From cbaf43e8eaf3f494b8a53160dfeef32f26d80654 Mon Sep 17 00:00:00 2001 From: Arne Welzel Date: Thu, 21 Sep 2023 15:06:50 +0200 Subject: [PATCH] VectorVal: Embed vector_val Similar motivation as for RecordVal, save an extra malloc/free and pointer indirection. This breaks the `auto& RawVec()` API which previously returned a reference to the std::vector*. It now returns a reference to the vector instead. It's commented as intended for internal and compiled code, so even though it's public API, The previous `std::vector>*&` return type was also very likely not intended (all consumers just dereference it anyhow). I'm certain this API was never meant to modify the actual pointer value. I've switched to explicit typing, too. --- src/EventTrace.cc | 4 +- src/Stmt.cc | 2 +- src/Val.cc | 81 +++++++++++++++++++------------------ src/Val.h | 34 +++++----------- src/logging/Manager.cc | 2 +- src/script_opt/ZAM/Ops.in | 16 ++++---- src/script_opt/ZAM/ZBody.cc | 12 +++--- 7 files changed, 70 insertions(+), 81 deletions(-) diff --git a/src/EventTrace.cc b/src/EventTrace.cc index baac235a57..d313f7c3ea 100644 --- a/src/EventTrace.cc +++ b/src/EventTrace.cc @@ -314,13 +314,13 @@ void ValTrace::TraceTable(const TableValPtr& tv) void ValTrace::TraceVector(const VectorValPtr& vv) { auto& vec = vv->RawVec(); - auto n = vec->size(); + auto n = vec.size(); auto& yt = vv->RawYieldType(); auto& yts = vv->RawYieldTypes(); for ( auto i = 0U; i < n; ++i ) { - auto& elem = (*vec)[i]; + auto& elem = vec[i]; if ( elem ) { auto& t = yts ? (*yts)[i] : yt; diff --git a/src/Stmt.cc b/src/Stmt.cc index 58728517fd..9bccb790cf 100644 --- a/src/Stmt.cc +++ b/src/Stmt.cc @@ -1411,7 +1411,7 @@ ValPtr ForStmt::DoExec(Frame* f, Val* v, StmtFlowType& flow) else if ( v->GetType()->Tag() == TYPE_VECTOR ) { VectorVal* vv = v->AsVectorVal(); - const auto& raw_vv = *vv->RawVec(); + const auto& raw_vv = vv->RawVec(); for ( auto i = 0u; i < vv->Size(); ++i ) { diff --git a/src/Val.cc b/src/Val.cc index cf99dcf0d1..332bf554b5 100644 --- a/src/Val.cc +++ b/src/Val.cc @@ -3487,11 +3487,8 @@ ValPtr TypeVal::DoClone(CloneState* state) return {NewRef{}, this}; } -VectorVal::VectorVal(VectorTypePtr t) : VectorVal(t, new vector>()) { } - -VectorVal::VectorVal(VectorTypePtr t, std::vector>* vals) : Val(t) +VectorVal::VectorVal(VectorTypePtr t) : Val(t) { - vector_val = vals; yield_type = t->Yield(); auto y_tag = yield_type->Tag(); @@ -3499,6 +3496,12 @@ VectorVal::VectorVal(VectorTypePtr t, std::vector>* vals) : managed_yield = ZVal::IsManagedType(yield_type); } +VectorVal::VectorVal(VectorTypePtr t, std::vector>* vals) : VectorVal(t) + { + if ( vals ) + vector_val = std::move(*vals); + } + VectorVal::~VectorVal() { if ( yield_types ) @@ -3506,7 +3509,7 @@ VectorVal::~VectorVal() int n = yield_types->size(); for ( auto i = 0; i < n; ++i ) { - auto& elem = (*vector_val)[i]; + auto& elem = vector_val[i]; if ( elem ) ZVal::DeleteIfManaged(*elem, (*yield_types)[i]); } @@ -3515,17 +3518,15 @@ VectorVal::~VectorVal() else if ( managed_yield ) { - for ( auto& elem : *vector_val ) + for ( auto& elem : vector_val ) if ( elem ) ZVal::DeleteManagedType(*elem); } - - delete vector_val; } ValPtr VectorVal::SizeVal() const { - return val_mgr->Count(uint32_t(vector_val->size())); + return val_mgr->Count(uint32_t(vector_val.size())); } bool VectorVal::CheckElementType(const ValPtr& element) @@ -3540,7 +3541,7 @@ bool VectorVal::CheckElementType(const ValPtr& element) if ( any_yield ) { - int n = vector_val->size(); + int n = vector_val.size(); if ( n == 0 ) { @@ -3574,14 +3575,14 @@ bool VectorVal::Assign(unsigned int index, ValPtr element) if ( ! CheckElementType(element) ) return false; - unsigned int n = vector_val->size(); + unsigned int n = vector_val.size(); if ( index >= n ) { if ( index > n ) AddHoles(index - n); - vector_val->resize(index + 1); + vector_val.resize(index + 1); if ( yield_types ) yield_types->resize(index + 1); } @@ -3590,14 +3591,14 @@ bool VectorVal::Assign(unsigned int index, ValPtr element) { const auto& t = element->GetType(); (*yield_types)[index] = t; - auto& elem = (*vector_val)[index]; + auto& elem = vector_val[index]; if ( elem ) ZVal::DeleteIfManaged(*elem, t); elem = ZVal(std::move(element), t); } else { - auto& elem = (*vector_val)[index]; + auto& elem = vector_val[index]; if ( managed_yield && elem ) ZVal::DeleteManagedType(*elem); @@ -3630,17 +3631,17 @@ bool VectorVal::Insert(unsigned int index, ValPtr element) vector>::iterator it; vector::iterator types_it; - auto n = vector_val->size(); + auto n = vector_val.size(); if ( index < n ) { // Find location within existing vector elements. - it = std::next(vector_val->begin(), index); + it = std::next(vector_val.begin(), index); if ( yield_types ) types_it = std::next(yield_types->begin(), index); } else { - it = vector_val->end(); + it = vector_val.end(); if ( yield_types ) types_it = yield_types->end(); @@ -3654,13 +3655,13 @@ bool VectorVal::Insert(unsigned int index, ValPtr element) { const auto& t = element->GetType(); yield_types->insert(types_it, t); - vector_val->insert(it, ZVal(std::move(element), t)); + vector_val.insert(it, ZVal(std::move(element), t)); } else - vector_val->insert(it, ZVal(std::move(element), yield_type)); + vector_val.insert(it, ZVal(std::move(element), yield_type)); } else - vector_val->insert(it, std::nullopt); + vector_val.insert(it, std::nullopt); Modified(); return true; @@ -3673,15 +3674,15 @@ void VectorVal::AddHoles(int nholes) fill_t = base_type(TYPE_ANY); for ( auto i = 0; i < nholes; ++i ) - vector_val->emplace_back(std::nullopt); + vector_val.emplace_back(std::nullopt); } bool VectorVal::Remove(unsigned int index) { - if ( index >= vector_val->size() ) + if ( index >= vector_val.size() ) return false; - auto it = std::next(vector_val->begin(), index); + auto it = std::next(vector_val.begin(), index); if ( yield_types ) { @@ -3697,7 +3698,7 @@ bool VectorVal::Remove(unsigned int index) ZVal::DeleteManagedType(**it); } - vector_val->erase(it); + vector_val.erase(it); Modified(); return true; @@ -3730,10 +3731,10 @@ bool VectorVal::AddTo(Val* val, bool /* is_first_init */) const ValPtr VectorVal::At(unsigned int index) const { - if ( index >= vector_val->size() ) + if ( index >= vector_val.size() ) return