Compiler Explorer

Source code

%"class.Eigen::symbolic::SymbolExpr" = type { i8 }
%"class.Eigen::symbolic::AddExpr" = type { %"class.Eigen::symbolic::SymbolExpr", %"class.Eigen::symbolic::ValueExpr" }
%"class.Eigen::symbolic::ValueExpr" = type { i8 }
%"class.Eigen::internal::FixedInt" = type { i8 }
%"struct.Eigen::internal::all_t" = type { i8 }
%"struct.std::__u::__function::__policy" = type { ptr, ptr, i8, ptr }
%"class.absl::log_internal::NullStream" = type { i8 }
%"class.Eigen::TensorMap" = type { ptr, %"struct.Eigen::DSizes" }
%"struct.Eigen::DSizes" = type { %"class.Eigen::array" }
%"class.Eigen::array" = type { [2 x i64] }
%"class.Eigen::TensorMap.3" = type { ptr, %"struct.Eigen::DSizes" }
%"class.Eigen::TensorMap.5" = type { ptr, %"struct.Eigen::DSizes" }
%"class.Eigen::TensorChippingOp" = type <{ %"class.Eigen::TensorMap", i64, %"struct.Eigen::internal::DimensionId", [7 x i8] }>
%"struct.Eigen::internal::DimensionId" = type { i8 }
%"class.Eigen::TensorChippingOp.9" = type <{ %"class.Eigen::TensorMap.5", i64, %"struct.Eigen::internal::DimensionId", [7 x i8] }>
%"class.std::initializer_list" = type { ptr, i64 }
%"class.Eigen::TensorDevice" = type { ptr, ptr }
%"class.Eigen::TensorCwiseBinaryOp" = type { %"class.Eigen::TensorChippingOp", %"class.Eigen::TensorChippingOp.9", %"struct.Eigen::internal::scalar_sum_op", [7 x i8] }
%"struct.Eigen::internal::scalar_sum_op" = type { i8 }
%"class.Eigen::TensorAssignOp" = type { ptr, ptr }
%"struct.Eigen::TensorEvaluator" = type { %"struct.Eigen::TensorEvaluator.14", %"struct.Eigen::TensorEvaluator.19" }
%"struct.Eigen::TensorEvaluator.14" = type { %"struct.Eigen::TensorEvaluator.15" }
%"struct.Eigen::TensorEvaluator.15" = type { %"struct.Eigen::DSizes.16", i64, i64, i64, %"struct.Eigen::TensorEvaluator.18", %"struct.Eigen::internal::DimensionId", ptr }
%"struct.Eigen::DSizes.16" = type { %"class.Eigen::array.17" }
%"class.Eigen::array.17" = type { [1 x i64] }
%"struct.Eigen::TensorEvaluator.18" = type { ptr, %"struct.Eigen::DSizes", ptr }
%"struct.Eigen::TensorEvaluator.19" = type { ptr, %"struct.Eigen::internal::scalar_sum_op", %"struct.Eigen::TensorEvaluator.15", %"struct.Eigen::TensorEvaluator.20" }
%"struct.Eigen::TensorEvaluator.20" = type { %"struct.Eigen::DSizes.16", i64, i64, i64, %"struct.Eigen::TensorEvaluator.21", %"struct.Eigen::internal::DimensionId", ptr }
%"struct.Eigen::TensorEvaluator.21" = type { ptr, %"struct.Eigen::DSizes", ptr }
%"class.Eigen::TensorOpCost" = type { double, double, double }
%"class.std::__u::function" = type { %"class.std::__u::__function::__policy_func" }
%"class.std::__u::__function::__policy_func" = type { %"union.std::__u::__function::__policy_storage", %"struct.std::__u::__function::__policy_invoker", ptr }
%"union.std::__u::__function::__policy_storage" = type { ptr, [8 x i8] }
%"struct.std::__u::__function::__policy_invoker" = type { ptr }
%"class.std::__u::function.22" = type { %"class.std::__u::__function::__policy_func.25" }
%"class.std::__u::__function::__policy_func.25" = type { %"union.std::__u::__function::__policy_storage", %"struct.std::__u::__function::__policy_invoker.26", ptr }
%"struct.std::__u::__function::__policy_invoker.26" = type { ptr }
%class.anon = type { ptr }
%"struct.Eigen::ThreadPoolDevice::ParallelForBlock" = type { i64, i64 }
%"class.Eigen::Barrier" = type <{ %"class.tsl::mutex", %"class.tsl::condition_variable", %"struct.std::__u::atomic.27", i8, [3 x i8] }>
%"class.tsl::mutex" = type { %"class.absl::Mutex" }
%"class.absl::Mutex" = type { %"struct.std::__u::atomic" }
%"struct.std::__u::atomic" = type { %"struct.std::__u::__atomic_base" }
%"struct.std::__u::__atomic_base" = type { %"struct.std::__u::__atomic_base.1" }
