[PowerPC] implement target hook getTgtMemIntrinsic

[shchenz]

Summary

Implement TargetTransformInfoImplBase::getTgtMemIntrinsic on PowerPC target, so that getTgtMemIntrinsic user can recognize PowerPC specific memory intrinsics.

For example, LSR uses getTgtMemIntrinsic to check if an intrinsic is a memory-related, so that it can mark operand as Address kind

[shchenz]

code format fix

[steven.zhang]

LGTM.

llvm/lib/Target/PowerPC/PPCTargetTransformInfo.h

public:
   TTI::PopcntSupportKind getPopcntSupport(unsigned TyWidth);
   bool isHardwareLoopProfitable(Loop *L, ScalarEvolution &SE,
                                 AssumptionCache &AC,
                                 TargetLibraryInfo *LibInfo,
                                 HardwareLoopInfo &HWLoopInfo);
   bool canSaveCmp(Loop *L, BranchInst **BI, ScalarEvolution *SE, LoopInfo *LI,
                   DominatorTree *DT, AssumptionCache *AC,
                   TargetLibraryInfo *LibInfo);
+  bool getTgtMemIntrinsic(IntrinsicInst *Inst, MemIntrinsicInfo &Info);
   void getUnrollingPreferences(Loop *L, ScalarEvolution &SE,
                                TTI::UnrollingPreferences &UP);
   void getPeelingPreferences(Loop *L, ScalarEvolution &SE,
                              TTI::PeelingPreferences &PP);
   bool isLSRCostLess(TargetTransformInfo::LSRCost &C1,
                      TargetTransformInfo::LSRCost &C2);
 
   /// @}

llvm/lib/Target/PowerPC/PPCTargetTransformInfo.cpp

bool PPCTTIImpl::isLSRCostLess(TargetTransformInfo::LSRCost &C1,
   if (!LsrNoInsnsCost)
     return std::tie(C1.Insns, C1.NumRegs, C1.AddRecCost, C1.NumIVMuls,
                     C1.NumBaseAdds, C1.ScaleCost, C1.ImmCost, C1.SetupCost) <
            std::tie(C2.Insns, C2.NumRegs, C2.AddRecCost, C2.NumIVMuls,
                     C2.NumBaseAdds, C2.ScaleCost, C2.ImmCost, C2.SetupCost);
   else
     return TargetTransformInfoImplBase::isLSRCostLess(C1, C2);
 }
+
+bool PPCTTIImpl::getTgtMemIntrinsic(IntrinsicInst *Inst,
+                                    MemIntrinsicInfo &Info) {
+  switch (Inst->getIntrinsicID()) {
+  case Intrinsic::ppc_altivec_lvx:
+  case Intrinsic::ppc_altivec_lvxl:
+  case Intrinsic::ppc_altivec_lvebx:
+  case Intrinsic::ppc_altivec_lvehx:
+  case Intrinsic::ppc_altivec_lvewx:
+  case Intrinsic::ppc_vsx_lxvd2x:
+  case Intrinsic::ppc_vsx_lxvw4x:
+  case Intrinsic::ppc_vsx_lxvd2x_be:
+  case Intrinsic::ppc_vsx_lxvw4x_be:
+  case Intrinsic::ppc_vsx_lxvl:
+  case Intrinsic::ppc_vsx_lxvll: {
+    Info.PtrVal = Inst->getArgOperand(0);
+    Info.ReadMem = true;
+    Info.WriteMem = false;
+    return true;
+  }
+  case Intrinsic::ppc_altivec_stvx:
+  case Intrinsic::ppc_altivec_stvxl:
+  case Intrinsic::ppc_altivec_stvebx:
+  case Intrinsic::ppc_altivec_stvehx:
+  case Intrinsic::ppc_altivec_stvewx:
+  case Intrinsic::ppc_vsx_stxvd2x:
+  case Intrinsic::ppc_vsx_stxvw4x:
+  case Intrinsic::ppc_vsx_stxvd2x_be:
+  case Intrinsic::ppc_vsx_stxvw4x_be:
+  case Intrinsic::ppc_vsx_stxvl:
+  case Intrinsic::ppc_vsx_stxvll: {
+    Info.PtrVal = Inst->getArgOperand(1);
+    Info.ReadMem = false;
+    Info.WriteMem = true;
+    return true;
+  }
+  default:
+    break;
+  }
+
+  return false;
+}
+

llvm/test/Transforms/LoopStrengthReduce/Power/memory-intrinsic.ll

+; REQUIRES: asserts
+; RUN: opt -loop-reduce -debug-only=loop-reduce -S  < %s 2>&1 | FileCheck %s
+
+target datalayout = "e-m:e-i64:64-n32:64"
+target triple = "powerpc64le-unknown-linux-gnu"
+
+; CHECK: LSR Use: Kind=Address
+; CHECK: LSR Use: Kind=Address
+; CHECK-NOT: LSR Use: Kind=Basic
+; CHECK-NOT: LSR Use: Kind=Basic
+
+declare <4 x i32> @llvm.ppc.altivec.lvx(i8*)
+declare void @llvm.ppc.altivec.stvx(<4 x i32>, i8*)
+
+; Function Attrs: nofree norecurse nounwind
+define void @foo(<4 x i32>* %0, <4 x i32>* %1, i32 signext %2) {
+  %4 = icmp sgt i32 %2, 0
+  br i1 %4, label %5, label %7
+
+5:                                                ; preds = %3
+  %6 = zext i32 %2 to i64
+  br label %8
+
+7:                                                ; preds = %8, %3
+  ret void
+
+8:                                                ; preds = %5, %8
+  %9 = phi i64 [ 0, %5 ], [ %15, %8 ]
+  %10 = getelementptr inbounds <4 x i32>, <4 x i32>* %1, i64 %9
+  %11 = bitcast <4 x i32>* %10 to i8*
+  %12 = call <4 x i32> @llvm.ppc.altivec.lvx(i8* %11)
+  %13 = getelementptr inbounds <4 x i32>, <4 x i32>* %0, i64 %9
+  %14 = bitcast <4 x i32>* %13 to i8*
+  call void @llvm.ppc.altivec.stvx(<4 x i32> %12, i8* %14)
+  %15 = add nuw nsw i64 %9, 10
+  %16 = icmp ult i64 %15, %6
+  br i1 %16, label %8, label %7
+}
+