8191998: C2: inlining through MH linkers drops speculative part of argument types
Reviewed-by: roland, thartmann, kvn
/*
* Copyright (c) 2000, 2019, Oracle and/or its affiliates. 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 Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
* or visit www.oracle.com if you need additional information or have any
* questions.
*
*/
#include "precompiled.hpp"
#include "ci/bcEscapeAnalyzer.hpp"
#include "ci/ciCallSite.hpp"
#include "ci/ciObjArray.hpp"
#include "ci/ciMemberName.hpp"
#include "ci/ciMethodHandle.hpp"
#include "classfile/javaClasses.hpp"
#include "compiler/compileLog.hpp"
#include "opto/addnode.hpp"
#include "opto/callGenerator.hpp"
#include "opto/callnode.hpp"
#include "opto/castnode.hpp"
#include "opto/cfgnode.hpp"
#include "opto/parse.hpp"
#include "opto/rootnode.hpp"
#include "opto/runtime.hpp"
#include "opto/subnode.hpp"
#include "runtime/sharedRuntime.hpp"
// Utility function.
const TypeFunc* CallGenerator::tf() const {
return TypeFunc::make(method());
}
bool CallGenerator::is_inlined_method_handle_intrinsic(JVMState* jvms, ciMethod* m) {
return is_inlined_method_handle_intrinsic(jvms->method(), jvms->bci(), m);
}
bool CallGenerator::is_inlined_method_handle_intrinsic(ciMethod* caller, int bci, ciMethod* m) {
ciMethod* symbolic_info = caller->get_method_at_bci(bci);
return is_inlined_method_handle_intrinsic(symbolic_info, m);
}
bool CallGenerator::is_inlined_method_handle_intrinsic(ciMethod* symbolic_info, ciMethod* m) {
return symbolic_info->is_method_handle_intrinsic() && !m->is_method_handle_intrinsic();
}
//-----------------------------ParseGenerator---------------------------------
// Internal class which handles all direct bytecode traversal.
class ParseGenerator : public InlineCallGenerator {
private:
bool _is_osr;
float _expected_uses;
public:
ParseGenerator(ciMethod* method, float expected_uses, bool is_osr = false)
: InlineCallGenerator(method)
{
_is_osr = is_osr;
_expected_uses = expected_uses;
assert(InlineTree::check_can_parse(method) == NULL, "parse must be possible");
}
virtual bool is_parse() const { return true; }
virtual JVMState* generate(JVMState* jvms);
int is_osr() { return _is_osr; }
};
JVMState* ParseGenerator::generate(JVMState* jvms) {
Compile* C = Compile::current();
C->print_inlining_update(this);
if (is_osr()) {
// The JVMS for a OSR has a single argument (see its TypeFunc).
assert(jvms->depth() == 1, "no inline OSR");
}
if (C->failing()) {
return NULL; // bailing out of the compile; do not try to parse
}
Parse parser(jvms, method(), _expected_uses);
// Grab signature for matching/allocation
#ifdef ASSERT
if (parser.tf() != (parser.depth() == 1 ? C->tf() : tf())) {
assert(C->env()->system_dictionary_modification_counter_changed(),
"Must invalidate if TypeFuncs differ");
}
#endif
GraphKit& exits = parser.exits();
if (C->failing()) {
while (exits.pop_exception_state() != NULL) ;
return NULL;
}
assert(exits.jvms()->same_calls_as(jvms), "sanity");
// Simply return the exit state of the parser,
// augmented by any exceptional states.
return exits.transfer_exceptions_into_jvms();
}
//---------------------------DirectCallGenerator------------------------------
// Internal class which handles all out-of-line calls w/o receiver type checks.
class DirectCallGenerator : public CallGenerator {
private:
CallStaticJavaNode* _call_node;
// Force separate memory and I/O projections for the exceptional
// paths to facilitate late inlinig.
bool _separate_io_proj;
public:
DirectCallGenerator(ciMethod* method, bool separate_io_proj)
: CallGenerator(method),
_separate_io_proj(separate_io_proj)
{
}
virtual JVMState* generate(JVMState* jvms);
CallStaticJavaNode* call_node() const { return _call_node; }
};
JVMState* DirectCallGenerator::generate(JVMState* jvms) {
GraphKit kit(jvms);
kit.C->print_inlining_update(this);
bool is_static = method()->is_static();
address target = is_static ? SharedRuntime::get_resolve_static_call_stub()
: SharedRuntime::get_resolve_opt_virtual_call_stub();
if (kit.C->log() != NULL) {
kit.C->log()->elem("direct_call bci='%d'", jvms->bci());
}
CallStaticJavaNode *call = new CallStaticJavaNode(kit.C, tf(), target, method(), kit.bci());
if (is_inlined_method_handle_intrinsic(jvms, method())) {
// To be able to issue a direct call and skip a call to MH.linkTo*/invokeBasic adapter,
// additional information about the method being invoked should be attached
// to the call site to make resolution logic work
// (see SharedRuntime::resolve_static_call_C).
call->set_override_symbolic_info(true);
}
_call_node = call; // Save the call node in case we need it later
if (!is_static) {
// Make an explicit receiver null_check as part of this call.
// Since we share a map with the caller, his JVMS gets adjusted.
kit.null_check_receiver_before_call(method());
if (kit.stopped()) {
// And dump it back to the caller, decorated with any exceptions:
return kit.transfer_exceptions_into_jvms();
}
// Mark the call node as virtual, sort of:
call->set_optimized_virtual(true);
if (method()->is_method_handle_intrinsic() ||
method()->is_compiled_lambda_form()) {
call->set_method_handle_invoke(true);
}
}
kit.set_arguments_for_java_call(call);
kit.set_edges_for_java_call(call, false, _separate_io_proj);
Node* ret = kit.set_results_for_java_call(call, _separate_io_proj);
kit.push_node(method()->return_type()->basic_type(), ret);
return kit.transfer_exceptions_into_jvms();
}
//--------------------------VirtualCallGenerator------------------------------
// Internal class which handles all out-of-line calls checking receiver type.
class VirtualCallGenerator : public CallGenerator {
private:
int _vtable_index;
public:
VirtualCallGenerator(ciMethod* method, int vtable_index)
: CallGenerator(method), _vtable_index(vtable_index)
{
assert(vtable_index == Method::invalid_vtable_index ||
vtable_index >= 0, "either invalid or usable");
}
virtual bool is_virtual() const { return true; }
virtual JVMState* generate(JVMState* jvms);
};
JVMState* VirtualCallGenerator::generate(JVMState* jvms) {
GraphKit kit(jvms);
Node* receiver = kit.argument(0);
kit.C->print_inlining_update(this);
if (kit.C->log() != NULL) {
kit.C->log()->elem("virtual_call bci='%d'", jvms->bci());
}
// If the receiver is a constant null, do not torture the system
// by attempting to call through it. The compile will proceed
// correctly, but may bail out in final_graph_reshaping, because
// the call instruction will have a seemingly deficient out-count.
