blob: 523500588d1539dfe3a99241559d194cffba7704 [file] [log] [blame]
// Copyright (c) 2013, the Dart project authors. Please see the AUTHORS file
// for details. All rights reserved. Use of this source code is governed by a
// BSD-style license that can be found in the LICENSE file.
#if !defined(DART_PRECOMPILED_RUNTIME)
#include "vm/compiler/backend/inliner.h"
#include "vm/compiler/aot/aot_call_specializer.h"
#include "vm/compiler/aot/precompiler.h"
#include "vm/compiler/backend/block_scheduler.h"
#include "vm/compiler/backend/branch_optimizer.h"
#include "vm/compiler/backend/flow_graph_checker.h"
#include "vm/compiler/backend/flow_graph_compiler.h"
#include "vm/compiler/backend/il_printer.h"
#include "vm/compiler/backend/type_propagator.h"
#include "vm/compiler/compiler_pass.h"
#include "vm/compiler/frontend/flow_graph_builder.h"
#include "vm/compiler/frontend/kernel_to_il.h"
#include "vm/compiler/jit/compiler.h"
#include "vm/compiler/jit/jit_call_specializer.h"
#include "vm/flags.h"
#include "vm/kernel.h"
#include "vm/longjump.h"
#include "vm/object.h"
#include "vm/object_store.h"
namespace dart {
DEFINE_FLAG(int,
deoptimization_counter_inlining_threshold,
12,
"How many times we allow deoptimization before we stop inlining.");
DEFINE_FLAG(bool, trace_inlining, false, "Trace inlining");
DEFINE_FLAG(charp, inlining_filter, NULL, "Inline only in named function");
// Flags for inlining heuristics.
DEFINE_FLAG(int,
inline_getters_setters_smaller_than,
10,
"Always inline getters and setters that have fewer instructions");
DEFINE_FLAG(int,
inlining_depth_threshold,
6,
"Inline function calls up to threshold nesting depth");
DEFINE_FLAG(
int,
inlining_size_threshold,
25,
"Always inline functions that have threshold or fewer instructions");
DEFINE_FLAG(int,
inlining_callee_call_sites_threshold,
1,
"Always inline functions containing threshold or fewer calls.");
DEFINE_FLAG(int,
inlining_callee_size_threshold,
80,
"Do not inline callees larger than threshold");
DEFINE_FLAG(int,
inlining_small_leaf_size_threshold,
50,
"Do not inline leaf callees larger than threshold");
DEFINE_FLAG(int,
inlining_caller_size_threshold,
50000,
"Stop inlining once caller reaches the threshold.");
DEFINE_FLAG(int,
inlining_constant_arguments_count,
1,
"Inline function calls with sufficient constant arguments "
"and up to the increased threshold on instructions");
DEFINE_FLAG(
int,
inlining_constant_arguments_max_size_threshold,
200,
"Do not inline callees larger than threshold if constant arguments");
DEFINE_FLAG(int,
inlining_constant_arguments_min_size_threshold,
60,
"Inline function calls with sufficient constant arguments "
"and up to the increased threshold on instructions");
DEFINE_FLAG(int,
inlining_hotness,
10,
"Inline only hotter calls, in percents (0 .. 100); "
"default 10%: calls above-equal 10% of max-count are inlined.");
DEFINE_FLAG(int,
inlining_recursion_depth_threshold,
1,
"Inline recursive function calls up to threshold recursion depth.");
DEFINE_FLAG(int,
max_inlined_per_depth,
500,
"Max. number of inlined calls per depth");
DEFINE_FLAG(bool, print_inlining_tree, false, "Print inlining tree");
DEFINE_FLAG(bool,
enable_inlining_annotations,
false,
"Enable inlining annotations");
DECLARE_FLAG(int, max_deoptimization_counter_threshold);
DECLARE_FLAG(bool, print_flow_graph);
DECLARE_FLAG(bool, print_flow_graph_optimized);
// Quick access to the current zone.
#define Z (zone())
#define I (isolate())
#define TRACE_INLINING(statement) \
do { \
if (trace_inlining()) statement; \
} while (false)
#define PRINT_INLINING_TREE(comment, caller, target, instance_call) \
do { \
if (FLAG_print_inlining_tree) { \
inlined_info_.Add(InlinedInfo(caller, target, inlining_depth_, \
instance_call, comment)); \
} \
} while (false)
// Test and obtain Smi value.
static bool IsSmiValue(Value* val, intptr_t* int_val) {
if (val->BindsToConstant() && val->BoundConstant().IsSmi()) {
*int_val = Smi::Cast(val->BoundConstant()).Value();
return true;
}
return false;
}
// Test if a call is recursive by looking in the deoptimization environment.
static bool IsCallRecursive(const Function& function, Definition* call) {
Environment* env = call->env();
while (env != NULL) {
if (function.raw() == env->function().raw()) {
return true;
}
env = env->outer();
}
return false;
}
// Helper to get the default value of a formal parameter.
static ConstantInstr* GetDefaultValue(intptr_t i,
const ParsedFunction& parsed_function) {
return new ConstantInstr(parsed_function.DefaultParameterValueAt(i));
}
// Pair of an argument name and its value.
struct NamedArgument {
String* name;
Value* value;
NamedArgument(String* name, Value* value) : name(name), value(value) {}
};
// Ensures we only inline callee graphs which are safe. There are certain
// instructions which cannot be inlined and we ensure here that we don't do
// that.
class CalleeGraphValidator : public AllStatic {
public:
static void Validate(FlowGraph* callee_graph) {
#ifdef DEBUG
for (BlockIterator block_it = callee_graph->reverse_postorder_iterator();
!block_it.Done(); block_it.Advance()) {
BlockEntryInstr* entry = block_it.Current();
for (ForwardInstructionIterator it(entry); !it.Done(); it.Advance()) {
Instruction* current = it.Current();
if (current->IsBranch()) {
current = current->AsBranch()->comparison();
}
// The following instructions are not safe to inline, since they make
// assumptions about the frame layout.
ASSERT(!current->IsTailCall());
ASSERT(!current->IsLoadIndexedUnsafe());
ASSERT(!current->IsStoreIndexedUnsafe());
}
}
#endif // DEBUG
}
};
// Helper to collect information about a callee graph when considering it for
// inlining.
class GraphInfoCollector : public ValueObject {
public:
GraphInfoCollector() : call_site_count_(0), instruction_count_(0) {}
void Collect(const FlowGraph& graph) {
call_site_count_ = 0;
instruction_count_ = 0;
for (BlockIterator block_it = graph.postorder_iterator(); !block_it.Done();
block_it.Advance()) {
// Skip any blocks from the prologue to make them not count towards the
// inlining instruction budget.
const intptr_t block_id = block_it.Current()->block_id();
if (graph.prologue_info().Contains(block_id)) {
continue;
}
for (ForwardInstructionIterator it(block_it.Current()); !it.Done();
it.Advance()) {
Instruction* current = it.Current();
// Don't count instructions that won't generate any code.
if (current->IsRedefinition()) {
continue;
}
++instruction_count_;
if (current->IsInstanceCall() || current->IsStaticCall() ||
current->IsClosureCall()) {
++call_site_count_;
continue;
}
if (current->IsPolymorphicInstanceCall()) {
PolymorphicInstanceCallInstr* call =
current->AsPolymorphicInstanceCall();
// These checks make sure that the number of call-sites counted does
// not change relative to the time when the current set of inlining
// parameters was fixed.
// TODO(fschneider): Determine new heuristic parameters that avoid
// these checks entirely.
if (!call->IsSureToCallSingleRecognizedTarget() &&
(call->instance_call()->token_kind() != Token::kEQ)) {
++call_site_count_;
}
}
}
}
}
intptr_t call_site_count() const { return call_site_count_; }
intptr_t instruction_count() const { return instruction_count_; }
private:
intptr_t call_site_count_;
intptr_t instruction_count_;
};
// Structure for collecting inline data needed to print inlining tree.
struct InlinedInfo {
const Function* caller;
const Function* inlined;
intptr_t inlined_depth;
const Definition* call_instr;
const char* bailout_reason;
InlinedInfo(const Function* caller_function,
const Function* inlined_function,
const intptr_t depth,
const Definition* call,
const char* reason)
: caller(caller_function),
inlined(inlined_function),
inlined_depth(depth),
call_instr(call),
bailout_reason(reason) {}
};
// A collection of call sites to consider for inlining.
