[PowerPC] using milicode call for strlen instead of lib call (#153600)

AIX has "millicode" routines, which are functions loaded at boot time
into fixed addresses in kernel memory. This allows them to be customized
for the processor. The __strlen routine is a millicode implementation;
we use millicode for the strlen function instead of a library call to
improve performance.

Author: diggerlin

llvm/include/llvm/CodeGen/SelectionDAG.h

@@ -1259,6 +1259,9 @@ class SelectionDAG {
   std::pair<SDValue, SDValue> getMemcmp(SDValue Chain, const SDLoc &dl,
                                         SDValue Dst, SDValue Src, SDValue Size,
                                         const CallInst *CI);
+  LLVM_ABI std::pair<SDValue, SDValue>
+  getStrlen(SDValue Chain, const SDLoc &dl, SDValue Src, const CallInst *CI);
+
   /* \p CI if not null is the memset call being lowered.
    * \p OverrideTailCall is an optional parameter that can be used to override
    * the tail call optimization decision. */

llvm/include/llvm/CodeGen/SelectionDAGTargetInfo.h

@@ -162,7 +162,7 @@ class SelectionDAGTargetInfo {
 
   virtual std::pair<SDValue, SDValue>
   EmitTargetCodeForStrlen(SelectionDAG &DAG, const SDLoc &DL, SDValue Chain,
-                          SDValue Src, MachinePointerInfo SrcPtrInfo) const {
+                          SDValue Src, const CallInst *CI) const {
     return std::make_pair(SDValue(), SDValue());
   }
 

llvm/include/llvm/IR/RuntimeLibcalls.td

@@ -297,6 +297,7 @@ def MEMMOVE : RuntimeLibcall;
 def MEMSET : RuntimeLibcall;
 def CALLOC : RuntimeLibcall;
 def BZERO : RuntimeLibcall;
+def STRLEN : RuntimeLibcall;
 
 // Element-wise unordered-atomic memory of different sizes
 foreach MemSize = [1, 2, 4, 8, 16] in {
@@ -2272,13 +2273,15 @@ defset list<RuntimeLibcallImpl> PPC64AIXCallList = {
   def ___memmove64 : RuntimeLibcallImpl<MEMCPY>;
   def ___memset64 : RuntimeLibcallImpl<MEMSET>;
   def ___bzero64 : RuntimeLibcallImpl<BZERO>;
+  def ___strlen64 : RuntimeLibcallImpl<STRLEN>;
 }
 
 defset list<RuntimeLibcallImpl> PPC32AIXCallList = {
   def ___memcmp : RuntimeLibcallImpl<MEMCMP>;
   def ___memmove : RuntimeLibcallImpl<MEMMOVE>;
   def ___memset : RuntimeLibcallImpl<MEMSET>;
   def ___bzero : RuntimeLibcallImpl<BZERO>;
+  def ___strlen : RuntimeLibcallImpl<STRLEN>;
 }
 
 defvar PPCOverrides = !foreach(entry, PPCRuntimeLibcalls, entry.Provides);

llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp

@@ -9140,6 +9140,33 @@ SelectionDAG::getMemcmp(SDValue Chain, const SDLoc &dl, SDValue Mem0,
   return TLI->LowerCallTo(CLI);
 }
 
+std::pair<SDValue, SDValue> SelectionDAG::getStrlen(SDValue Chain,
+                                                    const SDLoc &dl,
+                                                    SDValue Src,
+                                                    const CallInst *CI) {
+  const char *LibCallName = TLI->getLibcallName(RTLIB::STRLEN);
+  if (!LibCallName)
+    return {};
+
+  // Emit a library call.
+  TargetLowering::ArgListTy Args = {
+      {Src, PointerType::getUnqual(*getContext())}};
+
+  TargetLowering::CallLoweringInfo CLI(*this);
+  bool IsTailCall =
+      isInTailCallPositionWrapper(CI, this, /*AllowReturnsFirstArg*/ true);
+
+  CLI.setDebugLoc(dl)
+      .setChain(Chain)
+      .setLibCallee(TLI->getLibcallCallingConv(RTLIB::STRLEN), CI->getType(),
+                    getExternalSymbol(
+                        LibCallName, TLI->getProgramPointerTy(getDataLayout())),
+                    std::move(Args))
+      .setTailCall(IsTailCall);
+
+  return TLI->LowerCallTo(CLI);
+}
+
 SDValue SelectionDAG::getMemcpy(
     SDValue Chain, const SDLoc &dl, SDValue Dst, SDValue Src, SDValue Size,
     Align Alignment, bool isVol, bool AlwaysInline, const CallInst *CI,

llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp

@@ -9330,9 +9330,8 @@ bool SelectionDAGBuilder::visitStrLenCall(const CallInst &I) {
   const Value *Arg0 = I.getArgOperand(0);
 
   const SelectionDAGTargetInfo &TSI = DAG.getSelectionDAGInfo();
-  std::pair<SDValue, SDValue> Res =
-    TSI.EmitTargetCodeForStrlen(DAG, getCurSDLoc(), DAG.getRoot(),
-                                getValue(Arg0), MachinePointerInfo(Arg0));
+  std::pair<SDValue, SDValue> Res = TSI.EmitTargetCodeForStrlen(
+      DAG, getCurSDLoc(), DAG.getRoot(), getValue(Arg0), &I);
   if (Res.first.getNode()) {
     processIntegerCallValue(I, Res.first, false);
     PendingLoads.push_back(Res.second);

llvm/lib/Target/PowerPC/PPCSelectionDAGInfo.cpp

@@ -28,3 +28,10 @@ std::pair<SDValue, SDValue> PPCSelectionDAGInfo::EmitTargetCodeForMemcmp(
     SDValue Op3, const CallInst *CI) const {
   return DAG.getMemcmp(Chain, dl, Op1, Op2, Op3, CI);
 }
+
+std::pair<SDValue, SDValue>
+PPCSelectionDAGInfo::EmitTargetCodeForStrlen(SelectionDAG &DAG, const SDLoc &DL,
+                                             SDValue Chain, SDValue Src,
+                                             const CallInst *CI) const {
+  return DAG.getStrlen(Chain, DL, Src, CI);
+}

llvm/lib/Target/PowerPC/PPCSelectionDAGInfo.h

@@ -25,6 +25,9 @@ class PPCSelectionDAGInfo : public SelectionDAGTargetInfo {
   EmitTargetCodeForMemcmp(SelectionDAG &DAG, const SDLoc &dl, SDValue Chain,
                           SDValue Op1, SDValue Op2, SDValue Op3,
                           const CallInst *CI) const override;
+  std::pair<SDValue, SDValue>
+  EmitTargetCodeForStrlen(SelectionDAG &DAG, const SDLoc &DL, SDValue Chain,
+                          SDValue Src, const CallInst *CI) const override;
 };
 
 } // namespace llvm

llvm/lib/Target/SystemZ/SystemZSelectionDAGInfo.cpp

@@ -263,7 +263,7 @@ static std::pair<SDValue, SDValue> getBoundedStrlen(SelectionDAG &DAG,
 
 std::pair<SDValue, SDValue> SystemZSelectionDAGInfo::EmitTargetCodeForStrlen(
     SelectionDAG &DAG, const SDLoc &DL, SDValue Chain, SDValue Src,
-    MachinePointerInfo SrcPtrInfo) const {
+    const CallInst *CI) const {
   EVT PtrVT = Src.getValueType();
   return getBoundedStrlen(DAG, DL, Chain, Src, DAG.getConstant(0, DL, PtrVT));
 }

llvm/lib/Target/SystemZ/SystemZSelectionDAGInfo.h

@@ -61,8 +61,7 @@ class SystemZSelectionDAGInfo : public SelectionDAGTargetInfo {
 
   std::pair<SDValue, SDValue>
   EmitTargetCodeForStrlen(SelectionDAG &DAG, const SDLoc &DL, SDValue Chain,
-                          SDValue Src,
-                          MachinePointerInfo SrcPtrInfo) const override;
+                          SDValue Src, const CallInst *CI) const override;
 
   std::pair<SDValue, SDValue>
   EmitTargetCodeForStrnlen(SelectionDAG &DAG, const SDLoc &DL, SDValue Chain,

llvm/test/CodeGen/PowerPC/milicode32.ll

@@ -42,7 +42,7 @@ define i32 @strlen_test(ptr noundef %str) nounwind {
 ; CHECK-AIX-32-P9-NEXT:    stwu r1, -64(r1)
 ; CHECK-AIX-32-P9-NEXT:    stw r0, 72(r1)
 ; CHECK-AIX-32-P9-NEXT:    stw r3, 60(r1)
-; CHECK-AIX-32-P9-NEXT:    bl .strlen[PR]
+; CHECK-AIX-32-P9-NEXT:    bl .___strlen[PR]
 ; CHECK-AIX-32-P9-NEXT:    nop
 ; CHECK-AIX-32-P9-NEXT:    addi r1, r1, 64
 ; CHECK-AIX-32-P9-NEXT:    lwz r0, 8(r1)