// (The bailout says something misleading about an "infinite loop".)
if (kit.gvn().type(receiver)->higher_equal(TypePtr::NULL_PTR)) {
assert(Bytecodes::is_invoke(kit.java_bc()), "%d: %s", kit.java_bc(), Bytecodes::name(kit.java_bc()));
ciMethod* declared_method = kit.method()->get_method_at_bci(kit.bci());
int arg_size = declared_method->signature()->arg_size_for_bc(kit.java_bc());
kit.inc_sp(arg_size); // restore arguments
kit.uncommon_trap(Deoptimization::Reason_null_check,
Deoptimization::Action_none,
NULL, "null receiver");
return kit.transfer_exceptions_into_jvms();
}
// Ideally we would unconditionally do a null check here and let it
// be converted to an implicit check based on profile information.
// However currently the conversion to implicit null checks in
// Block::implicit_null_check() only looks for loads and stores, not calls.
ciMethod *caller = kit.method();
ciMethodData *caller_md = (caller == NULL) ? NULL : caller->method_data();
if (!UseInlineCaches || !ImplicitNullChecks || !os::zero_page_read_protected() ||
((ImplicitNullCheckThreshold > 0) && caller_md &&
(caller_md->trap_count(Deoptimization::Reason_null_check)
>= (uint)ImplicitNullCheckThreshold))) {
// Make an explicit receiver null_check as part of this call.
// Since we share a map with the caller, his JVMS gets adjusted.
receiver = kit.null_check_receiver_before_call(method());
if (kit.stopped()) {
// And dump it back to the caller, decorated with any exceptions:
return kit.transfer_exceptions_into_jvms();
}
}
assert(!method()->is_static(), "virtual call must not be to static");
assert(!method()->is_final(), "virtual call should not be to final");
assert(!method()->is_private(), "virtual call should not be to private");
assert(_vtable_index == Method::invalid_vtable_index || !UseInlineCaches,
"no vtable calls if +UseInlineCaches ");
address target = SharedRuntime::get_resolve_virtual_call_stub();
// Normal inline cache used for call
CallDynamicJavaNode *call = new CallDynamicJavaNode(tf(), target, method(), _vtable_index, kit.bci());
if (is_inlined_method_handle_intrinsic(jvms, method())) {
// To be able to issue a direct call (optimized virtual or virtual)
// and skip a call to MH.linkTo*/invokeBasic adapter, additional information
// about the method being invoked should be attached to the call site to
// make resolution logic work (see SharedRuntime::resolve_{virtual,opt_virtual}_call_C).
call->set_override_symbolic_info(true);
}
kit.set_arguments_for_java_call(call);
kit.set_edges_for_java_call(call);
Node* ret = kit.set_results_for_java_call(call);
kit.push_node(method()->return_type()->basic_type(), ret);
// Represent the effect of an implicit receiver null_check
// as part of this call. Since we share a map with the caller,
// his JVMS gets adjusted.
kit.cast_not_null(receiver);
return kit.transfer_exceptions_into_jvms();
}
CallGenerator* CallGenerator::for_inline(ciMethod* m, float expected_uses) {
if (InlineTree::check_can_parse(m) != NULL) return NULL;
return new ParseGenerator(m, expected_uses);
}
// As a special case, the JVMS passed to this CallGenerator is
// for the method execution already in progress, not just the JVMS
// of the caller. Thus, this CallGenerator cannot be mixed with others!
CallGenerator* CallGenerator::for_osr(ciMethod* m, int osr_bci) {
if (InlineTree::check_can_parse(m) != NULL) return NULL;
float past_uses = m->interpreter_invocation_count();
float expected_uses = past_uses;
return new ParseGenerator(m, expected_uses, true);
}
CallGenerator* CallGenerator::for_direct_call(ciMethod* m, bool separate_io_proj) {
assert(!m->is_abstract(), "for_direct_call mismatch");
return new DirectCallGenerator(m, separate_io_proj);
}
CallGenerator* CallGenerator::for_virtual_call(ciMethod* m, int vtable_index) {
assert(!m->is_static(), "for_virtual_call mismatch");
assert(!m->is_method_handle_intrinsic(), "should be a direct call");
return new VirtualCallGenerator(m, vtable_index);
}
// Allow inlining decisions to be delayed
class LateInlineCallGenerator : public DirectCallGenerator {
private:
// unique id for log compilation
jlong _unique_id;
protected:
CallGenerator* _inline_cg;
virtual bool do_late_inline_check(JVMState* jvms) { return true; }
public:
LateInlineCallGenerator(ciMethod* method, CallGenerator* inline_cg) :
DirectCallGenerator(method, true), _unique_id(0), _inline_cg(inline_cg) {}
virtual bool is_late_inline() const { return true; }
// Convert the CallStaticJava into an inline
virtual void do_late_inline();
virtual JVMState* generate(JVMState* jvms) {
Compile *C = Compile::current();
C->log_inline_id(this);
// Record that this call site should be revisited once the main
// parse is finished.
if (!is_mh_late_inline()) {
C->add_late_inline(this);
}
// Emit the CallStaticJava and request separate projections so
// that the late inlining logic can distinguish between fall
// through and exceptional uses of the memory and io projections
// as is done for allocations and macro expansion.