class CallSites : public ValueObject {
public:
explicit CallSites(intptr_t threshold)
: inlining_depth_threshold_(threshold),
static_calls_(),
closure_calls_(),
instance_calls_() {}
struct InstanceCallInfo {
PolymorphicInstanceCallInstr* call;
double ratio;
const FlowGraph* caller_graph;
intptr_t nesting_depth;
InstanceCallInfo(PolymorphicInstanceCallInstr* call_arg,
FlowGraph* flow_graph,
intptr_t depth)
: call(call_arg),
ratio(0.0),
caller_graph(flow_graph),
nesting_depth(depth) {}
const Function& caller() const { return caller_graph->function(); }
};
struct StaticCallInfo {
StaticCallInstr* call;
double ratio;
FlowGraph* caller_graph;
intptr_t nesting_depth;
StaticCallInfo(StaticCallInstr* value,
FlowGraph* flow_graph,
intptr_t depth)
: call(value),
ratio(0.0),
caller_graph(flow_graph),
nesting_depth(depth) {}
const Function& caller() const { return caller_graph->function(); }
};
struct ClosureCallInfo {
ClosureCallInstr* call;
FlowGraph* caller_graph;
ClosureCallInfo(ClosureCallInstr* value, FlowGraph* flow_graph)
: call(value), caller_graph(flow_graph) {}
const Function& caller() const { return caller_graph->function(); }
};
const GrowableArray<InstanceCallInfo>& instance_calls() const {
return instance_calls_;
}
const GrowableArray<StaticCallInfo>& static_calls() const {
return static_calls_;
}
const GrowableArray<ClosureCallInfo>& closure_calls() const {
return closure_calls_;
}
bool HasCalls() const {
return !(static_calls_.is_empty() && closure_calls_.is_empty() &&
instance_calls_.is_empty());
}
intptr_t NumCalls() const {
return instance_calls_.length() + static_calls_.length() +
closure_calls_.length();
}
void Clear() {
static_calls_.Clear();
closure_calls_.Clear();
instance_calls_.Clear();
}
// Heuristic that maps the loop nesting depth to a static estimate of number
// of times code at that depth is executed (code at each higher nesting
// depth is assumed to execute 10x more often up to depth 3).
static intptr_t AotCallCountApproximation(intptr_t nesting_depth) {
switch (nesting_depth) {
case 0:
// The value 1 makes most sense, but it may give a high ratio to call
// sites outside loops. Therefore, such call sites are subject to
// subsequent stricter heuristic to limit code size increase.
return 1;
case 1:
return 10;
case 2:
return 10 * 10;
default:
return 10 * 10 * 10;
}
}
// Computes the ratio for each call site in a method, defined as the
// number of times a call site is executed over the maximum number of
// times any call site is executed in the method. JIT uses actual call
// counts whereas AOT uses a static estimate based on nesting depth.
void ComputeCallSiteRatio(intptr_t static_call_start_ix,
intptr_t instance_call_start_ix) {
const intptr_t num_static_calls =
static_calls_.length() - static_call_start_ix;
const intptr_t num_instance_calls =
instance_calls_.length() - instance_call_start_ix;
intptr_t max_count = 0;
GrowableArray<intptr_t> instance_call_counts(num_instance_calls);
for (intptr_t i = 0; i < num_instance_calls; ++i) {
const InstanceCallInfo& info =
instance_calls_[i + instance_call_start_ix];
intptr_t aggregate_count =
FLAG_precompiled_mode ? AotCallCountApproximation(info.nesting_depth)
: info.call->CallCount();
instance_call_counts.Add(aggregate_count);
if (aggregate_count > max_count) max_count = aggregate_count;
}
GrowableArray<intptr_t> static_call_counts(num_static_calls);
for (intptr_t i = 0; i < num_static_calls; ++i) {
const StaticCallInfo& info = static_calls_[i + static_call_start_ix];
intptr_t aggregate_count =
FLAG_precompiled_mode ? AotCallCountApproximation(info.nesting_depth)
: info.call->CallCount();
static_call_counts.Add(aggregate_count);
if (aggregate_count > max_count) max_count = aggregate_count;
}
// Note that max_count can be 0 if none of the calls was executed.
for (intptr_t i = 0; i < num_instance_calls; ++i) {
const double ratio =
(max_count == 0)
? 0.0
: static_cast<double>(instance_call_counts[i]) / max_count;
instance_calls_[i + instance_call_start_ix].ratio = ratio;
}
for (intptr_t i = 0; i < num_static_calls; ++i) {
const double ratio =
(max_count == 0)
? 0.0
: static_cast<double>(static_call_counts[i]) / max_count;
static_calls_[i + static_call_start_ix].ratio = ratio;
}
}
static void RecordAllNotInlinedFunction(
FlowGraph* graph,
intptr_t depth,
GrowableArray<InlinedInfo>* inlined_info) {
const Function* caller = &graph->function();
Function& target = Function::ZoneHandle();
for (BlockIterator block_it = graph->postorder_iterator(); !block_it.Done();
block_it.Advance()) {
for (ForwardInstructionIterator it(block_it.Current()); !it.Done();
it.Advance()) {
Instruction* current = it.Current();
Definition* call = NULL;
if (current->IsPolymorphicInstanceCall()) {
PolymorphicInstanceCallInstr* instance_call =
current->AsPolymorphicInstanceCall();
target = instance_call->targets().FirstTarget().raw();
call = instance_call;
} else if (current->IsStaticCall()) {
StaticCallInstr* static_call = current->AsStaticCall();
target = static_call->function().raw();
call = static_call;
} else if (current->IsClosureCall()) {
// TODO(srdjan): Add data for closure calls.
}
if (call != NULL) {
inlined_info->Add(
InlinedInfo(caller, &target, depth + 1, call, "Too deep"));
}
}
}
}
void FindCallSites(FlowGraph* graph,
intptr_t depth,
GrowableArray<InlinedInfo>* inlined_info) {
ASSERT(graph != NULL);
if (depth > inlining_depth_threshold_) {
if (FLAG_print_inlining_tree) {
RecordAllNotInlinedFunction(graph, depth, inlined_info);
}
return;
}
// Recognized methods are not treated as normal calls. They don't have
// calls in themselves, so we keep adding those even when at the threshold.
const bool inline_only_recognized_methods =
(depth == inlining_depth_threshold_);
// In AOT, compute loop hierarchy.
if (FLAG_precompiled_mode) {
graph->GetLoopHierarchy();
}
const intptr_t instance_call_start_ix = instance_calls_.length();
const intptr_t static_call_start_ix = static_calls_.length();
for (BlockIterator block_it = graph->postorder_iterator(); !block_it.Done();
block_it.Advance()) {
BlockEntryInstr* entry = block_it.Current();
const intptr_t depth = entry->NestingDepth();
for (ForwardInstructionIterator it(entry); !it.Done(); it.Advance()) {
Instruction* current = it.Current();
if (current->IsPolymorphicInstanceCall()) {
PolymorphicInstanceCallInstr* instance_call =
current->AsPolymorphicInstanceCall();
if (!inline_only_recognized_methods ||
instance_call->IsSureToCallSingleRecognizedTarget() ||
instance_call->HasOnlyDispatcherOrImplicitAccessorTargets()) {
instance_calls_.Add(InstanceCallInfo(instance_call, graph, depth));
} else {
// Method not inlined because inlining too deep and method
// not recognized.
if (FLAG_print_inlining_tree) {
const Function* caller = &graph->function();
const Function* target = &instance_call->targets().FirstTarget();
inlined_info->Add(InlinedInfo(caller, target, depth + 1,
instance_call, "Too deep"));
}
}
} else if (current->IsStaticCall()) {
StaticCallInstr* static_call = current->AsStaticCall();
if (!inline_only_recognized_methods ||
static_call->function().IsRecognized() ||
static_call->function().IsDispatcherOrImplicitAccessor()) {
static_calls_.Add(StaticCallInfo(static_call, graph, depth));
} else {
// Method not inlined because inlining too deep and method
// not recognized.
if (FLAG_print_inlining_tree) {
const Function* caller = &graph->function();
const Function* target = &static_call->function();
inlined_info->Add(InlinedInfo(caller, target, depth + 1,
static_call, "Too deep"));
}
}
} else if (current->IsClosureCall()) {
if (!inline_only_recognized_methods) {
ClosureCallInstr* closure_call = current->AsClosureCall();
closure_calls_.Add(ClosureCallInfo(closure_call, graph));
}
}
}
}
ComputeCallSiteRatio(static_call_start_ix, instance_call_start_ix);
}
private:
intptr_t inlining_depth_threshold_;
GrowableArray<StaticCallInfo> static_calls_;
GrowableArray<ClosureCallInfo> closure_calls_;
GrowableArray<InstanceCallInfo> instance_calls_;
DISALLOW_COPY_AND_ASSIGN(CallSites);
};
// Determines if inlining this graph yields a small leaf node.
static bool IsSmallLeaf(FlowGraph* graph) {
intptr_t instruction_count = 0;
for (BlockIterator block_it = graph->postorder_iterator(); !block_it.Done();
block_it.Advance()) {
BlockEntryInstr* entry = block_it.Current();
for (ForwardInstructionIterator it(entry); !it.Done(); it.Advance()) {
Instruction* current = it.Current();
++instruction_count;
if (current->IsInstanceCall() || current->IsPolymorphicInstanceCall() ||
current->IsClosureCall()) {
return false;
} else if (current->IsStaticCall()) {
const Function& function = current->AsStaticCall()->function();
const intptr_t inl_size = function.optimized_instruction_count();
// Accept a static call is always inlined in some way and add the
// cached size to the total instruction count. A reasonable guess
// is made if the count has not been collected yet (listed methods
// are never very large).
if (!function.always_inline() && !function.IsRecognized()) {
return false;
}
if (!function.always_inline()) {
static constexpr intptr_t kAvgListedMethodSize = 20;
instruction_count +=
(inl_size == 0 ? kAvgListedMethodSize : inl_size);
}
}
}
}
return instruction_count <= FLAG_inlining_small_leaf_size_threshold;
}
struct InlinedCallData {
InlinedCallData(Definition* call,
const Array& arguments_descriptor,
intptr_t first_arg_index, // 1 if type args are passed.
