7122880: Extend vendor-specific command interface to include manageable switches
Summary: Add Flag::external_ext()/writable_ext(), both return false.
Reviewed-by: coleenp, zgu
/*
* Copyright (c) 2000, 2011, 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 "code/compiledIC.hpp"
#include "code/nmethod.hpp"
#include "code/scopeDesc.hpp"
#include "compiler/compilerOracle.hpp"
#include "interpreter/interpreter.hpp"
#include "oops/methodDataOop.hpp"
#include "oops/methodOop.hpp"
#include "oops/oop.inline.hpp"
#include "prims/nativeLookup.hpp"
#include "runtime/advancedThresholdPolicy.hpp"
#include "runtime/compilationPolicy.hpp"
#include "runtime/frame.hpp"
#include "runtime/handles.inline.hpp"
#include "runtime/rframe.hpp"
#include "runtime/simpleThresholdPolicy.hpp"
#include "runtime/stubRoutines.hpp"
#include "runtime/thread.hpp"
#include "runtime/timer.hpp"
#include "runtime/vframe.hpp"
#include "runtime/vm_operations.hpp"
#include "utilities/events.hpp"
#include "utilities/globalDefinitions.hpp"
CompilationPolicy* CompilationPolicy::_policy;
elapsedTimer CompilationPolicy::_accumulated_time;
bool CompilationPolicy::_in_vm_startup;
// Determine compilation policy based on command line argument
void compilationPolicy_init() {
CompilationPolicy::set_in_vm_startup(DelayCompilationDuringStartup);
switch(CompilationPolicyChoice) {
case 0:
CompilationPolicy::set_policy(new SimpleCompPolicy());
break;
case 1:
#ifdef COMPILER2
CompilationPolicy::set_policy(new StackWalkCompPolicy());
#else
Unimplemented();
#endif
break;
case 2:
#ifdef TIERED
CompilationPolicy::set_policy(new SimpleThresholdPolicy());
#else
Unimplemented();
#endif
break;
case 3:
#ifdef TIERED
CompilationPolicy::set_policy(new AdvancedThresholdPolicy());
#else
Unimplemented();
#endif
break;
default:
fatal("CompilationPolicyChoice must be in the range: [0-3]");
}
CompilationPolicy::policy()->initialize();
}
void CompilationPolicy::completed_vm_startup() {
if (TraceCompilationPolicy) {
tty->print("CompilationPolicy: completed vm startup.\n");
}
_in_vm_startup = false;
}
// Returns true if m must be compiled before executing it
// This is intended to force compiles for methods (usually for
// debugging) that would otherwise be interpreted for some reason.
bool CompilationPolicy::must_be_compiled(methodHandle m, int comp_level) {
if (m->has_compiled_code()) return false; // already compiled
if (!can_be_compiled(m, comp_level)) return false;
return !UseInterpreter || // must compile all methods
(UseCompiler && AlwaysCompileLoopMethods && m->has_loops() && CompileBroker::should_compile_new_jobs()); // eagerly compile loop methods
}
// Returns true if m is allowed to be compiled
bool CompilationPolicy::can_be_compiled(methodHandle m, int comp_level) {
if (m->is_abstract()) return false;
if (DontCompileHugeMethods && m->code_size() > HugeMethodLimit) return false;
// Math intrinsics should never be compiled as this can lead to
// monotonicity problems because the interpreter will prefer the
// compiled code to the intrinsic version. This can't happen in
// production because the invocation counter can't be incremented
// but we shouldn't expose the system to this problem in testing
// modes.
if (!AbstractInterpreter::can_be_compiled(m)) {
return false;
}
if (comp_level == CompLevel_all) {
return !m->is_not_compilable(CompLevel_simple) && !m->is_not_compilable(CompLevel_full_optimization);
} else {
return !m->is_not_compilable(comp_level);
}
}
bool CompilationPolicy::is_compilation_enabled() {
// NOTE: CompileBroker::should_compile_new_jobs() checks for UseCompiler
return !delay_compilation_during_startup() && CompileBroker::should_compile_new_jobs();
}
#ifndef PRODUCT
void CompilationPolicy::print_time() {
tty->print_cr ("Accumulated compilationPolicy times:");
tty->print_cr ("---------------------------");
tty->print_cr (" Total: %3.3f sec.", _accumulated_time.seconds());
}
void NonTieredCompPolicy::trace_osr_completion(nmethod* osr_nm) {
if (TraceOnStackReplacement) {
if (osr_nm == NULL) tty->print_cr("compilation failed");
else tty->print_cr("nmethod " INTPTR_FORMAT, osr_nm);
}
}
#endif // !PRODUCT
void NonTieredCompPolicy::initialize() {
// Setup the compiler thread numbers
if (CICompilerCountPerCPU) {
// Example: if CICompilerCountPerCPU is true, then we get
// max(log2(8)-1,1) = 2 compiler threads on an 8-way machine.