return DirectCallGenerator::generate(jvms);
}
virtual void print_inlining_late(const char* msg) {
CallNode* call = call_node();
Compile* C = Compile::current();
C->print_inlining_assert_ready();
C->print_inlining(method(), call->jvms()->depth()-1, call->jvms()->bci(), msg);
C->print_inlining_move_to(this);
C->print_inlining_update_delayed(this);
}
virtual void set_unique_id(jlong id) {
_unique_id = id;
}
virtual jlong unique_id() const {
return _unique_id;
}
};
void LateInlineCallGenerator::do_late_inline() {
// Can't inline it
CallStaticJavaNode* call = call_node();
if (call == NULL || call->outcnt() == 0 ||
call->in(0) == NULL || call->in(0)->is_top()) {
return;
}
const TypeTuple *r = call->tf()->domain();
for (int i1 = 0; i1 < method()->arg_size(); i1++) {
if (call->in(TypeFunc::Parms + i1)->is_top() && r->field_at(TypeFunc::Parms + i1) != Type::HALF) {
assert(Compile::current()->inlining_incrementally(), "shouldn't happen during parsing");
return;
}
}
if (call->in(TypeFunc::Memory)->is_top()) {
assert(Compile::current()->inlining_incrementally(), "shouldn't happen during parsing");
return;
}
// check for unreachable loop
CallProjections callprojs;
call->extract_projections(&callprojs, true);
if (callprojs.fallthrough_catchproj == call->in(0) ||
callprojs.catchall_catchproj == call->in(0) ||
callprojs.fallthrough_memproj == call->in(TypeFunc::Memory) ||
callprojs.catchall_memproj == call->in(TypeFunc::Memory) ||
callprojs.fallthrough_ioproj == call->in(TypeFunc::I_O) ||
callprojs.catchall_ioproj == call->in(TypeFunc::I_O) ||
(callprojs.resproj != NULL && call->find_edge(callprojs.resproj) != -1) ||
(callprojs.exobj != NULL && call->find_edge(callprojs.exobj) != -1)) {
return;
}
Compile* C = Compile::current();
// Remove inlined methods from Compiler's lists.
if (call->is_macro()) {
C->remove_macro_node(call);
}
// Make a clone of the JVMState that appropriate to use for driving a parse
JVMState* old_jvms = call->jvms();
JVMState* jvms = old_jvms->clone_shallow(C);
uint size = call->req();
SafePointNode* map = new SafePointNode(size, jvms);
for (uint i1 = 0; i1 < size; i1++) {
map->init_req(i1, call->in(i1));
}
// Make sure the state is a MergeMem for parsing.
if (!map->in(TypeFunc::Memory)->is_MergeMem()) {
Node* mem = MergeMemNode::make(map->in(TypeFunc::Memory));
C->initial_gvn()->set_type_bottom(mem);
map->set_req(TypeFunc::Memory, mem);
}
uint nargs = method()->arg_size();
// blow away old call arguments
Node* top = C->top();
for (uint i1 = 0; i1 < nargs; i1++) {
map->set_req(TypeFunc::Parms + i1, top);
}
jvms->set_map(map);
// Make enough space in the expression stack to transfer
// the incoming arguments and return value.
map->ensure_stack(jvms, jvms->method()->max_stack());
for (uint i1 = 0; i1 < nargs; i1++) {
map->set_argument(jvms, i1, call->in(TypeFunc::Parms + i1));
}
C->print_inlining_assert_ready();
C->print_inlining_move_to(this);
C->log_late_inline(this);
// This check is done here because for_method_handle_inline() method
// needs jvms for inlined state.
if (!do_late_inline_check(jvms)) {
map->disconnect_inputs(NULL, C);
return;
}
// Setup default node notes to be picked up by the inlining
Node_Notes* old_nn = C->node_notes_at(call->_idx);
if (old_nn != NULL) {
Node_Notes* entry_nn = old_nn->clone(C);
entry_nn->set_jvms(jvms);
C->set_default_node_notes(entry_nn);
}
// Now perform the inlining using the synthesized JVMState
JVMState* new_jvms = _inline_cg->generate(jvms);
if (new_jvms == NULL) return; // no change
if (C->failing()) return;
// Capture any exceptional control flow
GraphKit kit(new_jvms);
// Find the result object
Node* result = C->top();
int result_size = method()->return_type()->size();
if (result_size != 0 && !kit.stopped()) {
result = (result_size == 1) ? kit.pop() : kit.pop_pair();
}
C->set_has_loops(C->has_loops() || _inline_cg->method()->has_loops());
C->env()->notice_inlined_method(_inline_cg->method());
C->set_inlining_progress(true);
kit.replace_call(call, result, true);
}
CallGenerator* CallGenerator::for_late_inline(ciMethod* method, CallGenerator* inline_cg) {
return new LateInlineCallGenerator(method, inline_cg);
}
class LateInlineMHCallGenerator : public LateInlineCallGenerator {
ciMethod* _caller;
int _attempt;
bool _input_not_const;
virtual bool do_late_inline_check(JVMState* jvms);
virtual bool already_attempted() const { return _attempt > 0; }
public:
LateInlineMHCallGenerator(ciMethod* caller, ciMethod* callee, bool input_not_const) :
LateInlineCallGenerator(callee, NULL), _caller(caller), _attempt(0), _input_not_const(input_not_const) {}
virtual bool is_mh_late_inline() const { return true; }
virtual JVMState* generate(JVMState* jvms) {
JVMState* new_jvms = LateInlineCallGenerator::generate(jvms);
Compile* C = Compile::current();
if (_input_not_const) {
// inlining won't be possible so no need to enqueue right now.