GrowableArray<Value*>* arguments,
const Function& caller,
intptr_t caller_inlining_id)
: call(call),
arguments_descriptor(arguments_descriptor),
first_arg_index(first_arg_index),
arguments(arguments),
callee_graph(NULL),
parameter_stubs(NULL),
exit_collector(NULL),
caller(caller),
caller_inlining_id(caller_inlining_id) {}
Definition* call;
const Array& arguments_descriptor;
const intptr_t first_arg_index;
GrowableArray<Value*>* arguments;
FlowGraph* callee_graph;
ZoneGrowableArray<Definition*>* parameter_stubs;
InlineExitCollector* exit_collector;
const Function& caller;
const intptr_t caller_inlining_id;
};
class CallSiteInliner;
class PolymorphicInliner : public ValueObject {
public:
PolymorphicInliner(CallSiteInliner* owner,
PolymorphicInstanceCallInstr* call,
const Function& caller_function,
intptr_t caller_inlining_id);
bool Inline();
private:
bool CheckInlinedDuplicate(const Function& target);
bool CheckNonInlinedDuplicate(const Function& target);
bool TryInliningPoly(const TargetInfo& target);
bool TryInlineRecognizedMethod(intptr_t receiver_cid, const Function& target);
TargetEntryInstr* BuildDecisionGraph();
Isolate* isolate() const;
Zone* zone() const;
intptr_t AllocateBlockId() const;
inline bool trace_inlining() const;
CallSiteInliner* const owner_;
PolymorphicInstanceCallInstr* const call_;
const intptr_t num_variants_;
const CallTargets& variants_;
CallTargets inlined_variants_;
// The non_inlined_variants_ can be used in a long-lived instruction object,
// so they are not embedded into the shorter-lived PolymorphicInliner object.
CallTargets* non_inlined_variants_;
GrowableArray<BlockEntryInstr*> inlined_entries_;
InlineExitCollector* exit_collector_;
const Function& caller_function_;
const intptr_t caller_inlining_id_;
};
static bool HasAnnotation(const Function& function, const char* annotation) {
const Class& owner = Class::Handle(function.Owner());
const Library& library = Library::Handle(owner.library());
auto& metadata_or_error = Object::Handle(library.GetMetadata(function));
if (metadata_or_error.IsError()) {
Report::LongJump(Error::Cast(metadata_or_error));
}
const Array& metadata = Array::Cast(metadata_or_error);
if (metadata.Length() > 0) {
Object& val = Object::Handle();
for (intptr_t i = 0; i < metadata.Length(); i++) {
val = metadata.At(i);
if (val.IsString() && String::Cast(val).Equals(annotation)) {
return true;
}
}
}
return false;
}
static void ReplaceParameterStubs(Zone* zone,
FlowGraph* caller_graph,
InlinedCallData* call_data,
const TargetInfo* target_info) {
const bool is_polymorphic = call_data->call->IsPolymorphicInstanceCall();
ASSERT(is_polymorphic == (target_info != NULL));
FlowGraph* callee_graph = call_data->callee_graph;
auto callee_entry = callee_graph->graph_entry()->normal_entry();
// Replace each stub with the actual argument or the caller's constant.
// Nulls denote optional parameters for which no actual was given.
const intptr_t first_arg_index = call_data->first_arg_index;
// When first_arg_index > 0, the stub and actual argument processed in the
// first loop iteration represent a passed-in type argument vector.
GrowableArray<Value*>* arguments = call_data->arguments;
intptr_t first_arg_stub_index = 0;
if (arguments->length() != call_data->parameter_stubs->length()) {
ASSERT(arguments->length() == call_data->parameter_stubs->length() - 1);
ASSERT(first_arg_index == 0);
// The first parameter stub accepts an optional type argument vector, but
// none was provided in arguments.
first_arg_stub_index = 1;
}
for (intptr_t i = 0; i < arguments->length(); ++i) {
Value* actual = (*arguments)[i];
Definition* defn = NULL;
if (is_polymorphic && (i == first_arg_index)) {
// Replace the receiver argument with a redefinition to prevent code from
// the inlined body from being hoisted above the inlined entry.
RedefinitionInstr* redefinition =
new (zone) RedefinitionInstr(actual->Copy(zone));
redefinition->set_ssa_temp_index(caller_graph->alloc_ssa_temp_index());
if (target_info->IsSingleCid()) {
redefinition->UpdateType(CompileType::FromCid(target_info->cid_start));
}
redefinition->InsertAfter(callee_entry);
defn = redefinition;
} else if (actual != NULL) {
defn = actual->definition();
}
if (defn != NULL) {
call_data->parameter_stubs->At(first_arg_stub_index + i)
->ReplaceUsesWith(defn);
}
}
// Replace remaining constants with uses by constants in the caller's
// initial definitions.
auto defns = callee_graph->graph_entry()->initial_definitions();
for (intptr_t i = 0; i < defns->length(); ++i) {
ConstantInstr* constant = (*defns)[i]->AsConstant();
if (constant != NULL && constant->HasUses()) {
constant->ReplaceUsesWith(caller_graph->GetConstant(constant->value()));
}
}
defns = callee_graph->graph_entry()->normal_entry()->initial_definitions();
for (intptr_t i = 0; i < defns->length(); ++i) {
ConstantInstr* constant = (*defns)[i]->AsConstant();
if (constant != NULL && constant->HasUses()) {
constant->ReplaceUsesWith(caller_graph->GetConstant(constant->value()));
}
SpecialParameterInstr* param = (*defns)[i]->AsSpecialParameter();
if (param != NULL && param->HasUses()) {
switch (param->kind()) {
case SpecialParameterInstr::kContext: {
ASSERT(!is_polymorphic);
// We do not support polymorphic inlining of closure calls.
ASSERT(call_data->call->IsClosureCall());
LoadFieldInstr* context_load = new (zone) LoadFieldInstr(
new Value((*arguments)[first_arg_index]->definition()),
Slot::Closure_context(), call_data->call->token_pos());
context_load->set_ssa_temp_index(
caller_graph->alloc_ssa_temp_index());
context_load->InsertBefore(callee_entry->next());
param->ReplaceUsesWith(context_load);
break;
}
case SpecialParameterInstr::kTypeArgs: {
Definition* type_args;
if (first_arg_index > 0) {
type_args = (*arguments)[0]->definition();
} else {
type_args = caller_graph->constant_null();
}
param->ReplaceUsesWith(type_args);
break;
}
case SpecialParameterInstr::kArgDescriptor: {
param->ReplaceUsesWith(
caller_graph->GetConstant(call_data->arguments_descriptor));
break;
}
default: {
UNREACHABLE();
break;
}
}
}
}
}
class CallSiteInliner : public ValueObject {
public:
explicit CallSiteInliner(FlowGraphInliner* inliner, intptr_t threshold)
: inliner_(inliner),
caller_graph_(inliner->flow_graph()),
inlined_(false),
initial_size_(inliner->flow_graph()->InstructionCount()),
inlined_size_(0),
inlined_recursive_call_(false),
inlining_depth_(1),
inlining_recursion_depth_(0),
inlining_depth_threshold_(threshold),
collected_call_sites_(NULL),
inlining_call_sites_(NULL),
function_cache_(),
inlined_info_() {}
FlowGraph* caller_graph() const { return caller_graph_; }
Thread* thread() const { return caller_graph_->thread(); }
Isolate* isolate() const { return caller_graph_->isolate(); }
Zone* zone() const { return caller_graph_->zone(); }
bool trace_inlining() const { return inliner_->trace_inlining(); }
int inlining_depth() { return inlining_depth_; }
struct InliningDecision {
InliningDecision(bool b, const char* r) : value(b), reason(r) {}
bool value;
const char* reason;
static InliningDecision Yes(const char* reason) {
return InliningDecision(true, reason);
}
static InliningDecision No(const char* reason) {
return InliningDecision(false, reason);
}
};
// Inlining heuristics based on Cooper et al. 2008.