// May help big-app startup time.
_compiler_count = MAX2(log2_intptr(os::active_processor_count())-1,1);
} else {
_compiler_count = CICompilerCount;
}
}
// Note: this policy is used ONLY if TieredCompilation is off.
// compiler_count() behaves the following way:
// - with TIERED build (with both COMPILER1 and COMPILER2 defined) it should return
// zero for the c1 compilation levels, hence the particular ordering of the
// statements.
// - the same should happen when COMPILER2 is defined and COMPILER1 is not
// (server build without TIERED defined).
// - if only COMPILER1 is defined (client build), zero should be returned for
// the c2 level.
// - if neither is defined - always return zero.
int NonTieredCompPolicy::compiler_count(CompLevel comp_level) {
assert(!TieredCompilation, "This policy should not be used with TieredCompilation");
#ifdef COMPILER2
if (is_c2_compile(comp_level)) {
return _compiler_count;
} else {
return 0;
}
#endif
#ifdef COMPILER1
if (is_c1_compile(comp_level)) {
return _compiler_count;
} else {
return 0;
}
#endif
return 0;
}
void NonTieredCompPolicy::reset_counter_for_invocation_event(methodHandle m) {
// Make sure invocation and backedge counter doesn't overflow again right away
// as would be the case for native methods.
// BUT also make sure the method doesn't look like it was never executed.
// Set carry bit and reduce counter's value to min(count, CompileThreshold/2).
m->invocation_counter()->set_carry();
m->backedge_counter()->set_carry();
assert(!m->was_never_executed(), "don't reset to 0 -- could be mistaken for never-executed");
}
void NonTieredCompPolicy::reset_counter_for_back_branch_event(methodHandle m) {
// Delay next back-branch event but pump up invocation counter to triger
// whole method compilation.
InvocationCounter* i = m->invocation_counter();
InvocationCounter* b = m->backedge_counter();
// Don't set invocation_counter's value too low otherwise the method will
// look like immature (ic < ~5300) which prevents the inlining based on
// the type profiling.
i->set(i->state(), CompileThreshold);
// Don't reset counter too low - it is used to check if OSR method is ready.
b->set(b->state(), CompileThreshold / 2);
}
//
// CounterDecay
//
// Interates through invocation counters and decrements them. This
// is done at each safepoint.
//
class CounterDecay : public AllStatic {
static jlong _last_timestamp;
static void do_method(methodOop m) {
m->invocation_counter()->decay();
}
public:
static void decay();
static bool is_decay_needed() {
return (os::javaTimeMillis() - _last_timestamp) > CounterDecayMinIntervalLength;
}
};
jlong CounterDecay::_last_timestamp = 0;
void CounterDecay::decay() {
_last_timestamp = os::javaTimeMillis();
// This operation is going to be performed only at the end of a safepoint
// and hence GC's will not be going on, all Java mutators are suspended
// at this point and hence SystemDictionary_lock is also not needed.
assert(SafepointSynchronize::is_at_safepoint(), "can only be executed at a safepoint");
int nclasses = SystemDictionary::number_of_classes();
double classes_per_tick = nclasses * (CounterDecayMinIntervalLength * 1e-3 /
CounterHalfLifeTime);
for (int i = 0; i < classes_per_tick; i++) {
klassOop k = SystemDictionary::try_get_next_class();
if (k != NULL && k->klass_part()->oop_is_instance()) {
instanceKlass::cast(k)->methods_do(do_method);
}
}
}
// Called at the end of the safepoint
void NonTieredCompPolicy::do_safepoint_work() {
if(UseCounterDecay && CounterDecay::is_decay_needed()) {
CounterDecay::decay();
}
}
void NonTieredCompPolicy::reprofile(ScopeDesc* trap_scope, bool is_osr) {
ScopeDesc* sd = trap_scope;
for (; !sd->is_top(); sd = sd->sender()) {
// Reset ICs of inlined methods, since they can trigger compilations also.
sd->method()->invocation_counter()->reset();
}
InvocationCounter* c = sd->method()->invocation_counter();
if (is_osr) {
// It was an OSR method, so bump the count higher.
c->set(c->state(), CompileThreshold);
} else {
c->reset();
}
sd->method()->backedge_counter()->reset();
}
// This method can be called by any component of the runtime to notify the policy
// that it's recommended to delay the complation of this method.
