Mercurial > hg > graal-compiler
view src/share/vm/opto/escape.cpp @ 0:a61af66fc99e jdk7-b24
Initial load
author | duke |
---|---|
date | Sat, 01 Dec 2007 00:00:00 +0000 |
parents | |
children | b789bcaf2dd9 |
line wrap: on
line source
/* * Copyright 2005-2006 Sun Microsystems, Inc. All Rights Reserved. * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * * This code is free software; you can redistribute it and/or modify it * under the terms of the GNU General Public License version 2 only, as * published by the Free Software Foundation. * * This code is distributed in the hope that it will be useful, but WITHOUT * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License * version 2 for more details (a copy is included in the LICENSE file that * accompanied this code). * * You should have received a copy of the GNU General Public License version * 2 along with this work; if not, write to the Free Software Foundation, * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. * * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, * CA 95054 USA or visit www.sun.com if you need additional information or * have any questions. * */ #include "incls/_precompiled.incl" #include "incls/_escape.cpp.incl" uint PointsToNode::edge_target(uint e) const { assert(_edges != NULL && e < (uint)_edges->length(), "valid edge index"); return (_edges->at(e) >> EdgeShift); } PointsToNode::EdgeType PointsToNode::edge_type(uint e) const { assert(_edges != NULL && e < (uint)_edges->length(), "valid edge index"); return (EdgeType) (_edges->at(e) & EdgeMask); } void PointsToNode::add_edge(uint targIdx, PointsToNode::EdgeType et) { uint v = (targIdx << EdgeShift) + ((uint) et); if (_edges == NULL) { Arena *a = Compile::current()->comp_arena(); _edges = new(a) GrowableArray<uint>(a, INITIAL_EDGE_COUNT, 0, 0); } _edges->append_if_missing(v); } void PointsToNode::remove_edge(uint targIdx, PointsToNode::EdgeType et) { uint v = (targIdx << EdgeShift) + ((uint) et); _edges->remove(v); } #ifndef PRODUCT static char *node_type_names[] = { "UnknownType", "JavaObject", "LocalVar", "Field" }; static char *esc_names[] = { "UnknownEscape", "NoEscape ", "ArgEscape ", "GlobalEscape " }; static char *edge_type_suffix[] = { "?", // UnknownEdge "P", // PointsToEdge "D", // DeferredEdge "F" // FieldEdge }; void PointsToNode::dump() const { NodeType nt = node_type(); EscapeState es = escape_state(); tty->print("%s %s [[", node_type_names[(int) nt], esc_names[(int) es]); for (uint i = 0; i < edge_count(); i++) { tty->print(" %d%s", edge_target(i), edge_type_suffix[(int) edge_type(i)]); } tty->print("]] "); if (_node == NULL) tty->print_cr("<null>"); else _node->dump(); } #endif ConnectionGraph::ConnectionGraph(Compile * C) : _processed(C->comp_arena()), _node_map(C->comp_arena()) { _collecting = true; this->_compile = C; const PointsToNode &dummy = PointsToNode(); _nodes = new(C->comp_arena()) GrowableArray<PointsToNode>(C->comp_arena(), (int) INITIAL_NODE_COUNT, 0, dummy); _phantom_object = C->top()->_idx; PointsToNode *phn = ptnode_adr(_phantom_object); phn->set_node_type(PointsToNode::JavaObject); phn->set_escape_state(PointsToNode::GlobalEscape); } void ConnectionGraph::add_pointsto_edge(uint from_i, uint to_i) { PointsToNode *f = ptnode_adr(from_i); PointsToNode *t = ptnode_adr(to_i); assert(f->node_type() != PointsToNode::UnknownType && t->node_type() != PointsToNode::UnknownType, "node types must be set"); assert(f->node_type() == PointsToNode::LocalVar || f->node_type() == PointsToNode::Field, "invalid source of PointsTo edge"); assert(t->node_type() == PointsToNode::JavaObject, "invalid destination of PointsTo edge"); f->add_edge(to_i, PointsToNode::PointsToEdge); } void ConnectionGraph::add_deferred_edge(uint from_i, uint to_i) { PointsToNode *f = ptnode_adr(from_i); PointsToNode *t = ptnode_adr(to_i); assert(f->node_type() != PointsToNode::UnknownType && t->node_type() != PointsToNode::UnknownType, "node types must be set"); assert(f->node_type() == PointsToNode::LocalVar || f->node_type() == PointsToNode::Field, "invalid source of Deferred edge"); assert(t->node_type() == PointsToNode::LocalVar || t->node_type() == PointsToNode::Field, "invalid destination of Deferred edge"); // don't add a self-referential edge, this can occur during removal of // deferred edges if (from_i != to_i) f->add_edge(to_i, PointsToNode::DeferredEdge); } int ConnectionGraph::type_to_offset(const Type *t) { const TypePtr *t_ptr = t->isa_ptr(); assert(t_ptr != NULL, "must be a pointer type"); return t_ptr->offset(); } void ConnectionGraph::add_field_edge(uint from_i, uint to_i, int offset) { PointsToNode *f = ptnode_adr(from_i); PointsToNode *t = ptnode_adr(to_i); assert(f->node_type() != PointsToNode::UnknownType && t->node_type() != PointsToNode::UnknownType, "node types must be set"); assert(f->node_type() == PointsToNode::JavaObject, "invalid destination of Field edge"); assert(t->node_type() == PointsToNode::Field, "invalid destination of Field edge"); assert (t->offset() == -1 || t->offset() == offset, "conflicting field offsets"); t->set_offset(offset); f->add_edge(to_i, PointsToNode::FieldEdge); } void ConnectionGraph::set_escape_state(uint ni, PointsToNode::EscapeState es) { PointsToNode *npt = ptnode_adr(ni); PointsToNode::EscapeState old_es = npt->escape_state(); if (es > old_es) npt->set_escape_state(es); } PointsToNode::EscapeState ConnectionGraph::escape_state(Node *n, PhaseTransform *phase) { uint idx = n->_idx; PointsToNode::EscapeState es; // If we are still collecting we don't know the answer yet if (_collecting) return PointsToNode::UnknownEscape; // if the node was created after the escape computation, return // UnknownEscape if (idx >= (uint)_nodes->length()) return PointsToNode::UnknownEscape; es = _nodes->at_grow(idx).escape_state(); // if we have already computed a value, return it if (es != PointsToNode::UnknownEscape) return es; // compute max escape state of anything this node could point to VectorSet ptset(Thread::current()->resource_area()); PointsTo(ptset, n, phase); for( VectorSetI i(&ptset); i.test() && es != PointsToNode::GlobalEscape; ++i ) { uint pt = i.elem; PointsToNode::EscapeState pes = _nodes->at(pt).escape_state(); if (pes > es) es = pes; } // cache the computed escape state assert(es != PointsToNode::UnknownEscape, "should have computed an escape state"); _nodes->adr_at(idx)->set_escape_state(es); return es; } void ConnectionGraph::PointsTo(VectorSet &ptset, Node * n, PhaseTransform *phase) { VectorSet visited(Thread::current()->resource_area()); GrowableArray<uint> worklist; n = skip_casts(n); PointsToNode npt = _nodes->at_grow(n->_idx); // If we have a JavaObject, return just that object if (npt.node_type() == PointsToNode::JavaObject) { ptset.set(n->_idx); return; } // we may have a Phi which has not been processed if (npt._node == NULL) { assert(n->is_Phi(), "unprocessed node must be a Phi"); record_for_escape_analysis(n); npt = _nodes->at(n->_idx); } worklist.push(n->_idx); while(worklist.length() > 0) { int ni = worklist.pop(); PointsToNode pn = _nodes->at_grow(ni); if (!visited.test(ni)) { visited.set(ni); // ensure that all inputs of a Phi have been processed if (_collecting && pn._node->is_Phi()) { PhiNode *phi = pn._node->as_Phi(); process_phi_escape(phi, phase); } int edges_processed = 0; for (uint e = 0; e < pn.edge_count(); e++) { PointsToNode::EdgeType et = pn.edge_type(e); if (et == PointsToNode::PointsToEdge) { ptset.set(pn.edge_target(e)); edges_processed++; } else if (et == PointsToNode::DeferredEdge) { worklist.push(pn.edge_target(e)); edges_processed++; } } if (edges_processed == 0) { // no deferred or pointsto edges found. Assume the value was set outside // this method. Add the phantom object to the pointsto set. ptset.set(_phantom_object); } } } } void ConnectionGraph::remove_deferred(uint ni) { VectorSet visited(Thread::current()->resource_area()); uint i = 0; PointsToNode *ptn = ptnode_adr(ni); while(i < ptn->edge_count()) { if (ptn->edge_type(i) != PointsToNode::DeferredEdge) { i++; } else { uint t = ptn->edge_target(i); PointsToNode *ptt = ptnode_adr(t); ptn->remove_edge(t, PointsToNode::DeferredEdge); if(!visited.test(t)) { visited.set(t); for (uint j = 0; j < ptt->edge_count(); j++) { uint n1 = ptt->edge_target(j); PointsToNode *pt1 = ptnode_adr(n1); switch(ptt->edge_type(j)) { case PointsToNode::PointsToEdge: add_pointsto_edge(ni, n1); break; case PointsToNode::DeferredEdge: add_deferred_edge(ni, n1); break; case PointsToNode::FieldEdge: assert(false, "invalid connection graph"); break; } } } } } } // Add an edge to node given by "to_i" from any field of adr_i whose offset // matches "offset" A deferred edge is added if to_i is a LocalVar, and // a pointsto edge is added if it is a JavaObject void ConnectionGraph::add_edge_from_fields(uint adr_i, uint to_i, int offs) { PointsToNode an = _nodes->at_grow(adr_i); PointsToNode to = _nodes->at_grow(to_i); bool deferred = (to.node_type() == PointsToNode::LocalVar); for (uint fe = 0; fe < an.edge_count(); fe++) { assert(an.edge_type(fe) == PointsToNode::FieldEdge, "expecting a field edge"); int fi = an.edge_target(fe); PointsToNode pf = _nodes->at_grow(fi); int po = pf.offset(); if (po == offs || po == Type::OffsetBot || offs == Type::OffsetBot) { if (deferred) add_deferred_edge(fi, to_i); else add_pointsto_edge(fi, to_i); } } } // Add a deferred edge from node given by "from_i" to any field of adr_i whose offset // matches "offset" void ConnectionGraph::add_deferred_edge_to_fields(uint from_i, uint adr_i, int offs) { PointsToNode an = _nodes->at_grow(adr_i); for (uint fe = 0; fe < an.edge_count(); fe++) { assert(an.edge_type(fe) == PointsToNode::FieldEdge, "expecting a field edge"); int fi = an.edge_target(fe); PointsToNode pf = _nodes->at_grow(fi); int po = pf.offset(); if (pf.edge_count() == 0) { // we have not seen any stores to this field, assume it was set outside this method add_pointsto_edge(fi, _phantom_object); } if (po == offs || po == Type::OffsetBot || offs == Type::OffsetBot) { add_deferred_edge(from_i, fi); } } } // // Search memory chain of "mem" to find a MemNode whose address // is the specified alias index. Returns the MemNode found or the // first non-MemNode encountered. // Node *ConnectionGraph::find_mem(Node *mem, int alias_idx, PhaseGVN *igvn) { if (mem == NULL) return mem; while (mem->is_Mem()) { const Type *at = igvn->type(mem->in(MemNode::Address)); if (at != Type::TOP) { assert (at->isa_ptr() != NULL, "pointer type required."); int idx = _compile->get_alias_index(at->is_ptr()); if (idx == alias_idx) break; } mem = mem->in(MemNode::Memory); } return mem; } // // Adjust the type and inputs of an AddP which computes the // address of a field of an instance // void ConnectionGraph::split_AddP(Node *addp, Node *base, PhaseGVN *igvn) { const TypeOopPtr *t = igvn->type(addp)->isa_oopptr(); const TypeOopPtr *base_t = igvn->type(base)->isa_oopptr(); assert(t != NULL, "expecting oopptr"); assert(base_t != NULL && base_t->is_instance(), "expecting instance oopptr"); uint inst_id = base_t->instance_id(); assert(!t->is_instance() || t->instance_id() == inst_id, "old type must be non-instance or match new type"); const TypeOopPtr *tinst = base_t->add_offset(t->offset())->is_oopptr(); // ensure an alias index is allocated for the instance type int alias_idx = _compile->get_alias_index(tinst); igvn->set_type(addp, tinst); // record the allocation in the node map set_map(addp->_idx, get_map(base->_idx)); // if the Address input is not the appropriate instance type (due to intervening // casts,) insert a cast Node *adr = addp->in(AddPNode::Address); const TypeOopPtr *atype = igvn->type(adr)->isa_oopptr(); if (atype->instance_id() != inst_id) { assert(!atype->is_instance(), "no conflicting instances"); const TypeOopPtr *new_atype = base_t->add_offset(atype->offset())->isa_oopptr(); Node *acast = new (_compile, 2) CastPPNode(adr, new_atype); acast->set_req(0, adr->in(0)); igvn->set_type(acast, new_atype); record_for_optimizer(acast); Node *bcast = acast; Node *abase = addp->in(AddPNode::Base); if (abase != adr) { bcast = new (_compile, 2) CastPPNode(abase, base_t); bcast->set_req(0, abase->in(0)); igvn->set_type(bcast, base_t); record_for_optimizer(bcast); } igvn->hash_delete(addp); addp->set_req(AddPNode::Base, bcast); addp->set_req(AddPNode::Address, acast); igvn->hash_insert(addp); record_for_optimizer(addp); } } // // Create a new version of orig_phi if necessary. Returns either the newly // created phi or an existing phi. Sets create_new to indicate wheter a new // phi was created. Cache the last newly created phi in the node map. // PhiNode *ConnectionGraph::create_split_phi(PhiNode *orig_phi, int alias_idx, GrowableArray<PhiNode *> &orig_phi_worklist, PhaseGVN *igvn, bool &new_created) { Compile *C = _compile; new_created = false; int phi_alias_idx = C->get_alias_index(orig_phi->adr_type()); // nothing to do if orig_phi is bottom memory or matches alias_idx if (phi_alias_idx == Compile::AliasIdxBot || phi_alias_idx == alias_idx) { return orig_phi; } // have we already created a Phi for this alias index? PhiNode *result = get_map_phi(orig_phi->_idx); const TypePtr *atype = C->get_adr_type(alias_idx); if (result != NULL && C->get_alias_index(result->adr_type()) == alias_idx) { return result; } orig_phi_worklist.append_if_missing(orig_phi); result = PhiNode::make(orig_phi->in(0), NULL, Type::MEMORY, atype); set_map_phi(orig_phi->_idx, result); igvn->set_type(result, result->bottom_type()); record_for_optimizer(result); new_created = true; return result; } // // Return a new version of Memory Phi "orig_phi" with the inputs having the // specified alias index. // PhiNode *ConnectionGraph::split_memory_phi(PhiNode *orig_phi, int alias_idx, GrowableArray<PhiNode *> &orig_phi_worklist, PhaseGVN *igvn) { assert(alias_idx != Compile::AliasIdxBot, "can't split out bottom memory"); Compile *C = _compile; bool new_phi_created; PhiNode *result = create_split_phi(orig_phi, alias_idx, orig_phi_worklist, igvn, new_phi_created); if (!new_phi_created) { return result; } GrowableArray<PhiNode *> phi_list; GrowableArray<uint> cur_input; PhiNode *phi = orig_phi; uint idx = 1; bool finished = false; while(!finished) { while (idx < phi->req()) { Node *mem = find_mem(phi->in(idx), alias_idx, igvn); if (mem != NULL && mem->is_Phi()) { PhiNode *nphi = create_split_phi(mem->as_Phi(), alias_idx, orig_phi_worklist, igvn, new_phi_created); if (new_phi_created) { // found an phi for which we created a new split, push current one on worklist and begin // processing new one phi_list.push(phi); cur_input.push(idx); phi = mem->as_Phi(); result = nphi; idx = 1; continue; } else { mem = nphi; } } result->set_req(idx++, mem); } #ifdef ASSERT // verify that the new Phi has an input for each input of the original assert( phi->req() == result->req(), "must have same number of inputs."); assert( result->in(0) != NULL && result->in(0) == phi->in(0), "regions must match"); for (uint i = 1; i < phi->req(); i++) { assert((phi->in(i) == NULL) == (result->in(i) == NULL), "inputs must correspond."); } #endif // we have finished processing a Phi, see if there are any more to do finished = (phi_list.length() == 0 ); if (!finished) { phi = phi_list.pop(); idx = cur_input.pop(); PhiNode *prev_phi = get_map_phi(phi->_idx); prev_phi->set_req(idx++, result); result = prev_phi; } } return result; } // // Convert the types of unescaped object to instance types where possible, // propagate the new type information through the graph, and update memory // edges and MergeMem inputs to reflect the new type. // // We start with allocations (and calls which may be allocations) on