%"struct.std::__u::__atomic_base.1" = type { %"struct.std::__u::__cxx_atomic_impl" }
%"struct.std::__u::__cxx_atomic_impl" = type { %"struct.std::__u::__cxx_atomic_base_impl" }
%"struct.std::__u::__cxx_atomic_base_impl" = type { i64 }
%"class.tsl::condition_variable" = type { %"class.absl::CondVar" }
%"class.absl::CondVar" = type { %"struct.std::__u::atomic" }
%"struct.std::__u::atomic.27" = type { %"struct.std::__u::__atomic_base.28" }
%"struct.std::__u::__atomic_base.28" = type { %"struct.std::__u::__atomic_base.29" }
%"struct.std::__u::__atomic_base.29" = type { %"struct.std::__u::__cxx_atomic_impl.30" }
%"struct.std::__u::__cxx_atomic_impl.30" = type { %"struct.std::__u::__cxx_atomic_base_impl.31" }
%"struct.std::__u::__cxx_atomic_base_impl.31" = type { i32 }
%class.anon.46 = type { ptr, ptr, ptr, %"struct.Eigen::ThreadPoolDevice::ParallelForBlock", ptr }
%"class.std::__u::function.34" = type { %"class.std::__u::__function::__policy_func.37" }
%"class.std::__u::__function::__policy_func.37" = type { %"union.std::__u::__function::__policy_storage", %"struct.std::__u::__function::__policy_invoker.38", ptr }
%"struct.std::__u::__function::__policy_invoker.38" = type { ptr }
%class.anon.47 = type { ptr, i64 }
%"struct.Eigen::ThreadPoolDevice" = type { ptr, i32, ptr }
%class.anon.32 = type { i64, i64, ptr, ptr, i64 }
%"class.std::__u::__bind" = type { ptr, %"class.std::__u::tuple" }
%"class.std::__u::tuple" = type { %"struct.std::__u::__tuple_impl" }
%"struct.std::__u::__tuple_impl" = type { %"class.std::__u::__tuple_leaf", %"class.std::__u::__tuple_leaf.40" }
%"class.std::__u::__tuple_leaf" = type { ptr }
%"class.std::__u::__tuple_leaf.40" = type { %class.anon.32 }
%"class.tsl::mutex_lock" = type { ptr }
%"struct.std::__u::__less" = type { i8 }
%"struct.std::__u::__less.33" = type { i8 }
%"struct.std::__u::__tuple_indices" = type { i8 }
%"struct.std::__u::__tuple_types" = type { i8 }
%"struct.std::__u::__tuple_indices.41" = type { i8 }
%"struct.std::__u::__tuple_types.42" = type { i8 }
%"struct.std::__u::__less.43" = type { i8 }
%"class.std::__u::unique_ptr" = type { %"class.std::__u::__compressed_pair" }
%"class.std::__u::__compressed_pair" = type { %"struct.std::__u::__compressed_pair_elem", %"struct.std::__u::__compressed_pair_elem.44" }
%"struct.std::__u::__compressed_pair_elem" = type { ptr }
%"struct.std::__u::__compressed_pair_elem.44" = type { %"struct.std::__u::__builtin_new_allocator::__builtin_new_deleter" }
%"struct.std::__u::__builtin_new_allocator::__builtin_new_deleter" = type { i64, i64 }
%"struct.std::__u::integral_constant" = type { i8 }
%"struct.std::__u::__function::__use_small_storage" = type { i8 }
%"class.std::__u::__function::__default_alloc_func" = type { %"class.std::__u::__bind" }
%"class.std::__u::tuple.45" = type { i8 }
%"struct.std::__u::__less.48" = type { i8 }
%"struct.std::__u::__function::__use_small_storage.53" = type { i8 }
%"class.std::__u::__function::__default_alloc_func.49" = type { %class.anon.46 }
%class.anon.50 = type { ptr, i64, i64 }
%"struct.std::__u::__function::__use_small_storage.52" = type { i8 }
%"class.std::__u::__function::__default_alloc_func.51" = type { %class.anon.50 }
%"struct.std::__u::integral_constant.55" = type { i8 }
%"struct.std::__u::__function::__use_small_storage.56" = type { i8 }
%"class.std::__u::__function::__default_alloc_func.54" = type { %class.anon.47 }
%"struct.std::__u::__function::__use_small_storage.58" = type { i8 }
%"class.std::__u::__function::__default_alloc_func.57" = type { ptr }
%"struct.std::__u::__function::__use_small_storage.60" = type { i8 }
%"class.std::__u::__function::__default_alloc_func.59" = type { %class.anon }
%struct.__storeu_ps = type { <4 x float> }
%struct.__loadu_ps = type { <4 x float> }