InliningDecision ShouldWeInline(const Function& callee,
intptr_t instr_count,
intptr_t call_site_count,
intptr_t const_arg_count) {
if (inliner_->AlwaysInline(callee)) {
return InliningDecision::Yes("AlwaysInline");
}
if (inlined_size_ > FLAG_inlining_caller_size_threshold) {
// Prevent methods becoming humongous and thus slow to compile.
return InliningDecision::No("--inlining-caller-size-threshold");
}
if (const_arg_count > 0) {
if (instr_count > FLAG_inlining_constant_arguments_max_size_threshold) {
return InliningDecision(
false, "--inlining-constant-arguments-max-size-threshold");
}
} else if (instr_count > FLAG_inlining_callee_size_threshold) {
return InliningDecision::No("--inlining-callee-size-threshold");
}
int callee_inlining_depth = callee.inlining_depth();
if (callee_inlining_depth > 0 && callee_inlining_depth + inlining_depth_ >
FLAG_inlining_depth_threshold) {
return InliningDecision::No("--inlining-depth-threshold");
}
// 'instr_count' can be 0 if it was not computed yet.
if ((instr_count != 0) && (instr_count <= FLAG_inlining_size_threshold)) {
return InliningDecision::Yes("--inlining-size-threshold");
}
if (call_site_count <= FLAG_inlining_callee_call_sites_threshold) {
return InliningDecision::Yes("--inlining-callee-call-sites-threshold");
}
if ((const_arg_count >= FLAG_inlining_constant_arguments_count) &&
(instr_count <= FLAG_inlining_constant_arguments_min_size_threshold)) {
return InliningDecision(true,
"--inlining-constant-arguments-count and "
"inlining-constant-arguments-min-size-threshold");
}
return InliningDecision::No("default");
}
void InlineCalls() {
// If inlining depth is less than one abort.
if (inlining_depth_threshold_ < 1) return;
if (caller_graph_->function().deoptimization_counter() >=
FLAG_deoptimization_counter_inlining_threshold) {
return;
}
// Create two call site collections to swap between.
CallSites sites1(inlining_depth_threshold_);
CallSites sites2(inlining_depth_threshold_);
CallSites* call_sites_temp = NULL;
collected_call_sites_ = &sites1;
inlining_call_sites_ = &sites2;
// Collect initial call sites.
collected_call_sites_->FindCallSites(caller_graph_, inlining_depth_,
&inlined_info_);
while (collected_call_sites_->HasCalls()) {
TRACE_INLINING(
THR_Print(" Depth %" Pd " ----------\n", inlining_depth_));
if (FLAG_print_inlining_tree) {
THR_Print("**Depth % " Pd " calls to inline %" Pd " (threshold % " Pd
")\n",
inlining_depth_, collected_call_sites_->NumCalls(),
static_cast<intptr_t>(FLAG_max_inlined_per_depth));
}
if (collected_call_sites_->NumCalls() > FLAG_max_inlined_per_depth) {
break;
}
// Swap collected and inlining arrays and clear the new collecting array.
call_sites_temp = collected_call_sites_;
collected_call_sites_ = inlining_call_sites_;
inlining_call_sites_ = call_sites_temp;
collected_call_sites_->Clear();
// Inline call sites at the current depth.
bool inlined_instance = InlineInstanceCalls();
bool inlined_statics = InlineStaticCalls();
bool inlined_closures = InlineClosureCalls();
if (inlined_instance || inlined_statics || inlined_closures) {
// Increment the inlining depths. Checked before subsequent inlining.
++inlining_depth_;
if (inlined_recursive_call_) {
++inlining_recursion_depth_;
inlined_recursive_call_ = false;
}
thread()->CheckForSafepoint();
}
}
collected_call_sites_ = NULL;
inlining_call_sites_ = NULL;
}
bool inlined() const { return inlined_; }
double GrowthFactor() const {
return static_cast<double>(inlined_size_) /
static_cast<double>(initial_size_);
}
// Helper to create a parameter stub from an actual argument.
Definition* CreateParameterStub(intptr_t i,
Value* argument,
FlowGraph* graph) {
ConstantInstr* constant = argument->definition()->AsConstant();
if (constant != NULL) {
return new (Z) ConstantInstr(constant->value());
} else {
ParameterInstr* param = new (Z) ParameterInstr(i, graph->graph_entry());
param->UpdateType(*argument->Type());
return param;
}
}
bool TryInlining(const Function& function,
const Array& argument_names,
InlinedCallData* call_data,
bool stricter_heuristic) {
if (trace_inlining()) {
String& name = String::Handle(function.QualifiedUserVisibleName());
THR_Print(" => %s (deopt count %d)\n", name.ToCString(),
function.deoptimization_counter());
}
// Abort if the inlinable bit on the function is low.
if (!function.CanBeInlined()) {
TRACE_INLINING(THR_Print(" Bailout: not inlinable\n"));
PRINT_INLINING_TREE("Not inlinable", &call_data->caller, &function,
call_data->call);
return false;
}
// Don't inline any intrinsified functions in precompiled mode
// to reduce code size and make sure we use the intrinsic code.
if (FLAG_precompiled_mode && function.is_intrinsic() &&
!inliner_->AlwaysInline(function)) {
TRACE_INLINING(THR_Print(" Bailout: intrinisic\n"));
PRINT_INLINING_TREE("intrinsic", &call_data->caller, &function,
call_data->call);
return false;
}
// Do not rely on function type feedback or presence of code to determine
// if a function was compiled.
if (!FLAG_precompiled_mode && !function.WasCompiled()) {
TRACE_INLINING(THR_Print(" Bailout: not compiled yet\n"));
PRINT_INLINING_TREE("Not compiled", &call_data->caller, &function,
call_data->call);
return false;
}
// Type feedback may have been cleared for this function (ClearICDataArray),
// but we need it for inlining.
if (!FLAG_precompiled_mode && (function.ic_data_array() == Array::null())) {
TRACE_INLINING(THR_Print(" Bailout: type feedback cleared\n"));
PRINT_INLINING_TREE("Not compiled", &call_data->caller, &function,
call_data->call);
return false;
}
// Abort if this function has deoptimized too much.
if (function.deoptimization_counter() >=
FLAG_max_deoptimization_counter_threshold) {
function.set_is_inlinable(false);
TRACE_INLINING(THR_Print(" Bailout: deoptimization threshold\n"));
PRINT_INLINING_TREE("Deoptimization threshold exceeded",
&call_data->caller, &function, call_data->call);
return false;
}
const char* kNeverInlineAnnotation = "NeverInline";
if (FLAG_enable_inlining_annotations &&
HasAnnotation(function, kNeverInlineAnnotation)) {
TRACE_INLINING(THR_Print(" Bailout: NeverInline annotation\n"));
return false;
}
GrowableArray<Value*>* arguments = call_data->arguments;
const intptr_t constant_arguments = CountConstants(*arguments);
InliningDecision decision = ShouldWeInline(
function, function.optimized_instruction_count(),
function.optimized_call_site_count(), constant_arguments);
if (!decision.value) {
TRACE_INLINING(
THR_Print(" Bailout: early heuristics (%s) with "
"code size: %" Pd ", "
"call sites: %" Pd ", "
"inlining depth of callee: %d, "
"const args: %" Pd "\n",
decision.reason, function.optimized_instruction_count(),
function.optimized_call_site_count(),
function.inlining_depth(), constant_arguments));
PRINT_INLINING_TREE("Early heuristic", &call_data->caller, &function,
call_data->call);
return false;
}
// Abort if this is a recursive occurrence.
Definition* call = call_data->call;
// Added 'volatile' works around a possible GCC 4.9 compiler bug.
volatile bool is_recursive_call = IsCallRecursive(function, call);
if (is_recursive_call &&
inlining_recursion_depth_ >= FLAG_inlining_recursion_depth_threshold) {
TRACE_INLINING(THR_Print(" Bailout: recursive function\n"));
PRINT_INLINING_TREE("Recursive function", &call_data->caller, &function,
call_data->call);
return false;
}
Error& error = Error::Handle();
{
// Save and clear deopt id.
DeoptIdScope deopt_id_scope(thread(), 0);
// Install bailout jump.
LongJumpScope jump;
if (setjmp(*jump.Set()) == 0) {
Isolate* isolate = Isolate::Current();
// Makes sure no classes are loaded during parsing in background.
const intptr_t loading_invalidation_gen_at_start =
isolate->loading_invalidation_gen();
if (Compiler::IsBackgroundCompilation()) {
if ((loading_invalidation_gen_at_start !=
isolate->loading_invalidation_gen())) {
// Loading occured while parsing. We need to abort here because
// state changed while compiling.
Compiler::AbortBackgroundCompilation(
DeoptId::kNone, "Loading occured while parsing in inliner");
}
}
// Load IC data for the callee.
ZoneGrowableArray<const ICData*>* ic_data_array =
new (Z) ZoneGrowableArray<const ICData*>();
const bool clone_ic_data = Compiler::IsBackgroundCompilation();
function.RestoreICDataMap(ic_data_array, clone_ic_data);
if (Compiler::IsBackgroundCompilation() &&
(function.ic_data_array() == Array::null())) {
Compiler::AbortBackgroundCompilation(DeoptId::kNone,
"ICData cleared while inlining");
}
// Parse the callee function.
bool in_cache;
ParsedFunction* parsed_function;
{
parsed_function = GetParsedFunction(function, &in_cache);
if (!function.CanBeInlined()) {
// As a side effect of parsing the function, it may be marked
// as not inlinable. This happens for async and async* functions
// when causal stack traces are being tracked.
return false;
}
}
// Build the callee graph.