alloc_worklist. // The processing is done in 4 phases: // // Phase 1: Process possible allocations from alloc_worklist. Create instance // types for the CheckCastPP for allocations where possible. // Propagate the the new types through users as follows: // casts and Phi: push users on alloc_worklist // AddP: cast Base and Address inputs to the instance type // push any AddP users on alloc_worklist and push any memnode // users onto memnode_worklist. // Phase 2: Process MemNode's from memnode_worklist. compute new address type and // search the Memory chain for a store with the appropriate type // address type. If a Phi is found, create a new version with // the approriate memory slices from each of the Phi inputs. // For stores, process the users as follows: // MemNode: push on memnode_worklist // MergeMem: push on mergemem_worklist // Phase 3: Process MergeMem nodes from mergemem_worklist. Walk each memory slice // moving the first node encountered of each instance type to the // the input corresponding to its alias index. // appropriate memory slice. // Phase 4: Update the inputs of non-instance memory Phis and the Memory input of memnodes. // // In the following example, the CheckCastPP nodes are the cast of allocation // results and the allocation of node 29 is unescaped and eligible to be an // instance type. // // We start with: // // 7 Parm #memory // 10 ConI "12" // 19 CheckCastPP "Foo" // 20 AddP _ 19 19 10 Foo+12 alias_index=4 // 29 CheckCastPP "Foo" // 30 AddP _ 29 29 10 Foo+12 alias_index=4 // // 40 StoreP 25 7 20 ... alias_index=4 // 50 StoreP 35 40 30 ... alias_index=4 // 60 StoreP 45 50 20 ... alias_index=4 // 70 LoadP _ 60 30 ... alias_index=4 // 80 Phi 75 50 60 Memory alias_index=4 // 90 LoadP _ 80 30 ... alias_index=4 // 100 LoadP _ 80 20 ... alias_index=4 // // // Phase 1 creates an instance type for node 29 assigning it an instance id of 24 // and creating a new alias index for node 30. This gives: // // 7 Parm #memory // 10 ConI "12" // 19 CheckCastPP "Foo" // 20 AddP _ 19 19 10 Foo+12 alias_index=4 // 29 CheckCastPP "Foo" iid=24 // 30 AddP _ 29 29 10 Foo+12 alias_index=6 iid=24 // // 40 StoreP 25 7 20 ... alias_index=4 // 50 StoreP 35 40 30 ... alias_index=6 // 60 StoreP 45 50 20 ... alias_index=4 // 70 LoadP _ 60 30 ... alias_index=6 // 80 Phi 75 50 60 Memory alias_index=4 // 90 LoadP _ 80 30 ... alias_index=6 // 100 LoadP _ 80 20 ... alias_index=4 // // In phase 2, new memory inputs are computed for the loads and stores, // And a new version of the phi is created. In phase 4, the inputs to // node 80 are updated and then the memory nodes are updated with the // values computed in phase 2. This results in: // // 7 Parm #memory // 10 ConI "12" // 19 CheckCastPP "Foo" // 20 AddP _ 19 19 10 Foo+12 alias_index=4 // 29 CheckCastPP "Foo" iid=24 // 30 AddP _ 29 29 10 Foo+12 alias_index=6 iid=24 // // 40 StoreP 25 7 20 ... alias_index=4 // 50 StoreP 35 7 30 ... alias_index=6 // 60 StoreP 45 40 20 ... alias_index=4 // 70 LoadP _ 50 30 ... alias_index=6 // 80 Phi 75 40 60 Memory alias_index=4 // 120 Phi 75 50 50 Memory alias_index=6 // 90 LoadP _ 120 30 ... alias_index=6 // 100 LoadP _ 80 20 ... alias_index=4 // void ConnectionGraph::split_unique_types(GrowableArray<Node *> &alloc_worklist) { GrowableArray<Node *> memnode_worklist; GrowableArray<Node *> mergemem_worklist; GrowableArray<PhiNode *> orig_phis; PhaseGVN *igvn = _compile->initial_gvn(); uint new_index_start = (uint) _compile->num_alias_types(); VectorSet visited(Thread::current()->resource_area()); VectorSet ptset(Thread::current()->resource_area()); // Phase 1: Process possible allocations from alloc_worklist. Create instance // types for the CheckCastPP for allocations where possible. while (alloc_worklist.length() != 0) { Node *n = alloc_worklist.pop(); uint ni = n->_idx; if (n->is_Call()) { CallNode *alloc = n->as_Call(); // copy escape information to call node PointsToNode ptn = _nodes->at(alloc->_idx); PointsToNode::EscapeState es = escape_state(alloc, igvn); alloc->_escape_state = es; // find CheckCastPP of call return value n = alloc->proj_out(TypeFunc::Parms); if (n != NULL && n->outcnt() == 1) { n = n->unique_out(); if (n->Opcode() != Op_CheckCastPP) { continue; } } else { continue; } // we have an allocation or call which returns a Java object, see if it is unescaped if (es != PointsToNode::NoEscape || !ptn._unique_type) { continue; // can't make a unique type } set_map(alloc->_idx, n); set_map(n->_idx, alloc); const TypeInstPtr *t = igvn->type(n)->isa_instptr(); // Unique types which are arrays are not currently supported. // The check for AllocateArray is needed in case an array // allocation is immediately cast to Object if (t == NULL || alloc->is_AllocateArray()) continue; // not a TypeInstPtr const TypeOopPtr *tinst = t->cast_to_instance(ni); igvn->hash_delete(n); igvn->set_type(n, tinst); n->raise_bottom_type(tinst); igvn->hash_insert(n); } else if (n->is_AddP()) { ptset.Clear(); PointsTo(ptset, n->in(AddPNode::Address), igvn); assert(ptset.Size() == 1, "AddP address is unique"); Node *base = get_map(ptset.getelem()); split_AddP(n, base, igvn); } else if (n->is_Phi() || n->Opcode() == Op_CastPP || n->Opcode() == Op_CheckCastPP) { if (visited.test_set(n->_idx)) { assert(n->is_Phi(), "loops only through Phi's"); continue; // already processed } ptset.Clear(); PointsTo(ptset, n, igvn); if (ptset.Size() == 1) { TypeNode *tn = n->as_Type(); Node *val = get_map(ptset.getelem()); const TypeInstPtr *val_t = igvn->type(val)->isa_instptr();; assert(val_t != NULL && val_t->is_instance(), "instance type expected."); const TypeInstPtr *tn_t = igvn->type(tn)->isa_instptr();; if (tn_t != NULL && val_t->cast_to_instance(TypeOopPtr::UNKNOWN_INSTANCE)->higher_equal(tn_t)) { igvn->hash_delete(tn); igvn->set_type(tn, val_t); tn->set_type(val_t); igvn->hash_insert(tn); } } } else { continue; } // push users on appropriate worklist for (DUIterator_Fast imax, i = n->fast_outs(imax); i < imax; i++) { Node *use = n->fast_out(i); if(use->is_Mem() && use->in(MemNode::Address) == n) { memnode_worklist.push(use); } else if (use->is_AddP() || use->is_Phi() || use->Opcode() == Op_CastPP || use->Opcode() == Op_CheckCastPP) { alloc_worklist.push(use); } } } uint new_index_end = (uint) _compile->num_alias_types(); // Phase 2: Process MemNode's from memnode_worklist. compute new address type and // compute new values for Memory inputs (the Memory inputs are not // actually updated until phase 4.) if (memnode_worklist.length() == 0) return; // nothing to do while (memnode_worklist.length() != 0) { Node *n = memnode_worklist.pop(); if (n->is_Phi()) { assert(n->as_Phi()->adr_type() != TypePtr::BOTTOM, "narrow memory slice required"); // we don't need to do anything, but the users must be pushed if we haven't processed // this Phi before if (visited.test_set(n->_idx)) continue; } else { assert(n->is_Mem(), "memory node required."); Node *addr = n->in(MemNode::Address); const Type *addr_t = igvn->type(addr); if (addr_t == Type::TOP) continue; assert (addr_t->isa_ptr() != NULL, "pointer type required."); int alias_idx = _compile->get_alias_index(addr_t->is_ptr()); Node *mem = find_mem(n->in(MemNode::Memory), alias_idx, igvn); if (mem->is_Phi()) { mem = split_memory_phi(mem->as_Phi(), alias_idx, orig_phis, igvn); } if (mem != n->in(MemNode::Memory)) set_map(n->_idx, mem); if (n->is_Load()) { continue; // don't push users } else if (n->is_LoadStore()) { // get the memory projection for (DUIterator_Fast imax, i = n->fast_outs(imax); i < imax; i++) { Node *use = n->fast_out(i); if (use->Opcode() == Op_SCMemProj) { n = use; break; } } assert(n->Opcode() == Op_SCMemProj, "memory projection required"); } } // push user on appropriate worklist for (DUIterator_Fast imax, i = n->fast_outs(imax); i < imax; i++) { Node *use = n->fast_out(i); if (use->is_Phi()) { memnode_worklist.push(use); } else if(use->is_Mem() && use->in(MemNode::Memory) == n) { memnode_worklist.push(use); } else if (use->is_MergeMem()) { mergemem_worklist.push(use); } } } // Phase 3: Process MergeMem nodes from mergemem_worklist. Walk each memory slice // moving the first node encountered of each instance type to the // the input corresponding to its alias index. while (mergemem_worklist.length() != 0) { Node *n = mergemem_worklist.pop(); assert(n->is_MergeMem(), "MergeMem node required."); MergeMemNode *nmm = n->as_MergeMem(); // Note: we don't want to use MergeMemStream here because we only want to // scan inputs which exist at the start, not ones we add during processing uint nslices = nmm->req(); igvn->hash_delete(nmm); for (uint i = Compile::AliasIdxRaw+1; i < nslices; i++) { Node * mem = nmm->in(i); Node * cur = NULL; if (mem == NULL || mem->is_top()) continue; while (mem->is_Mem()) { const Type *at = igvn->type(mem->in(MemNode::Address)); if (at != Type::TOP) { assert (at->isa_ptr() != NULL, "pointer type required."); uint idx = (uint)_compile->get_alias_index(at->is_ptr()); if (idx == i) { if (cur == NULL) cur = mem; } else { if (idx >= nmm->req() || nmm->is_empty_memory(nmm->in(idx))) { nmm->set_memory_at(idx, mem); } } } mem = mem->in(MemNode::Memory); } nmm->set_memory_at(i, (cur != NULL) ? cur : mem); if (mem->is_Phi()) { // We have encountered a Phi, we need to split the Phi for // any instance of the current type if we haven't encountered // a value of the instance along the chain. for (uint ni = new_index_start; ni < new_index_end; ni++) { if((uint)_compile->get_general_index(ni) == i) { Node *m = (ni >= nmm->req()) ? nmm->empty_memory() : nmm->in(ni); if (nmm->is_empty_memory(m)) { nmm->set_memory_at(ni, split_memory_phi(mem->as_Phi(), ni, orig_phis, igvn)); } } } } } igvn->hash_insert(nmm); record_for_optimizer(nmm); } // Phase 4: Update the inputs of non-instance memory Phis and the Memory input of memnodes // // First update the inputs of any non-instance Phi's from // which we split out an instance Phi. Note we don't have // to recursively process Phi's encounted on the input memory // chains as is done in split_memory_phi() since they will // also be processed here. while (orig_phis.length() != 0) { PhiNode *phi = orig_phis.pop(); int alias_idx = _compile->get_alias_index(phi->adr_type()); igvn->hash_delete(phi); for (uint i = 1; i < phi->req(); i++) { Node *mem = phi->in(i); Node *new_mem = find_mem(mem, alias_idx, igvn); if (mem != new_mem) { phi->set_req(i, new_mem); } } igvn->hash_insert(phi); record_for_optimizer(phi); } // Update