declare i64 @_ZN5Eigen8internal9EvalRangeINS_15TensorEvaluatorIKNS_14TensorAssignOpINS_16TensorChippingOpILl0ENS_9TensorMapINS_6TensorIfLi2ELi1ElEELi16ENS_11MakePointerEEEEEKNS_19TensorCwiseBinaryOpINS0_13scalar_sum_opIKfSD_EEKSA_KNS4_ILl0EKNS5_INS6_ISD_Li2ELi1ElEELi16ES8_EEEEEEEENS_16ThreadPoolDeviceEEElLb1EE14alignBlockSizeEl(i64 noundef %0)
declare void @_ZN5Eigen15TensorEvaluatorIKNS_14TensorAssignOpINS_16TensorChippingOpILl0ENS_9TensorMapINS_6TensorIfLi2ELi1ElEELi16ENS_11MakePointerEEEEEKNS_19TensorCwiseBinaryOpINS_8internal13scalar_sum_opIKfSC_EEKS8_KNS2_ILl0EKNS3_INS4_ISC_Li2ELi1ElEELi16ES6_EEEEEEEENS_16ThreadPoolDeviceEEC2ERSN_RKSO_(ptr noundef nonnull align 8 dereferenceable(256) %0, ptr noundef nonnull align 8 dereferenceable(16) %1, ptr noundef nonnull align 8 dereferenceable(24) %2)
declare void @_ZNK5Eigen16ThreadPoolDevice11parallelForElRKNS_12TensorOpCostENSt3__u8functionIFllEEENS5_IFvllEEE(ptr noundef nonnull align 8 dereferenceable(24) %0, i64 noundef %1, ptr noundef nonnull align 8 dereferenceable(24) %2, ptr noundef %3, ptr noundef %4)
declare i64 @_ZNSt3__u10__function16__policy_invokerIFllEE11__call_implINS0_20__default_alloc_funcIPS2_S2_EEEElPKNS0_16__policy_storageEl(ptr noundef %0, i64 noundef %1)
declare void @_ZNSt3__u10__function16__policy_invokerIFvllEE11__call_implINS0_20__default_alloc_funcIZN5Eigen8internal14TensorExecutorIKNS6_14TensorAssignOpINS6_16TensorChippingOpILl0ENS6_9TensorMapINS6_6TensorIfLi2ELi1ElEELi16ENS6_11MakePointerEEEEEKNS6_19TensorCwiseBinaryOpINS7_13scalar_sum_opIKfSJ_EEKSG_KNSA_ILl0EKNSB_INSC_ISJ_Li2ELi1ElEELi16ESE_EEEEEEEENS6_16ThreadPoolDeviceELb1ELNS7_15TiledEvaluationE0EE3runERSU_RKSV_EUlllE_S2_EEEEvPKNS0_16__policy_storageEll(ptr noundef %0, i64 noundef %1, i64 noundef %2)
@_ZZNSt3__u10__function8__policy15__choose_policyINS0_20__default_alloc_funcIPFllES4_EEEEPKS1_NS_17integral_constantIbLb1EEEE8__policy = linkonce_odr hidden constant %"struct.std::__u::__function::__policy" { ptr null, ptr null, i8 0, ptr @_ZTIPFllE }, align 8
@_ZTVN10__cxxabiv119__pointer_type_infoE = external global ptr
@_ZTSPFllE = linkonce_odr constant [6 x i8] c"PFllE\00", align 1
@_ZTVN10__cxxabiv120__function_type_infoE = external global ptr
@_ZTSFllE = linkonce_odr constant [5 x i8] c"FllE\00", align 1
@_ZTIFllE = linkonce_odr constant { ptr, ptr } { ptr getelementptr inbounds (ptr, ptr @_ZTVN10__cxxabiv120__function_type_infoE, i64 2), ptr @_ZTSFllE }, align 8
@_ZTIPFllE = linkonce_odr constant { ptr, ptr, i32, ptr } { ptr getelementptr inbounds (ptr, ptr @_ZTVN10__cxxabiv119__pointer_type_infoE, i64 2), ptr @_ZTSPFllE, i32 0, ptr @_ZTIFllE }, align 8
