-; SKX: # %bb.0:
-; SKX-NEXT: vpackuswb %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SKX-NEXT: vpackuswb (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_packuswb:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpackuswb %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpackuswb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16> %a0, <16 x i16> %a1)
- %2 = bitcast <32 x i8> %1 to <16 x i16>
- %3 = load <16 x i16>, <16 x i16> *%a2, align 32
- %4 = call <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16> %2, <16 x i16> %3)
- ret <32 x i8> %4
-}
-declare <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16>, <16 x i16>) nounwind readnone
-
-define <32 x i8> @test_paddb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) {
-; GENERIC-LABEL: test_paddb:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpaddb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_paddb:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpaddb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_paddb:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpaddb (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_paddb:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: vpaddb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_paddb:
-; SKX: # %bb.0:
-; SKX-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: vpaddb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_paddb:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpaddb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = add <32 x i8> %a0, %a1
- %2 = load <32 x i8>, <32 x i8> *%a2, align 32
- %3 = add <32 x i8> %1, %2
- ret <32 x i8> %3
-}
-
-define <8 x i32> @test_paddd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
-; GENERIC-LABEL: test_paddd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpaddd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_paddd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpaddd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_paddd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpaddd (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_paddd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: vpaddd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_paddd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: vpaddd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_paddd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpaddd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = add <8 x i32> %a0, %a1
- %2 = load <8 x i32>, <8 x i32> *%a2, align 32
- %3 = add <8 x i32> %1, %2
- ret <8 x i32> %3
-}
-
-define <4 x i64> @test_paddq(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) {
-; GENERIC-LABEL: test_paddq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpaddq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_paddq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpaddq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_paddq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpaddq (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_paddq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: vpaddq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_paddq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: vpaddq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_paddq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpaddq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = add <4 x i64> %a0, %a1
- %2 = load <4 x i64>, <4 x i64> *%a2, align 32
- %3 = add <4 x i64> %1, %2
- ret <4 x i64> %3
-}
-
-define <32 x i8> @test_paddsb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) {
-; GENERIC-LABEL: test_paddsb:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpaddsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpaddsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_paddsb:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpaddsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpaddsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_paddsb:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpaddsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpaddsb (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_paddsb:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpaddsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpaddsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_paddsb:
-; SKX: # %bb.0:
-; SKX-NEXT: vpaddsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpaddsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_paddsb:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpaddsb %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpaddsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <32 x i8> @llvm.sadd.sat.v32i8(<32 x i8> %a0, <32 x i8> %a1)
- %2 = load <32 x i8>, <32 x i8> *%a2, align 32
- %3 = call <32 x i8> @llvm.sadd.sat.v32i8(<32 x i8> %1, <32 x i8> %2)
- ret <32 x i8> %3
-}
-declare <32 x i8> @llvm.sadd.sat.v32i8(<32 x i8>, <32 x i8>) nounwind readnone
-
-define <16 x i16> @test_paddsw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_paddsw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpaddsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpaddsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_paddsw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpaddsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpaddsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_paddsw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpaddsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpaddsw (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_paddsw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpaddsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpaddsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_paddsw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpaddsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpaddsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_paddsw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpaddsw %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpaddsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <16 x i16> @llvm.sadd.sat.v16i16(<16 x i16> %a0, <16 x i16> %a1)
- %2 = load <16 x i16>, <16 x i16> *%a2, align 32
- %3 = call <16 x i16> @llvm.sadd.sat.v16i16(<16 x i16> %1, <16 x i16> %2)
- ret <16 x i16> %3
-}
-declare <16 x i16> @llvm.sadd.sat.v16i16(<16 x i16>, <16 x i16>) nounwind readnone
-
-define <32 x i8> @test_paddusb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) {
-; GENERIC-LABEL: test_paddusb:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpaddusb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpaddusb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_paddusb:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpaddusb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpaddusb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_paddusb:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpaddusb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpaddusb (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_paddusb:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpaddusb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpaddusb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_paddusb:
-; SKX: # %bb.0:
-; SKX-NEXT: vpaddusb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpaddusb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_paddusb:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpaddusb %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpaddusb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <32 x i8> @llvm.uadd.sat.v32i8(<32 x i8> %a0, <32 x i8> %a1)
- %2 = load <32 x i8>, <32 x i8> *%a2, align 32
- %3 = call <32 x i8> @llvm.uadd.sat.v32i8(<32 x i8> %1, <32 x i8> %2)
- ret <32 x i8> %3
-}
-declare <32 x i8> @llvm.uadd.sat.v32i8(<32 x i8>, <32 x i8>) nounwind readnone
-
-define <16 x i16> @test_paddusw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_paddusw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpaddusw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpaddusw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_paddusw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpaddusw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpaddusw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_paddusw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpaddusw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpaddusw (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_paddusw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpaddusw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpaddusw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_paddusw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpaddusw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpaddusw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_paddusw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpaddusw %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpaddusw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <16 x i16> @llvm.uadd.sat.v16i16(<16 x i16> %a0, <16 x i16> %a1)
- %2 = load <16 x i16>, <16 x i16> *%a2, align 32
- %3 = call <16 x i16> @llvm.uadd.sat.v16i16(<16 x i16> %1, <16 x i16> %2)
- ret <16 x i16> %3
-}
-declare <16 x i16> @llvm.uadd.sat.v16i16(<16 x i16>, <16 x i16>) nounwind readnone
-
-define <16 x i16> @test_paddw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_paddw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpaddw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_paddw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpaddw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_paddw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpaddw (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_paddw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: vpaddw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_paddw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: vpaddw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_paddw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpaddw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = add <16 x i16> %a0, %a1
- %2 = load <16 x i16>, <16 x i16> *%a2, align 32
- %3 = add <16 x i16> %1, %2
- ret <16 x i16> %3
-}
-
-define <32 x i8> @test_palignr(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) {
-; GENERIC-LABEL: test_palignr:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpalignr {{.*#+}} ymm1 = ymm1[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0],ymm1[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16] sched: [1:1.00]
-; GENERIC-NEXT: vpalignr {{.*#+}} ymm0 = ymm0[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm1[0],ymm0[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm1[16] sched: [1:1.00]
-; GENERIC-NEXT: vpaddb %ymm0, %ymm1, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_palignr:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpalignr {{.*#+}} ymm1 = ymm1[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0],ymm1[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16] sched: [1:1.00]
-; HASWELL-NEXT: vpalignr {{.*#+}} ymm0 = ymm0[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm1[0],ymm0[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm1[16] sched: [1:1.00]
-; HASWELL-NEXT: vpaddb %ymm0, %ymm1, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_palignr:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpalignr {{.*#+}} ymm1 = ymm1[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0],ymm1[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16] sched: [1:1.00]
-; BROADWELL-NEXT: vpalignr {{.*#+}} ymm0 = ymm0[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm1[0],ymm0[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm1[16] sched: [1:1.00]
-; BROADWELL-NEXT: vpaddb %ymm0, %ymm1, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_palignr:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpalignr {{.*#+}} ymm1 = ymm1[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0],ymm1[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16] sched: [1:1.00]
-; SKYLAKE-NEXT: vpalignr {{.*#+}} ymm0 = ymm0[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm1[0],ymm0[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm1[16] sched: [1:1.00]
-; SKYLAKE-NEXT: vpaddb %ymm0, %ymm1, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_palignr:
-; SKX: # %bb.0:
-; SKX-NEXT: vpalignr {{.*#+}} ymm1 = ymm1[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0],ymm1[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16] sched: [1:1.00]
-; SKX-NEXT: vpalignr {{.*#+}} ymm0 = ymm0[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm1[0],ymm0[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm1[16] sched: [1:1.00]
-; SKX-NEXT: vpaddb %ymm0, %ymm1, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_palignr:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpalignr {{.*#+}} ymm1 = ymm1[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0],ymm1[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16] sched: [1:0.25]
-; ZNVER1-NEXT: vpalignr {{.*#+}} ymm0 = ymm0[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm1[0],ymm0[17,18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm1[16] sched: [1:0.25]
-; ZNVER1-NEXT: vpaddb %ymm0, %ymm1, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <32 x i8> %a1, <32 x i8> %a0, <32 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 32, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 48>
- %2 = load <32 x i8>, <32 x i8> *%a2, align 32
- %3 = shufflevector <32 x i8> %a0, <32 x i8> %1, <32 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 32, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 48>
- %4 = add <32 x i8> %1, %3
- ret <32 x i8> %4
-}
-
-define <4 x i64> @test_pand(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) {
-; GENERIC-LABEL: test_pand:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpand %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; GENERIC-NEXT: vpand (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pand:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpand %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; HASWELL-NEXT: vpand (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pand:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpand %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; BROADWELL-NEXT: vpand (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pand:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpand %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: vpand (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pand:
-; SKX: # %bb.0:
-; SKX-NEXT: vpand %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: vpand (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pand:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpand %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpand (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = and <4 x i64> %a0, %a1
- %2 = load <4 x i64>, <4 x i64> *%a2, align 32
- %3 = and <4 x i64> %1, %2
- %4 = add <4 x i64> %3, %a1
- ret <4 x i64> %4
-}
-
-define <4 x i64> @test_pandn(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) {
-; GENERIC-LABEL: test_pandn:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpandn %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; GENERIC-NEXT: vpandn (%rdi), %ymm0, %ymm1 # sched: [8:0.50]
-; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pandn:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpandn %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; HASWELL-NEXT: vpandn (%rdi), %ymm0, %ymm1 # sched: [8:0.50]
-; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pandn:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpandn %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; BROADWELL-NEXT: vpandn (%rdi), %ymm0, %ymm1 # sched: [7:0.50]
-; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pandn:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpandn %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: vpandn (%rdi), %ymm0, %ymm1 # sched: [8:0.50]
-; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pandn:
-; SKX: # %bb.0:
-; SKX-NEXT: vpandn %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: vpandn (%rdi), %ymm0, %ymm1 # sched: [8:0.50]
-; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pandn:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpandn %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpandn (%rdi), %ymm0, %ymm1 # sched: [8:0.50]
-; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = xor <4 x i64> %a0, <i64 -1, i64 -1, i64 -1, i64 -1>
- %2 = and <4 x i64> %a1, %1
- %3 = load <4 x i64>, <4 x i64> *%a2, align 32
- %4 = xor <4 x i64> %2, <i64 -1, i64 -1, i64 -1, i64 -1>
- %5 = and <4 x i64> %3, %4
- %6 = add <4 x i64> %2, %5
- ret <4 x i64> %6
-}
-
-define <32 x i8> @test_pavgb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) {
-; GENERIC-LABEL: test_pavgb:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpavgb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpavgb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pavgb:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpavgb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpavgb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pavgb:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpavgb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpavgb (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pavgb:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpavgb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpavgb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pavgb:
-; SKX: # %bb.0:
-; SKX-NEXT: vpavgb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpavgb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pavgb:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpavgb %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpavgb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = zext <32 x i8> %a0 to <32 x i16>
- %2 = zext <32 x i8> %a1 to <32 x i16>
- %3 = add <32 x i16> %1, %2
- %4 = add <32 x i16> %3, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
- %5 = lshr <32 x i16> %4, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
- %6 = trunc <32 x i16> %5 to <32 x i8>
- %7 = load <32 x i8>, <32 x i8> *%a2, align 32
- %8 = zext <32 x i8> %6 to <32 x i16>
- %9 = zext <32 x i8> %7 to <32 x i16>
- %10 = add <32 x i16> %8, %9
- %11 = add <32 x i16> %10, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
- %12 = lshr <32 x i16> %11, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
- %13 = trunc <32 x i16> %12 to <32 x i8>
- ret <32 x i8> %13
-}
-
-define <16 x i16> @test_pavgw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_pavgw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpavgw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpavgw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pavgw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpavgw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpavgw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pavgw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpavgw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpavgw (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pavgw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpavgw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpavgw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pavgw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpavgw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpavgw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pavgw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpavgw %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpavgw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = zext <16 x i16> %a0 to <16 x i32>
- %2 = zext <16 x i16> %a1 to <16 x i32>
- %3 = add <16 x i32> %1, %2
- %4 = add <16 x i32> %3, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
- %5 = lshr <16 x i32> %4, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
- %6 = trunc <16 x i32> %5 to <16 x i16>
- %7 = load <16 x i16>, <16 x i16> *%a2, align 32
- %8 = zext <16 x i16> %6 to <16 x i32>
- %9 = zext <16 x i16> %7 to <16 x i32>
- %10 = add <16 x i32> %8, %9
- %11 = add <16 x i32> %10, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
- %12 = lshr <16 x i32> %11, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
- %13 = trunc <16 x i32> %12 to <16 x i16>
- ret <16 x i16> %13
-}
-
-define <4 x i32> @test_pblendd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {
-; GENERIC-LABEL: test_pblendd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3] sched: [1:0.50]
-; GENERIC-NEXT: vpblendd {{.*#+}} xmm1 = mem[0],xmm1[1],mem[2],xmm1[3] sched: [7:0.50]
-; GENERIC-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pblendd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3] sched: [1:0.33]
-; HASWELL-NEXT: vpblendd {{.*#+}} xmm1 = mem[0],xmm1[1],mem[2],xmm1[3] sched: [7:0.50]
-; HASWELL-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pblendd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3] sched: [1:0.33]
-; BROADWELL-NEXT: vpblendd {{.*#+}} xmm1 = mem[0],xmm1[1],mem[2],xmm1[3] sched: [6:0.50]
-; BROADWELL-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pblendd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3] sched: [1:0.33]
-; SKYLAKE-NEXT: vpblendd {{.*#+}} xmm1 = mem[0],xmm1[1],mem[2],xmm1[3] sched: [7:0.50]
-; SKYLAKE-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pblendd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3] sched: [1:0.33]
-; SKX-NEXT: vpblendd {{.*#+}} xmm1 = mem[0],xmm1[1],mem[2],xmm1[3] sched: [7:0.50]
-; SKX-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pblendd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpblendd {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[3] sched: [1:0.50]
-; ZNVER1-NEXT: vpblendd {{.*#+}} xmm1 = mem[0],xmm1[1],mem[2],xmm1[3] sched: [8:1.00]
-; ZNVER1-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <4 x i32> %a0, <4 x i32> %a1, <4 x i32> <i32 4, i32 5, i32 6, i32 3>
- %2 = load <4 x i32>, <4 x i32> *%a2, align 16
- %3 = shufflevector <4 x i32> %a1, <4 x i32> %2, <4 x i32> <i32 4, i32 1, i32 6, i32 3>
- %4 = add <4 x i32> %1, %3
- ret <4 x i32> %4
-}
-
-define <8 x i32> @test_pblendd_ymm(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
-; GENERIC-LABEL: test_pblendd_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7] sched: [1:0.50]
-; GENERIC-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0],mem[1,2],ymm1[3,4,5,6,7] sched: [8:0.50]
-; GENERIC-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pblendd_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7] sched: [1:0.33]
-; HASWELL-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0],mem[1,2],ymm1[3,4,5,6,7] sched: [8:0.50]
-; HASWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pblendd_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7] sched: [1:0.33]
-; BROADWELL-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0],mem[1,2],ymm1[3,4,5,6,7] sched: [7:0.50]
-; BROADWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pblendd_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7] sched: [1:0.33]
-; SKYLAKE-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0],mem[1,2],ymm1[3,4,5,6,7] sched: [8:0.50]
-; SKYLAKE-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pblendd_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7] sched: [1:0.33]
-; SKX-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0],mem[1,2],ymm1[3,4,5,6,7] sched: [8:0.50]
-; SKX-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pblendd_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6],ymm1[7] sched: [1:0.50]
-; ZNVER1-NEXT: vpblendd {{.*#+}} ymm1 = ymm1[0],mem[1,2],ymm1[3,4,5,6,7] sched: [9:1.50]
-; ZNVER1-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <8 x i32> %a0, <8 x i32> %a1, <8 x i32> <i32 8, i32 9, i32 10, i32 3, i32 4, i32 5, i32 6, i32 15>
- %2 = load <8 x i32>, <8 x i32> *%a2, align 32
- %3 = shufflevector <8 x i32> %a1, <8 x i32> %2, <8 x i32> <i32 0, i32 9, i32 10, i32 3, i32 4, i32 5, i32 6, i32 7>
- %4 = add <8 x i32> %1, %3
- ret <8 x i32> %4
-}
-
-define <32 x i8> @test_pblendvb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> %a2, <32 x i8> *%a3, <32 x i8> %a4) {
-; GENERIC-LABEL: test_pblendvb:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpblendvb %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:1.00]
-; GENERIC-NEXT: vpblendvb %ymm3, (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pblendvb:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpblendvb %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:2.00]
-; HASWELL-NEXT: vpblendvb %ymm3, (%rdi), %ymm0, %ymm0 # sched: [9:2.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pblendvb:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpblendvb %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:2.00]
-; BROADWELL-NEXT: vpblendvb %ymm3, (%rdi), %ymm0, %ymm0 # sched: [8:2.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pblendvb:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpblendvb %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:0.67]
-; SKYLAKE-NEXT: vpblendvb %ymm3, (%rdi), %ymm0, %ymm0 # sched: [8:0.67]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pblendvb:
-; SKX: # %bb.0:
-; SKX-NEXT: vpblendvb %ymm2, %ymm1, %ymm0, %ymm0 # sched: [2:0.67]
-; SKX-NEXT: vpblendvb %ymm3, (%rdi), %ymm0, %ymm0 # sched: [8:0.67]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pblendvb:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpblendvb %ymm2, %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; ZNVER1-NEXT: vpblendvb %ymm3, (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <32 x i8> @llvm.x86.avx2.pblendvb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> %a2)
- %2 = load <32 x i8>, <32 x i8> *%a3, align 32
- %3 = call <32 x i8> @llvm.x86.avx2.pblendvb(<32 x i8> %1, <32 x i8> %2, <32 x i8> %a4)
- ret <32 x i8> %3
-}
-declare <32 x i8> @llvm.x86.avx2.pblendvb(<32 x i8>, <32 x i8>, <32 x i8>) nounwind readnone
-
-define <16 x i16> @test_pblendw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_pblendw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7,8,9],ymm1[10,11,12],ymm0[13,14,15] sched: [1:0.50]
-; GENERIC-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm1[1],mem[2],ymm1[3],mem[4],ymm1[5],mem[6],ymm1[7],mem[8],ymm1[9],mem[10],ymm1[11],mem[12],ymm1[13],mem[14],ymm1[15] sched: [8:0.50]
-; GENERIC-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pblendw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7,8,9],ymm1[10,11,12],ymm0[13,14,15] sched: [1:1.00]
-; HASWELL-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm1[1],mem[2],ymm1[3],mem[4],ymm1[5],mem[6],ymm1[7],mem[8],ymm1[9],mem[10],ymm1[11],mem[12],ymm1[13],mem[14],ymm1[15] sched: [8:1.00]
-; HASWELL-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pblendw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7,8,9],ymm1[10,11,12],ymm0[13,14,15] sched: [1:1.00]
-; BROADWELL-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm1[1],mem[2],ymm1[3],mem[4],ymm1[5],mem[6],ymm1[7],mem[8],ymm1[9],mem[10],ymm1[11],mem[12],ymm1[13],mem[14],ymm1[15] sched: [7:1.00]
-; BROADWELL-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pblendw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7,8,9],ymm1[10,11,12],ymm0[13,14,15] sched: [1:1.00]
-; SKYLAKE-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm1[1],mem[2],ymm1[3],mem[4],ymm1[5],mem[6],ymm1[7],mem[8],ymm1[9],mem[10],ymm1[11],mem[12],ymm1[13],mem[14],ymm1[15] sched: [8:1.00]
-; SKYLAKE-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pblendw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7,8,9],ymm1[10,11,12],ymm0[13,14,15] sched: [1:1.00]
-; SKX-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm1[1],mem[2],ymm1[3],mem[4],ymm1[5],mem[6],ymm1[7],mem[8],ymm1[9],mem[10],ymm1[11],mem[12],ymm1[13],mem[14],ymm1[15] sched: [8:1.00]
-; SKX-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pblendw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3,4],ymm0[5,6,7,8,9],ymm1[10,11,12],ymm0[13,14,15] sched: [2:0.33]
-; ZNVER1-NEXT: vpblendw {{.*#+}} ymm1 = mem[0],ymm1[1],mem[2],ymm1[3],mem[4],ymm1[5],mem[6],ymm1[7],mem[8],ymm1[9],mem[10],ymm1[11],mem[12],ymm1[13],mem[14],ymm1[15] sched: [9:0.50]
-; ZNVER1-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <16 x i16> %a0, <16 x i16> %a1, <16 x i32> <i32 0, i32 1, i32 18, i32 19, i32 20, i32 5, i32 6, i32 7, i32 8, i32 9, i32 26, i32 27, i32 28, i32 13, i32 14, i32 15>
- %2 = load <16 x i16>, <16 x i16> *%a2, align 32
- %3 = shufflevector <16 x i16> %a1, <16 x i16> %2, <16 x i32> <i32 16, i32 1, i32 18, i32 3, i32 20, i32 5, i32 22, i32 7, i32 24, i32 9, i32 26, i32 11, i32 28, i32 13, i32 30, i32 15>
- %4 = add <16 x i16> %1, %3
- ret <16 x i16> %4
-}
-
-define <16 x i8> @test_pbroadcastb(<16 x i8> %a0, <16 x i8> *%a1) {
-; GENERIC-LABEL: test_pbroadcastb:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpbroadcastb %xmm0, %xmm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpbroadcastb (%rdi), %xmm1 # sched: [7:0.50]