void NonTieredCompPolicy::delay_compilation(methodOop method) {
method->invocation_counter()->decay();
method->backedge_counter()->decay();
}
void NonTieredCompPolicy::disable_compilation(methodOop method) {
method->invocation_counter()->set_state(InvocationCounter::wait_for_nothing);
method->backedge_counter()->set_state(InvocationCounter::wait_for_nothing);
}
CompileTask* NonTieredCompPolicy::select_task(CompileQueue* compile_queue) {
return compile_queue->first();
}
bool NonTieredCompPolicy::is_mature(methodOop method) {
methodDataOop mdo = method->method_data();
assert(mdo != NULL, "Should be");
uint current = mdo->mileage_of(method);
uint initial = mdo->creation_mileage();
if (current < initial)
return true; // some sort of overflow
uint target;
if (ProfileMaturityPercentage <= 0)
target = (uint) -ProfileMaturityPercentage; // absolute value
else
target = (uint)( (ProfileMaturityPercentage * CompileThreshold) / 100 );
return (current >= initial + target);
}
nmethod* NonTieredCompPolicy::event(methodHandle method, methodHandle inlinee, int branch_bci, int bci, CompLevel comp_level, nmethod* nm, TRAPS) {
assert(comp_level == CompLevel_none, "This should be only called from the interpreter");
NOT_PRODUCT(trace_frequency_counter_overflow(method, branch_bci, bci));
if (JvmtiExport::can_post_interpreter_events()) {
assert(THREAD->is_Java_thread(), "Wrong type of thread");
if (((JavaThread*)THREAD)->is_interp_only_mode()) {
// If certain JVMTI events (e.g. frame pop event) are requested then the
// thread is forced to remain in interpreted code. This is
// implemented partly by a check in the run_compiled_code
// section of the interpreter whether we should skip running
// compiled code, and partly by skipping OSR compiles for
// interpreted-only threads.
if (bci != InvocationEntryBci) {
reset_counter_for_back_branch_event(method);
return NULL;
}
}
}
if (bci == InvocationEntryBci) {
// when code cache is full, compilation gets switched off, UseCompiler
// is set to false
if (!method->has_compiled_code() && UseCompiler) {
method_invocation_event(method, CHECK_NULL);
} else {
// Force counter overflow on method entry, even if no compilation
// happened. (The method_invocation_event call does this also.)
reset_counter_for_invocation_event(method);
}
// compilation at an invocation overflow no longer goes and retries test for
// compiled method. We always run the loser of the race as interpreted.
// so return NULL
return NULL;
} else {
// counter overflow in a loop => try to do on-stack-replacement
nmethod* osr_nm = method->lookup_osr_nmethod_for(bci, CompLevel_highest_tier, true);
NOT_PRODUCT(trace_osr_request(method, osr_nm, bci));
// when code cache is full, we should not compile any more...
if (osr_nm == NULL && UseCompiler) {
method_back_branch_event(method, bci, CHECK_NULL);
osr_nm = method->lookup_osr_nmethod_for(bci, CompLevel_highest_tier, true);
}
if (osr_nm == NULL) {
reset_counter_for_back_branch_event(method);
return NULL;
}
return osr_nm;
}
return NULL;
}
#ifndef PRODUCT
void NonTieredCompPolicy::trace_frequency_counter_overflow(methodHandle m, int branch_bci, int bci) {
if (TraceInvocationCounterOverflow) {
InvocationCounter* ic = m->invocation_counter();
InvocationCounter* bc = m->backedge_counter();
ResourceMark rm;
const char* msg =
bci == InvocationEntryBci
? "comp-policy cntr ovfl @ %d in entry of "
: "comp-policy cntr ovfl @ %d in loop of ";
tty->print(msg, bci);
m->print_value();
tty->cr();
ic->print();
bc->print();
if (ProfileInterpreter) {
if (bci != InvocationEntryBci) {
methodDataOop mdo = m->method_data();
if (mdo != NULL) {
int count = mdo->bci_to_data(branch_bci)->as_JumpData()->taken();
tty->print_cr("back branch count = %d", count);
}
}
}
}
}
void NonTieredCompPolicy::trace_osr_request(methodHandle method, nmethod* osr, int bci) {
if (TraceOnStackReplacement) {
ResourceMark rm;
tty->print(osr != NULL ? "Reused OSR entry for " : "Requesting OSR entry for ");
method->print_short_name(tty);
tty->print_cr(" at bci %d", bci);
}
}
#endif // !PRODUCT
// SimpleCompPolicy - compile current method
void SimpleCompPolicy::method_invocation_event( methodHandle m, TRAPS) {
int hot_count = m->invocation_count();
reset_counter_for_invocation_event(m);