@_ZZNSt3__u10__function8__policy15__choose_policyINS0_20__default_alloc_funcIZN5Eigen8internal14TensorExecutorIKNS4_14TensorAssignOpINS4_16TensorChippingOpILl0ENS4_9TensorMapINS4_6TensorIfLi2ELi1ElEELi16ENS4_11MakePointerEEEEEKNS4_19TensorCwiseBinaryOpINS5_13scalar_sum_opIKfSH_EEKSE_KNS8_ILl0EKNS9_INSA_ISH_Li2ELi1ElEELi16ESC_EEEEEEEENS4_16ThreadPoolDeviceELb1ELNS5_15TiledEvaluationE0EE3runERSS_RKST_EUlllE_FvllEEEEEPKS1_NS_17integral_constantIbLb1EEEE8__policy = linkonce_odr hidden constant %"struct.std::__u::__function::__policy" { ptr null, ptr null, i8 0, ptr @_ZTIZN5Eigen8internal14TensorExecutorIKNS_14TensorAssignOpINS_16TensorChippingOpILl0ENS_9TensorMapINS_6TensorIfLi2ELi1ElEELi16ENS_11MakePointerEEEEEKNS_19TensorCwiseBinaryOpINS0_13scalar_sum_opIKfSC_EEKS9_KNS3_ILl0EKNS4_INS5_ISC_Li2ELi1ElEELi16ES7_EEEEEEEENS_16ThreadPoolDeviceELb1ELNS0_15TiledEvaluationE0EE3runERSN_RKSO_EUlllE_ }, align 8
@_ZTSZN5Eigen8internal14TensorExecutorIKNS_14TensorAssignOpINS_16TensorChippingOpILl0ENS_9TensorMapINS_6TensorIfLi2ELi1ElEELi16ENS_11MakePointerEEEEEKNS_19TensorCwiseBinaryOpINS0_13scalar_sum_opIKfSC_EEKS9_KNS3_ILl0EKNS4_INS5_ISC_Li2ELi1ElEELi16ES7_EEEEEEEENS_16ThreadPoolDeviceELb1ELNS0_15TiledEvaluationE0EE3runERSN_RKSO_EUlllE_ = linkonce_odr constant [326 x i8] c"ZN5Eigen8internal14TensorExecutorIKNS_14TensorAssignOpINS_16TensorChippingOpILl0ENS_9TensorMapINS_6TensorIfLi2ELi1ElEELi16ENS_11MakePointerEEEEEKNS_19TensorCwiseBinaryOpINS0_13scalar_sum_opIKfSC_EEKS9_KNS3_ILl0EKNS4_INS5_ISC_Li2ELi1ElEELi16ES7_EEEEEEEENS_16ThreadPoolDeviceELb1ELNS0_15TiledEvaluationE0EE3runERSN_RKSO_EUlllE_\00", align 1
@_ZTIZN5Eigen8internal14TensorExecutorIKNS_14TensorAssignOpINS_16TensorChippingOpILl0ENS_9TensorMapINS_6TensorIfLi2ELi1ElEELi16ENS_11MakePointerEEEEEKNS_19TensorCwiseBinaryOpINS0_13scalar_sum_opIKfSC_EEKS9_KNS3_ILl0EKNS4_INS5_ISC_Li2ELi1ElEELi16ES7_EEEEEEEENS_16ThreadPoolDeviceELb1ELNS0_15TiledEvaluationE0EE3runERSN_RKSO_EUlllE_ = linkonce_odr constant { ptr, ptr } { ptr getelementptr inbounds (ptr, ptr @_ZTVN10__cxxabiv117__class_type_infoE, i64 2), ptr @_ZTSZN5Eigen8internal14TensorExecutorIKNS_14TensorAssignOpINS_16TensorChippingOpILl0ENS_9TensorMapINS_6TensorIfLi2ELi1ElEELi16ENS_11MakePointerEEEEEKNS_19TensorCwiseBinaryOpINS0_13scalar_sum_opIKfSC_EEKS9_KNS3_ILl0EKNS4_INS5_ISC_Li2ELi1ElEELi16ES7_EEEEEEEENS_16ThreadPoolDeviceELb1ELNS0_15TiledEvaluationE0EE3runERSN_RKSO_EUlllE_ }, align 8
@llvm.global_ctors = appending global [1 x { i32, ptr, ptr }] [{ i32, ptr, ptr } { i32 65535, ptr @_GLOBAL__sub_I_q.ii, ptr null }]
@_ZTVN10__cxxabiv117__class_type_infoE = external global ptr