-; GENERIC-NEXT: vpaddb %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pbroadcastb:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpbroadcastb %xmm0, %xmm0 # sched: [3:1.00]
-; HASWELL-NEXT: vpbroadcastb (%rdi), %xmm1 # sched: [9:1.00]
-; HASWELL-NEXT: vpaddb %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pbroadcastb:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpbroadcastb (%rdi), %xmm1 # sched: [9:1.00]
-; BROADWELL-NEXT: vpbroadcastb %xmm0, %xmm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vpaddb %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pbroadcastb:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpbroadcastb %xmm0, %xmm0 # sched: [3:1.00]
-; SKYLAKE-NEXT: vpbroadcastb (%rdi), %xmm1 # sched: [7:1.00]
-; SKYLAKE-NEXT: vpaddb %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pbroadcastb:
-; SKX: # %bb.0:
-; SKX-NEXT: vpbroadcastb %xmm0, %xmm0 # sched: [3:1.00]
-; SKX-NEXT: vpbroadcastb (%rdi), %xmm1 # sched: [7:1.00]
-; SKX-NEXT: vpaddb %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pbroadcastb:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpbroadcastb (%rdi), %xmm1 # sched: [8:1.00]
-; ZNVER1-NEXT: vpbroadcastb %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpaddb %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <16 x i8> %a0, <16 x i8> undef, <16 x i32> zeroinitializer
- %2 = load <16 x i8>, <16 x i8> *%a1, align 16
- %3 = shufflevector <16 x i8> %2, <16 x i8> undef, <16 x i32> zeroinitializer
- %4 = add <16 x i8> %1, %3
- ret <16 x i8> %4
-}
-
-define <32 x i8> @test_pbroadcastb_ymm(<32 x i8> %a0, <32 x i8> *%a1) {
-; GENERIC-LABEL: test_pbroadcastb_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpbroadcastb %xmm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpbroadcastb (%rdi), %ymm1 # sched: [7:0.50]
-; GENERIC-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pbroadcastb_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpbroadcastb %xmm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vpbroadcastb (%rdi), %ymm1 # sched: [9:1.00]
-; HASWELL-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pbroadcastb_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpbroadcastb (%rdi), %ymm1 # sched: [9:1.00]
-; BROADWELL-NEXT: vpbroadcastb %xmm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pbroadcastb_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpbroadcastb %xmm0, %ymm0 # sched: [3:1.00]
-; SKYLAKE-NEXT: vpbroadcastb (%rdi), %ymm1 # sched: [8:1.00]
-; SKYLAKE-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pbroadcastb_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vpbroadcastb %xmm0, %ymm0 # sched: [3:1.00]
-; SKX-NEXT: vpbroadcastb (%rdi), %ymm1 # sched: [8:1.00]
-; SKX-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pbroadcastb_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpbroadcastb (%rdi), %ymm1 # sched: [8:2.00]
-; ZNVER1-NEXT: vpbroadcastb %xmm0, %ymm0 # sched: [2:0.25]
-; ZNVER1-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <32 x i8> %a0, <32 x i8> undef, <32 x i32> zeroinitializer
- %2 = load <32 x i8>, <32 x i8> *%a1, align 32
- %3 = shufflevector <32 x i8> %2, <32 x i8> undef, <32 x i32> zeroinitializer
- %4 = add <32 x i8> %1, %3
- ret <32 x i8> %4
-}
-
-define <4 x i32> @test_pbroadcastd(<4 x i32> %a0, <4 x i32> *%a1) {
-; GENERIC-LABEL: test_pbroadcastd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpbroadcastd %xmm0, %xmm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpbroadcastd (%rdi), %xmm1 # sched: [7:0.50]
-; GENERIC-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pbroadcastd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpbroadcastd %xmm0, %xmm0 # sched: [1:1.00]
-; HASWELL-NEXT: vpbroadcastd (%rdi), %xmm1 # sched: [6:0.50]
-; HASWELL-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pbroadcastd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpbroadcastd %xmm0, %xmm0 # sched: [1:1.00]
-; BROADWELL-NEXT: vpbroadcastd (%rdi), %xmm1 # sched: [5:0.50]
-; BROADWELL-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pbroadcastd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpbroadcastd %xmm0, %xmm0 # sched: [1:1.00]
-; SKYLAKE-NEXT: vpbroadcastd (%rdi), %xmm1 # sched: [6:0.50]
-; SKYLAKE-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pbroadcastd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpbroadcastd %xmm0, %xmm0 # sched: [1:1.00]
-; SKX-NEXT: vpbroadcastd (%rdi), %xmm1 # sched: [6:0.50]
-; SKX-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pbroadcastd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpbroadcastd (%rdi), %xmm1 # sched: [8:0.50]
-; ZNVER1-NEXT: vpbroadcastd %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpaddd %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <4 x i32> %a0, <4 x i32> undef, <4 x i32> zeroinitializer
- %2 = load <4 x i32>, <4 x i32> *%a1, align 16
- %3 = shufflevector <4 x i32> %2, <4 x i32> undef, <4 x i32> zeroinitializer
- %4 = add <4 x i32> %1, %3
- ret <4 x i32> %4
-}
-
-define <8 x i32> @test_pbroadcastd_ymm(<8 x i32> %a0, <8 x i32> *%a1) {
-; GENERIC-LABEL: test_pbroadcastd_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpbroadcastd %xmm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpbroadcastd (%rdi), %ymm1 # sched: [7:0.50]
-; GENERIC-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pbroadcastd_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpbroadcastd %xmm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vpbroadcastd (%rdi), %ymm1 # sched: [7:0.50]
-; HASWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pbroadcastd_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpbroadcastd %xmm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vpbroadcastd (%rdi), %ymm1 # sched: [6:0.50]
-; BROADWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pbroadcastd_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpbroadcastd %xmm0, %ymm0 # sched: [3:1.00]
-; SKYLAKE-NEXT: vpbroadcastd (%rdi), %ymm1 # sched: [7:0.50]
-; SKYLAKE-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pbroadcastd_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vpbroadcastd %xmm0, %ymm0 # sched: [3:1.00]
-; SKX-NEXT: vpbroadcastd (%rdi), %ymm1 # sched: [7:0.50]
-; SKX-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pbroadcastd_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpbroadcastd (%rdi), %ymm1 # sched: [8:0.50]
-; ZNVER1-NEXT: vpbroadcastd %xmm0, %ymm0 # sched: [2:0.25]
-; ZNVER1-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <8 x i32> %a0, <8 x i32> undef, <8 x i32> zeroinitializer
- %2 = load <8 x i32>, <8 x i32> *%a1, align 32
- %3 = shufflevector <8 x i32> %2, <8 x i32> undef, <8 x i32> zeroinitializer
- %4 = add <8 x i32> %1, %3
- ret <8 x i32> %4
-}
-
-define <2 x i64> @test_pbroadcastq(<2 x i64> %a0, <2 x i64> *%a1) {
-; GENERIC-LABEL: test_pbroadcastq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpbroadcastq %xmm0, %xmm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpbroadcastq (%rdi), %xmm1 # sched: [7:0.50]
-; GENERIC-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pbroadcastq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpbroadcastq %xmm0, %xmm0 # sched: [1:1.00]
-; HASWELL-NEXT: vpbroadcastq (%rdi), %xmm1 # sched: [6:0.50]
-; HASWELL-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pbroadcastq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpbroadcastq %xmm0, %xmm0 # sched: [1:1.00]
-; BROADWELL-NEXT: vpbroadcastq (%rdi), %xmm1 # sched: [5:0.50]
-; BROADWELL-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pbroadcastq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpbroadcastq %xmm0, %xmm0 # sched: [1:1.00]
-; SKYLAKE-NEXT: vpbroadcastq (%rdi), %xmm1 # sched: [6:0.50]
-; SKYLAKE-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pbroadcastq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpbroadcastq %xmm0, %xmm0 # sched: [1:1.00]
-; SKX-NEXT: vpbroadcastq (%rdi), %xmm1 # sched: [6:0.50]
-; SKX-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pbroadcastq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpbroadcastq (%rdi), %xmm1 # sched: [8:0.50]
-; ZNVER1-NEXT: vpbroadcastq %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpaddq %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <2 x i64> %a0, <2 x i64> undef, <2 x i32> zeroinitializer
- %2 = load <2 x i64>, <2 x i64> *%a1, align 16
- %3 = shufflevector <2 x i64> %2, <2 x i64> undef, <2 x i32> zeroinitializer
- %4 = add <2 x i64> %1, %3
- ret <2 x i64> %4
-}
-
-define <4 x i64> @test_pbroadcastq_ymm(<4 x i64> %a0, <4 x i64> *%a1) {
-; GENERIC-LABEL: test_pbroadcastq_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpbroadcastq %xmm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpbroadcastq (%rdi), %ymm1 # sched: [7:0.50]
-; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pbroadcastq_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpbroadcastq %xmm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vpbroadcastq (%rdi), %ymm1 # sched: [7:0.50]
-; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pbroadcastq_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpbroadcastq %xmm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vpbroadcastq (%rdi), %ymm1 # sched: [6:0.50]
-; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pbroadcastq_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpbroadcastq %xmm0, %ymm0 # sched: [3:1.00]
-; SKYLAKE-NEXT: vpbroadcastq (%rdi), %ymm1 # sched: [7:0.50]
-; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pbroadcastq_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vpbroadcastq %xmm0, %ymm0 # sched: [3:1.00]
-; SKX-NEXT: vpbroadcastq (%rdi), %ymm1 # sched: [7:0.50]
-; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pbroadcastq_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpbroadcastq (%rdi), %ymm1 # sched: [8:0.50]
-; ZNVER1-NEXT: vpbroadcastq %xmm0, %ymm0 # sched: [2:0.25]
-; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <4 x i64> %a0, <4 x i64> undef, <4 x i32> zeroinitializer
- %2 = load <4 x i64>, <4 x i64> *%a1, align 32
- %3 = shufflevector <4 x i64> %2, <4 x i64> undef, <4 x i32> zeroinitializer
- %4 = add <4 x i64> %1, %3
- ret <4 x i64> %4
-}
-
-define <8 x i16> @test_pbroadcastw(<8 x i16> %a0, <8 x i16> *%a1) {
-; GENERIC-LABEL: test_pbroadcastw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpbroadcastw %xmm0, %xmm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpbroadcastw (%rdi), %xmm1 # sched: [7:0.50]
-; GENERIC-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pbroadcastw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpbroadcastw %xmm0, %xmm0 # sched: [3:1.00]
-; HASWELL-NEXT: vpbroadcastw (%rdi), %xmm1 # sched: [9:1.00]
-; HASWELL-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pbroadcastw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpbroadcastw (%rdi), %xmm1 # sched: [9:1.00]
-; BROADWELL-NEXT: vpbroadcastw %xmm0, %xmm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pbroadcastw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpbroadcastw %xmm0, %xmm0 # sched: [3:1.00]
-; SKYLAKE-NEXT: vpbroadcastw (%rdi), %xmm1 # sched: [7:1.00]
-; SKYLAKE-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pbroadcastw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpbroadcastw %xmm0, %xmm0 # sched: [3:1.00]
-; SKX-NEXT: vpbroadcastw (%rdi), %xmm1 # sched: [7:1.00]
-; SKX-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pbroadcastw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpbroadcastw (%rdi), %xmm1 # sched: [8:1.00]
-; ZNVER1-NEXT: vpbroadcastw %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <8 x i16> %a0, <8 x i16> undef, <8 x i32> zeroinitializer
- %2 = load <8 x i16>, <8 x i16> *%a1, align 16
- %3 = shufflevector <8 x i16> %2, <8 x i16> undef, <8 x i32> zeroinitializer
- %4 = add <8 x i16> %1, %3
- ret <8 x i16> %4
-}
-
-define <16 x i16> @test_pbroadcastw_ymm(<16 x i16> %a0, <16 x i16> *%a1) {
-; GENERIC-LABEL: test_pbroadcastw_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpbroadcastw %xmm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpbroadcastw (%rdi), %ymm1 # sched: [7:0.50]
-; GENERIC-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pbroadcastw_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpbroadcastw %xmm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vpbroadcastw (%rdi), %ymm1 # sched: [9:1.00]
-; HASWELL-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pbroadcastw_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpbroadcastw (%rdi), %ymm1 # sched: [9:1.00]
-; BROADWELL-NEXT: vpbroadcastw %xmm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pbroadcastw_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpbroadcastw %xmm0, %ymm0 # sched: [3:1.00]
-; SKYLAKE-NEXT: vpbroadcastw (%rdi), %ymm1 # sched: [8:1.00]
-; SKYLAKE-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pbroadcastw_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vpbroadcastw %xmm0, %ymm0 # sched: [3:1.00]
-; SKX-NEXT: vpbroadcastw (%rdi), %ymm1 # sched: [8:1.00]
-; SKX-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pbroadcastw_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpbroadcastw (%rdi), %ymm1 # sched: [8:2.00]
-; ZNVER1-NEXT: vpbroadcastw %xmm0, %ymm0 # sched: [2:0.25]
-; ZNVER1-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <16 x i16> %a0, <16 x i16> undef, <16 x i32> zeroinitializer
- %2 = load <16 x i16>, <16 x i16> *%a1, align 32
- %3 = shufflevector <16 x i16> %2, <16 x i16> undef, <16 x i32> zeroinitializer
- %4 = add <16 x i16> %1, %3
- ret <16 x i16> %4
-}
-
-define <32 x i8> @test_pcmpeqb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) {
-; GENERIC-LABEL: test_pcmpeqb:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpcmpeqb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpcmpeqb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pcmpeqb:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpcmpeqb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpcmpeqb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pcmpeqb:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpcmpeqb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpcmpeqb (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pcmpeqb:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpcmpeqb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpcmpeqb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pcmpeqb:
-; SKX: # %bb.0:
-; SKX-NEXT: vpcmpeqb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpcmpeqb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pcmpeqb:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpcmpeqb %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpcmpeqb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = icmp eq <32 x i8> %a0, %a1
- %2 = sext <32 x i1> %1 to <32 x i8>
- %3 = load <32 x i8>, <32 x i8> *%a2, align 32
- %4 = icmp eq <32 x i8> %2, %3
- %5 = sext <32 x i1> %4 to <32 x i8>
- ret <32 x i8> %5
-}
-
-define <8 x i32> @test_pcmpeqd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
-; GENERIC-LABEL: test_pcmpeqd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpcmpeqd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpcmpeqd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pcmpeqd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpcmpeqd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpcmpeqd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pcmpeqd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpcmpeqd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpcmpeqd (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pcmpeqd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpcmpeqd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpcmpeqd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pcmpeqd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpcmpeqd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpcmpeqd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pcmpeqd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpcmpeqd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpcmpeqd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = icmp eq <8 x i32> %a0, %a1
- %2 = sext <8 x i1> %1 to <8 x i32>
- %3 = load <8 x i32>, <8 x i32> *%a2, align 32
- %4 = icmp eq <8 x i32> %2, %3
- %5 = sext <8 x i1> %4 to <8 x i32>
- ret <8 x i32> %5
-}
-
-define <4 x i64> @test_pcmpeqq(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) {
-; GENERIC-LABEL: test_pcmpeqq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpcmpeqq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpcmpeqq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pcmpeqq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpcmpeqq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpcmpeqq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pcmpeqq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpcmpeqq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpcmpeqq (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pcmpeqq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpcmpeqq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpcmpeqq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pcmpeqq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpcmpeqq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpcmpeqq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pcmpeqq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpcmpeqq %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpcmpeqq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = icmp eq <4 x i64> %a0, %a1
- %2 = sext <4 x i1> %1 to <4 x i64>
- %3 = load <4 x i64>, <4 x i64> *%a2, align 32
- %4 = icmp eq <4 x i64> %2, %3
- %5 = sext <4 x i1> %4 to <4 x i64>
- ret <4 x i64> %5
-}
-
-define <16 x i16> @test_pcmpeqw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_pcmpeqw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpcmpeqw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpcmpeqw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pcmpeqw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpcmpeqw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpcmpeqw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pcmpeqw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpcmpeqw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpcmpeqw (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pcmpeqw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpcmpeqw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpcmpeqw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pcmpeqw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpcmpeqw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpcmpeqw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pcmpeqw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpcmpeqw %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpcmpeqw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = icmp eq <16 x i16> %a0, %a1
- %2 = sext <16 x i1> %1 to <16 x i16>
- %3 = load <16 x i16>, <16 x i16> *%a2, align 32
- %4 = icmp eq <16 x i16> %2, %3
- %5 = sext <16 x i1> %4 to <16 x i16>
- ret <16 x i16> %5
-}
-
-define <32 x i8> @test_pcmpgtb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) {
-; GENERIC-LABEL: test_pcmpgtb:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpcmpgtb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpcmpgtb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pcmpgtb:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpcmpgtb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpcmpgtb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pcmpgtb:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpcmpgtb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpcmpgtb (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pcmpgtb:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpcmpgtb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpcmpgtb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pcmpgtb:
-; SKX: # %bb.0:
-; SKX-NEXT: vpcmpgtb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpcmpgtb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pcmpgtb:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpcmpgtb %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpcmpgtb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = icmp sgt <32 x i8> %a0, %a1
- %2 = sext <32 x i1> %1 to <32 x i8>
- %3 = load <32 x i8>, <32 x i8> *%a2, align 32
- %4 = icmp sgt <32 x i8> %2, %3
- %5 = sext <32 x i1> %4 to <32 x i8>
- ret <32 x i8> %5
-}
-
-define <8 x i32> @test_pcmpgtd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
-; GENERIC-LABEL: test_pcmpgtd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpcmpgtd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpcmpgtd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pcmpgtd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpcmpgtd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpcmpgtd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pcmpgtd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpcmpgtd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpcmpgtd (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pcmpgtd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpcmpgtd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpcmpgtd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pcmpgtd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpcmpgtd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpcmpgtd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pcmpgtd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpcmpgtd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpcmpgtd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = icmp sgt <8 x i32> %a0, %a1
- %2 = sext <8 x i1> %1 to <8 x i32>
- %3 = load <8 x i32>, <8 x i32> *%a2, align 32
- %4 = icmp sgt <8 x i32> %2, %3
- %5 = sext <8 x i1> %4 to <8 x i32>
- ret <8 x i32> %5
-}
-
-define <4 x i64> @test_pcmpgtq(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) {
-; GENERIC-LABEL: test_pcmpgtq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpcmpgtq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpcmpgtq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pcmpgtq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpcmpgtq %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; HASWELL-NEXT: vpcmpgtq (%rdi), %ymm0, %ymm0 # sched: [12:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pcmpgtq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpcmpgtq %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; BROADWELL-NEXT: vpcmpgtq (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pcmpgtq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpcmpgtq %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SKYLAKE-NEXT: vpcmpgtq (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pcmpgtq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpcmpgtq %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SKX-NEXT: vpcmpgtq (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pcmpgtq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpcmpgtq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; ZNVER1-NEXT: vpcmpgtq (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = icmp sgt <4 x i64> %a0, %a1
- %2 = sext <4 x i1> %1 to <4 x i64>
- %3 = load <4 x i64>, <4 x i64> *%a2, align 32
- %4 = icmp sgt <4 x i64> %2, %3
- %5 = sext <4 x i1> %4 to <4 x i64>
- ret <4 x i64> %5
-}
-
-define <16 x i16> @test_pcmpgtw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_pcmpgtw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpcmpgtw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpcmpgtw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pcmpgtw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpcmpgtw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpcmpgtw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pcmpgtw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpcmpgtw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpcmpgtw (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pcmpgtw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpcmpgtw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpcmpgtw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pcmpgtw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpcmpgtw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpcmpgtw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pcmpgtw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpcmpgtw %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpcmpgtw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = icmp sgt <16 x i16> %a0, %a1
- %2 = sext <16 x i1> %1 to <16 x i16>
- %3 = load <16 x i16>, <16 x i16> *%a2, align 32
- %4 = icmp sgt <16 x i16> %2, %3
- %5 = sext <16 x i1> %4 to <16 x i16>
- ret <16 x i16> %5
-}
-
-define <4 x i64> @test_perm2i128(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) {
-; GENERIC-LABEL: test_perm2i128:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vperm2i128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1] sched: [1:1.00]
-; GENERIC-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],mem[0,1] sched: [8:1.00]
-; GENERIC-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_perm2i128:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vperm2i128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1] sched: [3:1.00]
-; HASWELL-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],mem[0,1] sched: [10:1.00]
-; HASWELL-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_perm2i128:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vperm2i128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1] sched: [3:1.00]
-; BROADWELL-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],mem[0,1] sched: [9:1.00]
-; BROADWELL-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_perm2i128:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vperm2i128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1] sched: [3:1.00]
-; SKYLAKE-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],mem[0,1] sched: [10:1.00]
-; SKYLAKE-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_perm2i128:
-; SKX: # %bb.0:
-; SKX-NEXT: vperm2i128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1] sched: [3:1.00]
-; SKX-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],mem[0,1] sched: [10:1.00]
-; SKX-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_perm2i128:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vperm2i128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[0,1] sched: [2:0.25]
-; ZNVER1-NEXT: vperm2i128 {{.*#+}} ymm0 = ymm0[2,3],mem[0,1] sched: [9:0.50]
-; ZNVER1-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <4 x i64> %a0, <4 x i64> %a1, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
- %2 = load <4 x i64>, <4 x i64> *%a2, align 32
- %3 = shufflevector <4 x i64> %a0, <4 x i64> %2, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
- %4 = add <4 x i64> %1, %3
- ret <4 x i64> %4
-}
-
-define <8 x i32> @test_permd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
-; GENERIC-LABEL: test_permd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpermd %ymm1, %ymm0, %ymm1 # sched: [1:1.00]
-; GENERIC-NEXT: vpermd (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; GENERIC-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_permd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpermd %ymm1, %ymm0, %ymm1 # sched: [3:1.00]
-; HASWELL-NEXT: vpermd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; HASWELL-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_permd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpermd %ymm1, %ymm0, %ymm1 # sched: [3:1.00]
-; BROADWELL-NEXT: vpermd (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; BROADWELL-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_permd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpermd %ymm1, %ymm0, %ymm1 # sched: [3:1.00]
-; SKYLAKE-NEXT: vpermd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; SKYLAKE-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_permd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpermd %ymm1, %ymm0, %ymm1 # sched: [3:1.00]
-; SKX-NEXT: vpermd (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; SKX-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_permd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpermd %ymm1, %ymm0, %ymm1 # sched: [2:0.25]
-; ZNVER1-NEXT: vpermd (%rdi), %ymm0, %ymm0 # sched: [9:0.50]
-; ZNVER1-NEXT: vpaddd %ymm0, %ymm1, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x i32> @llvm.x86.avx2.permd(<8 x i32> %a1, <8 x i32> %a0)
- %2 = load <8 x i32>, <8 x i32> *%a2, align 32
- %3 = call <8 x i32> @llvm.x86.avx2.permd(<8 x i32> %2, <8 x i32> %a0)
- %4 = add <8 x i32> %1, %3
- ret <8 x i32> %4
-}
-declare <8 x i32> @llvm.x86.avx2.permd(<8 x i32>, <8 x i32>) nounwind readonly
-
-define <4 x double> @test_permpd(<4 x double> %a0, <4 x double> *%a1) {
-; GENERIC-LABEL: test_permpd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,2,2,3] sched: [1:1.00]