call_node()->set_generator(this);
} else {
C->add_late_inline(this);
}
return new_jvms;
}
};
bool LateInlineMHCallGenerator::do_late_inline_check(JVMState* jvms) {
CallGenerator* cg = for_method_handle_inline(jvms, _caller, method(), _input_not_const);
Compile::current()->print_inlining_update_delayed(this);
if (!_input_not_const) {
_attempt++;
}
if (cg != NULL && cg->is_inline()) {
assert(!cg->is_late_inline(), "we're doing late inlining");
_inline_cg = cg;
Compile::current()->dec_number_of_mh_late_inlines();
return true;
}
call_node()->set_generator(this);
return false;
}
CallGenerator* CallGenerator::for_mh_late_inline(ciMethod* caller, ciMethod* callee, bool input_not_const) {
Compile::current()->inc_number_of_mh_late_inlines();
CallGenerator* cg = new LateInlineMHCallGenerator(caller, callee, input_not_const);
return cg;
}
class LateInlineStringCallGenerator : public LateInlineCallGenerator {
public:
LateInlineStringCallGenerator(ciMethod* method, CallGenerator* inline_cg) :
LateInlineCallGenerator(method, inline_cg) {}
virtual JVMState* generate(JVMState* jvms) {
Compile *C = Compile::current();
C->log_inline_id(this);
C->add_string_late_inline(this);
JVMState* new_jvms = DirectCallGenerator::generate(jvms);
return new_jvms;
}
virtual bool is_string_late_inline() const { return true; }
};
CallGenerator* CallGenerator::for_string_late_inline(ciMethod* method, CallGenerator* inline_cg) {
return new LateInlineStringCallGenerator(method, inline_cg);
}
class LateInlineBoxingCallGenerator : public LateInlineCallGenerator {
public:
LateInlineBoxingCallGenerator(ciMethod* method, CallGenerator* inline_cg) :
LateInlineCallGenerator(method, inline_cg) {}
virtual JVMState* generate(JVMState* jvms) {
Compile *C = Compile::current();
C->log_inline_id(this);
C->add_boxing_late_inline(this);
JVMState* new_jvms = DirectCallGenerator::generate(jvms);
return new_jvms;
}
};
CallGenerator* CallGenerator::for_boxing_late_inline(ciMethod* method, CallGenerator* inline_cg) {
return new LateInlineBoxingCallGenerator(method, inline_cg);
}
//---------------------------WarmCallGenerator--------------------------------
// Internal class which handles initial deferral of inlining decisions.
class WarmCallGenerator : public CallGenerator {
WarmCallInfo* _call_info;
CallGenerator* _if_cold;
CallGenerator* _if_hot;
bool _is_virtual; // caches virtuality of if_cold
bool _is_inline; // caches inline-ness of if_hot
public:
WarmCallGenerator(WarmCallInfo* ci,
CallGenerator* if_cold,
CallGenerator* if_hot)
: CallGenerator(if_cold->method())
{
assert(method() == if_hot->method(), "consistent choices");
_call_info = ci;
_if_cold = if_cold;
_if_hot = if_hot;
_is_virtual = if_cold->is_virtual();
_is_inline = if_hot->is_inline();
}
virtual bool is_inline() const { return _is_inline; }
virtual bool is_virtual() const { return _is_virtual; }
virtual bool is_deferred() const { return true; }
virtual JVMState* generate(JVMState* jvms);
};
CallGenerator* CallGenerator::for_warm_call(WarmCallInfo* ci,
CallGenerator* if_cold,
CallGenerator* if_hot) {
return new WarmCallGenerator(ci, if_cold, if_hot);
}
JVMState* WarmCallGenerator::generate(JVMState* jvms) {
Compile* C = Compile::current();
C->print_inlining_update(this);
if (C->log() != NULL) {
C->log()->elem("warm_call bci='%d'", jvms->bci());
}
jvms = _if_cold->generate(jvms);
if (jvms != NULL) {
Node* m = jvms->map()->control();
if (m->is_CatchProj()) m = m->in(0); else m = C->top();
if (m->is_Catch()) m = m->in(0); else m = C->top();
if (m->is_Proj()) m = m->in(0); else m = C->top();
if (m->is_CallJava()) {
_call_info->set_call(m->as_Call());
_call_info->set_hot_cg(_if_hot);
#ifndef PRODUCT
if (PrintOpto || PrintOptoInlining) {
tty->print_cr("Queueing for warm inlining at bci %d:", jvms->bci());
tty->print("WCI: ");
_call_info->print();
}
#endif
_call_info->set_heat(_call_info->compute_heat());
C->set_warm_calls(_call_info->insert_into(C->warm_calls()));
}
}
return jvms;
}
void WarmCallInfo::make_hot() {
Unimplemented();
}
void WarmCallInfo::make_cold() {
// No action: Just dequeue.
}
//------------------------PredictedCallGenerator------------------------------
// Internal class which handles all out-of-line calls checking receiver type.
class PredictedCallGenerator : public CallGenerator {
ciKlass* _predicted_receiver;
CallGenerator* _if_missed;
CallGenerator* _if_hit;
float _hit_prob;
public:
PredictedCallGenerator(ciKlass* predicted_receiver,
CallGenerator* if_missed,
CallGenerator* if_hit, float hit_prob)
: CallGenerator(if_missed->method())
{
// The call profile data may predict the hit_prob as extreme as 0 or 1.
// Remove the extremes values from the range.
if (hit_prob > PROB_MAX) hit_prob = PROB_MAX;
if (hit_prob < PROB_MIN) hit_prob = PROB_MIN;
_predicted_receiver = predicted_receiver;
_if_missed = if_missed;
_if_hit = if_hit;
_hit_prob = hit_prob;
}
virtual bool is_virtual() const { return true; }
virtual bool is_inline() const { return _if_hit->is_inline(); }
virtual bool is_deferred() const { return _if_hit->is_deferred(); }
virtual JVMState* generate(JVMState* jvms);
};
CallGenerator* CallGenerator::for_predicted_call(ciKlass* predicted_receiver,
CallGenerator* if_missed,
CallGenerator* if_hit,
float hit_prob) {
return new PredictedCallGenerator(predicted_receiver, if_missed, if_hit, hit_prob);
}
JVMState* PredictedCallGenerator::generate(JVMState* jvms) {
GraphKit kit(jvms);
kit.C->print_inlining_update(this);
PhaseGVN& gvn = kit.gvn();
// We need an explicit receiver null_check before checking its type.
// We share a map with the caller, so his JVMS gets adjusted.
Node* receiver = kit.argument(0);
CompileLog* log = kit.C->log();
if (log != NULL) {
log->elem("predicted_call bci='%d' klass='%d'",
jvms->bci(), log->identify(_predicted_receiver));
}
receiver = kit.null_check_receiver_before_call(method());
if (kit.stopped()) {
return kit.transfer_exceptions_into_jvms();
}
// Make a copy of the replaced nodes in case we need to restore them
ReplacedNodes replaced_nodes = kit.map()->replaced_nodes();
replaced_nodes.clone();
Node* exact_receiver = receiver; // will get updated in place...