declare void @_GLOBAL__sub_I_q.ii()

declare void @llvm.lifetime.start.p0(i64, ptr)
declare void @llvm.lifetime.end.p0(i64, ptr)
declare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg)

; Function Attrs: nounwind optsize uwtable
define i64 @_ZN10tensorflow7functor16ScatterNdFunctorIN5Eigen16ThreadPoolDeviceEflLNS_13scatter_nd_op8UpdateOpE1ELi2EEclERKS3_lNS2_5arrayIlLm2EEENS2_9TensorMapINS2_6TensorIfLi2ELi1ElEELi16ENS2_11MakePointerEEENSB_INSC_IKlLi2ELi1ElEELi16ESE_EENSB_INSC_IKfLi2ELi1ElEELi16ESE_EESF_(ptr noundef nonnull align 1 dereferenceable(1) %0, ptr noundef nonnull align 8 dereferenceable(24) %1, i64 noundef %2, i64 %3, i64 %4, ptr noundef byval(%"class.Eigen::TensorMap") align 8 %5, ptr noundef byval(%"class.Eigen::TensorMap.3") align 8 %6, ptr noundef byval(%"class.Eigen::TensorMap.5") align 8 %7, ptr noundef byval(%"class.Eigen::TensorMap") align 8 %8) {
  %10 = alloca %"struct.Eigen::TensorEvaluator", align 8
  %11 = alloca %"class.Eigen::TensorOpCost", align 8
  %12 = alloca %"class.std::__u::function", align 8
  %13 = alloca %"class.std::__u::function.22", align 8
  %14 = alloca %"class.Eigen::TensorCwiseBinaryOp", align 8
  %15 = alloca %"class.Eigen::TensorAssignOp", align 8
  %16 = alloca %"class.Eigen::TensorChippingOp", align 8
  %17 = alloca %"class.Eigen::array", align 8
  %18 = alloca [2 x i64], align 16
  store i64 %3, ptr %17, align 8
  %19 = getelementptr inbounds { i64, i64 }, ptr %17, i64 0, i32 1
  store i64 %4, ptr %19, align 8
  %20 = getelementptr inbounds %"class.Eigen::TensorMap.3", ptr %6, i64 0, i32 1
  %21 = load i64, ptr %20, align 8
  call void @llvm.lifetime.start.p0(i64 16, ptr nonnull %18) #11
  %22 = getelementptr inbounds [2 x i64], ptr %18, i64 0, i64 1
  br label %33

.preheader55:                                     ; preds = %33
  %.lcssa78 = phi i64 [ %36, %33 ]
  store i64 1, ptr %22, align 8
  store i64 %.lcssa78, ptr %18, align 16
  %23 = icmp sgt i64 %21, 0
  br i1 %23, label %.preheader.lr.ph, label %.thread