-; GENERIC-NEXT: vpermpd {{.*#+}} ymm1 = mem[0,2,2,3] sched: [8:1.00]
-; GENERIC-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_permpd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,2,2,3] sched: [3:1.00]
-; HASWELL-NEXT: vpermpd {{.*#+}} ymm1 = mem[0,2,2,3] sched: [10:1.00]
-; HASWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_permpd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,2,2,3] sched: [3:1.00]
-; BROADWELL-NEXT: vpermpd {{.*#+}} ymm1 = mem[0,2,2,3] sched: [9:1.00]
-; BROADWELL-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_permpd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,2,2,3] sched: [3:1.00]
-; SKYLAKE-NEXT: vpermpd {{.*#+}} ymm1 = mem[0,2,2,3] sched: [10:1.00]
-; SKYLAKE-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_permpd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,2,2,3] sched: [3:1.00]
-; SKX-NEXT: vpermpd {{.*#+}} ymm1 = mem[0,2,2,3] sched: [10:1.00]
-; SKX-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_permpd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpermpd {{.*#+}} ymm1 = mem[0,2,2,3] sched: [107:0.50]
-; ZNVER1-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[3,2,2,3] sched: [100:0.25]
-; ZNVER1-NEXT: vaddpd %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <4 x double> %a0, <4 x double> undef, <4 x i32> <i32 3, i32 2, i32 2, i32 3>
- %2 = load <4 x double>, <4 x double> *%a1, align 32
- %3 = shufflevector <4 x double> %2, <4 x double> undef, <4 x i32> <i32 0, i32 2, i32 2, i32 3>
- %4 = fadd <4 x double> %1, %3
- ret <4 x double> %4
-}
-
-define <8 x float> @test_permps(<8 x i32> %a0, <8 x float> %a1, <8 x float> *%a2) {
-; GENERIC-LABEL: test_permps:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpermps %ymm1, %ymm0, %ymm1 # sched: [1:1.00]
-; GENERIC-NEXT: vpermps (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; GENERIC-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_permps:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpermps %ymm1, %ymm0, %ymm1 # sched: [3:1.00]
-; HASWELL-NEXT: vpermps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; HASWELL-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_permps:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpermps %ymm1, %ymm0, %ymm1 # sched: [3:1.00]
-; BROADWELL-NEXT: vpermps (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; BROADWELL-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_permps:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpermps %ymm1, %ymm0, %ymm1 # sched: [3:1.00]
-; SKYLAKE-NEXT: vpermps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; SKYLAKE-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_permps:
-; SKX: # %bb.0:
-; SKX-NEXT: vpermps %ymm1, %ymm0, %ymm1 # sched: [3:1.00]
-; SKX-NEXT: vpermps (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; SKX-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_permps:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpermps %ymm1, %ymm0, %ymm1 # sched: [100:0.25]
-; ZNVER1-NEXT: vpermps (%rdi), %ymm0, %ymm0 # sched: [107:0.50]
-; ZNVER1-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x float> @llvm.x86.avx2.permps(<8 x float> %a1, <8 x i32> %a0)
- %2 = load <8 x float>, <8 x float> *%a2, align 32
- %3 = call <8 x float> @llvm.x86.avx2.permps(<8 x float> %2, <8 x i32> %a0)
- %4 = fadd <8 x float> %1, %3
- ret <8 x float> %4
-}
-declare <8 x float> @llvm.x86.avx2.permps(<8 x float>, <8 x i32>) nounwind readonly
-
-define <4 x i64> @test_permq(<4 x i64> %a0, <4 x i64> *%a1) {
-; GENERIC-LABEL: test_permq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpermq {{.*#+}} ymm0 = ymm0[3,2,2,3] sched: [1:1.00]
-; GENERIC-NEXT: vpermq {{.*#+}} ymm1 = mem[0,2,2,3] sched: [8:1.00]
-; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_permq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[3,2,2,3] sched: [3:1.00]
-; HASWELL-NEXT: vpermq {{.*#+}} ymm1 = mem[0,2,2,3] sched: [10:1.00]
-; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_permq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpermq {{.*#+}} ymm0 = ymm0[3,2,2,3] sched: [3:1.00]
-; BROADWELL-NEXT: vpermq {{.*#+}} ymm1 = mem[0,2,2,3] sched: [9:1.00]
-; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_permq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpermq {{.*#+}} ymm0 = ymm0[3,2,2,3] sched: [3:1.00]
-; SKYLAKE-NEXT: vpermq {{.*#+}} ymm1 = mem[0,2,2,3] sched: [10:1.00]
-; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_permq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpermq {{.*#+}} ymm0 = ymm0[3,2,2,3] sched: [3:1.00]
-; SKX-NEXT: vpermq {{.*#+}} ymm1 = mem[0,2,2,3] sched: [10:1.00]
-; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_permq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpermq {{.*#+}} ymm1 = mem[0,2,2,3] sched: [9:0.50]
-; ZNVER1-NEXT: vpermq {{.*#+}} ymm0 = ymm0[3,2,2,3] sched: [2:0.25]
-; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <4 x i64> %a0, <4 x i64> undef, <4 x i32> <i32 3, i32 2, i32 2, i32 3>
- %2 = load <4 x i64>, <4 x i64> *%a1, align 32
- %3 = shufflevector <4 x i64> %2, <4 x i64> undef, <4 x i32> <i32 0, i32 2, i32 2, i32 3>
- %4 = add <4 x i64> %1, %3
- ret <4 x i64> %4
-}
-
-define <4 x i32> @test_pgatherdd(<4 x i32> %a0, i8* %a1, <4 x i32> %a2, <4 x i32> %a3) {
-; GENERIC-LABEL: test_pgatherdd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpgatherdd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [5:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pgatherdd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpgatherdd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [26:2.67]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pgatherdd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpgatherdd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [5:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pgatherdd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpgatherdd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pgatherdd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpgatherdd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pgatherdd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpgatherdd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x i32> @llvm.x86.avx2.gather.d.d(<4 x i32> %a0, i8* %a1, <4 x i32> %a2, <4 x i32> %a3, i8 2)
- ret <4 x i32> %1
-}
-declare <4 x i32> @llvm.x86.avx2.gather.d.d(<4 x i32>, i8*, <4 x i32>, <4 x i32>, i8) nounwind readonly
-
-define <8 x i32> @test_pgatherdd_ymm(<8 x i32> %a0, i8* %a1, <8 x i32> %a2, <8 x i32> %a3) {
-; GENERIC-LABEL: test_pgatherdd_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpgatherdd %ymm2, (%rdi,%ymm1,2), %ymm0 # sched: [5:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pgatherdd_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpgatherdd %ymm2, (%rdi,%ymm1,2), %ymm0 # sched: [27:6.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pgatherdd_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpgatherdd %ymm2, (%rdi,%ymm1,2), %ymm0 # sched: [5:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pgatherdd_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpgatherdd %ymm2, (%rdi,%ymm1,2), %ymm0 # sched: [25:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pgatherdd_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vpgatherdd %ymm2, (%rdi,%ymm1,2), %ymm0 # sched: [25:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pgatherdd_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpgatherdd %ymm2, (%rdi,%ymm1,2), %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x i32> @llvm.x86.avx2.gather.d.d.256(<8 x i32> %a0, i8* %a1, <8 x i32> %a2, <8 x i32> %a3, i8 2)
- ret <8 x i32> %1
-}
-declare <8 x i32> @llvm.x86.avx2.gather.d.d.256(<8 x i32>, i8*, <8 x i32>, <8 x i32>, i8) nounwind readonly
-
-define <2 x i64> @test_pgatherdq(<2 x i64> %a0, i8* %a1, <4 x i32> %a2, <2 x i64> %a3) {
-; GENERIC-LABEL: test_pgatherdq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpgatherdq %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [5:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pgatherdq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpgatherdq %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [26:2.67]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pgatherdq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpgatherdq %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [5:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pgatherdq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpgatherdq %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pgatherdq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpgatherdq %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pgatherdq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpgatherdq %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <2 x i64> @llvm.x86.avx2.gather.d.q(<2 x i64> %a0, i8* %a1, <4 x i32> %a2, <2 x i64> %a3, i8 2)
- ret <2 x i64> %1
-}
-declare <2 x i64> @llvm.x86.avx2.gather.d.q(<2 x i64>, i8*, <4 x i32>, <2 x i64>, i8) nounwind readonly
-
-define <4 x i64> @test_pgatherdq_ymm(<4 x i64> %a0, i8* %a1, <4 x i32> %a2, <4 x i64> %a3) {
-; GENERIC-LABEL: test_pgatherdq_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpgatherdq %ymm2, (%rdi,%xmm1,2), %ymm0 # sched: [5:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pgatherdq_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpgatherdq %ymm2, (%rdi,%xmm1,2), %ymm0 # sched: [27:4.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pgatherdq_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpgatherdq %ymm2, (%rdi,%xmm1,2), %ymm0 # sched: [5:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pgatherdq_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpgatherdq %ymm2, (%rdi,%xmm1,2), %ymm0 # sched: [25:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pgatherdq_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vpgatherdq %ymm2, (%rdi,%xmm1,2), %ymm0 # sched: [25:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pgatherdq_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpgatherdq %ymm2, (%rdi,%xmm1,2), %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x i64> @llvm.x86.avx2.gather.d.q.256(<4 x i64> %a0, i8* %a1, <4 x i32> %a2, <4 x i64> %a3, i8 2)
- ret <4 x i64> %1
-}
-declare <4 x i64> @llvm.x86.avx2.gather.d.q.256(<4 x i64>, i8*, <4 x i32>, <4 x i64>, i8) nounwind readonly
-
-define <4 x i32> @test_pgatherqd(<4 x i32> %a0, i8* %a1, <2 x i64> %a2, <4 x i32> %a3) {
-; GENERIC-LABEL: test_pgatherqd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpgatherqd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [5:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pgatherqd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpgatherqd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [25:5.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pgatherqd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpgatherqd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [5:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pgatherqd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpgatherqd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pgatherqd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpgatherqd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pgatherqd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpgatherqd %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x i32> @llvm.x86.avx2.gather.q.d(<4 x i32> %a0, i8* %a1, <2 x i64> %a2, <4 x i32> %a3, i8 2)
- ret <4 x i32> %1
-}
-declare <4 x i32> @llvm.x86.avx2.gather.q.d(<4 x i32>, i8*, <2 x i64>, <4 x i32>, i8) nounwind readonly
-
-define <4 x i32> @test_pgatherqd_ymm(<4 x i32> %a0, i8* %a1, <4 x i64> %a2, <4 x i32> %a3) {
-; GENERIC-LABEL: test_pgatherqd_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpgatherqd %xmm2, (%rdi,%ymm1,2), %xmm0 # sched: [5:0.50]
-; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pgatherqd_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpgatherqd %xmm2, (%rdi,%ymm1,2), %xmm0 # sched: [28:5.00]
-; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pgatherqd_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpgatherqd %xmm2, (%rdi,%ymm1,2), %xmm0 # sched: [5:0.50]
-; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pgatherqd_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpgatherqd %xmm2, (%rdi,%ymm1,2), %xmm0 # sched: [25:1.00]
-; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pgatherqd_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vpgatherqd %xmm2, (%rdi,%ymm1,2), %xmm0 # sched: [25:1.00]
-; SKX-NEXT: vzeroupper # sched: [0:0.67]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pgatherqd_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpgatherqd %xmm2, (%rdi,%ymm1,2), %xmm0 # sched: [100:0.25]
-; ZNVER1-NEXT: vzeroupper # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x i32> @llvm.x86.avx2.gather.q.d.256(<4 x i32> %a0, i8* %a1, <4 x i64> %a2, <4 x i32> %a3, i8 2)
- ret <4 x i32> %1
-}
-declare <4 x i32> @llvm.x86.avx2.gather.q.d.256(<4 x i32>, i8*, <4 x i64>, <4 x i32>, i8) nounwind readonly
-
-define <2 x i64> @test_pgatherqq(<2 x i64> %a0, i8 *%a1, <2 x i64> %a2, <2 x i64> %a3) {
-; GENERIC-LABEL: test_pgatherqq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpgatherqq %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [5:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pgatherqq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpgatherqq %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [23:3.33]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pgatherqq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpgatherqq %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [5:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pgatherqq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpgatherqq %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pgatherqq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpgatherqq %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [22:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pgatherqq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpgatherqq %xmm2, (%rdi,%xmm1,2), %xmm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <2 x i64> @llvm.x86.avx2.gather.q.q(<2 x i64> %a0, i8* %a1, <2 x i64> %a2, <2 x i64> %a3, i8 2)
- ret <2 x i64> %1
-}
-declare <2 x i64> @llvm.x86.avx2.gather.q.q(<2 x i64>, i8*, <2 x i64>, <2 x i64>, i8) nounwind readonly
-
-define <4 x i64> @test_pgatherqq_ymm(<4 x i64> %a0, i8 *%a1, <4 x i64> %a2, <4 x i64> %a3) {
-; GENERIC-LABEL: test_pgatherqq_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpgatherqq %ymm2, (%rdi,%ymm1,2), %ymm0 # sched: [5:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pgatherqq_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpgatherqq %ymm2, (%rdi,%ymm1,2), %ymm0 # sched: [24:5.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pgatherqq_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpgatherqq %ymm2, (%rdi,%ymm1,2), %ymm0 # sched: [5:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pgatherqq_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpgatherqq %ymm2, (%rdi,%ymm1,2), %ymm0 # sched: [25:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pgatherqq_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vpgatherqq %ymm2, (%rdi,%ymm1,2), %ymm0 # sched: [25:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pgatherqq_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpgatherqq %ymm2, (%rdi,%ymm1,2), %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x i64> @llvm.x86.avx2.gather.q.q.256(<4 x i64> %a0, i8* %a1, <4 x i64> %a2, <4 x i64> %a3, i8 2)
- ret <4 x i64> %1
-}
-declare <4 x i64> @llvm.x86.avx2.gather.q.q.256(<4 x i64>, i8*, <4 x i64>, <4 x i64>, i8) nounwind readonly
-
-define <8 x i32> @test_phaddd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
-; GENERIC-LABEL: test_phaddd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vphaddd %ymm1, %ymm0, %ymm0 # sched: [3:1.50]
-; GENERIC-NEXT: vphaddd (%rdi), %ymm0, %ymm0 # sched: [10:1.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_phaddd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vphaddd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; HASWELL-NEXT: vphaddd (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_phaddd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vphaddd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BROADWELL-NEXT: vphaddd (%rdi), %ymm0, %ymm0 # sched: [9:2.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_phaddd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vphaddd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; SKYLAKE-NEXT: vphaddd (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_phaddd:
-; SKX: # %bb.0:
-; SKX-NEXT: vphaddd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; SKX-NEXT: vphaddd (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_phaddd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vphaddd %ymm1, %ymm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: vphaddd (%rdi), %ymm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x i32> @llvm.x86.avx2.phadd.d(<8 x i32> %a0, <8 x i32> %a1)
- %2 = load <8 x i32>, <8 x i32> *%a2, align 32
- %3 = call <8 x i32> @llvm.x86.avx2.phadd.d(<8 x i32> %1, <8 x i32> %2)
- ret <8 x i32> %3
-}
-declare <8 x i32> @llvm.x86.avx2.phadd.d(<8 x i32>, <8 x i32>) nounwind readnone
-
-define <16 x i16> @test_phaddsw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_phaddsw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vphaddsw %ymm1, %ymm0, %ymm0 # sched: [3:1.50]
-; GENERIC-NEXT: vphaddsw (%rdi), %ymm0, %ymm0 # sched: [10:1.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_phaddsw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vphaddsw %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; HASWELL-NEXT: vphaddsw (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_phaddsw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vphaddsw %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BROADWELL-NEXT: vphaddsw (%rdi), %ymm0, %ymm0 # sched: [9:2.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_phaddsw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vphaddsw %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; SKYLAKE-NEXT: vphaddsw (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_phaddsw:
-; SKX: # %bb.0:
-; SKX-NEXT: vphaddsw %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; SKX-NEXT: vphaddsw (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_phaddsw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vphaddsw %ymm1, %ymm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: vphaddsw (%rdi), %ymm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <16 x i16> @llvm.x86.avx2.phadd.sw(<16 x i16> %a0, <16 x i16> %a1)
- %2 = load <16 x i16>, <16 x i16> *%a2, align 32
- %3 = call <16 x i16> @llvm.x86.avx2.phadd.sw(<16 x i16> %1, <16 x i16> %2)
- ret <16 x i16> %3
-}
-declare <16 x i16> @llvm.x86.avx2.phadd.sw(<16 x i16>, <16 x i16>) nounwind readnone
-
-define <16 x i16> @test_phaddw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_phaddw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vphaddw %ymm1, %ymm0, %ymm0 # sched: [3:1.50]
-; GENERIC-NEXT: vphaddw (%rdi), %ymm0, %ymm0 # sched: [10:1.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_phaddw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vphaddw %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; HASWELL-NEXT: vphaddw (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_phaddw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vphaddw %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BROADWELL-NEXT: vphaddw (%rdi), %ymm0, %ymm0 # sched: [9:2.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_phaddw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vphaddw %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; SKYLAKE-NEXT: vphaddw (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_phaddw:
-; SKX: # %bb.0:
-; SKX-NEXT: vphaddw %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; SKX-NEXT: vphaddw (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_phaddw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vphaddw %ymm1, %ymm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: vphaddw (%rdi), %ymm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <16 x i16> @llvm.x86.avx2.phadd.w(<16 x i16> %a0, <16 x i16> %a1)
- %2 = load <16 x i16>, <16 x i16> *%a2, align 32
- %3 = call <16 x i16> @llvm.x86.avx2.phadd.w(<16 x i16> %1, <16 x i16> %2)
- ret <16 x i16> %3
-}
-declare <16 x i16> @llvm.x86.avx2.phadd.w(<16 x i16>, <16 x i16>) nounwind readnone
-
-define <8 x i32> @test_phsubd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
-; GENERIC-LABEL: test_phsubd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vphsubd %ymm1, %ymm0, %ymm0 # sched: [3:1.50]
-; GENERIC-NEXT: vphsubd (%rdi), %ymm0, %ymm0 # sched: [10:1.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_phsubd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vphsubd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; HASWELL-NEXT: vphsubd (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_phsubd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vphsubd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BROADWELL-NEXT: vphsubd (%rdi), %ymm0, %ymm0 # sched: [9:2.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_phsubd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vphsubd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; SKYLAKE-NEXT: vphsubd (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_phsubd:
-; SKX: # %bb.0:
-; SKX-NEXT: vphsubd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; SKX-NEXT: vphsubd (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_phsubd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vphsubd %ymm1, %ymm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: vphsubd (%rdi), %ymm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x i32> @llvm.x86.avx2.phsub.d(<8 x i32> %a0, <8 x i32> %a1)
- %2 = load <8 x i32>, <8 x i32> *%a2, align 32
- %3 = call <8 x i32> @llvm.x86.avx2.phsub.d(<8 x i32> %1, <8 x i32> %2)
- ret <8 x i32> %3
-}
-declare <8 x i32> @llvm.x86.avx2.phsub.d(<8 x i32>, <8 x i32>) nounwind readnone
-
-define <16 x i16> @test_phsubsw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_phsubsw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vphsubsw %ymm1, %ymm0, %ymm0 # sched: [3:1.50]
-; GENERIC-NEXT: vphsubsw (%rdi), %ymm0, %ymm0 # sched: [10:1.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_phsubsw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vphsubsw %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; HASWELL-NEXT: vphsubsw (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_phsubsw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vphsubsw %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BROADWELL-NEXT: vphsubsw (%rdi), %ymm0, %ymm0 # sched: [9:2.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_phsubsw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vphsubsw %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; SKYLAKE-NEXT: vphsubsw (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_phsubsw:
-; SKX: # %bb.0:
-; SKX-NEXT: vphsubsw %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; SKX-NEXT: vphsubsw (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_phsubsw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vphsubsw %ymm1, %ymm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: vphsubsw (%rdi), %ymm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <16 x i16> @llvm.x86.avx2.phsub.sw(<16 x i16> %a0, <16 x i16> %a1)
- %2 = load <16 x i16>, <16 x i16> *%a2, align 32
- %3 = call <16 x i16> @llvm.x86.avx2.phsub.sw(<16 x i16> %1, <16 x i16> %2)
- ret <16 x i16> %3
-}
-declare <16 x i16> @llvm.x86.avx2.phsub.sw(<16 x i16>, <16 x i16>) nounwind readnone
-
-define <16 x i16> @test_phsubw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_phsubw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vphsubw %ymm1, %ymm0, %ymm0 # sched: [3:1.50]
-; GENERIC-NEXT: vphsubw (%rdi), %ymm0, %ymm0 # sched: [10:1.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_phsubw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vphsubw %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; HASWELL-NEXT: vphsubw (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_phsubw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vphsubw %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BROADWELL-NEXT: vphsubw (%rdi), %ymm0, %ymm0 # sched: [9:2.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_phsubw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vphsubw %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; SKYLAKE-NEXT: vphsubw (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_phsubw:
-; SKX: # %bb.0:
-; SKX-NEXT: vphsubw %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; SKX-NEXT: vphsubw (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_phsubw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vphsubw %ymm1, %ymm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: vphsubw (%rdi), %ymm0, %ymm0 # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <16 x i16> @llvm.x86.avx2.phsub.w(<16 x i16> %a0, <16 x i16> %a1)
- %2 = load <16 x i16>, <16 x i16> *%a2, align 32
- %3 = call <16 x i16> @llvm.x86.avx2.phsub.w(<16 x i16> %1, <16 x i16> %2)
- ret <16 x i16> %3
-}
-declare <16 x i16> @llvm.x86.avx2.phsub.w(<16 x i16>, <16 x i16>) nounwind readnone
-
-define <16 x i16> @test_pmaddubsw(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) {
-; GENERIC-LABEL: test_pmaddubsw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmaddubsw %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; GENERIC-NEXT: vpmaddubsw (%rdi), %ymm0, %ymm0 # sched: [12:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmaddubsw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmaddubsw %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; HASWELL-NEXT: vpmaddubsw (%rdi), %ymm0, %ymm0 # sched: [12:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmaddubsw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmaddubsw %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; BROADWELL-NEXT: vpmaddubsw (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmaddubsw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmaddubsw %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vpmaddubsw (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmaddubsw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmaddubsw %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vpmaddubsw (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmaddubsw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmaddubsw %ymm1, %ymm0, %ymm0 # sched: [4:1.00]
-; ZNVER1-NEXT: vpmaddubsw (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <16 x i16> @llvm.x86.avx2.pmadd.ub.sw(<32 x i8> %a0, <32 x i8> %a1)
- %2 = bitcast <16 x i16> %1 to <32 x i8>
- %3 = load <32 x i8>, <32 x i8> *%a2, align 32
- %4 = call <16 x i16> @llvm.x86.avx2.pmadd.ub.sw(<32 x i8> %2, <32 x i8> %3)
- ret <16 x i16> %4
-}
-declare <16 x i16> @llvm.x86.avx2.pmadd.ub.sw(<32 x i8>, <32 x i8>) nounwind readnone
-
-define <8 x i32> @test_pmaddwd(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_pmaddwd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmaddwd %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; GENERIC-NEXT: vpmaddwd (%rdi), %ymm0, %ymm0 # sched: [12:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmaddwd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmaddwd %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; HASWELL-NEXT: vpmaddwd (%rdi), %ymm0, %ymm0 # sched: [12:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmaddwd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmaddwd %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; BROADWELL-NEXT: vpmaddwd (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmaddwd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmaddwd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vpmaddwd (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmaddwd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmaddwd %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vpmaddwd (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmaddwd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmaddwd %ymm1, %ymm0, %ymm0 # sched: [4:1.00]
-; ZNVER1-NEXT: vpmaddwd (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x i32> @llvm.x86.avx2.pmadd.wd(<16 x i16> %a0, <16 x i16> %a1)
- %2 = bitcast <8 x i32> %1 to <16 x i16>
- %3 = load <16 x i16>, <16 x i16> *%a2, align 32
- %4 = call <8 x i32> @llvm.x86.avx2.pmadd.wd(<16 x i16> %2, <16 x i16> %3)
- ret <8 x i32> %4
-}
-declare <8 x i32> @llvm.x86.avx2.pmadd.wd(<16 x i16>, <16 x i16>) nounwind readnone
-
-define <4 x i32> @test_pmaskmovd(i8* %a0, <4 x i32> %a1, <4 x i32> %a2) {
-; GENERIC-LABEL: test_pmaskmovd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmaskmovd (%rdi), %xmm0, %xmm2 # sched: [8:1.00]
-; GENERIC-NEXT: vpmaskmovd %xmm1, %xmm0, (%rdi) # sched: [5:1.00]