the memory inputs of MemNodes with the value we computed // in Phase 2. for (int i = 0; i < _nodes->length(); i++) { Node *nmem = get_map(i); if (nmem != NULL) { Node *n = _nodes->at(i)._node; if (n != NULL && n->is_Mem()) { igvn->hash_delete(n); n->set_req(MemNode::Memory, nmem); igvn->hash_insert(n); record_for_optimizer(n); } } } } void ConnectionGraph::compute_escape() { GrowableArray<int> worklist; GrowableArray<Node *> alloc_worklist; VectorSet visited(Thread::current()->resource_area()); PhaseGVN *igvn = _compile->initial_gvn(); // process Phi nodes from the deferred list, they may not have while(_deferred.size() > 0) { Node * n = _deferred.pop(); PhiNode * phi = n->as_Phi(); process_phi_escape(phi, igvn); } VectorSet ptset(Thread::current()->resource_area()); // remove deferred edges from the graph and collect // information we will need for type splitting for (uint ni = 0; ni < (uint)_nodes->length(); ni++) { PointsToNode * ptn = _nodes->adr_at(ni); PointsToNode::NodeType nt = ptn->node_type(); if (nt == PointsToNode::UnknownType) { continue; // not a node we are interested in } Node *n = ptn->_node; if (nt == PointsToNode::LocalVar || nt == PointsToNode::Field) { remove_deferred(ni); if (n->is_AddP()) { // if this AddP computes an address which may point to more that one // object, nothing the address points to can be a unique type. Node *base = n->in(AddPNode::Base); ptset.Clear(); PointsTo(ptset, base, igvn); if (ptset.Size() > 1) { for( VectorSetI j(&ptset); j.test(); ++j ) { PointsToNode *ptaddr = _nodes->adr_at(j.elem); ptaddr->_unique_type = false; } } } } else if (n->is_Call()) { // initialize _escape_state of calls to GlobalEscape n->as_Call()->_escape_state = PointsToNode::GlobalEscape; // push call on alloc_worlist (alocations are calls) // for processing by split_unique_types() alloc_worklist.push(n); } } // push all GlobalEscape nodes on the worklist for (uint nj = 0; nj < (uint)_nodes->length(); nj++) { if (_nodes->at(nj).escape_state() == PointsToNode::GlobalEscape) { worklist.append(nj); } } // mark all node reachable from GlobalEscape nodes while(worklist.length() > 0) { PointsToNode n = _nodes->at(worklist.pop()); for (uint ei = 0; ei < n.edge_count(); ei++) { uint npi = n.edge_target(ei); PointsToNode *np = ptnode_adr(npi); if (np->escape_state() != PointsToNode::GlobalEscape) { np->set_escape_state(PointsToNode::GlobalEscape); worklist.append_if_missing(npi); } } } // push all ArgEscape nodes on the worklist for (uint nk = 0; nk < (uint)_nodes->length(); nk++) { if (_nodes->at(nk).escape_state() == PointsToNode::ArgEscape) worklist.push(nk); } // mark all node reachable from ArgEscape nodes while(worklist.length() > 0) { PointsToNode n = _nodes->at(worklist.pop()); for (uint ei = 0; ei < n.edge_count(); ei++) { uint npi = n.edge_target(ei); PointsToNode *np = ptnode_adr(npi); if (np->escape_state() != PointsToNode::ArgEscape) { np->set_escape_state(PointsToNode::ArgEscape); worklist.append_if_missing(npi); } } } _collecting = false; // Now use the escape information to create unique types for // unescaped objects split_unique_types(alloc_worklist); } Node * ConnectionGraph::skip_casts(Node *n) { while(n->Opcode() == Op_CastPP || n->Opcode() == Op_CheckCastPP) { n = n->in(1); } return n; } void ConnectionGraph::process_phi_escape(PhiNode *phi, PhaseTransform *phase) { if (phi->type()->isa_oopptr() == NULL) return; // nothing to do if not an oop PointsToNode *ptadr = ptnode_adr(phi->_idx); int incount = phi->req(); int non_null_inputs = 0; for (int i = 1; i < incount ; i++) { if (phi->in(i) != NULL) non_null_inputs++; } if (non_null_inputs == ptadr->_inputs_processed) return; // no new inputs since the last time this node was processed, // the current information is valid ptadr->_inputs_processed = non_null_inputs; // prevent recursive processing of this node for (int j = 1; j < incount ; j++) { Node * n = phi->in(j); if (n == NULL) continue; // ignore NULL n = skip_casts(n); if (n->is_top() || n == phi) continue; // ignore top or inputs which go back this node int nopc = n->Opcode(); PointsToNode npt = _nodes->at(n->_idx); if (_nodes->at(n->_idx).node_type() == PointsToNode::JavaObject) { add_pointsto_edge(phi->_idx, n->_idx); } else { add_deferred_edge(phi->_idx, n->_idx); } } } void ConnectionGraph::process_call_arguments(CallNode *call, PhaseTransform *phase) { _processed.set(call->_idx); switch (call->Opcode()) { // arguments to allocation and locking don't escape case Op_Allocate: case Op_AllocateArray: case Op_Lock: case Op_Unlock: break; case Op_CallStaticJava: // For a static call, we know exactly what method is being called. // Use bytecode estimator to record the call's escape affects { ciMethod *meth = call->as_CallJava()->method(); if (meth != NULL) { const TypeTuple * d = call->tf()->domain(); BCEscapeAnalyzer call_analyzer(meth); VectorSet ptset(Thread::current()->resource_area()); for (uint i = TypeFunc::Parms; i < d->cnt(); i++) { const Type* at = d->field_at(i); int k = i - TypeFunc::Parms; if (at->isa_oopptr() != NULL) { Node *arg = skip_casts(call->in(i)); if (!call_analyzer.is_arg_stack(k)) { // The argument global escapes, mark everything it could point to