Val::nil; - auto& elem = (*vector_val)[index]; + auto& elem = vector_val[index]; if ( ! elem ) return Val::nil; @@ -3850,7 +3851,7 @@ void VectorVal::Sort(Func* cmp_func) } } - sort(vector_val->begin(), vector_val->end(), sort_func); + sort(vector_val.begin(), vector_val.end(), sort_func); } VectorValPtr VectorVal::Order(Func* cmp_func) @@ -3895,7 +3896,7 @@ VectorValPtr VectorVal::Order(Func* cmp_func) for ( i = 0; i < n; ++i ) { ind_vv[i] = i; - index_map.emplace_back(&(*vector_val)[i]); + index_map.emplace_back(&vector_val[i]); } sort(ind_vv.begin(), ind_vv.end(), sort_func); @@ -3920,14 +3921,14 @@ bool VectorVal::Concretize(const TypePtr& t) // shouldn't happen in any case. return yield_type->Tag() == t->Tag(); - if ( ! vector_val ) + if ( vector_val.empty() ) // Trivially concretized. return true; - auto n = vector_val->size(); + auto n = vector_val.size(); for ( auto i = 0U; i < n; ++i ) { - auto& v = (*vector_val)[i]; + auto& v = vector_val[i]; if ( ! v ) // Vector hole does not require concretization. continue; @@ -3960,7 +3961,7 @@ bool VectorVal::Concretize(const TypePtr& t) unsigned int VectorVal::ComputeFootprint(std::unordered_set* analyzed_vals) const { - auto n = vector_val->size(); + auto n = vector_val.size(); unsigned int fp = n; for ( auto i = 0U; i < n; ++i ) @@ -3975,9 +3976,9 @@ unsigned int VectorVal::ComputeFootprint(std::unordered_set* analyze unsigned int VectorVal::Resize(unsigned int new_num_elements) { - unsigned int oldsize = vector_val->size(); - vector_val->reserve(new_num_elements); - vector_val->resize(new_num_elements); + unsigned int oldsize = vector_val.size(); + vector_val.reserve(new_num_elements); + vector_val.resize(new_num_elements); if ( yield_types ) { @@ -3990,7 +3991,7 @@ unsigned int VectorVal::Resize(unsigned int new_num_elements) unsigned int VectorVal::ResizeAtLeast(unsigned int new_num_elements) { - unsigned int old_size = vector_val->size(); + unsigned int old_size = vector_val.size(); if ( new_num_elements <= old_size ) return old_size; @@ -3999,7 +4000,7 @@ unsigned int VectorVal::ResizeAtLeast(unsigned int new_num_elements) void VectorVal::Reserve(unsigned int num_elements) { - vector_val->reserve(num_elements); + vector_val.reserve(num_elements); if ( yield_types ) yield_types->reserve(num_elements); @@ -4008,10 +4009,10 @@ void VectorVal::Reserve(unsigned int num_elements) ValPtr VectorVal::DoClone(CloneState* state) { auto vv = make_intrusive(GetType()); - vv->Reserve(vector_val->size()); + vv->Reserve(vector_val.size()); state->NewClone(this, vv); - int n = vector_val->size(); + int n = vector_val.size(); for ( auto i = 0; i < n; ++i ) { @@ -4026,7 +4027,7 @@ void VectorVal::ValDescribe(ODesc* d) const { d->Add("["); - size_t vector_size = vector_val->size(); + size_t vector_size = vector_val.size(); if ( vector_size != 0 ) { diff --git a/src/Val.h b/src/Val.h index 7e90f9ad03..31acb6e68c 100644 --- a/src/Val.h +++ b/src/Val.h @@ -1592,7 +1592,7 @@ public: // Returns true if successful. bool AddTo(Val* v, bool is_first_init) const override; - unsigned int Size() const { return vector_val->size(); } + unsigned