.preheader.lr.ph:                                 ; preds = %.preheader55
  %24 = getelementptr inbounds %"class.Eigen::TensorMap.3", ptr %6, i64 0, i32 1, i32 0, i32 0, i64 1
  %.sroa.243.0..sroa_idx = getelementptr inbounds i8, ptr %16, i64 24
  %25 = getelementptr inbounds %"class.Eigen::TensorCwiseBinaryOp", ptr %14, i64 0, i32 1
  %.sroa.452.0..sroa_idx = getelementptr inbounds %"class.Eigen::TensorCwiseBinaryOp", ptr %14, i64 0, i32 1, i32 1
  %26 = getelementptr inbounds %"class.Eigen::TensorAssignOp", ptr %15, i64 0, i32 1
  %27 = getelementptr inbounds %"struct.Eigen::TensorEvaluator", ptr %10, i64 0, i32 1, i32 2
  %.sroa.617.0..sroa_idx.i.i.i.i = getelementptr inbounds i8, ptr %11, i64 8
  %.sroa.918.0..sroa_idx.i.i.i.i = getelementptr inbounds i8, ptr %11, i64 16
  %28 = getelementptr inbounds %"class.std::__u::__function::__policy_func", ptr %12, i64 0, i32 1
  %29 = getelementptr inbounds %"class.std::__u::__function::__policy_func", ptr %12, i64 0, i32 2
  %30 = getelementptr inbounds %"class.std::__u::__function::__policy_func.25", ptr %13, i64 0, i32 1
  %31 = getelementptr inbounds %"class.std::__u::__function::__policy_func.25", ptr %13, i64 0, i32 2
  %32 = ptrtoint ptr %10 to i64
  %.pre = load i64, ptr %24, align 8
  %.pre74 = load ptr, ptr %6, align 8
  br label %.preheader

33:                                               ; preds = %9, %33
  %.03256 = phi i32 [ 1, %9 ], [ %37, %33 ]
  %34 = phi i64 [ undef, %9 ], [ %36, %33 ]
  %35 = icmp eq i32 %.03256, 1
  %36 = select i1 %35, i64 %34, i64 %4
  %37 = add nsw i32 %.03256, -1
  %.not = icmp eq i32 %.03256, 0
  br i1 %.not, label %.preheader55, label %33

.preheader:                                       ; preds = %.preheader.lr.ph, %65
  %.03164 = phi i64 [ 0, %.preheader.lr.ph ], [ %66, %65 ]
  %38 = mul nsw i64 %.pre, %.03164
  br label %40

39:                                               ; preds = %40
  %.lcssa77 = phi i1 [ %48, %40 ]
  %.lcssa = phi i64 [ %52, %40 ]
  br i1 %.lcssa77, label %.thread.loopexit, label %53

40:                                               ; preds = %.preheader, %40
  %41 = phi i1 [ true, %.preheader ], [ false, %40 ]
  %indvars.iv = phi i64 [ 0, %.preheader ], [ 1, %40 ]
  %.02860 = phi i1 [ false, %.preheader ], [ %48, %40 ]
  %.02959 = phi i64 [ 0, %.preheader ], [ %52, %40 ]
  %42 = add nsw i64 %38, %indvars.iv
  %43 = getelementptr inbounds i64, ptr %.pre74, i64 %42
  %44 = load volatile i64, ptr %43, align 8
  %45 = getelementptr inbounds [2 x i64], ptr %17, i64 0, i64 %indvars.iv
  %46 = load i64, ptr %45, align 8
  %47 = icmp uge i64 %44, %46
  %48 = or i1 %.02860, %47
  %49 = getelementptr inbounds [2 x i64], ptr %18, i64 0, i64 %indvars.iv
  %50 = load i64, ptr %49, align 8
  %51 = mul nsw i64 %50, %44
  %52 = add nsw i64 %51, %.02959
  br i1 %41, label %40, label %39