InlineExitCollector* exit_collector =
new (Z) InlineExitCollector(caller_graph_, call);
FlowGraph* callee_graph;
Code::EntryKind entry_kind = Code::EntryKind::kNormal;
if (StaticCallInstr* instr = call_data->call->AsStaticCall()) {
entry_kind = instr->entry_kind();
} else if (InstanceCallInstr* instr =
call_data->call->AsInstanceCall()) {
entry_kind = instr->entry_kind();
} else if (PolymorphicInstanceCallInstr* instr =
call_data->call->AsPolymorphicInstanceCall()) {
entry_kind = instr->instance_call()->entry_kind();
} else if (ClosureCallInstr* instr = call_data->call->AsClosureCall()) {
entry_kind = instr->entry_kind();
}
kernel::FlowGraphBuilder builder(
parsed_function, ic_data_array, /* not building var desc */ NULL,
exit_collector,
/* optimized = */ true, Compiler::kNoOSRDeoptId,
caller_graph_->max_block_id() + 1,
entry_kind == Code::EntryKind::kUnchecked);
{
callee_graph = builder.BuildGraph();
#if defined(DEBUG)
FlowGraphChecker(callee_graph).Check("Builder (callee)");
#endif
CalleeGraphValidator::Validate(callee_graph);
}
#if defined(DART_PRECOMPILER) && !defined(TARGET_ARCH_DBC) && \
!defined(TARGET_ARCH_IA32)
if (FLAG_precompiled_mode) {
callee_graph->PopulateWithICData(parsed_function->function());
}
#endif
// If we inline a function which is intrinsified without a fall-through
// to IR code, we will not have any ICData attached, so we do it
// manually here.
if (!FLAG_precompiled_mode && function.is_intrinsic()) {
callee_graph->PopulateWithICData(parsed_function->function());
}
// The parameter stubs are a copy of the actual arguments providing
// concrete information about the values, for example constant values,
// without linking between the caller and callee graphs.
// TODO(zerny): Put more information in the stubs, eg, type information.
const intptr_t first_actual_param_index = call_data->first_arg_index;
const intptr_t inlined_type_args_param = function.IsGeneric() ? 1 : 0;
const intptr_t num_inlined_params =
inlined_type_args_param + function.NumParameters();
ZoneGrowableArray<Definition*>* param_stubs =
new (Z) ZoneGrowableArray<Definition*>(num_inlined_params);
// Create a ConstantInstr as Definition for the type arguments, if any.
if (first_actual_param_index > 0) {
// A type argument vector is explicitly passed.
param_stubs->Add(
CreateParameterStub(-1, (*arguments)[0], callee_graph));
} else if (inlined_type_args_param > 0) {
// No type argument vector is passed to the generic function,
// pass a null vector, which is the same as a vector of dynamic types.
param_stubs->Add(callee_graph->GetConstant(Object::ZoneHandle()));
}
// Create a parameter stub for each fixed positional parameter.
for (intptr_t i = 0; i < function.num_fixed_parameters(); ++i) {
param_stubs->Add(CreateParameterStub(
i, (*arguments)[first_actual_param_index + i], callee_graph));
}
// If the callee has optional parameters, rebuild the argument and stub
// arrays so that actual arguments are in one-to-one with the formal
// parameters.
if (function.HasOptionalParameters()) {
TRACE_INLINING(THR_Print(" adjusting for optional parameters\n"));
if (!AdjustForOptionalParameters(
*parsed_function, first_actual_param_index, argument_names,
arguments, param_stubs, callee_graph)) {
function.set_is_inlinable(false);
TRACE_INLINING(THR_Print(" Bailout: optional arg mismatch\n"));
PRINT_INLINING_TREE("Optional arg mismatch", &call_data->caller,
&function, call_data->call);
return false;
}
}
// After treating optional parameters the actual/formal count must
// match.
ASSERT(arguments->length() ==
first_actual_param_index + function.NumParameters());
// Update try-index of the callee graph.
BlockEntryInstr* call_block = call_data->call->GetBlock();
if (call_block->InsideTryBlock()) {
intptr_t try_index = call_block->try_index();
for (BlockIterator it = callee_graph->reverse_postorder_iterator();
!it.Done(); it.Advance()) {
BlockEntryInstr* block = it.Current();
block->set_try_index(try_index);
}
}
BlockScheduler block_scheduler(callee_graph);
block_scheduler.AssignEdgeWeights();
{
// Compute SSA on the callee graph, catching bailouts.
callee_graph->ComputeSSA(caller_graph_->max_virtual_register_number(),
param_stubs);
#if defined(DEBUG)
FlowGraphChecker(callee_graph).Check("SSA (callee)");
#endif
}
if (FLAG_support_il_printer && trace_inlining() &&
(FLAG_print_flow_graph || FLAG_print_flow_graph_optimized)) {
THR_Print("Callee graph for inlining %s (unoptimized)\n",
function.ToFullyQualifiedCString());
FlowGraphPrinter printer(*callee_graph);
printer.PrintBlocks();
}
{
// TODO(fschneider): Improve suppression of speculative inlining.
// Deopt-ids overlap between caller and callee.
if (FLAG_precompiled_mode) {
#if defined(DART_PRECOMPILER) && !defined(TARGET_ARCH_DBC) && \
!defined(TARGET_ARCH_IA32)
AotCallSpecializer call_specializer(inliner_->precompiler_,
callee_graph,
inliner_->speculative_policy_);
CompilerPassState state(Thread::Current(), callee_graph,
inliner_->speculative_policy_);
state.call_specializer = &call_specializer;
CompilerPass::RunInliningPipeline(CompilerPass::kAOT, &state);
#else
UNREACHABLE();
#endif // defined(DART_PRECOMPILER) && !defined(TARGET_ARCH_DBC) && \
// !defined(TARGET_ARCH_IA32)
} else {
JitCallSpecializer call_specializer(callee_graph,
inliner_->speculative_policy_);
CompilerPassState state(Thread::Current(), callee_graph,
inliner_->speculative_policy_);
state.call_specializer = &call_specializer;
CompilerPass::RunInliningPipeline(CompilerPass::kJIT, &state);
}
}
if (FLAG_support_il_printer && trace_inlining() &&
(FLAG_print_flow_graph || FLAG_print_flow_graph_optimized)) {
THR_Print("Callee graph for inlining %s (optimized)\n",
function.ToFullyQualifiedCString());
FlowGraphPrinter printer(*callee_graph);
printer.PrintBlocks();
}
// Collect information about the call site and caller graph.
// TODO(zerny): Do this after CP and dead code elimination.
intptr_t constants_count = 0;
for (intptr_t i = 0; i < param_stubs->length(); ++i) {
if ((*param_stubs)[i]->IsConstant()) ++constants_count;
}
FlowGraphInliner::CollectGraphInfo(callee_graph);
const intptr_t size = function.optimized_instruction_count();
const intptr_t call_site_count = function.optimized_call_site_count();
// Use heuristics do decide if this call should be inlined.
InliningDecision decision =
ShouldWeInline(function, size, call_site_count, constants_count);
if (!decision.value) {
// If size is larger than all thresholds, don't consider it again.
if ((size > FLAG_inlining_size_threshold) &&
(call_site_count > FLAG_inlining_callee_call_sites_threshold) &&
(size > FLAG_inlining_constant_arguments_min_size_threshold) &&
(size > FLAG_inlining_constant_arguments_max_size_threshold)) {
function.set_is_inlinable(false);
}
TRACE_INLINING(
THR_Print(" Bailout: heuristics (%s) with "
"code size: %" Pd ", "
"call sites: %" Pd ", "
"inlining depth of callee: %d, "
"const args: %" Pd "\n",
decision.reason, size, call_site_count,
function.inlining_depth(), constants_count));
PRINT_INLINING_TREE("Heuristic fail", &call_data->caller, &function,
call_data->call);
return false;
}
// If requested, a stricter heuristic is applied to this inlining. This
// heuristic always scans the method (rather than possibly reusing
// cached results) to make sure all specializations are accounted for.
if (stricter_heuristic) {
if (!IsSmallLeaf(callee_graph)) {
TRACE_INLINING(
THR_Print(" Bailout: heuristics (no small leaf)\n"));
PRINT_INLINING_TREE("Heuristic fail (no small leaf)",
&call_data->caller, &function, call_data->call);
return false;
}
}
// Inline dispatcher methods regardless of the current depth.
const intptr_t depth =
function.IsDispatcherOrImplicitAccessor() ? 0 : inlining_depth_;
collected_call_sites_->FindCallSites(callee_graph, depth,
&inlined_info_);
// Add the function to the cache.
if (!in_cache) {
function_cache_.Add(parsed_function);
}
// Build succeeded so we restore the bailout jump.
inlined_ = true;
inlined_size_ += size;
if (is_recursive_call) {
inlined_recursive_call_ = true;
}
call_data->callee_graph = callee_graph;
call_data->parameter_stubs = param_stubs;
call_data->exit_collector = exit_collector;
// When inlined, we add the guarded fields of the callee to the caller's
// list of guarded fields.
const ZoneGrowableArray<const Field*>& callee_guarded_fields =
*callee_graph->parsed_function().guarded_fields();
for (intptr_t i = 0; i < callee_guarded_fields.length(); ++i) {
caller_graph()->parsed_function().AddToGuardedFields(
callee_guarded_fields[i]);
}
// When inlined, we add the deferred prefixes of the callee to the
// caller's list of deferred prefixes.
caller_graph()->AddToDeferredPrefixes(
callee_graph->deferred_prefixes());
FlowGraphInliner::SetInliningId(
callee_graph,
inliner_->NextInlineId(callee_graph->function(),
call_data->call->token_pos(),
call_data->caller_inlining_id));
TRACE_INLINING(THR_Print(" Success\n"));
TRACE_INLINING(THR_Print(
" with reason %s, code size %" Pd ", call sites: %" Pd "\n",
decision.reason, function.optimized_instruction_count(),
call_site_count));
PRINT_INLINING_TREE(NULL, &call_data->caller, &function, call);
return true;
} else {
error = thread()->StealStickyError();
if (error.IsLanguageError() &&
(LanguageError::Cast(error).kind() == Report::kBailout)) {
if (error.raw() == Object::background_compilation_error().raw()) {
// Fall through to exit the compilation, and retry it later.