int Size() const { return vector_val.size(); } // Is there any way to reclaim previously-allocated memory when you // shrink a vector? The return value is the old size. @@ -1662,10 +1662,7 @@ public: ValPtr ValAt(unsigned int index) const { return At(index); } - bool Has(unsigned int index) const - { - return index < vector_val->size() && (*vector_val)[index]; - } + bool Has(unsigned int index) const { return index < vector_val.size() && vector_val[index]; } /** * Returns the given element in a given underlying representation. @@ -1675,26 +1672,17 @@ public: * @param index The position in the vector of the element to return. * @return The element's underlying value. */ - zeek_int_t IntAt(unsigned int index) const { return (*vector_val)[index]->int_val; } - zeek_uint_t CountAt(unsigned int index) const { return (*vector_val)[index]->uint_val; } - double DoubleAt(unsigned int index) const { return (*vector_val)[index]->double_val; } - const RecordVal* RecordValAt(unsigned int index) const - { - return (*vector_val)[index]->record_val; - } - bool BoolAt(unsigned int index) const - { - return static_cast((*vector_val)[index]->uint_val); - } - const StringVal* StringValAt(unsigned int index) const - { - return (*vector_val)[index]->string_val; - } + zeek_int_t IntAt(unsigned int index) const { return vector_val[index]->int_val; } + zeek_uint_t CountAt(unsigned int index) const { return vector_val[index]->uint_val; } + double DoubleAt(unsigned int index) const { return vector_val[index]->double_val; } + const RecordVal* RecordValAt(unsigned int index) const { return vector_val[index]->record_val; } + bool BoolAt(unsigned int index) const { return static_cast(vector_val[index]->uint_val); } + const StringVal* StringValAt(unsigned int index) const { return vector_val[index]->string_val; } const String* StringAt(unsigned int index) const { return StringValAt(index)->AsString(); } // Only intended for low-level access by internal or compiled code. - const auto& RawVec() const { return vector_val; } - auto& RawVec() { return vector_val; } + const std::vector>& RawVec() const { return vector_val; } + std::vector>& RawVec() { return vector_val; } const auto& RawYieldType() const { return yield_type; } const auto& RawYieldTypes() const { return yield_types; } @@ -1730,7 +1718,7 @@ private: // Add the given number of "holes" to the end of a vector. void AddHoles(int nholes); - std::vector>* vector_val; + std::vector> vector_val; // For homogeneous vectors (the usual case), the type of the // elements. Will be TYPE_VOID for empty vectors created using diff --git a/src/logging/Manager.cc b/src/logging/Manager.cc index e938a3f140..28db8f41cf 100644 --- a/src/logging/Manager.cc +++ b/src/logging/Manager.cc @@ -1113,7 +1113,7 @@ threading::Value* Manager::ValToLogVal(std::optional& val, Type* ty) for ( zeek_int_t i = 0; i < lval->val.vector_val.size; i++ ) { - lval->val.vector_val.vals[i] = ValToLogVal((*vv)[i], vt.get()); + lval->val.vector_val.vals[i] = ValToLogVal(vv[i], vt.get()); } break; diff --git a/src/script_opt/ZAM/Ops.in b/src/script_opt/ZAM/Ops.in index 2d555aaaee..09df3828b4 