53:                                               ; preds = %39
  call void @llvm.lifetime.start.p0(i64 40, ptr nonnull %16)
  call void @llvm.memcpy.p0.p0.i64(ptr noundef nonnull align 8 dereferenceable(24) %16, ptr noundef nonnull align 8 dereferenceable(24) %8, i64 24, i1 false)
  call void @llvm.lifetime.start.p0(i64 88, ptr nonnull %14) #11
  call void @llvm.memcpy.p0.p0.i64(ptr noundef nonnull align 8 dereferenceable(24) %25, ptr noundef nonnull align 8 dereferenceable(24) %7, i64 24, i1 false)
  store i64 %.lcssa, ptr %.sroa.243.0..sroa_idx, align 8
  call void @llvm.memcpy.p0.p0.i64(ptr noundef nonnull align 8 dereferenceable(40) %14, ptr noundef nonnull align 8 dereferenceable(40) %16, i64 40, i1 false)
  store i64 %.03164, ptr %.sroa.452.0..sroa_idx, align 8
  call void @llvm.lifetime.start.p0(i64 16, ptr nonnull %15) #11
  store ptr %16, ptr %15, align 8
  store ptr %14, ptr %26, align 8
  call void @llvm.lifetime.start.p0(i64 32, ptr nonnull %12)
  call void @llvm.lifetime.start.p0(i64 32, ptr nonnull %13)
  call void @llvm.lifetime.start.p0(i64 256, ptr nonnull %10) #11
  call void @_ZN5Eigen15TensorEvaluatorIKNS_14TensorAssignOpINS_16TensorChippingOpILl0ENS_9TensorMapINS_6TensorIfLi2ELi1ElEELi16ENS_11MakePointerEEEEEKNS_19TensorCwiseBinaryOpINS_8internal13scalar_sum_opIKfSC_EEKS8_KNS2_ILl0EKNS3_INS4_ISC_Li2ELi1ElEELi16ES6_EEEEEEEENS_16ThreadPoolDeviceEEC2ERSN_RKSO_(ptr noundef nonnull align 8 dereferenceable(256) %10, ptr noundef nonnull align 8 dereferenceable(16) %15, ptr noundef nonnull align 8 dereferenceable(24) %1) #12
  %54 = load i64, ptr %27, align 8
  call void @llvm.lifetime.start.p0(i64 24, ptr nonnull %11) #11
  store double 8.000000e+00, ptr %11, align 8
  store double 4.000000e+00, ptr %.sroa.617.0..sroa_idx.i.i.i.i, align 8
  store double 1.000000e+00, ptr %.sroa.918.0..sroa_idx.i.i.i.i, align 8
  store ptr @_ZNSt3__u10__function16__policy_invokerIFllEE11__call_implINS0_20__default_alloc_funcIPS2_S2_EEEElPKNS0_16__policy_storageEl, ptr %28, align 8
  store ptr @_ZZNSt3__u10__function8__policy15__choose_policyINS0_20__default_alloc_funcIPFllES4_EEEEPKS1_NS_17integral_constantIbLb1EEEE8__policy, ptr %29, align 8
  store ptr @_ZN5Eigen8internal9EvalRangeINS_15TensorEvaluatorIKNS_14TensorAssignOpINS_16TensorChippingOpILl0ENS_9TensorMapINS_6TensorIfLi2ELi1ElEELi16ENS_11MakePointerEEEEEKNS_19TensorCwiseBinaryOpINS0_13scalar_sum_opIKfSD_EEKSA_KNS4_ILl0EKNS5_INS6_ISD_Li2ELi1ElEELi16ES8_EEEEEEEENS_16ThreadPoolDeviceEEElLb1EE14alignBlockSizeEl, ptr %12, align 8
  store ptr @_ZNSt3__u10__function16__policy_invokerIFvllEE11__call_implINS0_20__default_alloc_funcIZN5Eigen8internal14TensorExecutorIKNS6_14TensorAssignOpINS6_16TensorChippingOpILl0ENS6_9TensorMapINS6_6TensorIfLi2ELi1ElEELi16ENS6_11MakePointerEEEEEKNS6_19TensorCwiseBinaryOpINS7_13scalar_sum_opIKfSJ_EEKSG_KNSA_ILl0EKNSB_INSC_ISJ_Li2ELi1ElEELi16ESE_EEEEEEEENS6_16ThreadPoolDeviceELb1ELNS7_15TiledEvaluationE0EE3runERSU_RKSV_EUlllE_S2_EEEEvPKNS0_16__policy_storageEll, ptr %30, align 8
  store ptr @_ZZNSt3__u10__function8__policy15__choose_policyINS0_20__default_alloc_funcIZN5Eigen8internal14TensorExecutorIKNS4_14TensorAssignOpINS4_16TensorChippingOpILl0ENS4_9TensorMapINS4_6TensorIfLi2ELi1ElEELi16ENS4_11MakePointerEEEEEKNS4_19TensorCwiseBinaryOpINS5_13scalar_sum_opIKfSH_EEKSE_KNS8_ILl0EKNS9_INSA_ISH_Li2ELi1ElEELi16ESC_EEEEEEEENS4_16ThreadPoolDeviceELb1ELNS5_15TiledEvaluationE0EE3runERSS_RKST_EUlllE_FvllEEEEEPKS1_NS_17integral_constantIbLb1EEEE8__policy, ptr %31, align 8
  store i64 %32, ptr %13, align 8
  call void @_ZNK5Eigen16ThreadPoolDevice11parallelForElRKNS_12TensorOpCostENSt3__u8functionIFllEEENS5_IFvllEEE(ptr noundef nonnull align 8 dereferenceable(24) %1, i64 noundef %54, ptr noundef nonnull align 8 dereferenceable(24) %11, ptr noundef nonnull %12, ptr noundef nonnull %13) #12
  %55 = load ptr, ptr %31, align 8
  %56 = getelementptr inbounds %"struct.std::__u::__function::__policy", ptr %55, i64 0, i32 1
  %57 = load ptr, ptr %56, align 8
  %.not.i.i.i.i.i = icmp eq ptr %57, null
  br i1 %.not.i.i.i.i.i, label %_ZNSt3__u8functionIFvllEED2Ev.exit.i.i.i, label %58