-; GENERIC-NEXT: vmovdqa %xmm2, %xmm0 # sched: [1:0.33]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmaskmovd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmaskmovd (%rdi), %xmm0, %xmm2 # sched: [8:2.00]
-; HASWELL-NEXT: vpmaskmovd %xmm1, %xmm0, (%rdi) # sched: [5:1.00]
-; HASWELL-NEXT: vmovdqa %xmm2, %xmm0 # sched: [1:0.33]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmaskmovd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmaskmovd (%rdi), %xmm0, %xmm2 # sched: [7:2.00]
-; BROADWELL-NEXT: vpmaskmovd %xmm1, %xmm0, (%rdi) # sched: [5:1.00]
-; BROADWELL-NEXT: vmovdqa %xmm2, %xmm0 # sched: [1:0.33]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmaskmovd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmaskmovd (%rdi), %xmm0, %xmm2 # sched: [7:0.50]
-; SKYLAKE-NEXT: vpmaskmovd %xmm1, %xmm0, (%rdi) # sched: [2:1.00]
-; SKYLAKE-NEXT: vmovdqa %xmm2, %xmm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmaskmovd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmaskmovd (%rdi), %xmm0, %xmm2 # sched: [7:0.50]
-; SKX-NEXT: vpmaskmovd %xmm1, %xmm0, (%rdi) # sched: [2:1.00]
-; SKX-NEXT: vmovdqa %xmm2, %xmm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmaskmovd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmaskmovd (%rdi), %xmm0, %xmm2 # sched: [100:0.25]
-; ZNVER1-NEXT: vpmaskmovd %xmm1, %xmm0, (%rdi) # sched: [100:0.25]
-; ZNVER1-NEXT: vmovdqa %xmm2, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x i32> @llvm.x86.avx2.maskload.d(i8* %a0, <4 x i32> %a1)
- call void @llvm.x86.avx2.maskstore.d(i8* %a0, <4 x i32> %a1, <4 x i32> %a2)
- ret <4 x i32> %1
-}
-declare <4 x i32> @llvm.x86.avx2.maskload.d(i8*, <4 x i32>) nounwind readonly
-declare void @llvm.x86.avx2.maskstore.d(i8*, <4 x i32>, <4 x i32>) nounwind
-
-define <8 x i32> @test_pmaskmovd_ymm(i8* %a0, <8 x i32> %a1, <8 x i32> %a2) {
-; GENERIC-LABEL: test_pmaskmovd_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmaskmovd (%rdi), %ymm0, %ymm2 # sched: [9:1.00]
-; GENERIC-NEXT: vpmaskmovd %ymm1, %ymm0, (%rdi) # sched: [5:1.00]
-; GENERIC-NEXT: vmovdqa %ymm2, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmaskmovd_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmaskmovd (%rdi), %ymm0, %ymm2 # sched: [9:2.00]
-; HASWELL-NEXT: vpmaskmovd %ymm1, %ymm0, (%rdi) # sched: [5:1.00]
-; HASWELL-NEXT: vmovdqa %ymm2, %ymm0 # sched: [1:0.33]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmaskmovd_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmaskmovd (%rdi), %ymm0, %ymm2 # sched: [8:2.00]
-; BROADWELL-NEXT: vpmaskmovd %ymm1, %ymm0, (%rdi) # sched: [5:1.00]
-; BROADWELL-NEXT: vmovdqa %ymm2, %ymm0 # sched: [1:0.33]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmaskmovd_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmaskmovd (%rdi), %ymm0, %ymm2 # sched: [8:0.50]
-; SKYLAKE-NEXT: vpmaskmovd %ymm1, %ymm0, (%rdi) # sched: [2:1.00]
-; SKYLAKE-NEXT: vmovdqa %ymm2, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmaskmovd_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmaskmovd (%rdi), %ymm0, %ymm2 # sched: [8:0.50]
-; SKX-NEXT: vpmaskmovd %ymm1, %ymm0, (%rdi) # sched: [2:1.00]
-; SKX-NEXT: vmovdqa %ymm2, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmaskmovd_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmaskmovd (%rdi), %ymm0, %ymm2 # sched: [100:0.25]
-; ZNVER1-NEXT: vpmaskmovd %ymm1, %ymm0, (%rdi) # sched: [100:0.25]
-; ZNVER1-NEXT: vmovdqa %ymm2, %ymm0 # sched: [2:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x i32> @llvm.x86.avx2.maskload.d.256(i8* %a0, <8 x i32> %a1)
- call void @llvm.x86.avx2.maskstore.d.256(i8* %a0, <8 x i32> %a1, <8 x i32> %a2)
- ret <8 x i32> %1
-}
-declare <8 x i32> @llvm.x86.avx2.maskload.d.256(i8*, <8 x i32>) nounwind readonly
-declare void @llvm.x86.avx2.maskstore.d.256(i8*, <8 x i32>, <8 x i32>) nounwind
-
-define <2 x i64> @test_pmaskmovq(i8* %a0, <2 x i64> %a1, <2 x i64> %a2) {
-; GENERIC-LABEL: test_pmaskmovq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmaskmovq (%rdi), %xmm0, %xmm2 # sched: [8:1.00]
-; GENERIC-NEXT: vpmaskmovq %xmm1, %xmm0, (%rdi) # sched: [5:1.00]
-; GENERIC-NEXT: vmovdqa %xmm2, %xmm0 # sched: [1:0.33]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmaskmovq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmaskmovq (%rdi), %xmm0, %xmm2 # sched: [8:2.00]
-; HASWELL-NEXT: vpmaskmovq %xmm1, %xmm0, (%rdi) # sched: [5:1.00]
-; HASWELL-NEXT: vmovdqa %xmm2, %xmm0 # sched: [1:0.33]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmaskmovq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmaskmovq (%rdi), %xmm0, %xmm2 # sched: [7:2.00]
-; BROADWELL-NEXT: vpmaskmovq %xmm1, %xmm0, (%rdi) # sched: [5:1.00]
-; BROADWELL-NEXT: vmovdqa %xmm2, %xmm0 # sched: [1:0.33]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmaskmovq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmaskmovq (%rdi), %xmm0, %xmm2 # sched: [7:0.50]
-; SKYLAKE-NEXT: vpmaskmovq %xmm1, %xmm0, (%rdi) # sched: [2:1.00]
-; SKYLAKE-NEXT: vmovdqa %xmm2, %xmm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmaskmovq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmaskmovq (%rdi), %xmm0, %xmm2 # sched: [7:0.50]
-; SKX-NEXT: vpmaskmovq %xmm1, %xmm0, (%rdi) # sched: [2:1.00]
-; SKX-NEXT: vmovdqa %xmm2, %xmm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmaskmovq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmaskmovq (%rdi), %xmm0, %xmm2 # sched: [8:1.00]
-; ZNVER1-NEXT: vpmaskmovq %xmm1, %xmm0, (%rdi) # sched: [100:0.25]
-; ZNVER1-NEXT: vmovdqa %xmm2, %xmm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <2 x i64> @llvm.x86.avx2.maskload.q(i8* %a0, <2 x i64> %a1)
- call void @llvm.x86.avx2.maskstore.q(i8* %a0, <2 x i64> %a1, <2 x i64> %a2)
- ret <2 x i64> %1
-}
-declare <2 x i64> @llvm.x86.avx2.maskload.q(i8*, <2 x i64>) nounwind readonly
-declare void @llvm.x86.avx2.maskstore.q(i8*, <2 x i64>, <2 x i64>) nounwind
-
-define <4 x i64> @test_pmaskmovq_ymm(i8* %a0, <4 x i64> %a1, <4 x i64> %a2) {
-; GENERIC-LABEL: test_pmaskmovq_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmaskmovq (%rdi), %ymm0, %ymm2 # sched: [9:1.00]
-; GENERIC-NEXT: vpmaskmovq %ymm1, %ymm0, (%rdi) # sched: [5:1.00]
-; GENERIC-NEXT: vmovdqa %ymm2, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmaskmovq_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmaskmovq (%rdi), %ymm0, %ymm2 # sched: [9:2.00]
-; HASWELL-NEXT: vpmaskmovq %ymm1, %ymm0, (%rdi) # sched: [5:1.00]
-; HASWELL-NEXT: vmovdqa %ymm2, %ymm0 # sched: [1:0.33]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmaskmovq_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmaskmovq (%rdi), %ymm0, %ymm2 # sched: [8:2.00]
-; BROADWELL-NEXT: vpmaskmovq %ymm1, %ymm0, (%rdi) # sched: [5:1.00]
-; BROADWELL-NEXT: vmovdqa %ymm2, %ymm0 # sched: [1:0.33]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmaskmovq_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmaskmovq (%rdi), %ymm0, %ymm2 # sched: [8:0.50]
-; SKYLAKE-NEXT: vpmaskmovq %ymm1, %ymm0, (%rdi) # sched: [2:1.00]
-; SKYLAKE-NEXT: vmovdqa %ymm2, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmaskmovq_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmaskmovq (%rdi), %ymm0, %ymm2 # sched: [8:0.50]
-; SKX-NEXT: vpmaskmovq %ymm1, %ymm0, (%rdi) # sched: [2:1.00]
-; SKX-NEXT: vmovdqa %ymm2, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmaskmovq_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmaskmovq (%rdi), %ymm0, %ymm2 # sched: [9:1.50]
-; ZNVER1-NEXT: vpmaskmovq %ymm1, %ymm0, (%rdi) # sched: [100:0.25]
-; ZNVER1-NEXT: vmovdqa %ymm2, %ymm0 # sched: [2:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x i64> @llvm.x86.avx2.maskload.q.256(i8* %a0, <4 x i64> %a1)
- call void @llvm.x86.avx2.maskstore.q.256(i8* %a0, <4 x i64> %a1, <4 x i64> %a2)
- ret <4 x i64> %1
-}
-declare <4 x i64> @llvm.x86.avx2.maskload.q.256(i8*, <4 x i64>) nounwind readonly
-declare void @llvm.x86.avx2.maskstore.q.256(i8*, <4 x i64>, <4 x i64>) nounwind
-
-define <32 x i8> @test_pmaxsb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) {
-; GENERIC-LABEL: test_pmaxsb:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmaxsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpmaxsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmaxsb:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmaxsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpmaxsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmaxsb:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmaxsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpmaxsb (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmaxsb:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmaxsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpmaxsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmaxsb:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmaxsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpmaxsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmaxsb:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmaxsb %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpmaxsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <32 x i8> @llvm.x86.avx2.pmaxs.b(<32 x i8> %a0, <32 x i8> %a1)
- %2 = load <32 x i8>, <32 x i8> *%a2, align 32
- %3 = call <32 x i8> @llvm.x86.avx2.pmaxs.b(<32 x i8> %1, <32 x i8> %2)
- ret <32 x i8> %3
-}
-declare <32 x i8> @llvm.x86.avx2.pmaxs.b(<32 x i8>, <32 x i8>) nounwind readnone
-
-define <8 x i32> @test_pmaxsd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
-; GENERIC-LABEL: test_pmaxsd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmaxsd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpmaxsd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmaxsd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmaxsd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpmaxsd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmaxsd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmaxsd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpmaxsd (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmaxsd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmaxsd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpmaxsd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmaxsd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmaxsd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpmaxsd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmaxsd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmaxsd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpmaxsd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x i32> @llvm.x86.avx2.pmaxs.d(<8 x i32> %a0, <8 x i32> %a1)
- %2 = load <8 x i32>, <8 x i32> *%a2, align 32
- %3 = call <8 x i32> @llvm.x86.avx2.pmaxs.d(<8 x i32> %1, <8 x i32> %2)
- ret <8 x i32> %3
-}
-declare <8 x i32> @llvm.x86.avx2.pmaxs.d(<8 x i32>, <8 x i32>) nounwind readnone
-
-define <16 x i16> @test_pmaxsw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_pmaxsw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmaxsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpmaxsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmaxsw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmaxsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpmaxsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmaxsw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmaxsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpmaxsw (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmaxsw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmaxsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpmaxsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmaxsw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmaxsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpmaxsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmaxsw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmaxsw %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpmaxsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <16 x i16> @llvm.x86.avx2.pmaxs.w(<16 x i16> %a0, <16 x i16> %a1)
- %2 = load <16 x i16>, <16 x i16> *%a2, align 32
- %3 = call <16 x i16> @llvm.x86.avx2.pmaxs.w(<16 x i16> %1, <16 x i16> %2)
- ret <16 x i16> %3
-}
-declare <16 x i16> @llvm.x86.avx2.pmaxs.w(<16 x i16>, <16 x i16>) nounwind readnone
-
-define <32 x i8> @test_pmaxub(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) {
-; GENERIC-LABEL: test_pmaxub:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmaxub %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpmaxub (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmaxub:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmaxub %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpmaxub (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmaxub:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmaxub %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpmaxub (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmaxub:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmaxub %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpmaxub (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmaxub:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmaxub %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpmaxub (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmaxub:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmaxub %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpmaxub (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <32 x i8> @llvm.x86.avx2.pmaxu.b(<32 x i8> %a0, <32 x i8> %a1)
- %2 = load <32 x i8>, <32 x i8> *%a2, align 32
- %3 = call <32 x i8> @llvm.x86.avx2.pmaxu.b(<32 x i8> %1, <32 x i8> %2)
- ret <32 x i8> %3
-}
-declare <32 x i8> @llvm.x86.avx2.pmaxu.b(<32 x i8>, <32 x i8>) nounwind readnone
-
-define <8 x i32> @test_pmaxud(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
-; GENERIC-LABEL: test_pmaxud:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmaxud %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpmaxud (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmaxud:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmaxud %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpmaxud (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmaxud:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmaxud %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpmaxud (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmaxud:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmaxud %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpmaxud (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmaxud:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmaxud %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpmaxud (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmaxud:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmaxud %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpmaxud (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x i32> @llvm.x86.avx2.pmaxu.d(<8 x i32> %a0, <8 x i32> %a1)
- %2 = load <8 x i32>, <8 x i32> *%a2, align 32
- %3 = call <8 x i32> @llvm.x86.avx2.pmaxu.d(<8 x i32> %1, <8 x i32> %2)
- ret <8 x i32> %3
-}
-declare <8 x i32> @llvm.x86.avx2.pmaxu.d(<8 x i32>, <8 x i32>) nounwind readnone
-
-define <16 x i16> @test_pmaxuw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_pmaxuw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmaxuw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpmaxuw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmaxuw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmaxuw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpmaxuw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmaxuw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmaxuw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpmaxuw (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmaxuw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmaxuw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpmaxuw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmaxuw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmaxuw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpmaxuw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmaxuw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmaxuw %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpmaxuw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <16 x i16> @llvm.x86.avx2.pmaxu.w(<16 x i16> %a0, <16 x i16> %a1)
- %2 = load <16 x i16>, <16 x i16> *%a2, align 32
- %3 = call <16 x i16> @llvm.x86.avx2.pmaxu.w(<16 x i16> %1, <16 x i16> %2)
- ret <16 x i16> %3
-}
-declare <16 x i16> @llvm.x86.avx2.pmaxu.w(<16 x i16>, <16 x i16>) nounwind readnone
-
-define <32 x i8> @test_pminsb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) {
-; GENERIC-LABEL: test_pminsb:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpminsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpminsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pminsb:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpminsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpminsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pminsb:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpminsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpminsb (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pminsb:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpminsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpminsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pminsb:
-; SKX: # %bb.0:
-; SKX-NEXT: vpminsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpminsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pminsb:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpminsb %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpminsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <32 x i8> @llvm.x86.avx2.pmins.b(<32 x i8> %a0, <32 x i8> %a1)
- %2 = load <32 x i8>, <32 x i8> *%a2, align 32
- %3 = call <32 x i8> @llvm.x86.avx2.pmins.b(<32 x i8> %1, <32 x i8> %2)
- ret <32 x i8> %3
-}
-declare <32 x i8> @llvm.x86.avx2.pmins.b(<32 x i8>, <32 x i8>) nounwind readnone
-
-define <8 x i32> @test_pminsd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
-; GENERIC-LABEL: test_pminsd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpminsd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpminsd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pminsd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpminsd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpminsd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pminsd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpminsd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpminsd (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pminsd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpminsd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpminsd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pminsd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpminsd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpminsd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pminsd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpminsd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpminsd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x i32> @llvm.x86.avx2.pmins.d(<8 x i32> %a0, <8 x i32> %a1)
- %2 = load <8 x i32>, <8 x i32> *%a2, align 32
- %3 = call <8 x i32> @llvm.x86.avx2.pmins.d(<8 x i32> %1, <8 x i32> %2)
- ret <8 x i32> %3
-}
-declare <8 x i32> @llvm.x86.avx2.pmins.d(<8 x i32>, <8 x i32>) nounwind readnone
-
-define <16 x i16> @test_pminsw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_pminsw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpminsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpminsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pminsw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpminsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpminsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pminsw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpminsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpminsw (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pminsw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpminsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpminsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pminsw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpminsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpminsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pminsw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpminsw %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpminsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <16 x i16> @llvm.x86.avx2.pmins.w(<16 x i16> %a0, <16 x i16> %a1)
- %2 = load <16 x i16>, <16 x i16> *%a2, align 32
- %3 = call <16 x i16> @llvm.x86.avx2.pmins.w(<16 x i16> %1, <16 x i16> %2)
- ret <16 x i16> %3
-}
-declare <16 x i16> @llvm.x86.avx2.pmins.w(<16 x i16>, <16 x i16>) nounwind readnone
-
-define <32 x i8> @test_pminub(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) {
-; GENERIC-LABEL: test_pminub:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpminub %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpminub (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pminub:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpminub %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpminub (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pminub:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpminub %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpminub (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pminub:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpminub %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpminub (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pminub:
-; SKX: # %bb.0:
-; SKX-NEXT: vpminub %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpminub (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pminub:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpminub %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpminub (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <32 x i8> @llvm.x86.avx2.pminu.b(<32 x i8> %a0, <32 x i8> %a1)
- %2 = load <32 x i8>, <32 x i8> *%a2, align 32
- %3 = call <32 x i8> @llvm.x86.avx2.pminu.b(<32 x i8> %1, <32 x i8> %2)
- ret <32 x i8> %3
-}
-declare <32 x i8> @llvm.x86.avx2.pminu.b(<32 x i8>, <32 x i8>) nounwind readnone
-
-define <8 x i32> @test_pminud(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
-; GENERIC-LABEL: test_pminud:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpminud %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpminud (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pminud:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpminud %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpminud (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pminud:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpminud %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpminud (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pminud:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpminud %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpminud (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pminud:
-; SKX: # %bb.0:
-; SKX-NEXT: vpminud %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpminud (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pminud:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpminud %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpminud (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x i32> @llvm.x86.avx2.pminu.d(<8 x i32> %a0, <8 x i32> %a1)
- %2 = load <8 x i32>, <8 x i32> *%a2, align 32
- %3 = call <8 x i32> @llvm.x86.avx2.pminu.d(<8 x i32> %1, <8 x i32> %2)
- ret <8 x i32> %3
-}
-declare <8 x i32> @llvm.x86.avx2.pminu.d(<8 x i32>, <8 x i32>) nounwind readnone
-
-define <16 x i16> @test_pminuw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_pminuw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpminuw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpminuw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pminuw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpminuw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpminuw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pminuw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpminuw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpminuw (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pminuw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpminuw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpminuw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pminuw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpminuw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpminuw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pminuw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpminuw %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpminuw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <16 x i16> @llvm.x86.avx2.pminu.w(<16 x i16> %a0, <16 x i16> %a1)
- %2 = load <16 x i16>, <16 x i16> *%a2, align 32
- %3 = call <16 x i16> @llvm.x86.avx2.pminu.w(<16 x i16> %1, <16 x i16> %2)
- ret <16 x i16> %3
-}
-declare <16 x i16> @llvm.x86.avx2.pminu.w(<16 x i16>, <16 x i16>) nounwind readnone
-
-define i32 @test_pmovmskb(<32 x i8> %a0) {
-; GENERIC-LABEL: test_pmovmskb:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmovmskb %ymm0, %eax # sched: [2:1.00]
-; GENERIC-NEXT: vzeroupper # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmovmskb:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmovmskb %ymm0, %eax # sched: [3:1.00]
-; HASWELL-NEXT: vzeroupper # sched: [0:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmovmskb:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmovmskb %ymm0, %eax # sched: [3:1.00]
-; BROADWELL-NEXT: vzeroupper # sched: [0:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmovmskb:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmovmskb %ymm0, %eax # sched: [2:1.00]
-; SKYLAKE-NEXT: vzeroupper # sched: [0:0.67]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmovmskb:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmovmskb %ymm0, %eax # sched: [2:1.00]
-; SKX-NEXT: vzeroupper # sched: [0:0.67]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmovmskb:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmovmskb %ymm0, %eax # sched: [2:2.00]
-; ZNVER1-NEXT: vzeroupper # sched: [100:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call i32 @llvm.x86.avx2.pmovmskb(<32 x i8> %a0)
- ret i32 %1
-}
-declare i32 @llvm.x86.avx2.pmovmskb(<32 x i8>) nounwind readnone
-
-define <8 x i32> @test_pmovsxbd(<16 x i8> %a0, <16 x i8> *%a1) {
-; GENERIC-LABEL: test_pmovsxbd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmovsxbd %xmm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpmovsxbd (%rdi), %ymm1 # sched: [8:1.00]
-; GENERIC-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmovsxbd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmovsxbd %xmm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vpmovsxbd (%rdi), %ymm1 # sched: [8:1.00]
-; HASWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmovsxbd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmovsxbd %xmm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vpmovsxbd (%rdi), %ymm1 # sched: [8:1.00]
-; BROADWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmovsxbd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmovsxbd %xmm0, %ymm0 # sched: [3:1.00]
-; SKYLAKE-NEXT: vpmovsxbd (%rdi), %ymm1 # sched: [8:1.00]
-; SKYLAKE-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmovsxbd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmovsxbd %xmm0, %ymm0 # sched: [3:1.00]
-; SKX-NEXT: vpmovsxbd (%rdi), %ymm1 # sched: [8:1.00]
-; SKX-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmovsxbd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmovsxbd (%rdi), %ymm1 # sched: [8:0.50]
-; ZNVER1-NEXT: vpmovsxbd %xmm0, %ymm0 # sched: [1:0.50]
-; ZNVER1-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <16 x i8> %a0, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
- %2 = sext <8 x i8> %1 to <8 x i32>
- %3 = load <16 x i8>, <16 x i8> *%a1, align 16
- %4 = shufflevector <16 x i8> %3, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
- %5 = sext <8 x i8> %4 to <8 x i32>
- %6 = add <8 x i32> %2, %5
- ret <8 x i32> %6
-}
-
-define <4 x i64> @test_pmovsxbq(<16 x i8> %a0, <16 x i8> *%a1) {
-; GENERIC-LABEL: test_pmovsxbq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmovsxbq %xmm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpmovsxbq (%rdi), %ymm1 # sched: [8:1.00]
-; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmovsxbq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmovsxbq %xmm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vpmovsxbq (%rdi), %ymm1 # sched: [8:1.00]
-; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmovsxbq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmovsxbq %xmm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vpmovsxbq (%rdi), %ymm1 # sched: [8:1.00]
-; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmovsxbq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmovsxbq %xmm0, %ymm0 # sched: [3:1.00]
-; SKYLAKE-NEXT: vpmovsxbq (%rdi), %ymm1 # sched: [8:1.00]
-; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmovsxbq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmovsxbq %xmm0, %ymm0 # sched: [3:1.00]
-; SKX-NEXT: vpmovsxbq (%rdi), %ymm1 # sched: [8:1.00]
-; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmovsxbq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmovsxbq (%rdi), %ymm1 # sched: [8:0.50]
-; ZNVER1-NEXT: vpmovsxbq %xmm0, %ymm0 # sched: [1:0.50]
-; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <16 x i8> %a0, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
- %2 = sext <4 x i8> %1 to <4 x i64>
- %3 = load <16 x i8>, <16 x i8> *%a1, align 16
- %4 = shufflevector <16 x i8> %3, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
- %5 = sext <4 x i8> %4 to <4 x i64>
- %6 = add <4 x i64> %2, %5
- ret <4 x i64> %6
-}
-
-define <16 x i16> @test_pmovsxbw(<16 x i8> %a0, <16 x i8> *%a1) {
-; GENERIC-LABEL: test_pmovsxbw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmovsxbw %xmm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpmovsxbw (%rdi), %ymm1 # sched: [8:1.00]
-; GENERIC-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmovsxbw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmovsxbw %xmm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vpmovsxbw (%rdi), %ymm1 # sched: [9:1.00]
-; HASWELL-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmovsxbw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmovsxbw %xmm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vpmovsxbw (%rdi), %ymm1 # sched: [8:1.00]
-; BROADWELL-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmovsxbw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmovsxbw %xmm0, %ymm0 # sched: [3:1.00]
-; SKYLAKE-NEXT: vpmovsxbw (%rdi), %ymm1 # sched: [9:1.00]
-; SKYLAKE-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmovsxbw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmovsxbw %xmm0, %ymm0 # sched: [3:1.00]
-; SKX-NEXT: vpmovsxbw (%rdi), %ymm1 # sched: [9:1.00]
-; SKX-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmovsxbw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmovsxbw (%rdi), %ymm1 # sched: [8:0.50]
-; ZNVER1-NEXT: vpmovsxbw %xmm0, %ymm0 # sched: [1:0.50]
-; ZNVER1-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = sext <16 x i8> %a0 to <16 x i16>
- %2 = load <16 x i8>, <16 x i8> *%a1, align 16
- %3 = sext <16 x i8> %2 to <16 x i16>
- %4 = add <16 x i16> %1, %3
- ret <16 x i16> %4
-}
-
-define <4 x i64> @test_pmovsxdq(<4 x i32> %a0, <4 x i32> *%a1) {
-; GENERIC-LABEL: test_pmovsxdq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmovsxdq %xmm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpmovsxdq (%rdi), %ymm1 # sched: [8:1.00]
-; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmovsxdq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmovsxdq %xmm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vpmovsxdq (%rdi), %ymm1 # sched: [9:1.00]
-; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmovsxdq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmovsxdq %xmm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vpmovsxdq (%rdi), %ymm1 # sched: [8:1.00]
-; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmovsxdq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmovsxdq %xmm0, %ymm0 # sched: [3:1.00]
-; SKYLAKE-NEXT: vpmovsxdq (%rdi), %ymm1 # sched: [9:1.00]
-; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmovsxdq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmovsxdq %xmm0, %ymm0 # sched: [3:1.00]
-; SKX-NEXT: vpmovsxdq (%rdi), %ymm1 # sched: [9:1.00]
-; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmovsxdq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmovsxdq (%rdi), %ymm1 # sched: [8:0.50]
-; ZNVER1-NEXT: vpmovsxdq %xmm0, %ymm0 # sched: [1:0.50]
-; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = sext <4 x i32> %a0 to <4 x i64>
- %2 = load <4 x i32>, <4 x i32> *%a1, align 16
- %3 = sext <4 x i32> %2 to <4 x i64>
- %4 = add <4 x i64> %1, %3
- ret <4 x i64> %4
-}
-
-define <8 x i32> @test_pmovsxwd(<8 x i16> %a0, <8 x i16> *%a1) {
-; GENERIC-LABEL: test_pmovsxwd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmovsxwd %xmm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpmovsxwd (%rdi), %ymm1 # sched: [8:1.00]
-; GENERIC-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmovsxwd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmovsxwd %xmm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vpmovsxwd (%rdi), %ymm1 # sched: [9:1.00]
-; HASWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmovsxwd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmovsxwd %xmm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vpmovsxwd (%rdi), %ymm1 # sched: [8:1.00]
-; BROADWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmovsxwd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmovsxwd %xmm0, %ymm0 # sched: [3:1.00]
-; SKYLAKE-NEXT: vpmovsxwd (%rdi), %ymm1 # sched: [9:1.00]
-; SKYLAKE-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmovsxwd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmovsxwd %xmm0, %ymm0 # sched: [3:1.00]
-; SKX-NEXT: vpmovsxwd (%rdi), %ymm1 # sched: [9:1.00]
-; SKX-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmovsxwd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmovsxwd (%rdi), %ymm1 # sched: [8:0.50]
-; ZNVER1-NEXT: vpmovsxwd %xmm0, %ymm0 # sched: [1:0.50]
-; ZNVER1-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = sext <8 x i16> %a0 to <8 x i32>
- %2 = load <8 x i16>, <8 x i16> *%a1, align 16
- %3 = sext <8 x i16> %2 to <8 x i32>
- %4 = add <8 x i32> %1, %3
- ret <8 x i32> %4
-}
-
-define <4 x i64> @test_pmovsxwq(<8 x i16> %a0, <8 x i16> *%a1) {
-; GENERIC-LABEL: test_pmovsxwq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmovsxwq %xmm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpmovsxwq (%rdi), %ymm1 # sched: [8:1.00]
-; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmovsxwq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmovsxwq %xmm0, %ymm0 # sched: [3:1.00]
-; HASWELL-NEXT: vpmovsxwq (%rdi), %ymm1 # sched: [8:1.00]
-; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmovsxwq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmovsxwq %xmm0, %ymm0 # sched: [3:1.00]
-; BROADWELL-NEXT: vpmovsxwq (%rdi), %ymm1 # sched: [8:1.00]
-; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmovsxwq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmovsxwq %xmm0, %ymm0 # sched: [3:1.00]
-; SKYLAKE-NEXT: vpmovsxwq (%rdi), %ymm1 # sched: [8:1.00]
-; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmovsxwq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmovsxwq %xmm0, %ymm0 # sched: [3:1.00]
-; SKX-NEXT: vpmovsxwq (%rdi), %ymm1 # sched: [8:1.00]
-; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmovsxwq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmovsxwq (%rdi), %ymm1 # sched: [8:0.50]
-; ZNVER1-NEXT: vpmovsxwq %xmm0, %ymm0 # sched: [1:0.50]
-; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <8 x i16> %a0, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
- %2 = sext <4 x i16> %1 to <4 x i64>
- %3 = load <8 x i16>, <8 x i16> *%a1, align 16
- %4 = shufflevector <8 x i16> %3, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
- %5 = sext <4 x i16> %4 to <4 x i64>
- %6 = add <4 x i64> %2, %5
- ret <4 x i64> %6
-}
-
-define <8 x i32> @test_pmovzxbd(<16 x i8> %a0, <16 x i8> *%a1) {
-; GENERIC-LABEL: test_pmovzxbd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmovzxbd {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero sched: [1:1.00]
-; GENERIC-NEXT: vpmovzxbd {{.*#+}} ymm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero sched: [8:1.00]
-; GENERIC-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmovzxbd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmovzxbd {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero sched: [3:1.00]
-; HASWELL-NEXT: vpmovzxbd {{.*#+}} ymm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero sched: [10:1.00]
-; HASWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmovzxbd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmovzxbd {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero sched: [3:1.00]
-; BROADWELL-NEXT: vpmovzxbd {{.*#+}} ymm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero sched: [9:1.00]
-; BROADWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmovzxbd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmovzxbd {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero sched: [3:1.00]
-; SKYLAKE-NEXT: vpmovzxbd {{.*#+}} ymm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero sched: [10:1.00]
-; SKYLAKE-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmovzxbd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmovzxbd {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero sched: [3:1.00]
-; SKX-NEXT: vpmovzxbd {{.*#+}} ymm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero sched: [10:1.00]
-; SKX-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmovzxbd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmovzxbd {{.*#+}} ymm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero sched: [8:0.50]
-; ZNVER1-NEXT: vpmovzxbd {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero sched: [1:0.50]
-; ZNVER1-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <16 x i8> %a0, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
- %2 = zext <8 x i8> %1 to <8 x i32>
- %3 = load <16 x i8>, <16 x i8> *%a1, align 16
- %4 = shufflevector <16 x i8> %3, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
- %5 = zext <8 x i8> %4 to <8 x i32>
- %6 = add <8 x i32> %2, %5
- ret <8 x i32> %6
-}
-
-define <4 x i64> @test_pmovzxbq(<16 x i8> %a0, <16 x i8> *%a1) {
-; GENERIC-LABEL: test_pmovzxbq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmovzxbq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero sched: [1:1.00]
-; GENERIC-NEXT: vpmovzxbq {{.*#+}} ymm1 = mem[0],zero,zero,zero,zero,zero,zero,zero,mem[1],zero,zero,zero,zero,zero,zero,zero,mem[2],zero,zero,zero,zero,zero,zero,zero,mem[3],zero,zero,zero,zero,zero,zero,zero sched: [8:1.00]
-; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmovzxbq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmovzxbq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero sched: [3:1.00]
-; HASWELL-NEXT: vpmovzxbq {{.*#+}} ymm1 = mem[0],zero,zero,zero,zero,zero,zero,zero,mem[1],zero,zero,zero,zero,zero,zero,zero,mem[2],zero,zero,zero,zero,zero,zero,zero,mem[3],zero,zero,zero,zero,zero,zero,zero sched: [10:1.00]
-; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmovzxbq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmovzxbq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero sched: [3:1.00]
-; BROADWELL-NEXT: vpmovzxbq {{.*#+}} ymm1 = mem[0],zero,zero,zero,zero,zero,zero,zero,mem[1],zero,zero,zero,zero,zero,zero,zero,mem[2],zero,zero,zero,zero,zero,zero,zero,mem[3],zero,zero,zero,zero,zero,zero,zero sched: [9:1.00]
-; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmovzxbq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmovzxbq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero sched: [3:1.00]
-; SKYLAKE-NEXT: vpmovzxbq {{.*#+}} ymm1 = mem[0],zero,zero,zero,zero,zero,zero,zero,mem[1],zero,zero,zero,zero,zero,zero,zero,mem[2],zero,zero,zero,zero,zero,zero,zero,mem[3],zero,zero,zero,zero,zero,zero,zero sched: [10:1.00]
-; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmovzxbq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmovzxbq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero sched: [3:1.00]
-; SKX-NEXT: vpmovzxbq {{.*#+}} ymm1 = mem[0],zero,zero,zero,zero,zero,zero,zero,mem[1],zero,zero,zero,zero,zero,zero,zero,mem[2],zero,zero,zero,zero,zero,zero,zero,mem[3],zero,zero,zero,zero,zero,zero,zero sched: [10:1.00]
-; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmovzxbq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmovzxbq {{.*#+}} ymm1 = mem[0],zero,zero,zero,zero,zero,zero,zero,mem[1],zero,zero,zero,zero,zero,zero,zero,mem[2],zero,zero,zero,zero,zero,zero,zero,mem[3],zero,zero,zero,zero,zero,zero,zero sched: [8:0.50]
-; ZNVER1-NEXT: vpmovzxbq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,zero,zero,zero,zero,xmm0[1],zero,zero,zero,zero,zero,zero,zero,xmm0[2],zero,zero,zero,zero,zero,zero,zero,xmm0[3],zero,zero,zero,zero,zero,zero,zero sched: [1:0.50]
-; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <16 x i8> %a0, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
- %2 = zext <4 x i8> %1 to <4 x i64>
- %3 = load <16 x i8>, <16 x i8> *%a1, align 16
- %4 = shufflevector <16 x i8> %3, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
- %5 = zext <4 x i8> %4 to <4 x i64>
- %6 = add <4 x i64> %2, %5
- ret <4 x i64> %6
-}
-
-define <16 x i16> @test_pmovzxbw(<16 x i8> %a0, <16 x i8> *%a1) {
-; GENERIC-LABEL: test_pmovzxbw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero sched: [1:1.00]
-; GENERIC-NEXT: vpmovzxbw {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero sched: [8:1.00]
-; GENERIC-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmovzxbw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero sched: [3:1.00]
-; HASWELL-NEXT: vpmovzxbw {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero sched: [10:1.00]
-; HASWELL-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmovzxbw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero sched: [3:1.00]
-; BROADWELL-NEXT: vpmovzxbw {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero sched: [9:1.00]
-; BROADWELL-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmovzxbw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero sched: [3:1.00]
-; SKYLAKE-NEXT: vpmovzxbw {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero sched: [10:1.00]
-; SKYLAKE-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmovzxbw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero sched: [3:1.00]
-; SKX-NEXT: vpmovzxbw {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero sched: [10:1.00]
-; SKX-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmovzxbw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmovzxbw {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero sched: [8:0.50]
-; ZNVER1-NEXT: vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero sched: [1:0.50]
-; ZNVER1-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = zext <16 x i8> %a0 to <16 x i16>
- %2 = load <16 x i8>, <16 x i8> *%a1, align 16
- %3 = zext <16 x i8> %2 to <16 x i16>
- %4 = add <16 x i16> %1, %3
- ret <16 x i16> %4
-}
-
-define <4 x i64> @test_pmovzxdq(<4 x i32> %a0, <4 x i32> *%a1) {
-; GENERIC-LABEL: test_pmovzxdq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmovzxdq {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero sched: [1:1.00]
-; GENERIC-NEXT: vpmovzxdq {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero sched: [8:1.00]
-; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmovzxdq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmovzxdq {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero sched: [3:1.00]
-; HASWELL-NEXT: vpmovzxdq {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero sched: [10:1.00]
-; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmovzxdq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmovzxdq {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero sched: [3:1.00]
-; BROADWELL-NEXT: vpmovzxdq {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero sched: [9:1.00]
-; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmovzxdq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmovzxdq {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero sched: [3:1.00]
-; SKYLAKE-NEXT: vpmovzxdq {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero sched: [10:1.00]
-; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmovzxdq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmovzxdq {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero sched: [3:1.00]
-; SKX-NEXT: vpmovzxdq {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero sched: [10:1.00]
-; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmovzxdq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmovzxdq {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero sched: [8:0.50]
-; ZNVER1-NEXT: vpmovzxdq {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero sched: [1:0.50]
-; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = zext <4 x i32> %a0 to <4 x i64>
- %2 = load <4 x i32>, <4 x i32> *%a1, align 16
- %3 = zext <4 x i32> %2 to <4 x i64>
- %4 = add <4 x i64> %1, %3
- ret <4 x i64> %4
-}
-
-define <8 x i32> @test_pmovzxwd(<8 x i16> %a0, <8 x i16> *%a1) {
-; GENERIC-LABEL: test_pmovzxwd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero sched: [1:1.00]
-; GENERIC-NEXT: vpmovzxwd {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero sched: [8:1.00]
-; GENERIC-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmovzxwd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero sched: [3:1.00]
-; HASWELL-NEXT: vpmovzxwd {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero sched: [9:1.00]
-; HASWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmovzxwd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero sched: [3:1.00]
-; BROADWELL-NEXT: vpmovzxwd {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero sched: [8:1.00]
-; BROADWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmovzxwd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero sched: [3:1.00]
-; SKYLAKE-NEXT: vpmovzxwd {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero sched: [9:1.00]
-; SKYLAKE-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmovzxwd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero sched: [3:1.00]
-; SKX-NEXT: vpmovzxwd {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero sched: [9:1.00]
-; SKX-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmovzxwd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmovzxwd {{.*#+}} ymm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero sched: [8:0.50]
-; ZNVER1-NEXT: vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero sched: [1:0.50]
-; ZNVER1-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = zext <8 x i16> %a0 to <8 x i32>
- %2 = load <8 x i16>, <8 x i16> *%a1, align 16
- %3 = zext <8 x i16> %2 to <8 x i32>
- %4 = add <8 x i32> %1, %3
- ret <8 x i32> %4
-}
-
-define <4 x i64> @test_pmovzxwq(<8 x i16> %a0, <8 x i16> *%a1) {
-; GENERIC-LABEL: test_pmovzxwq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmovzxwq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero sched: [1:1.00]
-; GENERIC-NEXT: vpmovzxwq {{.*#+}} ymm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero sched: [8:1.00]
-; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmovzxwq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmovzxwq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero sched: [3:1.00]
-; HASWELL-NEXT: vpmovzxwq {{.*#+}} ymm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero sched: [10:1.00]
-; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmovzxwq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmovzxwq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero sched: [3:1.00]
-; BROADWELL-NEXT: vpmovzxwq {{.*#+}} ymm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero sched: [9:1.00]
-; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmovzxwq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmovzxwq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero sched: [3:1.00]
-; SKYLAKE-NEXT: vpmovzxwq {{.*#+}} ymm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero sched: [10:1.00]
-; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmovzxwq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmovzxwq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero sched: [3:1.00]
-; SKX-NEXT: vpmovzxwq {{.*#+}} ymm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero sched: [10:1.00]
-; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmovzxwq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmovzxwq {{.*#+}} ymm1 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero sched: [8:0.50]
-; ZNVER1-NEXT: vpmovzxwq {{.*#+}} ymm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero sched: [1:0.50]
-; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <8 x i16> %a0, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
- %2 = zext <4 x i16> %1 to <4 x i64>
- %3 = load <8 x i16>, <8 x i16> *%a1, align 16
- %4 = shufflevector <8 x i16> %3, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
- %5 = zext <4 x i16> %4 to <4 x i64>
- %6 = add <4 x i64> %2, %5
- ret <4 x i64> %6
-}
-
-define <4 x i64> @test_pmuldq(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> %a2, <8 x i32> *%a3) {
-; GENERIC-LABEL: test_pmuldq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmuldq %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; GENERIC-NEXT: vpmuldq (%rdi), %ymm2, %ymm1 # sched: [12:1.00]
-; GENERIC-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmuldq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmuldq %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; HASWELL-NEXT: vpmuldq (%rdi), %ymm2, %ymm1 # sched: [12:1.00]
-; HASWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmuldq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmuldq %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; BROADWELL-NEXT: vpmuldq (%rdi), %ymm2, %ymm1 # sched: [11:1.00]
-; BROADWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmuldq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmuldq %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vpmuldq (%rdi), %ymm2, %ymm1 # sched: [11:0.50]
-; SKYLAKE-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmuldq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmuldq %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vpmuldq (%rdi), %ymm2, %ymm1 # sched: [11:0.50]
-; SKX-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmuldq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmuldq (%rdi), %ymm2, %ymm2 # sched: [11:1.00]
-; ZNVER1-NEXT: vpmuldq %ymm1, %ymm0, %ymm0 # sched: [4:1.00]
-; ZNVER1-NEXT: vpor %ymm2, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x i64> @llvm.x86.avx2.pmul.dq(<8 x i32> %a0, <8 x i32> %a1)
- %2 = load <8 x i32>, <8 x i32> *%a3, align 32
- %3 = call <4 x i64> @llvm.x86.avx2.pmul.dq(<8 x i32> %a2, <8 x i32> %2)
- %4 = or <4 x i64> %1, %3
- ret <4 x i64> %4
-}
-declare <4 x i64> @llvm.x86.avx2.pmul.dq(<8 x i32>, <8 x i32>) nounwind readnone
-
-define <16 x i16> @test_pmulhrsw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_pmulhrsw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmulhrsw %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; GENERIC-NEXT: vpmulhrsw (%rdi), %ymm0, %ymm0 # sched: [12:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmulhrsw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmulhrsw %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; HASWELL-NEXT: vpmulhrsw (%rdi), %ymm0, %ymm0 # sched: [12:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmulhrsw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmulhrsw %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; BROADWELL-NEXT: vpmulhrsw (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmulhrsw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmulhrsw %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vpmulhrsw (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmulhrsw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmulhrsw %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vpmulhrsw (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmulhrsw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmulhrsw %ymm1, %ymm0, %ymm0 # sched: [4:1.00]
-; ZNVER1-NEXT: vpmulhrsw (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <16 x i16> @llvm.x86.avx2.pmul.hr.sw(<16 x i16> %a0, <16 x i16> %a1)
- %2 = load <16 x i16>, <16 x i16> *%a2, align 32
- %3 = call <16 x i16> @llvm.x86.avx2.pmul.hr.sw(<16 x i16> %1, <16 x i16> %2)
- ret <16 x i16> %3
-}
-declare <16 x i16> @llvm.x86.avx2.pmul.hr.sw(<16 x i16>, <16 x i16>) nounwind readnone
-
-define <16 x i16> @test_pmulhuw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_pmulhuw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmulhuw %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; GENERIC-NEXT: vpmulhuw (%rdi), %ymm0, %ymm0 # sched: [12:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmulhuw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmulhuw %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; HASWELL-NEXT: vpmulhuw (%rdi), %ymm0, %ymm0 # sched: [12:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmulhuw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmulhuw %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; BROADWELL-NEXT: vpmulhuw (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmulhuw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmulhuw %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vpmulhuw (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmulhuw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmulhuw %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vpmulhuw (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmulhuw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmulhuw %ymm1, %ymm0, %ymm0 # sched: [4:1.00]
-; ZNVER1-NEXT: vpmulhuw (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <16 x i16> @llvm.x86.avx2.pmulhu.w(<16 x i16> %a0, <16 x i16> %a1)
- %2 = load <16 x i16>, <16 x i16> *%a2, align 32
- %3 = call <16 x i16> @llvm.x86.avx2.pmulhu.w(<16 x i16> %1, <16 x i16> %2)
- ret <16 x i16> %3
-}
-declare <16 x i16> @llvm.x86.avx2.pmulhu.w(<16 x i16>, <16 x i16>) nounwind readnone
-
-define <16 x i16> @test_pmulhw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_pmulhw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmulhw %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; GENERIC-NEXT: vpmulhw (%rdi), %ymm0, %ymm0 # sched: [12:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmulhw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmulhw %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; HASWELL-NEXT: vpmulhw (%rdi), %ymm0, %ymm0 # sched: [12:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmulhw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmulhw %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; BROADWELL-NEXT: vpmulhw (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmulhw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmulhw %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vpmulhw (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmulhw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmulhw %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vpmulhw (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmulhw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmulhw %ymm1, %ymm0, %ymm0 # sched: [4:1.00]
-; ZNVER1-NEXT: vpmulhw (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <16 x i16> @llvm.x86.avx2.pmulh.w(<16 x i16> %a0, <16 x i16> %a1)
- %2 = load <16 x i16>, <16 x i16> *%a2, align 32
- %3 = call <16 x i16> @llvm.x86.avx2.pmulh.w(<16 x i16> %1, <16 x i16> %2)
- ret <16 x i16> %3
-}
-declare <16 x i16> @llvm.x86.avx2.pmulh.w(<16 x i16>, <16 x i16>) nounwind readnone
-
-define <8 x i32> @test_pmulld(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
-; GENERIC-LABEL: test_pmulld:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmulld %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; GENERIC-NEXT: vpmulld (%rdi), %ymm0, %ymm0 # sched: [12:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmulld:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmulld %ymm1, %ymm0, %ymm0 # sched: [10:2.00]
-; HASWELL-NEXT: vpmulld (%rdi), %ymm0, %ymm0 # sched: [17:2.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmulld:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmulld %ymm1, %ymm0, %ymm0 # sched: [10:2.00]
-; BROADWELL-NEXT: vpmulld (%rdi), %ymm0, %ymm0 # sched: [16:2.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmulld:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmulld %ymm1, %ymm0, %ymm0 # sched: [10:1.00]
-; SKYLAKE-NEXT: vpmulld (%rdi), %ymm0, %ymm0 # sched: [17:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmulld:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmulld %ymm1, %ymm0, %ymm0 # sched: [10:1.00]
-; SKX-NEXT: vpmulld (%rdi), %ymm0, %ymm0 # sched: [17:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmulld:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmulld %ymm1, %ymm0, %ymm0 # sched: [5:2.00]
-; ZNVER1-NEXT: vpmulld (%rdi), %ymm0, %ymm0 # sched: [12:2.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = mul <8 x i32> %a0, %a1
- %2 = load <8 x i32>, <8 x i32> *%a2, align 32
- %3 = mul <8 x i32> %1, %2
- ret <8 x i32> %3
-}
-