} else {
TRACE_INLINING(
THR_Print(" Bailout: %s\n", error.ToErrorCString()));
PRINT_INLINING_TREE("Bailout", &call_data->caller, &function, call);
return false;
}
} else {
// Fall through to exit long jump scope.
}
}
}
// Propagate a compile-time error. In precompilation we attempt to
// inline functions that have never been compiled before; when JITing we
// should only see language errors in unoptimized compilation.
// Otherwise, there can be an out-of-memory error (unhandled exception).
// In background compilation we may abort compilation as the state
// changes while compiling. Propagate that 'error' and retry compilation
// later.
ASSERT(FLAG_precompiled_mode || Compiler::IsBackgroundCompilation() ||
error.IsUnhandledException());
Thread::Current()->long_jump_base()->Jump(1, error);
UNREACHABLE();
return false;
}
void PrintInlinedInfo(const Function& top) {
if (inlined_info_.length() > 0) {
THR_Print("Inlining into: '%s'\n growth: %f (%" Pd " -> %" Pd ")\n",
top.ToFullyQualifiedCString(), GrowthFactor(), initial_size_,
inlined_size_);
PrintInlinedInfoFor(top, 1);
}
}
private:
friend class PolymorphicInliner;
static bool Contains(const GrowableArray<intptr_t>& a, intptr_t deopt_id) {
for (intptr_t i = 0; i < a.length(); i++) {
if (a[i] == deopt_id) return true;
}
return false;
}
void PrintInlinedInfoFor(const Function& caller, intptr_t depth) {
// Prevent duplicate printing as inlined_info aggregates all inlinining.
GrowableArray<intptr_t> call_instructions_printed;
// Print those that were inlined.
for (intptr_t i = 0; i < inlined_info_.length(); i++) {
const InlinedInfo& info = inlined_info_[i];
if (info.bailout_reason != NULL) {
continue;
}
if ((info.inlined_depth == depth) &&
(info.caller->raw() == caller.raw()) &&
!Contains(call_instructions_printed, info.call_instr->GetDeoptId())) {
for (int t = 0; t < depth; t++) {
THR_Print(" ");
}
THR_Print("%" Pd " %s\n", info.call_instr->GetDeoptId(),
info.inlined->ToQualifiedCString());
PrintInlinedInfoFor(*info.inlined, depth + 1);
call_instructions_printed.Add(info.call_instr->GetDeoptId());
}
}
call_instructions_printed.Clear();
// Print those that were not inlined.
for (intptr_t i = 0; i < inlined_info_.length(); i++) {
const InlinedInfo& info = inlined_info_[i];
if (info.bailout_reason == NULL) {
continue;
}
if ((info.inlined_depth == depth) &&
(info.caller->raw() == caller.raw()) &&
!Contains(call_instructions_printed, info.call_instr->GetDeoptId())) {
for (int t = 0; t < depth; t++) {
THR_Print(" ");
}
THR_Print("NO %" Pd " %s - %s\n", info.call_instr->GetDeoptId(),
info.inlined->ToQualifiedCString(), info.bailout_reason);
call_instructions_printed.Add(info.call_instr->GetDeoptId());
}
}
}
void InlineCall(InlinedCallData* call_data) {
FlowGraph* callee_graph = call_data->callee_graph;
auto callee_function_entry = callee_graph->graph_entry()->normal_entry();
// Plug result in the caller graph.
InlineExitCollector* exit_collector = call_data->exit_collector;
exit_collector->PrepareGraphs(callee_graph);
exit_collector->ReplaceCall(callee_function_entry);
ReplaceParameterStubs(zone(), caller_graph_, call_data, NULL);
// Remove push arguments of the call.
Definition* call = call_data->call;
for (intptr_t i = 0; i < call->ArgumentCount(); ++i) {
PushArgumentInstr* push = call->PushArgumentAt(i);
push->ReplaceUsesWith(push->value()->definition());
push->RemoveFromGraph();
}
}
static intptr_t CountConstants(const GrowableArray<Value*>& arguments) {
intptr_t count = 0;
for (intptr_t i = 0; i < arguments.length(); i++) {
if (arguments[i]->BindsToConstant()) count++;
}
return count;
}
// Parse a function reusing the cache if possible.
ParsedFunction* GetParsedFunction(const Function& function, bool* in_cache) {
// TODO(zerny): Use a hash map for the cache.
for (intptr_t i = 0; i < function_cache_.length(); ++i) {
ParsedFunction* parsed_function = function_cache_[i];
if (parsed_function->function().raw() == function.raw()) {
*in_cache = true;
return parsed_function;
}
}
*in_cache = false;
ParsedFunction* parsed_function =
new (Z) ParsedFunction(thread(), function);
return parsed_function;
}
bool InlineStaticCalls() {
bool inlined = false;
const GrowableArray<CallSites::StaticCallInfo>& call_info =
inlining_call_sites_->static_calls();
TRACE_INLINING(THR_Print(" Static Calls (%" Pd ")\n", call_info.length()));
for (intptr_t call_idx = 0; call_idx < call_info.length(); ++call_idx) {
StaticCallInstr* call = call_info[call_idx].call;
if (FlowGraphInliner::TryReplaceStaticCallWithInline(
inliner_->flow_graph(), NULL, call,
inliner_->speculative_policy_)) {
inlined = true;
continue;
}
const Function& target = call->function();
if (!inliner_->AlwaysInline(target) &&
(call_info[call_idx].ratio * 100) < FLAG_inlining_hotness) {
if (trace_inlining()) {
String& name = String::Handle(target.QualifiedUserVisibleName());
THR_Print(" => %s (deopt count %d)\n Bailout: cold %f\n",
name.ToCString(), target.deoptimization_counter(),
call_info[call_idx].ratio);
}
PRINT_INLINING_TREE("Too cold", &call_info[call_idx].caller(),
&call->function(), call);
continue;
}
GrowableArray<Value*> arguments(call->ArgumentCount());
for (int i = 0; i < call->ArgumentCount(); ++i) {
arguments.Add(call->PushArgumentAt(i)->value());
}
InlinedCallData call_data(
call, Array::ZoneHandle(Z, call->GetArgumentsDescriptor()),
call->FirstArgIndex(), &arguments, call_info[call_idx].caller(),
call_info[call_idx].caller_graph->inlining_id());
// Under AOT, calls outside loops may pass our regular heuristics due
// to a relatively high ratio. So, unless we are optimizing solely for
// speed, such call sites are subject to subsequent stricter heuristic
// to limit code size increase.
bool stricter_heuristic = FLAG_precompiled_mode &&
FLAG_optimization_level <= 2 &&
!inliner_->AlwaysInline(target) &&
call_info[call_idx].nesting_depth == 0;
if (TryInlining(call->function(), call->argument_names(), &call_data,
stricter_heuristic)) {
InlineCall(&call_data);
inlined = true;
}
}
return inlined;
}
bool InlineClosureCalls() {
bool inlined = false;
const GrowableArray<CallSites::ClosureCallInfo>& call_info =
inlining_call_sites_->closure_calls();
TRACE_INLINING(
THR_Print(" Closure Calls (%" Pd ")\n", call_info.length()));
for (intptr_t call_idx = 0; call_idx < call_info.length(); ++call_idx) {
ClosureCallInstr* call = call_info[call_idx].call;
// Find the closure of the callee.