ptset.Clear(); PointsTo(ptset, arg, phase); for( VectorSetI j(&ptset); j.test(); ++j ) { uint pt = j.elem; set_escape_state(pt, PointsToNode::GlobalEscape); } } else if (!call_analyzer.is_arg_local(k)) { // The argument itself doesn't escape, but any fields might ptset.Clear(); PointsTo(ptset, arg, phase); for( VectorSetI j(&ptset); j.test(); ++j ) { uint pt = j.elem; add_edge_from_fields(pt, _phantom_object, Type::OffsetBot); } } } } call_analyzer.copy_dependencies(C()->dependencies()); break; } // fall-through if not a Java method } default: // Some other type of call, assume the worst case: all arguments // globally escape. { // adjust escape state for outgoing arguments const TypeTuple * d = call->tf()->domain(); VectorSet ptset(Thread::current()->resource_area()); for (uint i = TypeFunc::Parms; i < d->cnt(); i++) { const Type* at = d->field_at(i); if (at->isa_oopptr() != NULL) { Node *arg = skip_casts(call->in(i)); ptset.Clear(); PointsTo(ptset, arg, phase); for( VectorSetI j(&ptset); j.test(); ++j ) { uint pt = j.elem; set_escape_state(pt, PointsToNode::GlobalEscape); } } } } } } void ConnectionGraph::process_call_result(ProjNode *resproj, PhaseTransform *phase) { CallNode *call = resproj->in(0)->as_Call(); PointsToNode *ptadr = ptnode_adr(resproj->_idx); ptadr->_node = resproj; ptadr->set_node_type(PointsToNode::LocalVar); set_escape_state(resproj->_idx, PointsToNode::UnknownEscape); _processed.set(resproj->_idx); switch (call->Opcode()) { case Op_Allocate: { Node *k = call->in(AllocateNode::KlassNode); const TypeKlassPtr *kt; if (k->Opcode() == Op_LoadKlass) { kt = k->as_Load()->type()->isa_klassptr(); } else { kt = k->as_Type()->type()->isa_klassptr(); } assert(kt != NULL, "TypeKlassPtr required."); ciKlass* cik = kt->klass(); ciInstanceKlass* ciik = cik->as_instance_klass(); PointsToNode *ptadr = ptnode_adr(call->_idx); ptadr->set_node_type(PointsToNode::JavaObject); if (cik->is_subclass_of(_compile->env()->Thread_klass()) || ciik->has_finalizer()) { set_escape_state(call->_idx, PointsToNode::GlobalEscape); add_pointsto_edge(resproj->_idx, _phantom_object); } else { set_escape_state(call->_idx, PointsToNode::NoEscape); add_pointsto_edge(resproj->_idx, call->_idx); } _processed.set(call->_idx); break; } case Op_AllocateArray: { PointsToNode *ptadr = ptnode_adr(call->_idx); ptadr->set_node_type(PointsToNode::JavaObject); set_escape_state(call->_idx, PointsToNode::NoEscape); _processed.set(call->_idx); add_pointsto_edge(resproj->_idx, call->_idx); break; } case Op_Lock: case Op_Unlock: break; case Op_CallStaticJava: // For a static call, we know exactly what method is being called. // Use bytecode estimator to record whether the call's return value escapes { const TypeTuple *r = call->tf()->range(); const Type* ret_type = NULL; if (r->cnt() > TypeFunc::Parms) ret_type = r->field_at(TypeFunc::Parms); // Note: we use isa_ptr() instead of isa_oopptr() here because the // _multianewarray functions return a TypeRawPtr. if (ret_type == NULL || ret_type->isa_ptr() == NULL) break; // doesn't return a pointer type ciMethod *meth = call->as_CallJava()->method(); if (meth == NULL) { // not a Java method, assume global escape set_escape_state(call->_idx, PointsToNode::GlobalEscape); if (resproj != NULL) add_pointsto_edge(resproj->_idx, _phantom_object); } else { BCEscapeAnalyzer call_analyzer(meth); VectorSet ptset(Thread::current()->resource_area()); if (call_analyzer.is_return_local() && resproj != NULL) { // determine whether any arguments are returned const TypeTuple * d = call->tf()->domain(); set_escape_state(call->_idx, PointsToNode::NoEscape); for (uint i = TypeFunc::Parms; i < d->cnt(); i++) { const Type* at = d->field_at(i); if (at->isa_oopptr() != NULL) { Node *arg = skip_casts(call->in(i)); if (call_analyzer.is_arg_returned(i - TypeFunc::Parms)) { PointsToNode *arg_esp = _nodes->adr_at(arg->_idx); if (arg_esp->node_type() == PointsToNode::JavaObject) add_pointsto_edge(resproj->_idx, arg->_idx); else add_deferred_edge(resproj->_idx, arg->_idx); arg_esp->_hidden_alias = true; } } } } else { set_escape_state(call->_idx, PointsToNode::GlobalEscape); if (resproj != NULL) add_pointsto_edge(resproj->_idx, _phantom_object); } call_analyzer.copy_dependencies(C()->dependencies()); } break; } default: // Some other type of call, assume the worst case that the // returned value, if any, globally escapes. { const TypeTuple *r = call->tf()->range(); if (r->cnt() > TypeFunc::Parms) { const Type* ret_type = r->field_at(TypeFunc::Parms); // Note: we use isa_ptr() instead of isa_oopptr() here because the // _multianewarray functions return a TypeRawPtr. if (ret_type->isa_ptr() != NULL) { PointsToNode *ptadr = ptnode_adr(call->_idx); ptadr->set_node_type(PointsToNode::JavaObject); set_escape_state(call->_idx, PointsToNode::GlobalEscape); if (resproj != NULL) add_pointsto_edge(resproj->_idx, _phantom_object); } } } } } void ConnectionGraph::record_for_escape_analysis(Node *n) { if (_collecting) { if (n->is_Phi()) { PhiNode *phi = n->as_Phi(); const Type *pt = phi->type(); if ((pt->isa_oopptr() != NULL) || pt == TypePtr::NULL_PTR) { PointsToNode *ptn = ptnode_adr(phi->_idx); ptn->set_node_type(PointsToNode::LocalVar); ptn->_node = n; _deferred.push(n); } } } } void