-define <16 x i16> @test_pmullw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_pmullw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmullw %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; GENERIC-NEXT: vpmullw (%rdi), %ymm0, %ymm0 # sched: [12:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmullw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmullw %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; HASWELL-NEXT: vpmullw (%rdi), %ymm0, %ymm0 # sched: [12:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmullw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmullw %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; BROADWELL-NEXT: vpmullw (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmullw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmullw %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vpmullw (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmullw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmullw %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vpmullw (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmullw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmullw %ymm1, %ymm0, %ymm0 # sched: [4:1.00]
-; ZNVER1-NEXT: vpmullw (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = mul <16 x i16> %a0, %a1
- %2 = load <16 x i16>, <16 x i16> *%a2, align 32
- %3 = mul <16 x i16> %1, %2
- ret <16 x i16> %3
-}
-
-define <4 x i64> @test_pmuludq(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
-; GENERIC-LABEL: test_pmuludq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpmuludq %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; GENERIC-NEXT: vpmuludq (%rdi), %ymm0, %ymm0 # sched: [12:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pmuludq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpmuludq %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; HASWELL-NEXT: vpmuludq (%rdi), %ymm0, %ymm0 # sched: [12:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pmuludq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpmuludq %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; BROADWELL-NEXT: vpmuludq (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pmuludq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpmuludq %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKYLAKE-NEXT: vpmuludq (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pmuludq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpmuludq %ymm1, %ymm0, %ymm0 # sched: [4:0.50]
-; SKX-NEXT: vpmuludq (%rdi), %ymm0, %ymm0 # sched: [11:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pmuludq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpmuludq %ymm1, %ymm0, %ymm0 # sched: [4:1.00]
-; ZNVER1-NEXT: vpmuludq (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x i64> @llvm.x86.avx2.pmulu.dq(<8 x i32> %a0, <8 x i32> %a1)
- %2 = bitcast <4 x i64> %1 to <8 x i32>
- %3 = load <8 x i32>, <8 x i32> *%a2, align 32
- %4 = call <4 x i64> @llvm.x86.avx2.pmulu.dq(<8 x i32> %2, <8 x i32> %3)
- ret <4 x i64> %4
-}
-declare <4 x i64> @llvm.x86.avx2.pmulu.dq(<8 x i32>, <8 x i32>) nounwind readnone
-
-define <4 x i64> @test_por(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) {
-; GENERIC-LABEL: test_por:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; GENERIC-NEXT: vpor (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_por:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; HASWELL-NEXT: vpor (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_por:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; BROADWELL-NEXT: vpor (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_por:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: vpor (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_por:
-; SKX: # %bb.0:
-; SKX-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: vpor (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_por:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpor (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = or <4 x i64> %a0, %a1
- %2 = load <4 x i64>, <4 x i64> *%a2, align 32
- %3 = or <4 x i64> %1, %2
- %4 = add <4 x i64> %3, %a1
- ret <4 x i64> %4
-}
-
-define <4 x i64> @test_psadbw(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) {
-; GENERIC-LABEL: test_psadbw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsadbw %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; GENERIC-NEXT: vpsadbw (%rdi), %ymm0, %ymm0 # sched: [12:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psadbw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsadbw %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; HASWELL-NEXT: vpsadbw (%rdi), %ymm0, %ymm0 # sched: [12:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psadbw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsadbw %ymm1, %ymm0, %ymm0 # sched: [5:1.00]
-; BROADWELL-NEXT: vpsadbw (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psadbw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsadbw %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SKYLAKE-NEXT: vpsadbw (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psadbw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsadbw %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; SKX-NEXT: vpsadbw (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psadbw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsadbw %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
-; ZNVER1-NEXT: vpsadbw (%rdi), %ymm0, %ymm0 # sched: [10:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x i64> @llvm.x86.avx2.psad.bw(<32 x i8> %a0, <32 x i8> %a1)
- %2 = bitcast <4 x i64> %1 to <32 x i8>
- %3 = load <32 x i8>, <32 x i8> *%a2, align 32
- %4 = call <4 x i64> @llvm.x86.avx2.psad.bw(<32 x i8> %2, <32 x i8> %3)
- ret <4 x i64> %4
-}
-declare <4 x i64> @llvm.x86.avx2.psad.bw(<32 x i8>, <32 x i8>) nounwind readnone
-
-define <32 x i8> @test_pshufb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) {
-; GENERIC-LABEL: test_pshufb:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpshufb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpshufb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pshufb:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpshufb %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: vpshufb (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pshufb:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpshufb %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: vpshufb (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pshufb:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpshufb %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SKYLAKE-NEXT: vpshufb (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pshufb:
-; SKX: # %bb.0:
-; SKX-NEXT: vpshufb %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; SKX-NEXT: vpshufb (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pshufb:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpshufb %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpshufb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %a0, <32 x i8> %a1)
- %2 = load <32 x i8>, <32 x i8> *%a2, align 32
- %3 = call <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8> %1, <32 x i8> %2)
- ret <32 x i8> %3
-}
-declare <32 x i8> @llvm.x86.avx2.pshuf.b(<32 x i8>, <32 x i8>) nounwind readnone
-
-define <8 x i32> @test_pshufd(<8 x i32> %a0, <8 x i32> *%a1) {
-; GENERIC-LABEL: test_pshufd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4] sched: [1:1.00]
-; GENERIC-NEXT: vpshufd {{.*#+}} ymm1 = mem[1,0,3,2,5,4,7,6] sched: [8:1.00]
-; GENERIC-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pshufd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4] sched: [1:1.00]
-; HASWELL-NEXT: vpshufd {{.*#+}} ymm1 = mem[1,0,3,2,5,4,7,6] sched: [8:1.00]
-; HASWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pshufd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4] sched: [1:1.00]
-; BROADWELL-NEXT: vpshufd {{.*#+}} ymm1 = mem[1,0,3,2,5,4,7,6] sched: [7:1.00]
-; BROADWELL-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pshufd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4] sched: [1:1.00]
-; SKYLAKE-NEXT: vpshufd {{.*#+}} ymm1 = mem[1,0,3,2,5,4,7,6] sched: [8:1.00]
-; SKYLAKE-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pshufd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4] sched: [1:1.00]
-; SKX-NEXT: vpshufd {{.*#+}} ymm1 = mem[1,0,3,2,5,4,7,6] sched: [8:1.00]
-; SKX-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pshufd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpshufd {{.*#+}} ymm1 = mem[1,0,3,2,5,4,7,6] sched: [8:0.50]
-; ZNVER1-NEXT: vpshufd {{.*#+}} ymm0 = ymm0[3,2,1,0,7,6,5,4] sched: [1:0.25]
-; ZNVER1-NEXT: vpaddd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <8 x i32> %a0, <8 x i32> undef, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4>
- %2 = load <8 x i32>, <8 x i32> *%a1, align 32
- %3 = shufflevector <8 x i32> %2, <8 x i32> undef, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6>
- %4 = add <8 x i32> %1, %3
- ret <8 x i32> %4
-}
-
-define <16 x i16> @test_pshufhw(<16 x i16> %a0, <16 x i16> *%a1) {
-; GENERIC-LABEL: test_pshufhw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,6,5,4,8,9,10,11,15,14,13,12] sched: [1:1.00]
-; GENERIC-NEXT: vpshufhw {{.*#+}} ymm1 = mem[0,1,2,3,5,4,7,6,8,9,10,11,13,12,15,14] sched: [8:1.00]
-; GENERIC-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pshufhw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,6,5,4,8,9,10,11,15,14,13,12] sched: [1:1.00]
-; HASWELL-NEXT: vpshufhw {{.*#+}} ymm1 = mem[0,1,2,3,5,4,7,6,8,9,10,11,13,12,15,14] sched: [8:1.00]
-; HASWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pshufhw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,6,5,4,8,9,10,11,15,14,13,12] sched: [1:1.00]
-; BROADWELL-NEXT: vpshufhw {{.*#+}} ymm1 = mem[0,1,2,3,5,4,7,6,8,9,10,11,13,12,15,14] sched: [7:1.00]
-; BROADWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pshufhw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,6,5,4,8,9,10,11,15,14,13,12] sched: [1:1.00]
-; SKYLAKE-NEXT: vpshufhw {{.*#+}} ymm1 = mem[0,1,2,3,5,4,7,6,8,9,10,11,13,12,15,14] sched: [8:1.00]
-; SKYLAKE-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pshufhw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,6,5,4,8,9,10,11,15,14,13,12] sched: [1:1.00]
-; SKX-NEXT: vpshufhw {{.*#+}} ymm1 = mem[0,1,2,3,5,4,7,6,8,9,10,11,13,12,15,14] sched: [8:1.00]
-; SKX-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pshufhw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpshufhw {{.*#+}} ymm1 = mem[0,1,2,3,5,4,7,6,8,9,10,11,13,12,15,14] sched: [8:0.50]
-; ZNVER1-NEXT: vpshufhw {{.*#+}} ymm0 = ymm0[0,1,2,3,7,6,5,4,8,9,10,11,15,14,13,12] sched: [1:0.25]
-; ZNVER1-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <16 x i16> %a0, <16 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 7, i32 6, i32 5, i32 4, i32 8, i32 9, i32 10, i32 11, i32 15, i32 14, i32 13, i32 12>
- %2 = load <16 x i16>, <16 x i16> *%a1, align 32
- %3 = shufflevector <16 x i16> %2, <16 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 5, i32 4, i32 7, i32 6, i32 8, i32 9, i32 10, i32 11, i32 13, i32 12, i32 15, i32 14>
- %4 = or <16 x i16> %1, %3
- ret <16 x i16> %4
-}
-
-define <16 x i16> @test_pshuflw(<16 x i16> %a0, <16 x i16> *%a1) {
-; GENERIC-LABEL: test_pshuflw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpshuflw {{.*#+}} ymm0 = ymm0[3,2,1,0,4,5,6,7,11,10,9,8,12,13,14,15] sched: [1:1.00]
-; GENERIC-NEXT: vpshuflw {{.*#+}} ymm1 = mem[1,0,3,2,4,5,6,7,9,8,11,10,12,13,14,15] sched: [8:1.00]
-; GENERIC-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pshuflw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpshuflw {{.*#+}} ymm0 = ymm0[3,2,1,0,4,5,6,7,11,10,9,8,12,13,14,15] sched: [1:1.00]
-; HASWELL-NEXT: vpshuflw {{.*#+}} ymm1 = mem[1,0,3,2,4,5,6,7,9,8,11,10,12,13,14,15] sched: [8:1.00]
-; HASWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pshuflw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpshuflw {{.*#+}} ymm0 = ymm0[3,2,1,0,4,5,6,7,11,10,9,8,12,13,14,15] sched: [1:1.00]
-; BROADWELL-NEXT: vpshuflw {{.*#+}} ymm1 = mem[1,0,3,2,4,5,6,7,9,8,11,10,12,13,14,15] sched: [7:1.00]
-; BROADWELL-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pshuflw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpshuflw {{.*#+}} ymm0 = ymm0[3,2,1,0,4,5,6,7,11,10,9,8,12,13,14,15] sched: [1:1.00]
-; SKYLAKE-NEXT: vpshuflw {{.*#+}} ymm1 = mem[1,0,3,2,4,5,6,7,9,8,11,10,12,13,14,15] sched: [8:1.00]
-; SKYLAKE-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pshuflw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpshuflw {{.*#+}} ymm0 = ymm0[3,2,1,0,4,5,6,7,11,10,9,8,12,13,14,15] sched: [1:1.00]
-; SKX-NEXT: vpshuflw {{.*#+}} ymm1 = mem[1,0,3,2,4,5,6,7,9,8,11,10,12,13,14,15] sched: [8:1.00]
-; SKX-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pshuflw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpshuflw {{.*#+}} ymm1 = mem[1,0,3,2,4,5,6,7,9,8,11,10,12,13,14,15] sched: [8:0.50]
-; ZNVER1-NEXT: vpshuflw {{.*#+}} ymm0 = ymm0[3,2,1,0,4,5,6,7,11,10,9,8,12,13,14,15] sched: [1:0.25]
-; ZNVER1-NEXT: vpor %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <16 x i16> %a0, <16 x i16> undef, <16 x i32> <i32 3, i32 2, i32 1, i32 0, i32 4, i32 5, i32 6, i32 7, i32 11, i32 10, i32 9, i32 8, i32 12, i32 13, i32 14, i32 15>
- %2 = load <16 x i16>, <16 x i16> *%a1, align 32
- %3 = shufflevector <16 x i16> %2, <16 x i16> undef, <16 x i32> <i32 1, i32 0, i32 3, i32 2, i32 4, i32 5, i32 6, i32 7, i32 9, i32 8, i32 11, i32 10, i32 12, i32 13, i32 14, i32 15>
- %4 = or <16 x i16> %1, %3
- ret <16 x i16> %4
-}
-
-define <32 x i8> @test_psignb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) {
-; GENERIC-LABEL: test_psignb:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsignb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpsignb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psignb:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsignb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpsignb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psignb:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsignb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpsignb (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psignb:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsignb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpsignb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psignb:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsignb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpsignb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psignb:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsignb %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpsignb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <32 x i8> @llvm.x86.avx2.psign.b(<32 x i8> %a0, <32 x i8> %a1)
- %2 = load <32 x i8>, <32 x i8> *%a2, align 32
- %3 = call <32 x i8> @llvm.x86.avx2.psign.b(<32 x i8> %1, <32 x i8> %2)
- ret <32 x i8> %3
-}
-declare <32 x i8> @llvm.x86.avx2.psign.b(<32 x i8>, <32 x i8>) nounwind readnone
-
-define <8 x i32> @test_psignd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
-; GENERIC-LABEL: test_psignd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsignd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpsignd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psignd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsignd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpsignd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psignd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsignd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpsignd (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psignd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsignd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpsignd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psignd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsignd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpsignd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psignd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsignd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpsignd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x i32> @llvm.x86.avx2.psign.d(<8 x i32> %a0, <8 x i32> %a1)
- %2 = load <8 x i32>, <8 x i32> *%a2, align 32
- %3 = call <8 x i32> @llvm.x86.avx2.psign.d(<8 x i32> %1, <8 x i32> %2)
- ret <8 x i32> %3
-}
-declare <8 x i32> @llvm.x86.avx2.psign.d(<8 x i32>, <8 x i32>) nounwind readnone
-
-define <16 x i16> @test_psignw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_psignw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsignw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpsignw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psignw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsignw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpsignw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psignw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsignw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpsignw (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psignw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsignw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpsignw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psignw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsignw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpsignw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psignw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsignw %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpsignw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <16 x i16> @llvm.x86.avx2.psign.w(<16 x i16> %a0, <16 x i16> %a1)
- %2 = load <16 x i16>, <16 x i16> *%a2, align 32
- %3 = call <16 x i16> @llvm.x86.avx2.psign.w(<16 x i16> %1, <16 x i16> %2)
- ret <16 x i16> %3
-}
-declare <16 x i16> @llvm.x86.avx2.psign.w(<16 x i16>, <16 x i16>) nounwind readnone
-
-define <8 x i32> @test_pslld(<8 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {
-; GENERIC-LABEL: test_pslld:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpslld %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; GENERIC-NEXT: vpslld (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; GENERIC-NEXT: vpslld $2, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pslld:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpslld %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; HASWELL-NEXT: vpslld (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: vpslld $2, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pslld:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpslld %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; BROADWELL-NEXT: vpslld (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: vpslld $2, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pslld:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpslld %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; SKYLAKE-NEXT: vpslld (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: vpslld $2, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pslld:
-; SKX: # %bb.0:
-; SKX-NEXT: vpslld %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; SKX-NEXT: vpslld (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: vpslld $2, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pslld:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpslld %xmm1, %ymm0, %ymm0 # sched: [2:1.00]
-; ZNVER1-NEXT: vpslld (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; ZNVER1-NEXT: vpslld $2, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x i32> @llvm.x86.avx2.psll.d(<8 x i32> %a0, <4 x i32> %a1)
- %2 = load <4 x i32>, <4 x i32> *%a2, align 16
- %3 = call <8 x i32> @llvm.x86.avx2.psll.d(<8 x i32> %1, <4 x i32> %2)
- %4 = shl <8 x i32> %3, <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>
- ret <8 x i32> %4
-}
-declare <8 x i32> @llvm.x86.avx2.psll.d(<8 x i32>, <4 x i32>) nounwind readnone
-
-define <32 x i8> @test_pslldq(<32 x i8> %a0) {
-; GENERIC-LABEL: test_pslldq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpslldq {{.*#+}} ymm0 = zero,zero,zero,ymm0[0,1,2,3,4,5,6,7,8,9,10,11,12],zero,zero,zero,ymm0[16,17,18,19,20,21,22,23,24,25,26,27,28] sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pslldq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpslldq {{.*#+}} ymm0 = zero,zero,zero,ymm0[0,1,2,3,4,5,6,7,8,9,10,11,12],zero,zero,zero,ymm0[16,17,18,19,20,21,22,23,24,25,26,27,28] sched: [1:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pslldq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpslldq {{.*#+}} ymm0 = zero,zero,zero,ymm0[0,1,2,3,4,5,6,7,8,9,10,11,12],zero,zero,zero,ymm0[16,17,18,19,20,21,22,23,24,25,26,27,28] sched: [1:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pslldq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpslldq {{.*#+}} ymm0 = zero,zero,zero,ymm0[0,1,2,3,4,5,6,7,8,9,10,11,12],zero,zero,zero,ymm0[16,17,18,19,20,21,22,23,24,25,26,27,28] sched: [1:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pslldq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpslldq {{.*#+}} ymm0 = zero,zero,zero,ymm0[0,1,2,3,4,5,6,7,8,9,10,11,12],zero,zero,zero,ymm0[16,17,18,19,20,21,22,23,24,25,26,27,28] sched: [1:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pslldq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpslldq {{.*#+}} ymm0 = zero,zero,zero,ymm0[0,1,2,3,4,5,6,7,8,9,10,11,12],zero,zero,zero,ymm0[16,17,18,19,20,21,22,23,24,25,26,27,28] sched: [2:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <32 x i8> zeroinitializer, <32 x i8> %a0, <32 x i32> <i32 13, i32 14, i32 15, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 29, i32 30, i32 31, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60>
- ret <32 x i8> %1
-}
-
-define <4 x i64> @test_psllq(<4 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) {
-; GENERIC-LABEL: test_psllq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsllq %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; GENERIC-NEXT: vpsllq (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; GENERIC-NEXT: vpsllq $2, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psllq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsllq %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; HASWELL-NEXT: vpsllq (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: vpsllq $2, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psllq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsllq %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; BROADWELL-NEXT: vpsllq (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: vpsllq $2, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psllq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsllq %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; SKYLAKE-NEXT: vpsllq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: vpsllq $2, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psllq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsllq %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; SKX-NEXT: vpsllq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: vpsllq $2, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psllq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsllq %xmm1, %ymm0, %ymm0 # sched: [2:1.00]
-; ZNVER1-NEXT: vpsllq (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; ZNVER1-NEXT: vpsllq $2, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x i64> @llvm.x86.avx2.psll.q(<4 x i64> %a0, <2 x i64> %a1)
- %2 = load <2 x i64>, <2 x i64> *%a2, align 16
- %3 = call <4 x i64> @llvm.x86.avx2.psll.q(<4 x i64> %1, <2 x i64> %2)
- %4 = shl <4 x i64> %3, <i64 2, i64 2, i64 2, i64 2>
- ret <4 x i64> %4
-}
-declare <4 x i64> @llvm.x86.avx2.psll.q(<4 x i64>, <2 x i64>) nounwind readnone
-
-define <4 x i32> @test_psllvd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {
-; GENERIC-LABEL: test_psllvd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsllvd %xmm1, %xmm0, %xmm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpsllvd (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psllvd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsllvd %xmm1, %xmm0, %xmm0 # sched: [3:2.00]
-; HASWELL-NEXT: vpsllvd (%rdi), %xmm0, %xmm0 # sched: [9:2.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psllvd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsllvd %xmm1, %xmm0, %xmm0 # sched: [3:2.00]
-; BROADWELL-NEXT: vpsllvd (%rdi), %xmm0, %xmm0 # sched: [8:2.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psllvd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsllvd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpsllvd (%rdi), %xmm0, %xmm0 # sched: [7:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psllvd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsllvd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SKX-NEXT: vpsllvd (%rdi), %xmm0, %xmm0 # sched: [7:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psllvd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsllvd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; ZNVER1-NEXT: vpsllvd (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x i32> @llvm.x86.avx2.psllv.d(<4 x i32> %a0, <4 x i32> %a1)
- %2 = load <4 x i32>, <4 x i32> *%a2, align 16
- %3 = call <4 x i32> @llvm.x86.avx2.psllv.d(<4 x i32> %1, <4 x i32> %2)
- ret <4 x i32> %3
-}
-declare <4 x i32> @llvm.x86.avx2.psllv.d(<4 x i32>, <4 x i32>) nounwind readnone
-
-define <8 x i32> @test_psllvd_ymm(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
-; GENERIC-LABEL: test_psllvd_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsllvd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpsllvd (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psllvd_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsllvd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; HASWELL-NEXT: vpsllvd (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psllvd_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsllvd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BROADWELL-NEXT: vpsllvd (%rdi), %ymm0, %ymm0 # sched: [9:2.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psllvd_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsllvd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpsllvd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psllvd_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsllvd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpsllvd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psllvd_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsllvd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; ZNVER1-NEXT: vpsllvd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x i32> @llvm.x86.avx2.psllv.d.256(<8 x i32> %a0, <8 x i32> %a1)
- %2 = load <8 x i32>, <8 x i32> *%a2, align 32
- %3 = call <8 x i32> @llvm.x86.avx2.psllv.d.256(<8 x i32> %1, <8 x i32> %2)
- ret <8 x i32> %3
-}
-declare <8 x i32> @llvm.x86.avx2.psllv.d.256(<8 x i32>, <8 x i32>) nounwind readnone
-
-define <2 x i64> @test_psllvq(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) {
-; GENERIC-LABEL: test_psllvq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsllvq %xmm1, %xmm0, %xmm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpsllvq (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psllvq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsllvq %xmm1, %xmm0, %xmm0 # sched: [1:1.00]
-; HASWELL-NEXT: vpsllvq (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psllvq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsllvq %xmm1, %xmm0, %xmm0 # sched: [1:1.00]
-; BROADWELL-NEXT: vpsllvq (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psllvq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsllvq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpsllvq (%rdi), %xmm0, %xmm0 # sched: [7:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psllvq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsllvq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SKX-NEXT: vpsllvq (%rdi), %xmm0, %xmm0 # sched: [7:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psllvq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsllvq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; ZNVER1-NEXT: vpsllvq (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <2 x i64> @llvm.x86.avx2.psllv.q(<2 x i64> %a0, <2 x i64> %a1)
- %2 = load <2 x i64>, <2 x i64> *%a2, align 16
- %3 = call <2 x i64> @llvm.x86.avx2.psllv.q(<2 x i64> %1, <2 x i64> %2)
- ret <2 x i64> %3
-}
-declare <2 x i64> @llvm.x86.avx2.psllv.q(<2 x i64>, <2 x i64>) nounwind readnone
-
-define <4 x i64> @test_psllvq_ymm(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) {