const char* comment = "count";
if (is_compilation_enabled() && can_be_compiled(m)) {
nmethod* nm = m->code();
if (nm == NULL ) {
const char* comment = "count";
CompileBroker::compile_method(m, InvocationEntryBci, CompLevel_highest_tier,
m, hot_count, comment, CHECK);
}
}
}
void SimpleCompPolicy::method_back_branch_event(methodHandle m, int bci, TRAPS) {
int hot_count = m->backedge_count();
const char* comment = "backedge_count";
if (is_compilation_enabled() && !m->is_not_osr_compilable() && can_be_compiled(m)) {
CompileBroker::compile_method(m, bci, CompLevel_highest_tier,
m, hot_count, comment, CHECK);
NOT_PRODUCT(trace_osr_completion(m->lookup_osr_nmethod_for(bci, CompLevel_highest_tier, true));)
}
}
// StackWalkCompPolicy - walk up stack to find a suitable method to compile
#ifdef COMPILER2
const char* StackWalkCompPolicy::_msg = NULL;
// Consider m for compilation
void StackWalkCompPolicy::method_invocation_event(methodHandle m, TRAPS) {
int hot_count = m->invocation_count();
reset_counter_for_invocation_event(m);
const char* comment = "count";
if (is_compilation_enabled() && m->code() == NULL && can_be_compiled(m)) {
ResourceMark rm(THREAD);
JavaThread *thread = (JavaThread*)THREAD;
frame fr = thread->last_frame();
assert(fr.is_interpreted_frame(), "must be interpreted");
assert(fr.interpreter_frame_method() == m(), "bad method");
if (TraceCompilationPolicy) {
tty->print("method invocation trigger: ");
m->print_short_name(tty);
tty->print(" ( interpreted " INTPTR_FORMAT ", size=%d ) ", (address)m(), m->code_size());
}
RegisterMap reg_map(thread, false);
javaVFrame* triggerVF = thread->last_java_vframe(®_map);
// triggerVF is the frame that triggered its counter
RFrame* first = new InterpretedRFrame(triggerVF->fr(), thread, m);
if (first->top_method()->code() != NULL) {
// called obsolete method/nmethod -- no need to recompile
if (TraceCompilationPolicy) tty->print_cr(" --> " INTPTR_FORMAT, first->top_method()->code());
} else {
if (TimeCompilationPolicy) accumulated_time()->start();
GrowableArray<RFrame*>* stack = new GrowableArray<RFrame*>(50);
stack->push(first);
RFrame* top = findTopInlinableFrame(stack);
if (TimeCompilationPolicy) accumulated_time()->stop();
assert(top != NULL, "findTopInlinableFrame returned null");
if (TraceCompilationPolicy) top->print();
CompileBroker::compile_method(top->top_method(), InvocationEntryBci, CompLevel_highest_tier,
m, hot_count, comment, CHECK);
}
}
}
void StackWalkCompPolicy::method_back_branch_event(methodHandle m, int bci, TRAPS) {
int hot_count = m->backedge_count();
const char* comment = "backedge_count";
if (is_compilation_enabled() && !m->is_not_osr_compilable() && can_be_compiled(m)) {
CompileBroker::compile_method(m, bci, CompLevel_highest_tier, m, hot_count, comment, CHECK);
NOT_PRODUCT(trace_osr_completion(m->lookup_osr_nmethod_for(bci, CompLevel_highest_tier, true));)
}
}
RFrame* StackWalkCompPolicy::findTopInlinableFrame(GrowableArray<RFrame*>* stack) {
// go up the stack until finding a frame that (probably) won't be inlined
// into its caller
RFrame* current = stack->at(0); // current choice for stopping
assert( current && !current->is_compiled(), "" );
const char* msg = NULL;
while (1) {
// before going up the stack further, check if doing so would get us into
// compiled code
RFrame* next = senderOf(current, stack);
if( !next ) // No next frame up the stack?
break; // Then compile with current frame
methodHandle m = current->top_method();
methodHandle next_m = next->top_method();
if (TraceCompilationPolicy && Verbose) {
tty->print("[caller: ");
next_m->print_short_name(tty);
tty->print("] ");
}
if( !Inline ) { // Inlining turned off
msg = "Inlining turned off";
break;
}
if (next_m->is_not_compilable()) { // Did fail to compile this before/
msg = "caller not compilable";
break;
}
if (next->num() > MaxRecompilationSearchLength) {
// don't go up too high when searching for recompilees
msg = "don't go up any further: > MaxRecompilationSearchLength";
break;
}
if (next->distance() > MaxInterpretedSearchLength) {
// don't go up too high when searching for recompilees
msg = "don't go up any further: next > MaxInterpretedSearchLength";
break;
}
// Compiled frame above already decided not to inline;
// do not recompile him.