ASSERT(call->ArgumentCount() > 0);
Function& target = Function::ZoneHandle();
Definition* receiver =
call->Receiver()->definition()->OriginalDefinition();
if (AllocateObjectInstr* alloc = receiver->AsAllocateObject()) {
if (!alloc->closure_function().IsNull()) {
target = alloc->closure_function().raw();
ASSERT(alloc->cls().IsClosureClass());
}
} else if (ConstantInstr* constant = receiver->AsConstant()) {
if (constant->value().IsClosure()) {
target = Closure::Cast(constant->value()).function();
}
}
if (target.IsNull()) {
TRACE_INLINING(THR_Print(" Bailout: non-closure operator\n"));
continue;
}
if (call->ArgumentCount() > target.NumParameters() ||
call->ArgumentCount() < target.num_fixed_parameters()) {
TRACE_INLINING(THR_Print(" Bailout: wrong parameter count\n"));
continue;
}
GrowableArray<Value*> arguments(call->ArgumentCount());
for (int i = 0; i < call->ArgumentCount(); ++i) {
arguments.Add(call->PushArgumentAt(i)->value());
}
const Array& arguments_descriptor =
Array::ZoneHandle(Z, call->GetArgumentsDescriptor());
InlinedCallData call_data(
call, arguments_descriptor, call->FirstArgIndex(), &arguments,
call_info[call_idx].caller(),
call_info[call_idx].caller_graph->inlining_id());
if (TryInlining(target, call->argument_names(), &call_data, false)) {
InlineCall(&call_data);
inlined = true;
}
}
return inlined;
}
bool InlineInstanceCalls() {
bool inlined = false;
const GrowableArray<CallSites::InstanceCallInfo>& call_info =
inlining_call_sites_->instance_calls();
TRACE_INLINING(THR_Print(" Polymorphic Instance Calls (%" Pd ")\n",
call_info.length()));
for (intptr_t call_idx = 0; call_idx < call_info.length(); ++call_idx) {
PolymorphicInstanceCallInstr* call = call_info[call_idx].call;
// PolymorphicInliner introduces deoptimization paths.
if (!call->complete() && !FLAG_polymorphic_with_deopt) {
TRACE_INLINING(
THR_Print(" => %s\n Bailout: call with checks\n",
call->instance_call()->function_name().ToCString()));
continue;
}
const Function& cl = call_info[call_idx].caller();
intptr_t caller_inlining_id =
call_info[call_idx].caller_graph->inlining_id();
PolymorphicInliner inliner(this, call, cl, caller_inlining_id);
if (inliner.Inline()) inlined = true;
}
return inlined;
}
bool AdjustForOptionalParameters(const ParsedFunction& parsed_function,
intptr_t first_arg_index,
const Array& argument_names,
GrowableArray<Value*>* arguments,
ZoneGrowableArray<Definition*>* param_stubs,
FlowGraph* callee_graph) {
const Function& function = parsed_function.function();
// The language and this code does not support both optional positional
// and optional named parameters for the same function.
ASSERT(!function.HasOptionalPositionalParameters() ||
!function.HasOptionalNamedParameters());
intptr_t arg_count = arguments->length();
intptr_t param_count = function.NumParameters();
intptr_t fixed_param_count = function.num_fixed_parameters();
ASSERT(fixed_param_count <= arg_count - first_arg_index);
ASSERT(arg_count - first_arg_index <= param_count);
if (function.HasOptionalPositionalParameters()) {
// Create a stub for each optional positional parameters with an actual.
for (intptr_t i = first_arg_index + fixed_param_count; i < arg_count;
++i) {
param_stubs->Add(CreateParameterStub(i, (*arguments)[i], callee_graph));
}
ASSERT(function.NumOptionalPositionalParameters() ==
(param_count - fixed_param_count));
// For each optional positional parameter without an actual, add its
// default value.
for (intptr_t i = arg_count - first_arg_index; i < param_count; ++i) {
const Instance& object =
parsed_function.DefaultParameterValueAt(i - fixed_param_count);
ConstantInstr* constant = new (Z) ConstantInstr(object);
arguments->Add(NULL);
param_stubs->Add(constant);
}
return true;
}
ASSERT(function.HasOptionalNamedParameters());
// Passed arguments (not counting optional type args) must match fixed
// parameters plus named arguments.
intptr_t argument_names_count =
(argument_names.IsNull()) ? 0 : argument_names.Length();
ASSERT((arg_count - first_arg_index) ==
(fixed_param_count + argument_names_count));
// Fast path when no optional named parameters are given.
if (argument_names_count == 0) {
for (intptr_t i = 0; i < param_count - fixed_param_count; ++i) {
arguments->Add(NULL);
param_stubs->Add(GetDefaultValue(i, parsed_function));
}
return true;
}
// Otherwise, build a collection of name/argument pairs.
GrowableArray<NamedArgument> named_args(argument_names_count);
for (intptr_t i = 0; i < argument_names.Length(); ++i) {
String& arg_name = String::Handle(caller_graph_->zone());
arg_name ^= argument_names.At(i);
named_args.Add(NamedArgument(
&arg_name, (*arguments)[first_arg_index + fixed_param_count + i]));
}
// Truncate the arguments array to just type args and fixed parameters.
arguments->TruncateTo(first_arg_index + fixed_param_count);
// For each optional named parameter, add the actual argument or its
// default if no argument is passed.
intptr_t match_count = 0;
for (intptr_t i = fixed_param_count; i < param_count; ++i) {
String& param_name = String::Handle(function.ParameterNameAt(i));
// Search for and add the named argument.
Value* arg = NULL;
for (intptr_t j = 0; j < named_args.length(); ++j) {
if (param_name.Equals(*named_args[j].name)) {
arg = named_args[j].value;
match_count++;
break;
}
}
arguments->Add(arg);
// Create a stub for the argument or use the parameter's default value.
if (arg != NULL) {
param_stubs->Add(
CreateParameterStub(first_arg_index + i, arg, callee_graph));
} else {
param_stubs->Add(
GetDefaultValue(i - fixed_param_count, parsed_function));
}
}
return argument_names_count == match_count;
}
FlowGraphInliner* inliner_;
FlowGraph* caller_graph_;
bool inlined_;
const intptr_t initial_size_;
intptr_t inlined_size_;
bool inlined_recursive_call_;
intptr_t inlining_depth_;
intptr_t inlining_recursion_depth_;
intptr_t inlining_depth_threshold_;
CallSites* collected_call_sites_;
CallSites* inlining_call_sites_;
GrowableArray<ParsedFunction*> function_cache_;
GrowableArray<InlinedInfo> inlined_info_;
DISALLOW_COPY_AND_ASSIGN(CallSiteInliner);
};
PolymorphicInliner::PolymorphicInliner(CallSiteInliner* owner,
PolymorphicInstanceCallInstr* call,
const Function& caller_function,
intptr_t caller_inlining_id)
: owner_(owner),
call_(call),
num_variants_(call->NumberOfChecks()),
variants_(call->targets_),
inlined_variants_(zone()),
non_inlined_variants_(new (zone()) CallTargets(zone())),
inlined_entries_(num_variants_),
exit_collector_(new (Z) InlineExitCollector(owner->caller_graph(), call)),
caller_function_(caller_function),
caller_inlining_id_(caller_inlining_id) {}
Isolate* PolymorphicInliner::isolate() const {
return owner_->caller_graph()->isolate();
}
Zone* PolymorphicInliner::zone() const {
return owner_->caller_graph()->zone();
}
intptr_t PolymorphicInliner::AllocateBlockId() const {
return owner_->caller_graph()->allocate_block_id();
}
// Inlined bodies are shared if two different class ids have the same
// inlined target. This sharing is represented by using three different
// types of entries in the inlined_entries_ array:
//
// * GraphEntry: the inlined body is not shared.
//
// * TargetEntry: the inlined body is shared and this is the first variant.
//
// * JoinEntry: the inlined body is shared and this is a subsequent variant.
bool PolymorphicInliner::CheckInlinedDuplicate(const Function& target) {
for (intptr_t i = 0; i < inlined_variants_.length(); ++i) {
if ((target.raw() == inlined_variants_.TargetAt(i)->target->raw()) &&
!MethodRecognizer::PolymorphicTarget(target)) {
// The call target is shared with a previous inlined variant. Share
// the graph. This requires a join block at the entry, and edge-split
// form requires a target for each branch.
//
// Represent the sharing by recording a fresh target for the first
// variant and the shared join for all later variants.
if (inlined_entries_[i]->IsGraphEntry()) {
// Convert the old target entry to a new join entry.
auto old_entry = inlined_entries_[i]->AsGraphEntry()->normal_entry();
BlockEntryInstr* old_target = old_entry;
// Unuse all inputs in the old graph entry since it is not part of
// the graph anymore. A new target be created instead.
inlined_entries_[i]->AsGraphEntry()->UnuseAllInputs();
JoinEntryInstr* new_join =
BranchSimplifier::ToJoinEntry(zone(), old_target);
old_target->ReplaceAsPredecessorWith(new_join);
for (intptr_t j = 0; j < old_target->dominated_blocks().length(); ++j) {
BlockEntryInstr* block = old_target->dominated_blocks()[j];
new_join->AddDominatedBlock(block);
}
// Since we are reusing the same inlined body across multiple cids,
// reset the type information on the redefinition of the receiver
// in case it was originally given a concrete type.
ASSERT(new_join->next()->IsRedefinition());
new_join->next()->AsRedefinition()->UpdateType(CompileType::Dynamic());
// Create a new target with the join as unconditional successor.