-; GENERIC-LABEL: test_psllvq_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsllvq %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpsllvq (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psllvq_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsllvq %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: vpsllvq (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psllvq_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsllvq %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: vpsllvq (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psllvq_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsllvq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpsllvq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psllvq_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsllvq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpsllvq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psllvq_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsllvq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; ZNVER1-NEXT: vpsllvq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x i64> @llvm.x86.avx2.psllv.q.256(<4 x i64> %a0, <4 x i64> %a1)
- %2 = load <4 x i64>, <4 x i64> *%a2, align 32
- %3 = call <4 x i64> @llvm.x86.avx2.psllv.q.256(<4 x i64> %1, <4 x i64> %2)
- ret <4 x i64> %3
-}
-declare <4 x i64> @llvm.x86.avx2.psllv.q.256(<4 x i64>, <4 x i64>) nounwind readnone
-
-define <16 x i16> @test_psllw(<16 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {
-; GENERIC-LABEL: test_psllw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsllw %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; GENERIC-NEXT: vpsllw (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; GENERIC-NEXT: vpsllw $2, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psllw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsllw %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; HASWELL-NEXT: vpsllw (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: vpsllw $2, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psllw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsllw %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; BROADWELL-NEXT: vpsllw (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: vpsllw $2, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psllw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsllw %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; SKYLAKE-NEXT: vpsllw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: vpsllw $2, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psllw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsllw %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; SKX-NEXT: vpsllw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: vpsllw $2, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psllw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsllw %xmm1, %ymm0, %ymm0 # sched: [2:1.00]
-; ZNVER1-NEXT: vpsllw (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; ZNVER1-NEXT: vpsllw $2, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <16 x i16> @llvm.x86.avx2.psll.w(<16 x i16> %a0, <8 x i16> %a1)
- %2 = load <8 x i16>, <8 x i16> *%a2, align 16
- %3 = call <16 x i16> @llvm.x86.avx2.psll.w(<16 x i16> %1, <8 x i16> %2)
- %4 = shl <16 x i16> %3, <i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2>
- ret <16 x i16> %4
-}
-declare <16 x i16> @llvm.x86.avx2.psll.w(<16 x i16>, <8 x i16>) nounwind readnone
-
-define <8 x i32> @test_psrad(<8 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {
-; GENERIC-LABEL: test_psrad:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsrad %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; GENERIC-NEXT: vpsrad (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; GENERIC-NEXT: vpsrad $2, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psrad:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsrad %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; HASWELL-NEXT: vpsrad (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: vpsrad $2, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psrad:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsrad %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; BROADWELL-NEXT: vpsrad (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: vpsrad $2, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psrad:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsrad %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; SKYLAKE-NEXT: vpsrad (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: vpsrad $2, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psrad:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsrad %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; SKX-NEXT: vpsrad (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: vpsrad $2, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psrad:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsrad %xmm1, %ymm0, %ymm0 # sched: [2:1.00]
-; ZNVER1-NEXT: vpsrad (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; ZNVER1-NEXT: vpsrad $2, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x i32> @llvm.x86.avx2.psra.d(<8 x i32> %a0, <4 x i32> %a1)
- %2 = load <4 x i32>, <4 x i32> *%a2, align 16
- %3 = call <8 x i32> @llvm.x86.avx2.psra.d(<8 x i32> %1, <4 x i32> %2)
- %4 = ashr <8 x i32> %3, <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>
- ret <8 x i32> %4
-}
-declare <8 x i32> @llvm.x86.avx2.psra.d(<8 x i32>, <4 x i32>) nounwind readnone
-
-define <4 x i32> @test_psravd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {
-; GENERIC-LABEL: test_psravd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsravd %xmm1, %xmm0, %xmm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpsravd (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psravd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsravd %xmm1, %xmm0, %xmm0 # sched: [3:2.00]
-; HASWELL-NEXT: vpsravd (%rdi), %xmm0, %xmm0 # sched: [9:2.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psravd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsravd %xmm1, %xmm0, %xmm0 # sched: [3:2.00]
-; BROADWELL-NEXT: vpsravd (%rdi), %xmm0, %xmm0 # sched: [8:2.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psravd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsravd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpsravd (%rdi), %xmm0, %xmm0 # sched: [7:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psravd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsravd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SKX-NEXT: vpsravd (%rdi), %xmm0, %xmm0 # sched: [7:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psravd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsravd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; ZNVER1-NEXT: vpsravd (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x i32> @llvm.x86.avx2.psrav.d(<4 x i32> %a0, <4 x i32> %a1)
- %2 = load <4 x i32>, <4 x i32> *%a2, align 16
- %3 = call <4 x i32> @llvm.x86.avx2.psrav.d(<4 x i32> %1, <4 x i32> %2)
- ret <4 x i32> %3
-}
-declare <4 x i32> @llvm.x86.avx2.psrav.d(<4 x i32>, <4 x i32>) nounwind readnone
-
-define <8 x i32> @test_psravd_ymm(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
-; GENERIC-LABEL: test_psravd_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsravd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpsravd (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psravd_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsravd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; HASWELL-NEXT: vpsravd (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psravd_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsravd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BROADWELL-NEXT: vpsravd (%rdi), %ymm0, %ymm0 # sched: [9:2.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psravd_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsravd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpsravd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psravd_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsravd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpsravd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psravd_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsravd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; ZNVER1-NEXT: vpsravd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x i32> @llvm.x86.avx2.psrav.d.256(<8 x i32> %a0, <8 x i32> %a1)
- %2 = load <8 x i32>, <8 x i32> *%a2, align 32
- %3 = call <8 x i32> @llvm.x86.avx2.psrav.d.256(<8 x i32> %1, <8 x i32> %2)
- ret <8 x i32> %3
-}
-declare <8 x i32> @llvm.x86.avx2.psrav.d.256(<8 x i32>, <8 x i32>) nounwind readnone
-
-define <16 x i16> @test_psraw(<16 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {
-; GENERIC-LABEL: test_psraw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsraw %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; GENERIC-NEXT: vpsraw (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; GENERIC-NEXT: vpsraw $2, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psraw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsraw %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; HASWELL-NEXT: vpsraw (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: vpsraw $2, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psraw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsraw %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; BROADWELL-NEXT: vpsraw (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: vpsraw $2, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psraw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsraw %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; SKYLAKE-NEXT: vpsraw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: vpsraw $2, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psraw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsraw %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; SKX-NEXT: vpsraw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: vpsraw $2, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psraw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsraw %xmm1, %ymm0, %ymm0 # sched: [2:1.00]
-; ZNVER1-NEXT: vpsraw (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; ZNVER1-NEXT: vpsraw $2, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <16 x i16> @llvm.x86.avx2.psra.w(<16 x i16> %a0, <8 x i16> %a1)
- %2 = load <8 x i16>, <8 x i16> *%a2, align 16
- %3 = call <16 x i16> @llvm.x86.avx2.psra.w(<16 x i16> %1, <8 x i16> %2)
- %4 = ashr <16 x i16> %3, <i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2>
- ret <16 x i16> %4
-}
-declare <16 x i16> @llvm.x86.avx2.psra.w(<16 x i16>, <8 x i16>) nounwind readnone
-
-define <8 x i32> @test_psrld(<8 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {
-; GENERIC-LABEL: test_psrld:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsrld %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; GENERIC-NEXT: vpsrld (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; GENERIC-NEXT: vpsrld $2, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psrld:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsrld %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; HASWELL-NEXT: vpsrld (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: vpsrld $2, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psrld:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsrld %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; BROADWELL-NEXT: vpsrld (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: vpsrld $2, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psrld:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsrld %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; SKYLAKE-NEXT: vpsrld (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: vpsrld $2, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psrld:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsrld %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; SKX-NEXT: vpsrld (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: vpsrld $2, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psrld:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsrld %xmm1, %ymm0, %ymm0 # sched: [2:1.00]
-; ZNVER1-NEXT: vpsrld (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; ZNVER1-NEXT: vpsrld $2, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x i32> @llvm.x86.avx2.psrl.d(<8 x i32> %a0, <4 x i32> %a1)
- %2 = load <4 x i32>, <4 x i32> *%a2, align 16
- %3 = call <8 x i32> @llvm.x86.avx2.psrl.d(<8 x i32> %1, <4 x i32> %2)
- %4 = lshr <8 x i32> %3, <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>
- ret <8 x i32> %4
-}
-declare <8 x i32> @llvm.x86.avx2.psrl.d(<8 x i32>, <4 x i32>) nounwind readnone
-
-define <32 x i8> @test_psrldq(<32 x i8> %a0) {
-; GENERIC-LABEL: test_psrldq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsrldq {{.*#+}} ymm0 = ymm0[3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero,zero,ymm0[19,20,21,22,23,24,25,26,27,28,29,30,31],zero,zero,zero sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psrldq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsrldq {{.*#+}} ymm0 = ymm0[3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero,zero,ymm0[19,20,21,22,23,24,25,26,27,28,29,30,31],zero,zero,zero sched: [1:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psrldq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsrldq {{.*#+}} ymm0 = ymm0[3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero,zero,ymm0[19,20,21,22,23,24,25,26,27,28,29,30,31],zero,zero,zero sched: [1:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psrldq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsrldq {{.*#+}} ymm0 = ymm0[3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero,zero,ymm0[19,20,21,22,23,24,25,26,27,28,29,30,31],zero,zero,zero sched: [1:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psrldq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsrldq {{.*#+}} ymm0 = ymm0[3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero,zero,ymm0[19,20,21,22,23,24,25,26,27,28,29,30,31],zero,zero,zero sched: [1:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psrldq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsrldq {{.*#+}} ymm0 = ymm0[3,4,5,6,7,8,9,10,11,12,13,14,15],zero,zero,zero,ymm0[19,20,21,22,23,24,25,26,27,28,29,30,31],zero,zero,zero sched: [2:1.00]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <32 x i8> %a0, <32 x i8> zeroinitializer, <32 x i32> <i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 32, i32 33, i32 34, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 48, i32 49, i32 50>
- ret <32 x i8> %1
-}
-
-define <4 x i64> @test_psrlq(<4 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) {
-; GENERIC-LABEL: test_psrlq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsrlq %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; GENERIC-NEXT: vpsrlq (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; GENERIC-NEXT: vpsrlq $2, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psrlq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsrlq %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; HASWELL-NEXT: vpsrlq (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: vpsrlq $2, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psrlq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsrlq %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; BROADWELL-NEXT: vpsrlq (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: vpsrlq $2, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psrlq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsrlq %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; SKYLAKE-NEXT: vpsrlq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: vpsrlq $2, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psrlq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsrlq %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; SKX-NEXT: vpsrlq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: vpsrlq $2, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psrlq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsrlq %xmm1, %ymm0, %ymm0 # sched: [2:1.00]
-; ZNVER1-NEXT: vpsrlq (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; ZNVER1-NEXT: vpsrlq $2, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x i64> @llvm.x86.avx2.psrl.q(<4 x i64> %a0, <2 x i64> %a1)
- %2 = load <2 x i64>, <2 x i64> *%a2, align 16
- %3 = call <4 x i64> @llvm.x86.avx2.psrl.q(<4 x i64> %1, <2 x i64> %2)
- %4 = lshr <4 x i64> %3, <i64 2, i64 2, i64 2, i64 2>
- ret <4 x i64> %4
-}
-declare <4 x i64> @llvm.x86.avx2.psrl.q(<4 x i64>, <2 x i64>) nounwind readnone
-
-define <4 x i32> @test_psrlvd(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> *%a2) {
-; GENERIC-LABEL: test_psrlvd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsrlvd %xmm1, %xmm0, %xmm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpsrlvd (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psrlvd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsrlvd %xmm1, %xmm0, %xmm0 # sched: [3:2.00]
-; HASWELL-NEXT: vpsrlvd (%rdi), %xmm0, %xmm0 # sched: [9:2.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psrlvd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsrlvd %xmm1, %xmm0, %xmm0 # sched: [3:2.00]
-; BROADWELL-NEXT: vpsrlvd (%rdi), %xmm0, %xmm0 # sched: [8:2.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psrlvd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsrlvd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpsrlvd (%rdi), %xmm0, %xmm0 # sched: [7:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psrlvd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsrlvd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SKX-NEXT: vpsrlvd (%rdi), %xmm0, %xmm0 # sched: [7:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psrlvd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsrlvd %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; ZNVER1-NEXT: vpsrlvd (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x i32> @llvm.x86.avx2.psrlv.d(<4 x i32> %a0, <4 x i32> %a1)
- %2 = load <4 x i32>, <4 x i32> *%a2, align 16
- %3 = call <4 x i32> @llvm.x86.avx2.psrlv.d(<4 x i32> %1, <4 x i32> %2)
- ret <4 x i32> %3
-}
-declare <4 x i32> @llvm.x86.avx2.psrlv.d(<4 x i32>, <4 x i32>) nounwind readnone
-
-define <8 x i32> @test_psrlvd_ymm(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
-; GENERIC-LABEL: test_psrlvd_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsrlvd %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpsrlvd (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psrlvd_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsrlvd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; HASWELL-NEXT: vpsrlvd (%rdi), %ymm0, %ymm0 # sched: [10:2.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psrlvd_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsrlvd %ymm1, %ymm0, %ymm0 # sched: [3:2.00]
-; BROADWELL-NEXT: vpsrlvd (%rdi), %ymm0, %ymm0 # sched: [9:2.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psrlvd_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsrlvd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpsrlvd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psrlvd_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsrlvd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpsrlvd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psrlvd_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsrlvd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; ZNVER1-NEXT: vpsrlvd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <8 x i32> @llvm.x86.avx2.psrlv.d.256(<8 x i32> %a0, <8 x i32> %a1)
- %2 = load <8 x i32>, <8 x i32> *%a2, align 32
- %3 = call <8 x i32> @llvm.x86.avx2.psrlv.d.256(<8 x i32> %1, <8 x i32> %2)
- ret <8 x i32> %3
-}
-declare <8 x i32> @llvm.x86.avx2.psrlv.d.256(<8 x i32>, <8 x i32>) nounwind readnone
-
-define <2 x i64> @test_psrlvq(<2 x i64> %a0, <2 x i64> %a1, <2 x i64> *%a2) {
-; GENERIC-LABEL: test_psrlvq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsrlvq %xmm1, %xmm0, %xmm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpsrlvq (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psrlvq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsrlvq %xmm1, %xmm0, %xmm0 # sched: [1:1.00]
-; HASWELL-NEXT: vpsrlvq (%rdi), %xmm0, %xmm0 # sched: [7:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psrlvq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsrlvq %xmm1, %xmm0, %xmm0 # sched: [1:1.00]
-; BROADWELL-NEXT: vpsrlvq (%rdi), %xmm0, %xmm0 # sched: [6:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psrlvq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsrlvq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpsrlvq (%rdi), %xmm0, %xmm0 # sched: [7:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psrlvq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsrlvq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; SKX-NEXT: vpsrlvq (%rdi), %xmm0, %xmm0 # sched: [7:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psrlvq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsrlvq %xmm1, %xmm0, %xmm0 # sched: [1:0.50]
-; ZNVER1-NEXT: vpsrlvq (%rdi), %xmm0, %xmm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <2 x i64> @llvm.x86.avx2.psrlv.q(<2 x i64> %a0, <2 x i64> %a1)
- %2 = load <2 x i64>, <2 x i64> *%a2, align 16
- %3 = call <2 x i64> @llvm.x86.avx2.psrlv.q(<2 x i64> %1, <2 x i64> %2)
- ret <2 x i64> %3
-}
-declare <2 x i64> @llvm.x86.avx2.psrlv.q(<2 x i64>, <2 x i64>) nounwind readnone
-
-define <4 x i64> @test_psrlvq_ymm(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) {
-; GENERIC-LABEL: test_psrlvq_ymm:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsrlvq %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: vpsrlvq (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psrlvq_ymm:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsrlvq %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: vpsrlvq (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psrlvq_ymm:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsrlvq %ymm1, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: vpsrlvq (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psrlvq_ymm:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsrlvq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpsrlvq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psrlvq_ymm:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsrlvq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpsrlvq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psrlvq_ymm:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsrlvq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; ZNVER1-NEXT: vpsrlvq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <4 x i64> @llvm.x86.avx2.psrlv.q.256(<4 x i64> %a0, <4 x i64> %a1)
- %2 = load <4 x i64>, <4 x i64> *%a2, align 32
- %3 = call <4 x i64> @llvm.x86.avx2.psrlv.q.256(<4 x i64> %1, <4 x i64> %2)
- ret <4 x i64> %3
-}
-declare <4 x i64> @llvm.x86.avx2.psrlv.q.256(<4 x i64>, <4 x i64>) nounwind readnone
-
-define <16 x i16> @test_psrlw(<16 x i16> %a0, <8 x i16> %a1, <8 x i16> *%a2) {
-; GENERIC-LABEL: test_psrlw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsrlw %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; GENERIC-NEXT: vpsrlw (%rdi), %ymm0, %ymm0 # sched: [11:1.00]
-; GENERIC-NEXT: vpsrlw $2, %ymm0, %ymm0 # sched: [1:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psrlw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsrlw %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; HASWELL-NEXT: vpsrlw (%rdi), %ymm0, %ymm0 # sched: [8:1.00]
-; HASWELL-NEXT: vpsrlw $2, %ymm0, %ymm0 # sched: [1:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psrlw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsrlw %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; BROADWELL-NEXT: vpsrlw (%rdi), %ymm0, %ymm0 # sched: [7:1.00]
-; BROADWELL-NEXT: vpsrlw $2, %ymm0, %ymm0 # sched: [1:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psrlw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsrlw %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; SKYLAKE-NEXT: vpsrlw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: vpsrlw $2, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psrlw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsrlw %xmm1, %ymm0, %ymm0 # sched: [4:1.00]
-; SKX-NEXT: vpsrlw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: vpsrlw $2, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psrlw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsrlw %xmm1, %ymm0, %ymm0 # sched: [2:1.00]
-; ZNVER1-NEXT: vpsrlw (%rdi), %ymm0, %ymm0 # sched: [9:1.00]
-; ZNVER1-NEXT: vpsrlw $2, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <16 x i16> @llvm.x86.avx2.psrl.w(<16 x i16> %a0, <8 x i16> %a1)
- %2 = load <8 x i16>, <8 x i16> *%a2, align 16
- %3 = call <16 x i16> @llvm.x86.avx2.psrl.w(<16 x i16> %1, <8 x i16> %2)
- %4 = lshr <16 x i16> %3, <i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2>
- ret <16 x i16> %4
-}
-declare <16 x i16> @llvm.x86.avx2.psrl.w(<16 x i16>, <8 x i16>) nounwind readnone
-
-define <32 x i8> @test_psubb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) {
-; GENERIC-LABEL: test_psubb:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsubb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpsubb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psubb:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsubb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpsubb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psubb:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsubb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpsubb (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psubb:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsubb %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: vpsubb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psubb:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsubb %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: vpsubb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psubb:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsubb %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpsubb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = sub <32 x i8> %a0, %a1
- %2 = load <32 x i8>, <32 x i8> *%a2, align 32
- %3 = sub <32 x i8> %1, %2
- ret <32 x i8> %3
-}
-
-define <8 x i32> @test_psubd(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
-; GENERIC-LABEL: test_psubd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpsubd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psubd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpsubd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psubd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpsubd (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psubd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: vpsubd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psubd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: vpsubd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psubd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpsubd (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = sub <8 x i32> %a0, %a1
- %2 = load <8 x i32>, <8 x i32> *%a2, align 32
- %3 = sub <8 x i32> %1, %2
- ret <8 x i32> %3
-}
-
-define <4 x i64> @test_psubq(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) {
-; GENERIC-LABEL: test_psubq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsubq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpsubq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psubq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsubq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpsubq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psubq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsubq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpsubq (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psubq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsubq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: vpsubq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psubq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsubq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: vpsubq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psubq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsubq %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpsubq (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = sub <4 x i64> %a0, %a1
- %2 = load <4 x i64>, <4 x i64> *%a2, align 32
- %3 = sub <4 x i64> %1, %2
- ret <4 x i64> %3
-}
-
-define <32 x i8> @test_psubsb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) {
-; GENERIC-LABEL: test_psubsb:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsubsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpsubsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psubsb:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsubsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpsubsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psubsb:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsubsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpsubsb (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psubsb:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsubsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpsubsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psubsb:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsubsb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpsubsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psubsb:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsubsb %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpsubsb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <32 x i8> @llvm.ssub.sat.v32i8(<32 x i8> %a0, <32 x i8> %a1)