Node* slow_ctl = kit.type_check_receiver(receiver,
_predicted_receiver, _hit_prob,
&exact_receiver);
SafePointNode* slow_map = NULL;
JVMState* slow_jvms = NULL;
{ PreserveJVMState pjvms(&kit);
kit.set_control(slow_ctl);
if (!kit.stopped()) {
slow_jvms = _if_missed->generate(kit.sync_jvms());
if (kit.failing())
return NULL; // might happen because of NodeCountInliningCutoff
assert(slow_jvms != NULL, "must be");
kit.add_exception_states_from(slow_jvms);
kit.set_map(slow_jvms->map());
if (!kit.stopped())
slow_map = kit.stop();
}
}
if (kit.stopped()) {
// Instance exactly does not matches the desired type.
kit.set_jvms(slow_jvms);
return kit.transfer_exceptions_into_jvms();
}
// fall through if the instance exactly matches the desired type
kit.replace_in_map(receiver, exact_receiver);
// Make the hot call:
JVMState* new_jvms = _if_hit->generate(kit.sync_jvms());
if (new_jvms == NULL) {
// Inline failed, so make a direct call.
assert(_if_hit->is_inline(), "must have been a failed inline");
CallGenerator* cg = CallGenerator::for_direct_call(_if_hit->method());
new_jvms = cg->generate(kit.sync_jvms());
}
kit.add_exception_states_from(new_jvms);
kit.set_jvms(new_jvms);
// Need to merge slow and fast?
if (slow_map == NULL) {
// The fast path is the only path remaining.
return kit.transfer_exceptions_into_jvms();
}
if (kit.stopped()) {
// Inlined method threw an exception, so it's just the slow path after all.
kit.set_jvms(slow_jvms);
return kit.transfer_exceptions_into_jvms();
}
// There are 2 branches and the replaced nodes are only valid on
// one: restore the replaced nodes to what they were before the
// branch.
kit.map()->set_replaced_nodes(replaced_nodes);
// Finish the diamond.
kit.C->set_has_split_ifs(true); // Has chance for split-if optimization
RegionNode* region = new RegionNode(3);
region->init_req(1, kit.control());
region->init_req(2, slow_map->control());
kit.set_control(gvn.transform(region));
Node* iophi = PhiNode::make(region, kit.i_o(), Type::ABIO);
iophi->set_req(2, slow_map->i_o());
kit.set_i_o(gvn.transform(iophi));
// Merge memory
kit.merge_memory(slow_map->merged_memory(), region, 2);
// Transform new memory Phis.
for (MergeMemStream mms(kit.merged_memory()); mms.next_non_empty();) {
Node* phi = mms.memory();
if (phi->is_Phi() && phi->in(0) == region) {
mms.set_memory(gvn.transform(phi));
}
}
uint tos = kit.jvms()->stkoff() + kit.sp();
uint limit = slow_map->req();
for (uint i = TypeFunc::Parms; i < limit; i++) {
// Skip unused stack slots; fast forward to monoff();
if (i == tos) {
i = kit.jvms()->monoff();
if( i >= limit ) break;
}
Node* m = kit.map()->in(i);
Node* n = slow_map->in(i);
if (m != n) {
const Type* t = gvn.type(m)->meet_speculative(gvn.type(n));
Node* phi = PhiNode::make(region, m, t);
phi->set_req(2, n);
kit.map()->set_req(i, gvn.transform(phi));
}
}
return kit.transfer_exceptions_into_jvms();
}
CallGenerator* CallGenerator::for_method_handle_call(JVMState* jvms, ciMethod* caller, ciMethod* callee, bool delayed_forbidden) {
assert(callee->is_method_handle_intrinsic(), "for_method_handle_call mismatch");
bool input_not_const;
CallGenerator* cg = CallGenerator::for_method_handle_inline(jvms, caller, callee, input_not_const);
Compile* C = Compile::current();
if (cg != NULL) {
if (!delayed_forbidden && AlwaysIncrementalInline) {
return CallGenerator::for_late_inline(callee, cg);
} else {
return cg;
}
}
int bci = jvms->bci();
ciCallProfile profile = caller->call_profile_at_bci(bci);
int call_site_count = caller->scale_count(profile.count());
if (IncrementalInline && call_site_count > 0 &&
(input_not_const || !C->inlining_incrementally() || C->over_inlining_cutoff())) {
return CallGenerator::for_mh_late_inline(caller, callee, input_not_const);
} else {
// Out-of-line call.
return CallGenerator::for_direct_call(callee);
}
}
CallGenerator* CallGenerator::for_method_handle_inline(JVMState* jvms, ciMethod* caller, ciMethod* callee, bool& input_not_const) {
GraphKit kit(jvms);
PhaseGVN& gvn = kit.gvn();
Compile* C = kit.C;
vmIntrinsics::ID iid = callee->intrinsic_id();
input_not_const = true;
switch (iid) {
case vmIntrinsics::_invokeBasic:
{
// Get MethodHandle receiver:
Node* receiver = kit.argument(0);
if (receiver->Opcode() == Op_ConP) {
input_not_const = false;
const TypeOopPtr* oop_ptr = receiver->bottom_type()->is_oopptr();
ciMethod* target = oop_ptr->const_oop()->as_method_handle()->get_vmtarget();
const int vtable_index = Method::invalid_vtable_index;
if (!ciMethod::is_consistent_info(callee, target)) {
print_inlining_failure(C, callee, jvms->depth() - 1, jvms->bci(),
"signatures mismatch");
return NULL;
}
CallGenerator* cg = C->call_generator(target, vtable_index,
false /* call_does_dispatch */,
jvms,
true /* allow_inline */,
PROB_ALWAYS);
return cg;
} else {
print_inlining_failure(C, callee, jvms->depth() - 1, jvms->bci(),
"receiver not constant");
}
}
break;
case vmIntrinsics::_linkToVirtual:
case vmIntrinsics::_linkToStatic:
case vmIntrinsics::_linkToSpecial:
case vmIntrinsics::_linkToInterface:
{
// Get MemberName argument:
Node* member_name = kit.argument(callee->arg_size() - 1);
if (member_name->Opcode() == Op_ConP) {
input_not_const = false;
const TypeOopPtr* oop_ptr = member_name->bottom_type()->is_oopptr();
ciMethod* target = oop_ptr->const_oop()->as_member_name()->get_vmtarget();
if (!ciMethod::is_consistent_info(callee, target)) {
print_inlining_failure(C, callee, jvms->depth() - 1, jvms->bci(),
"signatures mismatch");
return NULL;
}
// In lambda forms we erase signature types to avoid resolving issues
// involving class loaders. When we optimize a method handle invoke
// to a direct call we must cast the receiver and arguments to its
// actual types.