100644 --- a/src/script_opt/ZAM/Ops.in +++ b/src/script_opt/ZAM/Ops.in @@ -364,7 +364,7 @@ eval auto vv = frame[z.v1].vector_val; vv->Assign(0, $1.ToVal(z.t)); else { - vv->RawVec()->push_back(CopyVal($1)); + vv->RawVec().push_back(CopyVal($1)); vv->Modified(); } @@ -876,9 +876,9 @@ eval AssignV1(frame[z.v2].int_val ? CopyVal(z.c) : CopyVal(frame[z.v3])) op Bool-Vec-Cond type VVVV set-type $2 -eval auto& vsel = *frame[z.v2].vector_val->RawVec(); - auto& v1 = *frame[z.v3].vector_val->RawVec(); - auto& v2 = *frame[z.v4].vector_val->RawVec(); +eval auto& vsel = frame[z.v2].vector_val->RawVec(); + auto& v1 = frame[z.v3].vector_val->RawVec(); + auto& v2 = frame[z.v4].vector_val->RawVec(); auto n = v1.size(); auto res = new vector>(n); for ( auto i = 0U; i < n; ++i ) @@ -968,11 +968,11 @@ eval EvalIndexVec(frame[z.v3].uint_val) macro EvalIndexVec(index) auto& vv = frame[z.v2].vector_val->RawVec(); - const auto& vec = *vv; + const auto& vec = vv; zeek_int_t ind = index; if ( ind < 0 ) - ind += vv->size(); - if ( ind < 0 || ind >= vv->size() ) + ind += vv.size(); + if ( ind < 0 || ind >= vv.size() ) ZAM_run_time_error(z.loc, "no such index"); AssignV1(CopyVal(*vec[ind])) @@ -1847,7 +1847,7 @@ internal-op Init-Vector-Loop type VV op1-read eval auto& vv = frame[z.v1].vector_val->RawVec(); - step_iters[z.v2].InitLoop(vv); + step_iters[z.v2].InitLoop(&vv); macro NextVectorIterCore(info, branch) auto& si = step_iters[info]; diff --git a/src/script_opt/ZAM/ZBody.cc b/src/script_opt/ZAM/ZBody.cc index 63ebe9a7e4..a63361ec51 100644 --- a/src/script_opt/ZAM/ZBody.cc +++ b/src/script_opt/ZAM/ZBody.cc @@ -60,7 +60,7 @@ static bool copy_vec_elem(VectorVal* vv, zeek_uint_t ind, ZVal zv, const TypePtr if ( vv->Size() <= ind ) vv->Resize(ind + 1); - auto& elem = (*vv->RawVec())[ind]; + auto& elem = vv->RawVec()[ind]; if ( ! ZVal::IsManagedType(t) ) { @@ -96,12 +96,12 @@ static void vec_exec(ZOp op, TypePtr t, VectorVal*& v1, const VectorVal* v2, con #define VEC_COERCE(tag, lhs_type, cast, rhs_accessor, ov_check, ov_err) \ static VectorVal* vec_coerce_##tag(VectorVal* vec, const ZInst& z) \ { \ - auto& v = *vec->RawVec(); \ + auto& v = vec->RawVec(); \ auto yt = make_intrusive(base_type(lhs_type)); \ auto res_zv = new VectorVal(yt); \ auto n = v.size(); \ res_zv->Resize(n); \ - auto& res = *res_zv->RawVec(); \ + auto& res = res_zv->RawVec(); \ for ( auto i = 0U; i < n; ++i ) \ if ( v[i] ) \ { \ @@ -479,7 +479,7 @@ static void vec_exec(ZOp op, TypePtr t, VectorVal*& v1, const VectorVal* v2, con // well move the whole kit-and-caboodle into the Exec method). But // that seems like a lot of code bloat for only a very modest gain. - auto& vec2 = *v2->RawVec(); + auto& vec2 = v2->RawVec(); auto n = vec2.size(); auto vec1_ptr = new vector>(n); auto& vec1 = *vec1_ptr; @@ -511,8 +511,8 @@ static void vec_exec(ZOp op, TypePtr t, VectorVal*& v1, const VectorVal* v2, con { // See comment above re further speed-up. - auto& vec2 = *v2->RawVec(); - auto& vec3 = *v3->RawVec(); + auto& vec2 = v2->RawVec(); + auto& vec3 = v3->RawVec(); auto n = vec2.size(); auto vec1_ptr = new vector>(n); auto& vec1 = *vec1_ptr;