58:                                               ; preds = %53
  %59 = load ptr, ptr %13, align 8
  call void %57(ptr noundef %59) #13
  br label %_ZNSt3__u8functionIFvllEED2Ev.exit.i.i.i

_ZNSt3__u8functionIFvllEED2Ev.exit.i.i.i:         ; preds = %58, %53
  %60 = load ptr, ptr %29, align 8
  %61 = getelementptr inbounds %"struct.std::__u::__function::__policy", ptr %60, i64 0, i32 1
  %62 = load ptr, ptr %61, align 8
  %.not.i.i4.i.i.i = icmp eq ptr %62, null
  br i1 %.not.i.i4.i.i.i, label %65, label %63

63:                                               ; preds = %_ZNSt3__u8functionIFvllEED2Ev.exit.i.i.i
  %64 = load ptr, ptr %12, align 8
  call void %62(ptr noundef %64) #13
  br label %65

65:                                               ; preds = %_ZNSt3__u8functionIFvllEED2Ev.exit.i.i.i, %63
  call void @llvm.lifetime.end.p0(i64 24, ptr nonnull %11) #11
  call void @llvm.lifetime.end.p0(i64 256, ptr nonnull %10) #11
  call void @llvm.lifetime.end.p0(i64 32, ptr nonnull %12)
  call void @llvm.lifetime.end.p0(i64 32, ptr nonnull %13)
  call void @llvm.lifetime.end.p0(i64 16, ptr nonnull %15) #11
  call void @llvm.lifetime.end.p0(i64 88, ptr nonnull %14) #11
  call void @llvm.lifetime.end.p0(i64 40, ptr nonnull %16)
  %66 = add nuw nsw i64 %.03164, 1
  %exitcond.not = icmp eq i64 %66, %21
  br i1 %exitcond.not, label %.thread.loopexit, label %.preheader

.thread.loopexit:                                 ; preds = %39, %65
  %.2.ph = phi i64 [ -1, %65 ], [ %.03164, %39 ]
  br label %.thread

.thread:                                          ; preds = %.thread.loopexit, %.preheader55
  %.2 = phi i64 [ -1, %.preheader55 ], [ %.2.ph, %.thread.loopexit ]
  call void @llvm.lifetime.end.p0(i64 16, ptr nonnull %18) #11
  ret i64 %.2
}