ciSignature* signature = target->signature();
const int receiver_skip = target->is_static() ? 0 : 1;
// Cast receiver to its type.
if (!target->is_static()) {
Node* arg = kit.argument(0);
const TypeOopPtr* arg_type = arg->bottom_type()->isa_oopptr();
const Type* sig_type = TypeOopPtr::make_from_klass(signature->accessing_klass());
if (arg_type != NULL && !arg_type->higher_equal(sig_type)) {
const Type* recv_type = arg_type->join_speculative(sig_type); // keep speculative part
Node* cast_obj = gvn.transform(new CheckCastPPNode(kit.control(), arg, recv_type));
kit.set_argument(0, cast_obj);
}
}
// Cast reference arguments to its type.
for (int i = 0, j = 0; i < signature->count(); i++) {
ciType* t = signature->type_at(i);
if (t->is_klass()) {
Node* arg = kit.argument(receiver_skip + j);
const TypeOopPtr* arg_type = arg->bottom_type()->isa_oopptr();
const Type* sig_type = TypeOopPtr::make_from_klass(t->as_klass());
if (arg_type != NULL && !arg_type->higher_equal(sig_type)) {
const Type* narrowed_arg_type = arg_type->join_speculative(sig_type); // keep speculative part
Node* cast_obj = gvn.transform(new CheckCastPPNode(kit.control(), arg, narrowed_arg_type));
kit.set_argument(receiver_skip + j, cast_obj);
}
}
j += t->size(); // long and double take two slots
}
// Try to get the most accurate receiver type
const bool is_virtual = (iid == vmIntrinsics::_linkToVirtual);
const bool is_virtual_or_interface = (is_virtual || iid == vmIntrinsics::_linkToInterface);
int vtable_index = Method::invalid_vtable_index;
bool call_does_dispatch = false;
ciKlass* speculative_receiver_type = NULL;
if (is_virtual_or_interface) {
ciInstanceKlass* klass = target->holder();
Node* receiver_node = kit.argument(0);
const TypeOopPtr* receiver_type = gvn.type(receiver_node)->isa_oopptr();
// call_does_dispatch and vtable_index are out-parameters. They might be changed.
// optimize_virtual_call() takes 2 different holder
// arguments for a corner case that doesn't apply here (see
// Parse::do_call())
target = C->optimize_virtual_call(caller, jvms->bci(), klass, klass,
target, receiver_type, is_virtual,
call_does_dispatch, vtable_index, // out-parameters
false /* check_access */);
// We lack profiling at this call but type speculation may
// provide us with a type
speculative_receiver_type = (receiver_type != NULL) ? receiver_type->speculative_type() : NULL;
}
CallGenerator* cg = C->call_generator(target, vtable_index, call_does_dispatch, jvms,
!StressMethodHandleLinkerInlining /* allow_inline */,
PROB_ALWAYS,
speculative_receiver_type);
return cg;
} else {
print_inlining_failure(C, callee, jvms->depth() - 1, jvms->bci(),
"member_name not constant");
}
}
break;
default:
fatal("unexpected intrinsic %d: %s", iid, vmIntrinsics::name_at(iid));
break;
}
return NULL;
}
//------------------------PredicatedIntrinsicGenerator------------------------------
// Internal class which handles all predicated Intrinsic calls.
class PredicatedIntrinsicGenerator : public CallGenerator {
CallGenerator* _intrinsic;
CallGenerator* _cg;
public:
PredicatedIntrinsicGenerator(CallGenerator* intrinsic,
CallGenerator* cg)
: CallGenerator(cg->method())
{
_intrinsic = intrinsic;
_cg = cg;
}
virtual bool is_virtual() const { return true; }
virtual bool is_inlined() const { return true; }
virtual bool is_intrinsic() const { return true; }
virtual JVMState* generate(JVMState* jvms);
};
CallGenerator* CallGenerator::for_predicated_intrinsic(CallGenerator* intrinsic,
CallGenerator* cg) {
return new PredicatedIntrinsicGenerator(intrinsic, cg);
}
JVMState* PredicatedIntrinsicGenerator::generate(JVMState* jvms) {
// The code we want to generate here is:
// if (receiver == NULL)
// uncommon_Trap
// if (predicate(0))
// do_intrinsic(0)
// else
// if (predicate(1))
// do_intrinsic(1)
// ...
// else
// do_java_comp
GraphKit kit(jvms);
PhaseGVN& gvn = kit.gvn();
CompileLog* log = kit.C->log();
if (log != NULL) {
log->elem("predicated_intrinsic bci='%d' method='%d'",
jvms->bci(), log->identify(method()));
}
if (!method()->is_static()) {
// We need an explicit receiver null_check before checking its type in predicate.
// We share a map with the caller, so his JVMS gets adjusted.
Node* receiver = kit.null_check_receiver_before_call(method());
if (kit.stopped()) {
return kit.transfer_exceptions_into_jvms();
}
}
int n_predicates = _intrinsic->predicates_count();
assert(n_predicates > 0, "sanity");
JVMState** result_jvms = NEW_RESOURCE_ARRAY(JVMState*, (n_predicates+1));
// Region for normal compilation code if intrinsic failed.