ConnectionGraph::record_escape_work(Node *n, PhaseTransform *phase) { int opc = n->Opcode(); PointsToNode *ptadr = ptnode_adr(n->_idx); if (_processed.test(n->_idx)) return; ptadr->_node = n; if (n->is_Call()) { CallNode *call = n->as_Call(); process_call_arguments(call, phase); return; } switch (opc) { case Op_AddP: { Node *base = skip_casts(n->in(AddPNode::Base)); ptadr->set_node_type(PointsToNode::Field); // create a field edge to this node from everything adr could point to VectorSet ptset(Thread::current()->resource_area()); PointsTo(ptset, base, phase); for( VectorSetI i(&ptset); i.test(); ++i ) { uint pt = i.elem; add_field_edge(pt, n->_idx, type_to_offset(phase->type(n))); } break; } case Op_Parm: { ProjNode *nproj = n->as_Proj(); uint con = nproj->_con; if (con < TypeFunc::Parms) return; const Type *t = nproj->in(0)->as_Start()->_domain->field_at(con); if (t->isa_ptr() == NULL) return; ptadr->set_node_type(PointsToNode::JavaObject); if (t->isa_oopptr() != NULL) { set_escape_state(n->_idx, PointsToNode::ArgEscape); } else { // this must be the incoming state of an OSR compile, we have to assume anything // passed in globally escapes assert(_compile->is_osr_compilation(), "bad argument type for non-osr compilation"); set_escape_state(n->_idx, PointsToNode::GlobalEscape); } _processed.set(n->_idx); break; } case Op_Phi: { PhiNode *phi = n->as_Phi(); if (phi->type()->isa_oopptr() == NULL) return; // nothing to do if not an oop ptadr->set_node_type(PointsToNode::LocalVar); process_phi_escape(phi, phase); break; } case Op_CreateEx: { // assume that all exception objects globally escape ptadr->set_node_type(PointsToNode::JavaObject); set_escape_state(n->_idx, PointsToNode::GlobalEscape); _processed.set(n->_idx); break; } case Op_ConP: { const Type *t = phase->type(n); ptadr->set_node_type(PointsToNode::JavaObject); // assume all pointer constants globally escape except for null if (t == TypePtr::NULL_PTR) set_escape_state(n->_idx, PointsToNode::NoEscape); else set_escape_state(n->_idx, PointsToNode::GlobalEscape); _processed.set(n->_idx); break; } case Op_LoadKlass: { ptadr->set_node_type(PointsToNode::JavaObject); set_escape_state(n->_idx, PointsToNode::GlobalEscape); _processed.set(n->_idx); break; } case Op_LoadP: { const Type *t = phase->type(n); if (!t->isa_oopptr()) return; ptadr->set_node_type(PointsToNode::LocalVar); set_escape_state(n->_idx, PointsToNode::UnknownEscape); Node *adr = skip_casts(n->in(MemNode::Address)); const Type *adr_type = phase->type(adr); Node *adr_base = skip_casts((adr->Opcode() == Op_AddP) ? adr->in(AddPNode::Base) : adr); // For everything "adr" could point to, create a deferred edge from // this node to each field with the same offset as "adr_type" VectorSet ptset(Thread::current()->resource_area()); PointsTo(ptset, adr_base, phase); // If ptset is empty, then this value must have been set outside // this method, so we add the phantom node if (ptset.Size() == 0) ptset.set(_phantom_object); for( VectorSetI i(&ptset); i.test(); ++i ) { uint pt = i.elem; add_deferred_edge_to_fields(n->_idx, pt, type_to_offset(adr_type)); } break; } case Op_StoreP: case Op_StorePConditional: case Op_CompareAndSwapP: { Node *adr = n->in(MemNode::Address); Node *val = skip_casts(n->in(MemNode::ValueIn)); const Type *adr_type = phase->type(adr); if (!adr_type->isa_oopptr()) return; assert(adr->Opcode() == Op_AddP, "expecting an AddP"); Node *adr_base = adr->in(AddPNode::Base); // For everything "adr_base" could point to, create a deferred edge to "val" from each field // with the same offset as "adr_type" VectorSet ptset(Thread::current()->resource_area()); PointsTo(ptset, adr_base, phase); for( VectorSetI i(&ptset); i.test(); ++i ) { uint pt = i.elem; add_edge_from_fields(pt, val->_idx, type_to_offset(adr_type)); } break; } case Op_Proj: { ProjNode *nproj = n->as_Proj(); Node *n0 = nproj->in(0); // we are only interested in the result projection from a call if (nproj->_con == TypeFunc::Parms && n0->is_Call() ) { process_call_result(nproj, phase); } break; } case Op_CastPP: case Op_CheckCastPP: { ptadr->set_node_type(PointsToNode::LocalVar); int ti = n->in(1)->_idx; if (_nodes->at(ti).node_type() == PointsToNode::JavaObject) { add_pointsto_edge(n->_idx, ti); } else { add_deferred_edge(n->_idx, ti); } break; } default: ; // nothing to do } } void ConnectionGraph::record_escape(Node *n, PhaseTransform *phase) { if (_collecting) record_escape_work(n, phase); } #ifndef PRODUCT void ConnectionGraph::dump() { PhaseGVN *igvn = _compile->initial_gvn(); bool first = true; for (uint ni = 0; ni < (uint)_nodes->length(); ni++) { PointsToNode *esp = _nodes->adr_at(ni); if (esp->node_type() == PointsToNode::UnknownType || esp->_node == NULL) continue; PointsToNode::EscapeState es = escape_state(esp->_node, igvn); if (es == PointsToNode::NoEscape || (Verbose && (es != PointsToNode::UnknownEscape || esp->edge_count() != 0))) { // don't print null pointer node which almost every method has if (esp->_node->Opcode() != Op_ConP || igvn->type(esp->_node) != TypePtr::NULL_PTR) { if (first) { tty->print("======== Connection graph for "); C()->method()->print_short_name(); tty->cr(); first = false; } tty->print("%4d ", ni); esp->dump(); } } } } #endif