[AArch64] Lower READCYCLECOUNTER using MRS PMCCTNR_EL0.

[ab]

Summary

This matches the ARM behavior. However, here the lowering isn't guarded
by a "PerfMon" feature: while it is indeed optional on AArch64 as well,
all of the CPUs we support have it, so why pretend.

[jmolloy]

Hi Ahmed,

I'd really prefer this to be predicated on a new feature. I'm happy for the feature to be enabled by default though.

Cheers,

James

lib/Target/AArch64/AArch64ISelLowering.cpp

405 ↗

(On Diff #33391)

I'd really prefer that this is predicated on a feature.

[ab]

Fair enough; here's an updated patch.

[jmolloy]

Thanks for making that change Ahmed, LGTM.

James

[salvatoredipietro]

Hi @ab,
I was checking this code and I believe that it would be better for the readcyclecounter function to use the CNTVCT_EL0 register (Counter-timer Virtual Count register) instead of the PMCCNTR_EL0 (Performance Monitors Cycle Count Register) because the current one is a PMU register which, depending on PMU configuration, it might always return zeroes and it doesn't guaranteed to always be increased.

llvm/trunk/lib/Target/AArch64/AArch64.td

def FeatureNEON : SubtargetFeature<"neon", "HasNEON", "true",
   "Enable Advanced SIMD instructions", [FeatureFPARMv8]>;
 
 def FeatureCrypto : SubtargetFeature<"crypto", "HasCrypto", "true",
   "Enable cryptographic instructions">;
 
 def FeatureCRC : SubtargetFeature<"crc", "HasCRC", "true",
   "Enable ARMv8 CRC-32 checksum instructions">;
 
+def FeaturePerfMon : SubtargetFeature<"perfmon", "HasPerfMon", "true",
+  "Enable ARMv8 PMUv3 Performance Monitors extension">;
+
 /// Cyclone has register move instructions which are "free".
 def FeatureZCRegMove : SubtargetFeature<"zcm", "HasZeroCycleRegMove", "true",
                                         "Has zero-cycle register moves">;
 
 /// Cyclone has instructions which zero registers for "free".
 def FeatureZCZeroing : SubtargetFeature<"zcz", "HasZeroCycleZeroing", "true",
                                         "Has zero-cycle zeroing instructions">;
 

 include "AArch64SchedA57.td"
 include "AArch64SchedCyclone.td"
 
 def ProcA53     : SubtargetFeature<"a53", "ARMProcFamily", "CortexA53",
                                    "Cortex-A53 ARM processors",
                                    [FeatureFPARMv8,
                                    FeatureNEON,
                                    FeatureCrypto,
-                                   FeatureCRC]>;
+                                   FeatureCRC,
+                                   FeaturePerfMon]>;
 
 def ProcA57     : SubtargetFeature<"a57", "ARMProcFamily", "CortexA57",
                                    "Cortex-A57 ARM processors",
                                    [FeatureFPARMv8,
                                    FeatureNEON,
                                    FeatureCrypto,
-                                   FeatureCRC]>;
+                                   FeatureCRC,
+                                   FeaturePerfMon]>;
 
 def ProcCyclone : SubtargetFeature<"cyclone", "ARMProcFamily", "Cyclone",
                                    "Cyclone",
                                    [FeatureFPARMv8,
                                    FeatureNEON,
                                    FeatureCrypto,
                                    FeatureCRC,
+                                   FeaturePerfMon,
                                    FeatureZCRegMove, FeatureZCZeroing]>;
 
 def : ProcessorModel<"generic", NoSchedModel, [FeatureFPARMv8,
                                               FeatureNEON,
-                                              FeatureCRC]>;
+                                              FeatureCRC,
+                                              FeaturePerfMon]>;
 
 def : ProcessorModel<"cortex-a53", CortexA53Model, [ProcA53]>;
 def : ProcessorModel<"cortex-a57", CortexA57Model, [ProcA57]>;
 // FIXME: Cortex-A72 is currently modelled as an Cortex-A57.
 def : ProcessorModel<"cortex-a72", CortexA57Model, [ProcA57]>;
 def : ProcessorModel<"cyclone", CycloneModel, [ProcCyclone]>;
 
 //===----------------------------------------------------------------------===//

llvm/trunk/lib/Target/AArch64/AArch64ISelLowering.cpp

for (MVT Ty : {MVT::f32, MVT::f64}) {
     setOperationAction(ISD::FMINNUM, Ty, Legal);
     setOperationAction(ISD::FMAXNUM, Ty, Legal);
     setOperationAction(ISD::FMINNAN, Ty, Legal);
     setOperationAction(ISD::FMAXNAN, Ty, Legal);
   }
 
   setOperationAction(ISD::PREFETCH, MVT::Other, Custom);
 
+  // Lower READCYCLECOUNTER using an mrs from PMCCNTR_EL0.
+  // This requires the Performance Monitors extension.
+  if (Subtarget->hasPerfMon())
+    setOperationAction(ISD::READCYCLECOUNTER, MVT::i64, Legal);
+
   if (Subtarget->isTargetMachO()) {
     // For iOS, we don't want to the normal expansion of a libcall to
     // sincos. We want to issue a libcall to __sincos_stret to avoid memory
     // traffic.
     setOperationAction(ISD::FSINCOS, MVT::f64, Custom);
     setOperationAction(ISD::FSINCOS, MVT::f32, Custom);
   } else {
     setOperationAction(ISD::FSINCOS, MVT::f64, Expand);

llvm/trunk/lib/Target/AArch64/AArch64InstrInfo.td

 def HasFPARMv8       : Predicate<"Subtarget->hasFPARMv8()">,
                                AssemblerPredicate<"FeatureFPARMv8", "fp-armv8">;
 def HasNEON          : Predicate<"Subtarget->hasNEON()">,
                                  AssemblerPredicate<"FeatureNEON", "neon">;
 def HasCrypto        : Predicate<"Subtarget->hasCrypto()">,
                                  AssemblerPredicate<"FeatureCrypto", "crypto">;
 def HasCRC           : Predicate<"Subtarget->hasCRC()">,
                                  AssemblerPredicate<"FeatureCRC", "crc">;
+def HasPerfMon       : Predicate<"Subtarget->hasPerfMon()">;
 def IsLE             : Predicate<"Subtarget->isLittleEndian()">;
 def IsBE             : Predicate<"!Subtarget->isLittleEndian()">;
 def IsCyclone        : Predicate<"Subtarget->isCyclone()">;
 