Node* slow_region = new RegionNode(1);
int results = 0;
for (int predicate = 0; (predicate < n_predicates) && !kit.stopped(); predicate++) {
#ifdef ASSERT
JVMState* old_jvms = kit.jvms();
SafePointNode* old_map = kit.map();
Node* old_io = old_map->i_o();
Node* old_mem = old_map->memory();
Node* old_exc = old_map->next_exception();
#endif
Node* else_ctrl = _intrinsic->generate_predicate(kit.sync_jvms(), predicate);
#ifdef ASSERT
// Assert(no_new_memory && no_new_io && no_new_exceptions) after generate_predicate.
assert(old_jvms == kit.jvms(), "generate_predicate should not change jvm state");
SafePointNode* new_map = kit.map();
assert(old_io == new_map->i_o(), "generate_predicate should not change i_o");
assert(old_mem == new_map->memory(), "generate_predicate should not change memory");
assert(old_exc == new_map->next_exception(), "generate_predicate should not add exceptions");
#endif
if (!kit.stopped()) {
PreserveJVMState pjvms(&kit);
// Generate intrinsic code:
JVMState* new_jvms = _intrinsic->generate(kit.sync_jvms());
if (new_jvms == NULL) {
// Intrinsic failed, use normal compilation path for this predicate.
slow_region->add_req(kit.control());
} else {
kit.add_exception_states_from(new_jvms);
kit.set_jvms(new_jvms);
if (!kit.stopped()) {
result_jvms[results++] = kit.jvms();
}
}
}
if (else_ctrl == NULL) {
else_ctrl = kit.C->top();
}
kit.set_control(else_ctrl);
}
if (!kit.stopped()) {
// Final 'else' after predicates.
slow_region->add_req(kit.control());
}
if (slow_region->req() > 1) {
PreserveJVMState pjvms(&kit);
// Generate normal compilation code:
kit.set_control(gvn.transform(slow_region));
JVMState* new_jvms = _cg->generate(kit.sync_jvms());
if (kit.failing())
return NULL; // might happen because of NodeCountInliningCutoff
assert(new_jvms != NULL, "must be");
kit.add_exception_states_from(new_jvms);
kit.set_jvms(new_jvms);
if (!kit.stopped()) {
result_jvms[results++] = kit.jvms();
}
}
if (results == 0) {
// All paths ended in uncommon traps.
(void) kit.stop();
return kit.transfer_exceptions_into_jvms();
}
if (results == 1) { // Only one path
kit.set_jvms(result_jvms[0]);
return kit.transfer_exceptions_into_jvms();
}
// Merge all paths.
kit.C->set_has_split_ifs(true); // Has chance for split-if optimization
RegionNode* region = new RegionNode(results + 1);
Node* iophi = PhiNode::make(region, kit.i_o(), Type::ABIO);
for (int i = 0; i < results; i++) {
JVMState* jvms = result_jvms[i];
int path = i + 1;
SafePointNode* map = jvms->map();
region->init_req(path, map->control());
iophi->set_req(path, map->i_o());
if (i == 0) {
kit.set_jvms(jvms);
} else {
kit.merge_memory(map->merged_memory(), region, path);
}
}
kit.set_control(gvn.transform(region));
kit.set_i_o(gvn.transform(iophi));
// Transform new memory Phis.
for (MergeMemStream mms(kit.merged_memory()); mms.next_non_empty();) {
Node* phi = mms.memory();
if (phi->is_Phi() && phi->in(0) == region) {
mms.set_memory(gvn.transform(phi));
}
}
// Merge debug info.
Node** ins = NEW_RESOURCE_ARRAY(Node*, results);
uint tos = kit.jvms()->stkoff() + kit.sp();
Node* map = kit.map();
uint limit = map->req();
for (uint i = TypeFunc::Parms; i < limit; i++) {
// Skip unused stack slots; fast forward to monoff();
if (i == tos) {
i = kit.jvms()->monoff();
if( i >= limit ) break;
}
Node* n = map->in(i);
ins[0] = n;
const Type* t = gvn.type(n);
bool needs_phi = false;
for (int j = 1; j < results; j++) {
JVMState* jvms = result_jvms[j];
Node* jmap = jvms->map();
Node* m = NULL;
if (jmap->req() > i) {
m = jmap->in(i);
if (m != n) {
needs_phi = true;
t = t->meet_speculative(gvn.type(m));
}
}
ins[j] = m;
}
if (needs_phi) {
Node* phi = PhiNode::make(region, n, t);
for (int j = 1; j < results; j++) {
phi->set_req(j + 1, ins[j]);
}
map->set_req(i, gvn.transform(phi));
}
}
return kit.transfer_exceptions_into_jvms();
}
//-------------------------UncommonTrapCallGenerator-----------------------------
// Internal class which handles all out-of-line calls checking receiver type.
class UncommonTrapCallGenerator : public CallGenerator {
Deoptimization::DeoptReason _reason;
Deoptimization::DeoptAction _action;
public:
UncommonTrapCallGenerator(ciMethod* m,
Deoptimization::DeoptReason reason,
Deoptimization::DeoptAction action)
: CallGenerator(m)
{
_reason = reason;
_action = action;
}
virtual bool is_virtual() const { ShouldNotReachHere(); return false; }
virtual bool is_trap() const { return true; }
virtual JVMState* generate(JVMState* jvms);
};
CallGenerator*
CallGenerator::for_uncommon_trap(ciMethod* m,
Deoptimization::DeoptReason reason,
Deoptimization::DeoptAction action) {
return new UncommonTrapCallGenerator(m, reason, action);
}
JVMState* UncommonTrapCallGenerator::generate(JVMState* jvms) {
GraphKit kit(jvms);
kit.C->print_inlining_update(this);
// Take the trap with arguments pushed on the stack. (Cf. null_check_receiver).
// Callsite signature can be different from actual method being called (i.e _linkTo* sites).
// Use callsite signature always.
ciMethod* declared_method = kit.method()->get_method_at_bci(kit.bci());
int nargs = declared_method->arg_size();
kit.inc_sp(nargs);
assert(nargs <= kit.sp() && kit.sp() <= jvms->stk_size(), "sane sp w/ args pushed");
if (_reason == Deoptimization::Reason_class_check &&
_action == Deoptimization::Action_maybe_recompile) {
// Temp fix for 6529811
// Don't allow uncommon_trap to override our decision to recompile in the event
// of a class cast failure for a monomorphic call as it will never let us convert
// the call to either bi-morphic or megamorphic and can lead to unc-trap loops
bool keep_exact_action = true;
kit.uncommon_trap(_reason, _action, NULL, "monomorphic vcall checkcast", false, keep_exact_action);
} else {
kit.uncommon_trap(_reason, _action);
}
return kit.transfer_exceptions_into_jvms();
}
// (Note: Moved hook_up_call to GraphKit::set_edges_for_java_call.)