- %2 = load <32 x i8>, <32 x i8> *%a2, align 32
- %3 = call <32 x i8> @llvm.ssub.sat.v32i8(<32 x i8> %1, <32 x i8> %2)
- ret <32 x i8> %3
-}
-declare <32 x i8> @llvm.ssub.sat.v32i8(<32 x i8>, <32 x i8>) nounwind readnone
-
-define <16 x i16> @test_psubsw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_psubsw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsubsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpsubsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psubsw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsubsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpsubsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psubsw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsubsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpsubsw (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psubsw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsubsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpsubsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psubsw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsubsw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpsubsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psubsw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsubsw %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpsubsw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <16 x i16> @llvm.ssub.sat.v16i16(<16 x i16> %a0, <16 x i16> %a1)
- %2 = load <16 x i16>, <16 x i16> *%a2, align 32
- %3 = call <16 x i16> @llvm.ssub.sat.v16i16(<16 x i16> %1, <16 x i16> %2)
- ret <16 x i16> %3
-}
-declare <16 x i16> @llvm.ssub.sat.v16i16(<16 x i16>, <16 x i16>) nounwind readnone
-
-define <32 x i8> @test_psubusb(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) {
-; GENERIC-LABEL: test_psubusb:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsubusb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpsubusb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psubusb:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsubusb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpsubusb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psubusb:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsubusb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpsubusb (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psubusb:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsubusb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpsubusb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psubusb:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsubusb %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpsubusb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psubusb:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsubusb %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpsubusb (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <32 x i8> @llvm.usub.sat.v32i8(<32 x i8> %a0, <32 x i8> %a1)
- %2 = load <32 x i8>, <32 x i8> *%a2, align 32
- %3 = call <32 x i8> @llvm.usub.sat.v32i8(<32 x i8> %1, <32 x i8> %2)
- ret <32 x i8> %3
-}
-declare <32 x i8> @llvm.usub.sat.v32i8(<32 x i8>, <32 x i8>) nounwind readnone
-
-define <16 x i16> @test_psubusw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_psubusw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsubusw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpsubusw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psubusw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsubusw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpsubusw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psubusw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsubusw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpsubusw (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psubusw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsubusw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpsubusw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psubusw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsubusw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; SKX-NEXT: vpsubusw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psubusw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsubusw %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpsubusw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = call <16 x i16> @llvm.usub.sat.v16i16(<16 x i16> %a0, <16 x i16> %a1)
- %2 = load <16 x i16>, <16 x i16> *%a2, align 32
- %3 = call <16 x i16> @llvm.usub.sat.v16i16(<16 x i16> %1, <16 x i16> %2)
- ret <16 x i16> %3
-}
-declare <16 x i16> @llvm.usub.sat.v16i16(<16 x i16>, <16 x i16>) nounwind readnone
-
-define <16 x i16> @test_psubw(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_psubw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpsubw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: vpsubw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_psubw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpsubw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: vpsubw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_psubw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpsubw %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: vpsubw (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_psubw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpsubw %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: vpsubw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_psubw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpsubw %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: vpsubw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_psubw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpsubw %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpsubw (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = sub <16 x i16> %a0, %a1
- %2 = load <16 x i16>, <16 x i16> *%a2, align 32
- %3 = sub <16 x i16> %1, %2
- ret <16 x i16> %3
-}
-
-define <32 x i8> @test_punpckhbw(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) {
-; GENERIC-LABEL: test_punpckhbw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpunpckhbw {{.*#+}} ymm0 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31] sched: [1:1.00]
-; GENERIC-NEXT: vpunpckhbw {{.*#+}} ymm0 = ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11],ymm0[12],mem[12],ymm0[13],mem[13],ymm0[14],mem[14],ymm0[15],mem[15],ymm0[24],mem[24],ymm0[25],mem[25],ymm0[26],mem[26],ymm0[27],mem[27],ymm0[28],mem[28],ymm0[29],mem[29],ymm0[30],mem[30],ymm0[31],mem[31] sched: [8:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_punpckhbw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpunpckhbw {{.*#+}} ymm0 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31] sched: [1:1.00]
-; HASWELL-NEXT: vpunpckhbw {{.*#+}} ymm0 = ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11],ymm0[12],mem[12],ymm0[13],mem[13],ymm0[14],mem[14],ymm0[15],mem[15],ymm0[24],mem[24],ymm0[25],mem[25],ymm0[26],mem[26],ymm0[27],mem[27],ymm0[28],mem[28],ymm0[29],mem[29],ymm0[30],mem[30],ymm0[31],mem[31] sched: [8:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_punpckhbw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpunpckhbw {{.*#+}} ymm0 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31] sched: [1:1.00]
-; BROADWELL-NEXT: vpunpckhbw {{.*#+}} ymm0 = ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11],ymm0[12],mem[12],ymm0[13],mem[13],ymm0[14],mem[14],ymm0[15],mem[15],ymm0[24],mem[24],ymm0[25],mem[25],ymm0[26],mem[26],ymm0[27],mem[27],ymm0[28],mem[28],ymm0[29],mem[29],ymm0[30],mem[30],ymm0[31],mem[31] sched: [7:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_punpckhbw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpunpckhbw {{.*#+}} ymm0 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31] sched: [1:1.00]
-; SKYLAKE-NEXT: vpunpckhbw {{.*#+}} ymm0 = ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11],ymm0[12],mem[12],ymm0[13],mem[13],ymm0[14],mem[14],ymm0[15],mem[15],ymm0[24],mem[24],ymm0[25],mem[25],ymm0[26],mem[26],ymm0[27],mem[27],ymm0[28],mem[28],ymm0[29],mem[29],ymm0[30],mem[30],ymm0[31],mem[31] sched: [8:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_punpckhbw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpunpckhbw {{.*#+}} ymm0 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31] sched: [1:1.00]
-; SKX-NEXT: vpunpckhbw {{.*#+}} ymm0 = ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11],ymm0[12],mem[12],ymm0[13],mem[13],ymm0[14],mem[14],ymm0[15],mem[15],ymm0[24],mem[24],ymm0[25],mem[25],ymm0[26],mem[26],ymm0[27],mem[27],ymm0[28],mem[28],ymm0[29],mem[29],ymm0[30],mem[30],ymm0[31],mem[31] sched: [8:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_punpckhbw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpunpckhbw {{.*#+}} ymm0 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31] sched: [1:0.25]
-; ZNVER1-NEXT: vpunpckhbw {{.*#+}} ymm0 = ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11],ymm0[12],mem[12],ymm0[13],mem[13],ymm0[14],mem[14],ymm0[15],mem[15],ymm0[24],mem[24],ymm0[25],mem[25],ymm0[26],mem[26],ymm0[27],mem[27],ymm0[28],mem[28],ymm0[29],mem[29],ymm0[30],mem[30],ymm0[31],mem[31] sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <32 x i8> %a0, <32 x i8> %a1, <32 x i32> <i32 8, i32 40, i32 9, i32 41, i32 10, i32 42, i32 11, i32 43, i32 12, i32 44, i32 13, i32 45, i32 14, i32 46, i32 15, i32 47, i32 24, i32 56, i32 25, i32 57, i32 26, i32 58, i32 27, i32 59, i32 28, i32 60, i32 29, i32 61, i32 30, i32 62, i32 31, i32 63>
- %2 = load <32 x i8>, <32 x i8> *%a2, align 32
- %3 = shufflevector <32 x i8> %1, <32 x i8> %2, <32 x i32> <i32 8, i32 40, i32 9, i32 41, i32 10, i32 42, i32 11, i32 43, i32 12, i32 44, i32 13, i32 45, i32 14, i32 46, i32 15, i32 47, i32 24, i32 56, i32 25, i32 57, i32 26, i32 58, i32 27, i32 59, i32 28, i32 60, i32 29, i32 61, i32 30, i32 62, i32 31, i32 63>
- ret <32 x i8> %3
-}
-
-define <8 x i32> @test_punpckhdq(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
-; GENERIC-LABEL: test_punpckhdq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] sched: [1:1.00]
-; GENERIC-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7] sched: [8:1.00]
-; GENERIC-NEXT: vpcmpeqd %ymm1, %ymm1, %ymm1 # sched: [1:0.50]
-; GENERIC-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_punpckhdq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] sched: [1:1.00]
-; HASWELL-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7] sched: [8:1.00]
-; HASWELL-NEXT: vpcmpeqd %ymm1, %ymm1, %ymm1 # sched: [1:0.50]
-; HASWELL-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_punpckhdq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] sched: [1:1.00]
-; BROADWELL-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7] sched: [7:1.00]
-; BROADWELL-NEXT: vpcmpeqd %ymm1, %ymm1, %ymm1 # sched: [1:0.50]
-; BROADWELL-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_punpckhdq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] sched: [1:1.00]
-; SKYLAKE-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7] sched: [8:1.00]
-; SKYLAKE-NEXT: vpcmpeqd %ymm1, %ymm1, %ymm1 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_punpckhdq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] sched: [1:1.00]
-; SKX-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7] sched: [8:1.00]
-; SKX-NEXT: vpcmpeqd %ymm1, %ymm1, %ymm1 # sched: [1:0.50]
-; SKX-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_punpckhdq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7] sched: [1:0.25]
-; ZNVER1-NEXT: vpunpckhdq {{.*#+}} ymm0 = ymm0[2],mem[2],ymm0[3],mem[3],ymm0[6],mem[6],ymm0[7],mem[7] sched: [8:0.50]
-; ZNVER1-NEXT: vpcmpeqd %ymm1, %ymm1, %ymm1 # sched: [1:0.25]
-; ZNVER1-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <8 x i32> %a0, <8 x i32> %a1, <8 x i32> <i32 2, i32 10, i32 3, i32 11, i32 6, i32 14, i32 7, i32 15>
- %2 = load <8 x i32>, <8 x i32> *%a2, align 32
- %3 = shufflevector <8 x i32> %1, <8 x i32> %2, <8 x i32> <i32 2, i32 10, i32 3, i32 11, i32 6, i32 14, i32 7, i32 15>
- %4 = add <8 x i32> %3, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
- ret <8 x i32> %4
-}
-
-define <4 x i64> @test_punpckhqdq(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) {
-; GENERIC-LABEL: test_punpckhqdq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpunpckhqdq {{.*#+}} ymm1 = ymm0[1],ymm1[1],ymm0[3],ymm1[3] sched: [1:1.00]
-; GENERIC-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],mem[1],ymm0[3],mem[3] sched: [8:1.00]
-; GENERIC-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_punpckhqdq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpunpckhqdq {{.*#+}} ymm1 = ymm0[1],ymm1[1],ymm0[3],ymm1[3] sched: [1:1.00]
-; HASWELL-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],mem[1],ymm0[3],mem[3] sched: [8:1.00]
-; HASWELL-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_punpckhqdq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpunpckhqdq {{.*#+}} ymm1 = ymm0[1],ymm1[1],ymm0[3],ymm1[3] sched: [1:1.00]
-; BROADWELL-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],mem[1],ymm0[3],mem[3] sched: [7:1.00]
-; BROADWELL-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_punpckhqdq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpunpckhqdq {{.*#+}} ymm1 = ymm0[1],ymm1[1],ymm0[3],ymm1[3] sched: [1:1.00]
-; SKYLAKE-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],mem[1],ymm0[3],mem[3] sched: [8:1.00]
-; SKYLAKE-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_punpckhqdq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpunpckhqdq {{.*#+}} ymm1 = ymm0[1],ymm1[1],ymm0[3],ymm1[3] sched: [1:1.00]
-; SKX-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],mem[1],ymm0[3],mem[3] sched: [8:1.00]
-; SKX-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_punpckhqdq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpunpckhqdq {{.*#+}} ymm1 = ymm0[1],ymm1[1],ymm0[3],ymm1[3] sched: [1:0.25]
-; ZNVER1-NEXT: vpunpckhqdq {{.*#+}} ymm0 = ymm0[1],mem[1],ymm0[3],mem[3] sched: [8:0.50]
-; ZNVER1-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <4 x i64> %a0, <4 x i64> %a1, <4 x i32> <i32 1, i32 5, i32 3, i32 7>
- %2 = load <4 x i64>, <4 x i64> *%a2, align 32
- %3 = shufflevector <4 x i64> %a0, <4 x i64> %2, <4 x i32> <i32 1, i32 5, i32 3, i32 7>
- %4 = add <4 x i64> %1, %3
- ret <4 x i64> %4
-}
-
-define <16 x i16> @test_punpckhwd(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_punpckhwd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15] sched: [1:1.00]
-; GENERIC-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],mem[4],ymm0[5],mem[5],ymm0[6],mem[6],ymm0[7],mem[7],ymm0[12],mem[12],ymm0[13],mem[13],ymm0[14],mem[14],ymm0[15],mem[15] sched: [8:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_punpckhwd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15] sched: [1:1.00]
-; HASWELL-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],mem[4],ymm0[5],mem[5],ymm0[6],mem[6],ymm0[7],mem[7],ymm0[12],mem[12],ymm0[13],mem[13],ymm0[14],mem[14],ymm0[15],mem[15] sched: [8:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_punpckhwd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15] sched: [1:1.00]
-; BROADWELL-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],mem[4],ymm0[5],mem[5],ymm0[6],mem[6],ymm0[7],mem[7],ymm0[12],mem[12],ymm0[13],mem[13],ymm0[14],mem[14],ymm0[15],mem[15] sched: [7:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_punpckhwd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15] sched: [1:1.00]
-; SKYLAKE-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],mem[4],ymm0[5],mem[5],ymm0[6],mem[6],ymm0[7],mem[7],ymm0[12],mem[12],ymm0[13],mem[13],ymm0[14],mem[14],ymm0[15],mem[15] sched: [8:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_punpckhwd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15] sched: [1:1.00]
-; SKX-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],mem[4],ymm0[5],mem[5],ymm0[6],mem[6],ymm0[7],mem[7],ymm0[12],mem[12],ymm0[13],mem[13],ymm0[14],mem[14],ymm0[15],mem[15] sched: [8:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_punpckhwd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15] sched: [1:0.25]
-; ZNVER1-NEXT: vpunpckhwd {{.*#+}} ymm0 = ymm0[4],mem[4],ymm0[5],mem[5],ymm0[6],mem[6],ymm0[7],mem[7],ymm0[12],mem[12],ymm0[13],mem[13],ymm0[14],mem[14],ymm0[15],mem[15] sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <16 x i16> %a0, <16 x i16> %a1, <16 x i32> <i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
- %2 = load <16 x i16>, <16 x i16> *%a2, align 32
- %3 = shufflevector <16 x i16> %1, <16 x i16> %2, <16 x i32> <i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
- ret <16 x i16> %3
-}
-
-define <32 x i8> @test_punpcklbw(<32 x i8> %a0, <32 x i8> %a1, <32 x i8> *%a2) {
-; GENERIC-LABEL: test_punpcklbw:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[16],ymm1[16],ymm0[17],ymm1[17],ymm0[18],ymm1[18],ymm0[19],ymm1[19],ymm0[20],ymm1[20],ymm0[21],ymm1[21],ymm0[22],ymm1[22],ymm0[23],ymm1[23] sched: [1:1.00]
-; GENERIC-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[4],mem[4],ymm0[5],mem[5],ymm0[6],mem[6],ymm0[7],mem[7],ymm0[16],mem[16],ymm0[17],mem[17],ymm0[18],mem[18],ymm0[19],mem[19],ymm0[20],mem[20],ymm0[21],mem[21],ymm0[22],mem[22],ymm0[23],mem[23] sched: [8:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_punpcklbw:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[16],ymm1[16],ymm0[17],ymm1[17],ymm0[18],ymm1[18],ymm0[19],ymm1[19],ymm0[20],ymm1[20],ymm0[21],ymm1[21],ymm0[22],ymm1[22],ymm0[23],ymm1[23] sched: [1:1.00]
-; HASWELL-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[4],mem[4],ymm0[5],mem[5],ymm0[6],mem[6],ymm0[7],mem[7],ymm0[16],mem[16],ymm0[17],mem[17],ymm0[18],mem[18],ymm0[19],mem[19],ymm0[20],mem[20],ymm0[21],mem[21],ymm0[22],mem[22],ymm0[23],mem[23] sched: [8:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_punpcklbw:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[16],ymm1[16],ymm0[17],ymm1[17],ymm0[18],ymm1[18],ymm0[19],ymm1[19],ymm0[20],ymm1[20],ymm0[21],ymm1[21],ymm0[22],ymm1[22],ymm0[23],ymm1[23] sched: [1:1.00]
-; BROADWELL-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[4],mem[4],ymm0[5],mem[5],ymm0[6],mem[6],ymm0[7],mem[7],ymm0[16],mem[16],ymm0[17],mem[17],ymm0[18],mem[18],ymm0[19],mem[19],ymm0[20],mem[20],ymm0[21],mem[21],ymm0[22],mem[22],ymm0[23],mem[23] sched: [7:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_punpcklbw:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[16],ymm1[16],ymm0[17],ymm1[17],ymm0[18],ymm1[18],ymm0[19],ymm1[19],ymm0[20],ymm1[20],ymm0[21],ymm1[21],ymm0[22],ymm1[22],ymm0[23],ymm1[23] sched: [1:1.00]
-; SKYLAKE-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[4],mem[4],ymm0[5],mem[5],ymm0[6],mem[6],ymm0[7],mem[7],ymm0[16],mem[16],ymm0[17],mem[17],ymm0[18],mem[18],ymm0[19],mem[19],ymm0[20],mem[20],ymm0[21],mem[21],ymm0[22],mem[22],ymm0[23],mem[23] sched: [8:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_punpcklbw:
-; SKX: # %bb.0:
-; SKX-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[16],ymm1[16],ymm0[17],ymm1[17],ymm0[18],ymm1[18],ymm0[19],ymm1[19],ymm0[20],ymm1[20],ymm0[21],ymm1[21],ymm0[22],ymm1[22],ymm0[23],ymm1[23] sched: [1:1.00]
-; SKX-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[4],mem[4],ymm0[5],mem[5],ymm0[6],mem[6],ymm0[7],mem[7],ymm0[16],mem[16],ymm0[17],mem[17],ymm0[18],mem[18],ymm0[19],mem[19],ymm0[20],mem[20],ymm0[21],mem[21],ymm0[22],mem[22],ymm0[23],mem[23] sched: [8:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_punpcklbw:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[16],ymm1[16],ymm0[17],ymm1[17],ymm0[18],ymm1[18],ymm0[19],ymm1[19],ymm0[20],ymm1[20],ymm0[21],ymm1[21],ymm0[22],ymm1[22],ymm0[23],ymm1[23] sched: [1:0.25]
-; ZNVER1-NEXT: vpunpcklbw {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[4],mem[4],ymm0[5],mem[5],ymm0[6],mem[6],ymm0[7],mem[7],ymm0[16],mem[16],ymm0[17],mem[17],ymm0[18],mem[18],ymm0[19],mem[19],ymm0[20],mem[20],ymm0[21],mem[21],ymm0[22],mem[22],ymm0[23],mem[23] sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <32 x i8> %a0, <32 x i8> %a1, <32 x i32> <i32 0, i32 32, i32 1, i32 33, i32 2, i32 34, i32 3, i32 35, i32 4, i32 36, i32 5, i32 37, i32 6, i32 38, i32 7, i32 39, i32 16, i32 48, i32 17, i32 49, i32 18, i32 50, i32 19, i32 51, i32 20, i32 52, i32 21, i32 53, i32 22, i32 54, i32 23, i32 55>
- %2 = load <32 x i8>, <32 x i8> *%a2, align 32
- %3 = shufflevector <32 x i8> %1, <32 x i8> %2, <32 x i32> <i32 0, i32 32, i32 1, i32 33, i32 2, i32 34, i32 3, i32 35, i32 4, i32 36, i32 5, i32 37, i32 6, i32 38, i32 7, i32 39, i32 16, i32 48, i32 17, i32 49, i32 18, i32 50, i32 19, i32 51, i32 20, i32 52, i32 21, i32 53, i32 22, i32 54, i32 23, i32 55>
- ret <32 x i8> %3
-}
-
-define <8 x i32> @test_punpckldq(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> *%a2) {
-; GENERIC-LABEL: test_punpckldq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] sched: [1:1.00]
-; GENERIC-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [8:1.00]
-; GENERIC-NEXT: vpcmpeqd %ymm1, %ymm1, %ymm1 # sched: [1:0.50]
-; GENERIC-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_punpckldq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] sched: [1:1.00]
-; HASWELL-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [8:1.00]
-; HASWELL-NEXT: vpcmpeqd %ymm1, %ymm1, %ymm1 # sched: [1:0.50]
-; HASWELL-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_punpckldq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] sched: [1:1.00]
-; BROADWELL-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [7:1.00]
-; BROADWELL-NEXT: vpcmpeqd %ymm1, %ymm1, %ymm1 # sched: [1:0.50]
-; BROADWELL-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_punpckldq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] sched: [1:1.00]
-; SKYLAKE-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [8:1.00]
-; SKYLAKE-NEXT: vpcmpeqd %ymm1, %ymm1, %ymm1 # sched: [1:0.50]
-; SKYLAKE-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_punpckldq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] sched: [1:1.00]
-; SKX-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [8:1.00]
-; SKX-NEXT: vpcmpeqd %ymm1, %ymm1, %ymm1 # sched: [1:0.50]
-; SKX-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_punpckldq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5] sched: [1:0.25]
-; ZNVER1-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] sched: [8:0.50]
-; ZNVER1-NEXT: vpcmpeqd %ymm1, %ymm1, %ymm1 # sched: [1:0.25]
-; ZNVER1-NEXT: vpsubd %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <8 x i32> %a0, <8 x i32> %a1, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 4, i32 12, i32 5, i32 13>
- %2 = load <8 x i32>, <8 x i32> *%a2, align 32
- %3 = shufflevector <8 x i32> %1, <8 x i32> %2, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 4, i32 12, i32 5, i32 13>
- %4 = add <8 x i32> %3, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
- ret <8 x i32> %4
-}
-
-define <4 x i64> @test_punpcklqdq(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) {
-; GENERIC-LABEL: test_punpcklqdq:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm0[0],ymm1[0],ymm0[2],ymm1[2] sched: [1:1.00]
-; GENERIC-NEXT: vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[2],mem[2] sched: [8:1.00]
-; GENERIC-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_punpcklqdq:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm0[0],ymm1[0],ymm0[2],ymm1[2] sched: [1:1.00]
-; HASWELL-NEXT: vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[2],mem[2] sched: [8:1.00]
-; HASWELL-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_punpcklqdq:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm0[0],ymm1[0],ymm0[2],ymm1[2] sched: [1:1.00]
-; BROADWELL-NEXT: vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[2],mem[2] sched: [7:1.00]
-; BROADWELL-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_punpcklqdq:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm0[0],ymm1[0],ymm0[2],ymm1[2] sched: [1:1.00]
-; SKYLAKE-NEXT: vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[2],mem[2] sched: [8:1.00]
-; SKYLAKE-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_punpcklqdq:
-; SKX: # %bb.0:
-; SKX-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm0[0],ymm1[0],ymm0[2],ymm1[2] sched: [1:1.00]
-; SKX-NEXT: vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[2],mem[2] sched: [8:1.00]
-; SKX-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_punpcklqdq:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm0[0],ymm1[0],ymm0[2],ymm1[2] sched: [1:0.25]
-; ZNVER1-NEXT: vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[2],mem[2] sched: [8:0.50]
-; ZNVER1-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <4 x i64> %a0, <4 x i64> %a1, <4 x i32> <i32 0, i32 4, i32 2, i32 6>
- %2 = load <4 x i64>, <4 x i64> *%a2, align 32
- %3 = shufflevector <4 x i64> %a0, <4 x i64> %2, <4 x i32> <i32 0, i32 4, i32 2, i32 6>
- %4 = add <4 x i64> %1, %3
- ret <4 x i64> %4
-}
-
-define <16 x i16> @test_punpcklwd(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> *%a2) {
-; GENERIC-LABEL: test_punpcklwd:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11] sched: [1:1.00]
-; GENERIC-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11] sched: [8:1.00]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_punpcklwd:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11] sched: [1:1.00]
-; HASWELL-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11] sched: [8:1.00]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_punpcklwd:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11] sched: [1:1.00]
-; BROADWELL-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11] sched: [7:1.00]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_punpcklwd:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11] sched: [1:1.00]
-; SKYLAKE-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11] sched: [8:1.00]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_punpcklwd:
-; SKX: # %bb.0:
-; SKX-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11] sched: [1:1.00]
-; SKX-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11] sched: [8:1.00]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_punpcklwd:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11] sched: [1:0.25]
-; ZNVER1-NEXT: vpunpcklwd {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[2],mem[2],ymm0[3],mem[3],ymm0[8],mem[8],ymm0[9],mem[9],ymm0[10],mem[10],ymm0[11],mem[11] sched: [8:0.50]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = shufflevector <16 x i16> %a0, <16 x i16> %a1, <16 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27>
- %2 = load <16 x i16>, <16 x i16> *%a2, align 32
- %3 = shufflevector <16 x i16> %1, <16 x i16> %2, <16 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27>
- ret <16 x i16> %3
-}
-
-define <4 x i64> @test_pxor(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> *%a2) {
-; GENERIC-LABEL: test_pxor:
-; GENERIC: # %bb.0:
-; GENERIC-NEXT: vpxor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; GENERIC-NEXT: vpxor (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; GENERIC-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; GENERIC-NEXT: retq # sched: [1:1.00]
-;
-; HASWELL-LABEL: test_pxor:
-; HASWELL: # %bb.0:
-; HASWELL-NEXT: vpxor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; HASWELL-NEXT: vpxor (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; HASWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; HASWELL-NEXT: retq # sched: [7:1.00]
-;
-; BROADWELL-LABEL: test_pxor:
-; BROADWELL: # %bb.0:
-; BROADWELL-NEXT: vpxor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; BROADWELL-NEXT: vpxor (%rdi), %ymm0, %ymm0 # sched: [7:0.50]
-; BROADWELL-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.50]
-; BROADWELL-NEXT: retq # sched: [7:1.00]
-;
-; SKYLAKE-LABEL: test_pxor:
-; SKYLAKE: # %bb.0:
-; SKYLAKE-NEXT: vpxor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: vpxor (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKYLAKE-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKYLAKE-NEXT: retq # sched: [7:1.00]
-;
-; SKX-LABEL: test_pxor:
-; SKX: # %bb.0:
-; SKX-NEXT: vpxor %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: vpxor (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; SKX-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.33]
-; SKX-NEXT: retq # sched: [7:1.00]
-;
-; ZNVER1-LABEL: test_pxor:
-; ZNVER1: # %bb.0:
-; ZNVER1-NEXT: vpxor %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: vpxor (%rdi), %ymm0, %ymm0 # sched: [8:0.50]
-; ZNVER1-NEXT: vpaddq %ymm1, %ymm0, %ymm0 # sched: [1:0.25]
-; ZNVER1-NEXT: retq # sched: [1:0.50]
- %1 = xor <4 x i64> %a0, %a1
- %2 = load <4 x i64>, <4 x i64> *%a2, align 32
- %3 = xor <4 x i64> %1, %2
- %4 = add <4 x i64> %3, %a1
- ret <4 x i64> %4
-}
-
-!0 = !{i32 1}
More information about the llvm-commits
mailing list