 //===----------------------------------------------------------------------===//
 // AArch64-specific DAG Nodes.
 //
 

 def MRS    : MRSI;
 def MSR    : MSRI;
 def MSRpstate: MSRpstateI;
 
 // The thread pointer (on Linux, at least, where this has been implemented) is
 // TPIDR_EL0.
 def : Pat<(AArch64threadpointer), (MRS 0xde82)>;
 
+// The cycle counter PMC register is PMCCNTR_EL0.
+let Predicates = [HasPerfMon] in
+def : Pat<(readcyclecounter), (MRS 0xdce8)>;
+
 // Generic system instructions
 def SYSxt  : SystemXtI<0, "sys">;
 def SYSLxt : SystemLXtI<1, "sysl">;
 
 def : InstAlias<"sys $op1, $Cn, $Cm, $op2",
                 (SYSxt imm0_7:$op1, sys_cr_op:$Cn,
                  sys_cr_op:$Cm, imm0_7:$op2, XZR)>;
 

llvm/trunk/lib/Target/AArch64/AArch64Subtarget.h

protected:
   ARMProcFamilyEnum ARMProcFamily;
 
   bool HasV8_1aOps;
 
   bool HasFPARMv8;
   bool HasNEON;
   bool HasCrypto;
   bool HasCRC;
+  bool HasPerfMon;
 
   // HasZeroCycleRegMove - Has zero-cycle register mov instructions.
   bool HasZeroCycleRegMove;
 
   // HasZeroCycleZeroing - Has zero-cycle zeroing instructions.
   bool HasZeroCycleZeroing;
 
   // StrictAlign - Disallow unaligned memory accesses.
public:
 
   bool requiresStrictAlign() const { return StrictAlign; }
 
   bool isX18Reserved() const { return ReserveX18; }
   bool hasFPARMv8() const { return HasFPARMv8; }
   bool hasNEON() const { return HasNEON; }
   bool hasCrypto() const { return HasCrypto; }
   bool hasCRC() const { return HasCRC; }
+  bool hasPerfMon() const { return HasPerfMon; }
 
   bool isLittleEndian() const { return IsLittle; }
 
   bool isTargetDarwin() const { return TargetTriple.isOSDarwin(); }
   bool isTargetIOS() const { return TargetTriple.isiOS(); }
   bool isTargetLinux() const { return TargetTriple.isOSLinux(); }
   bool isTargetWindows() const { return TargetTriple.isOSWindows(); }
 

llvm/trunk/lib/Target/AArch64/AArch64Subtarget.cpp

AArch64Subtarget::initializeSubtargetDependencies(StringRef FS) {
   return *this;
 }
 
 AArch64Subtarget::AArch64Subtarget(const Triple &TT, const std::string &CPU,
                                    const std::string &FS,
                                    const TargetMachine &TM, bool LittleEndian)
     : AArch64GenSubtargetInfo(TT, CPU, FS), ARMProcFamily(Others),
       HasV8_1aOps(false), HasFPARMv8(false), HasNEON(false), HasCrypto(false),
-      HasCRC(false), HasZeroCycleRegMove(false), HasZeroCycleZeroing(false),
-      StrictAlign(false), ReserveX18(false), IsLittle(LittleEndian),
-      CPUString(CPU), TargetTriple(TT), FrameLowering(),
+      HasCRC(false), HasPerfMon(false), HasZeroCycleRegMove(false),
+      HasZeroCycleZeroing(false), StrictAlign(false), ReserveX18(false),
+      IsLittle(LittleEndian), CPUString(CPU), TargetTriple(TT), FrameLowering(),
       InstrInfo(initializeSubtargetDependencies(FS)), TSInfo(),
       TLInfo(TM, *this) {}
 
 /// ClassifyGlobalReference - Find the target operand flags that describe
 /// how a global value should be referenced for the current subtarget.
 unsigned char
 AArch64Subtarget::ClassifyGlobalReference(const GlobalValue *GV,
                                         const TargetMachine &TM) const {

llvm/trunk/test/CodeGen/AArch64/readcyclecounter.ll

+; RUN: llc -mtriple=aarch64-unknown-unknown -asm-verbose=false < %s |\
+; RUN:   FileCheck %s --check-prefix=CHECK --check-prefix=PERFMON
+; RUN: llc -mtriple=aarch64-unknown-unknown -mattr=-perfmon -asm-verbose=false < %s |\
+; RUN:   FileCheck %s --check-prefix=CHECK --check-prefix=NOPERFMON
+
+define i64 @test_readcyclecounter() nounwind {
+  ; CHECK-LABEL:   test_readcyclecounter:
+  ; PERFMON-NEXT:   mrs x0, PMCCNTR_EL0
+  ; NOPERFMON-NEXT: mov x0, xzr
+  ; CHECK-NEXT:     ret
+  %tmp0 = call i64 @llvm.readcyclecounter()
+  ret i64 %tmp0
+}
+
+declare i64 @llvm.readcyclecounter()