// (Node: Merged hook_up_exits into ParseGenerator::generate.)
#define NODES_OVERHEAD_PER_METHOD (30.0)
#define NODES_PER_BYTECODE (9.5)
void WarmCallInfo::init(JVMState* call_site, ciMethod* call_method, ciCallProfile& profile, float prof_factor) {
int call_count = profile.count();
int code_size = call_method->code_size();
// Expected execution count is based on the historical count:
_count = call_count < 0 ? 1 : call_site->method()->scale_count(call_count, prof_factor);
// Expected profit from inlining, in units of simple call-overheads.
_profit = 1.0;
// Expected work performed by the call in units of call-overheads.
// %%% need an empirical curve fit for "work" (time in call)
float bytecodes_per_call = 3;
_work = 1.0 + code_size / bytecodes_per_call;
// Expected size of compilation graph:
// -XX:+PrintParseStatistics once reported:
// Methods seen: 9184 Methods parsed: 9184 Nodes created: 1582391
// Histogram of 144298 parsed bytecodes:
// %%% Need an better predictor for graph size.
_size = NODES_OVERHEAD_PER_METHOD + (NODES_PER_BYTECODE * code_size);
}
// is_cold: Return true if the node should never be inlined.
// This is true if any of the key metrics are extreme.
bool WarmCallInfo::is_cold() const {
if (count() < WarmCallMinCount) return true;
if (profit() < WarmCallMinProfit) return true;
if (work() > WarmCallMaxWork) return true;
if (size() > WarmCallMaxSize) return true;
return false;
}
// is_hot: Return true if the node should be inlined immediately.
// This is true if any of the key metrics are extreme.
bool WarmCallInfo::is_hot() const {
assert(!is_cold(), "eliminate is_cold cases before testing is_hot");
if (count() >= HotCallCountThreshold) return true;
if (profit() >= HotCallProfitThreshold) return true;
if (work() <= HotCallTrivialWork) return true;
if (size() <= HotCallTrivialSize) return true;
return false;
}
// compute_heat:
float WarmCallInfo::compute_heat() const {
assert(!is_cold(), "compute heat only on warm nodes");
assert(!is_hot(), "compute heat only on warm nodes");
int min_size = MAX2(0, (int)HotCallTrivialSize);
int max_size = MIN2(500, (int)WarmCallMaxSize);
float method_size = (size() - min_size) / MAX2(1, max_size - min_size);
float size_factor;
if (method_size < 0.05) size_factor = 4; // 2 sigmas better than avg.
else if (method_size < 0.15) size_factor = 2; // 1 sigma better than avg.
else if (method_size < 0.5) size_factor = 1; // better than avg.
else size_factor = 0.5; // worse than avg.
return (count() * profit() * size_factor);
}
bool WarmCallInfo::warmer_than(WarmCallInfo* that) {
assert(this != that, "compare only different WCIs");
assert(this->heat() != 0 && that->heat() != 0, "call compute_heat 1st");
if (this->heat() > that->heat()) return true;
if (this->heat() < that->heat()) return false;
assert(this->heat() == that->heat(), "no NaN heat allowed");
// Equal heat. Break the tie some other way.
if (!this->call() || !that->call()) return (address)this > (address)that;
return this->call()->_idx > that->call()->_idx;
}
//#define UNINIT_NEXT ((WarmCallInfo*)badAddress)
#define UNINIT_NEXT ((WarmCallInfo*)NULL)
WarmCallInfo* WarmCallInfo::insert_into(WarmCallInfo* head) {
assert(next() == UNINIT_NEXT, "not yet on any list");
WarmCallInfo* prev_p = NULL;
WarmCallInfo* next_p = head;
while (next_p != NULL && next_p->warmer_than(this)) {
prev_p = next_p;
next_p = prev_p->next();
}
// Install this between prev_p and next_p.
this->set_next(next_p);
if (prev_p == NULL)
head = this;
else
prev_p->set_next(this);
return head;
}
WarmCallInfo* WarmCallInfo::remove_from(WarmCallInfo* head) {
WarmCallInfo* prev_p = NULL;
WarmCallInfo* next_p = head;
while (next_p != this) {
assert(next_p != NULL, "this must be in the list somewhere");
prev_p = next_p;
next_p = prev_p->next();
}
next_p = this->next();
debug_only(this->set_next(UNINIT_NEXT));
// Remove this from between prev_p and next_p.
if (prev_p == NULL)
head = next_p;
else
prev_p->set_next(next_p);
return head;
}
WarmCallInfo WarmCallInfo::_always_hot(WarmCallInfo::MAX_VALUE(), WarmCallInfo::MAX_VALUE(),
WarmCallInfo::MIN_VALUE(), WarmCallInfo::MIN_VALUE());
WarmCallInfo WarmCallInfo::_always_cold(WarmCallInfo::MIN_VALUE(), WarmCallInfo::MIN_VALUE(),
WarmCallInfo::MAX_VALUE(), WarmCallInfo::MAX_VALUE());
WarmCallInfo* WarmCallInfo::always_hot() {
assert(_always_hot.is_hot(), "must always be hot");
return &_always_hot;
}
WarmCallInfo* WarmCallInfo::always_cold() {
assert(_always_cold.is_cold(), "must always be cold");
return &_always_cold;
}
#ifndef PRODUCT
void WarmCallInfo::print() const {
tty->print("%s : C=%6.1f P=%6.1f W=%6.1f S=%6.1f H=%6.1f -> %p",
is_cold() ? "cold" : is_hot() ? "hot " : "warm",
count(), profit(), work(), size(), compute_heat(), next());
tty->cr();
if (call() != NULL) call()->dump();
}
void print_wci(WarmCallInfo* ci) {
ci->print();
}
void WarmCallInfo::print_all() const {
for (const WarmCallInfo* p = this; p != NULL; p = p->next())
p->print();
}
int WarmCallInfo::count_all() const {
int cnt = 0;
for (const WarmCallInfo* p = this; p != NULL; p = p->next())
cnt++;
return cnt;
}
#endif //PRODUCT