if (next->is_compiled()) {
msg = "not going up into optimized code";
break;
}
// Interpreted frame above us was already compiled. Do not force
// a recompile, although if the frame above us runs long enough an
// OSR might still happen.
if( current->is_interpreted() && next_m->has_compiled_code() ) {
msg = "not going up -- already compiled caller";
break;
}
// Compute how frequent this call site is. We have current method 'm'.
// We know next method 'next_m' is interpreted. Find the call site and
// check the various invocation counts.
int invcnt = 0; // Caller counts
if (ProfileInterpreter) {
invcnt = next_m->interpreter_invocation_count();
}
int cnt = 0; // Call site counts
if (ProfileInterpreter && next_m->method_data() != NULL) {
ResourceMark rm;
int bci = next->top_vframe()->bci();
ProfileData* data = next_m->method_data()->bci_to_data(bci);
if (data != NULL && data->is_CounterData())
cnt = data->as_CounterData()->count();
}
// Caller counts / call-site counts; i.e. is this call site
// a hot call site for method next_m?
int freq = (invcnt) ? cnt/invcnt : cnt;
// Check size and frequency limits
if ((msg = shouldInline(m, freq, cnt)) != NULL) {
break;
}
// Check inlining negative tests
if ((msg = shouldNotInline(m)) != NULL) {
break;
}
// If the caller method is too big or something then we do not want to
// compile it just to inline a method
if (!can_be_compiled(next_m)) {
msg = "caller cannot be compiled";
break;
}
if( next_m->name() == vmSymbols::class_initializer_name() ) {
msg = "do not compile class initializer (OSR ok)";
break;
}
if (TraceCompilationPolicy && Verbose) {
tty->print("\n\t check caller: ");
next_m->print_short_name(tty);
tty->print(" ( interpreted " INTPTR_FORMAT ", size=%d ) ", (address)next_m(), next_m->code_size());
}
current = next;
}
assert( !current || !current->is_compiled(), "" );
if (TraceCompilationPolicy && msg) tty->print("(%s)\n", msg);
return current;
}
RFrame* StackWalkCompPolicy::senderOf(RFrame* rf, GrowableArray<RFrame*>* stack) {
RFrame* sender = rf->caller();
if (sender && sender->num() == stack->length()) stack->push(sender);
return sender;
}
const char* StackWalkCompPolicy::shouldInline(methodHandle m, float freq, int cnt) {
// Allows targeted inlining
// positive filter: should send be inlined? returns NULL (--> yes)
// or rejection msg
int max_size = MaxInlineSize;
int cost = m->code_size();
// Check for too many throws (and not too huge)
if (m->interpreter_throwout_count() > InlineThrowCount && cost < InlineThrowMaxSize ) {
return NULL;
}
// bump the max size if the call is frequent
if ((freq >= InlineFrequencyRatio) || (cnt >= InlineFrequencyCount)) {
if (TraceFrequencyInlining) {
tty->print("(Inlined frequent method)\n");
m->print();
}
max_size = FreqInlineSize;
}
if (cost > max_size) {
return (_msg = "too big");
}
return NULL;
}
const char* StackWalkCompPolicy::shouldNotInline(methodHandle m) {
// negative filter: should send NOT be inlined? returns NULL (--> inline) or rejection msg
if (m->is_abstract()) return (_msg = "abstract method");
// note: we allow ik->is_abstract()
if (!instanceKlass::cast(m->method_holder())->is_initialized()) return (_msg = "method holder not initialized");
if (m->is_native()) return (_msg = "native method");
nmethod* m_code = m->code();
if (m_code != NULL && m_code->code_size() > InlineSmallCode)
return (_msg = "already compiled into a big method");
// use frequency-based objections only for non-trivial methods
if (m->code_size() <= MaxTrivialSize) return NULL;
if (UseInterpreter) { // don't use counts with -Xcomp
if ((m->code() == NULL) && m->was_never_executed()) return (_msg = "never executed");
if (!m->was_executed_more_than(MIN2(MinInliningThreshold, CompileThreshold >> 1))) return (_msg = "executed < MinInliningThreshold times");
}
if (methodOopDesc::has_unloaded_classes_in_signature(m, JavaThread::current())) return (_msg = "unloaded signature classes");
return NULL;
}
#endif // COMPILER2