TargetEntryInstr* new_target = new TargetEntryInstr(
AllocateBlockId(), old_target->try_index(), DeoptId::kNone);
new_target->InheritDeoptTarget(zone(), new_join);
GotoInstr* new_goto = new (Z) GotoInstr(new_join, DeoptId::kNone);
new_goto->InheritDeoptTarget(zone(), new_join);
new_target->LinkTo(new_goto);
new_target->set_last_instruction(new_goto);
new_join->predecessors_.Add(new_target);
// Record the new target for the first variant.
inlined_entries_[i] = new_target;
}
ASSERT(inlined_entries_[i]->IsTargetEntry());
// Record the shared join for this variant.
BlockEntryInstr* join =
inlined_entries_[i]->last_instruction()->SuccessorAt(0);
ASSERT(join->IsJoinEntry());
inlined_entries_.Add(join);
return true;
}
}
return false;
}
bool PolymorphicInliner::CheckNonInlinedDuplicate(const Function& target) {
for (intptr_t i = 0; i < non_inlined_variants_->length(); ++i) {
if (target.raw() == non_inlined_variants_->TargetAt(i)->target->raw()) {
return true;
}
}
return false;
}
bool PolymorphicInliner::TryInliningPoly(const TargetInfo& target_info) {
if ((!FLAG_precompiled_mode ||
owner_->inliner_->speculative_policy()->AllowsSpeculativeInlining()) &&
target_info.IsSingleCid() &&
TryInlineRecognizedMethod(target_info.cid_start, *target_info.target)) {
owner_->inlined_ = true;
return true;
}
GrowableArray<Value*> arguments(call_->ArgumentCount());
for (int i = 0; i < call_->ArgumentCount(); ++i) {
arguments.Add(call_->PushArgumentAt(i)->value());
}
const Array& arguments_descriptor =
Array::ZoneHandle(Z, call_->instance_call()->GetArgumentsDescriptor());
InlinedCallData call_data(call_, arguments_descriptor,
call_->instance_call()->FirstArgIndex(), &arguments,
caller_function_, caller_inlining_id_);
Function& target = Function::ZoneHandle(zone(), target_info.target->raw());
if (!owner_->TryInlining(target, call_->instance_call()->argument_names(),
&call_data, false)) {
return false;
}
FlowGraph* callee_graph = call_data.callee_graph;
call_data.exit_collector->PrepareGraphs(callee_graph);
inlined_entries_.Add(callee_graph->graph_entry());
exit_collector_->Union(call_data.exit_collector);
ReplaceParameterStubs(zone(), owner_->caller_graph(), &call_data,
&target_info);
return true;
}
static Instruction* AppendInstruction(Instruction* first, Instruction* second) {
for (intptr_t i = second->InputCount() - 1; i >= 0; --i) {
Value* input = second->InputAt(i);
input->definition()->AddInputUse(input);
}
first->LinkTo(second);
return second;
}
bool PolymorphicInliner::TryInlineRecognizedMethod(intptr_t receiver_cid,
const Function& target) {
auto temp_parsed_function = new (Z) ParsedFunction(Thread::Current(), target);
auto graph_entry =
new (Z) GraphEntryInstr(*temp_parsed_function, Compiler::kNoOSRDeoptId);
FunctionEntryInstr* entry = nullptr;
Instruction* last = nullptr;
Definition* result = nullptr;
// Replace the receiver argument with a redefinition to prevent code from
// the inlined body from being hoisted above the inlined entry.
GrowableArray<Definition*> arguments(call_->ArgumentCount());
Definition* receiver = call_->Receiver()->definition();
RedefinitionInstr* redefinition =
new (Z) RedefinitionInstr(new (Z) Value(receiver));
redefinition->set_ssa_temp_index(
owner_->caller_graph()->alloc_ssa_temp_index());
if (FlowGraphInliner::TryInlineRecognizedMethod(
owner_->caller_graph(), receiver_cid, target, call_, redefinition,
call_->instance_call()->token_pos(),
call_->instance_call()->ic_data(), graph_entry, &entry, &last,
&result, owner_->inliner_->speculative_policy())) {
// The empty Object constructor is the only case where the inlined body is
// empty and there is no result.
ASSERT((last != nullptr && result != nullptr) ||
MethodRecognizer::RecognizeKind(target) ==
MethodRecognizer::kObjectConstructor);
graph_entry->set_normal_entry(entry);
// Create a graph fragment.
redefinition->InsertAfter(entry);
InlineExitCollector* exit_collector =
new (Z) InlineExitCollector(owner_->caller_graph(), call_);
ReturnInstr* return_result =
new (Z) ReturnInstr(call_->instance_call()->token_pos(),
new (Z) Value(result), DeoptId::kNone);
owner_->caller_graph()->AppendTo(
last, return_result,
call_->env(), // Return can become deoptimization target.
FlowGraph::kEffect);
entry->set_last_instruction(return_result);
exit_collector->AddExit(return_result);
// Update polymorphic inliner state.
inlined_entries_.Add(graph_entry);
exit_collector_->Union(exit_collector);
return true;
}
return false;
}
// Build a DAG to dispatch to the inlined function bodies. Load the class
// id of the receiver and make explicit comparisons for each inlined body,
// in frequency order. If all variants are inlined, the entry to the last
// inlined body is guarded by a CheckClassId instruction which can deopt.
// If not all variants are inlined, we add a PolymorphicInstanceCall
// instruction to handle the non-inlined variants.
TargetEntryInstr* PolymorphicInliner::BuildDecisionGraph() {
const intptr_t try_idx = call_->GetBlock()->try_index();
// Start with a fresh target entry.
TargetEntryInstr* entry = new (Z) TargetEntryInstr(
AllocateBlockId(), try_idx, CompilerState::Current().GetNextDeoptId());
entry->InheritDeoptTarget(zone(), call_);
// This function uses a cursor (a pointer to the 'current' instruction) to
// build the graph. The next instruction will be inserted after the
// cursor.
BlockEntryInstr* current_block = entry;
Instruction* cursor = entry;
Definition* receiver = call_->Receiver()->definition();
// There are at least two variants including non-inlined ones, so we have
// at least one branch on the class id.
LoadClassIdInstr* load_cid =
new (Z) LoadClassIdInstr(new (Z) Value(receiver));
load_cid->set_ssa_temp_index(owner_->caller_graph()->alloc_ssa_temp_index());
cursor = AppendInstruction(cursor, load_cid);
for (intptr_t i = 0; i < inlined_variants_.length(); ++i) {
const CidRange& variant = inlined_variants_[i];
bool test_is_range = !variant.IsSingleCid();
bool is_last_test = (i == inlined_variants_.length() - 1);
// 1. Guard the body with a class id check. We don't need any check if
// it's the last test and global analysis has told us that the call is
// complete.
if (is_last_test && non_inlined_variants_->is_empty()) {
// If it is the last variant use a check class id instruction which can
// deoptimize, followed unconditionally by the body. Omit the check if
// we know that we have covered all possible classes.
if (!call_->complete()) {
RedefinitionInstr* cid_redefinition =
new RedefinitionInstr(new (Z) Value(load_cid));
cid_redefinition->set_ssa_temp_index(
owner_->caller_graph()->alloc_ssa_temp_index());
cursor = AppendInstruction(cursor, cid_redefinition);
CheckClassIdInstr* check_class_id = new (Z) CheckClassIdInstr(
new (Z) Value(cid_redefinition), variant, call_->deopt_id());
check_class_id->InheritDeoptTarget(zone(), call_);
cursor = AppendInstruction(cursor, check_class_id);
}
// The next instruction is the first instruction of the inlined body.
// Handle the two possible cases (unshared and shared subsequent
// predecessors) separately.
BlockEntryInstr* callee_entry = inlined_entries_[i];
if (callee_entry->IsGraphEntry()) {
// Unshared. Graft the normal entry on after the check class
// instruction.
auto target = callee_entry->AsGraphEntry()->normal_entry();
cursor->LinkTo(target->next());
target->ReplaceAsPredecessorWith(current_block);
// Unuse all inputs of the graph entry and the normal entry. They are
// not in the graph anymore.
callee_entry->UnuseAllInputs();
target->UnuseAllInputs();
// All blocks that were dominated by the normal entry are now
// dominated by the current block.
for (intptr_t j = 0; j < target->dominated_blocks().length(); ++j) {
BlockEntryInstr* block = target->dominated_blocks()[j];
current_block->AddDominatedBlock(block);
}
} else if (callee_entry->IsJoinEntry()) {
// Shared inlined body and this is a subsequent entry. We have
// already constructed a join and set its dominator. Add a jump to
// the join.
JoinEntryInstr* join = callee_entry->AsJoinEntry();
ASSERT(join->dominator() != NULL);
GotoInstr* goto_join = new GotoInstr(join, DeoptId::kNone);
goto_join->InheritDeoptTarget(zone(), join);
cursor->LinkTo(goto_join);
current_block->set_last_instruction(goto_join);
} else {
// There is no possibility of a TargetEntry (the first entry to a
// shared inlined body) because this is the last inlined entry.
UNREACHABLE();
}
cursor = NULL;
} else {
// For all variants except the last, use a branch on the loaded class
// id.
//
// TODO(ajcbik): see if this can use the NewDiamond() utility.
//
const Smi& cid = Smi::ZoneHandle(Smi::New(variant.cid_start));
ConstantInstr* cid_constant = owner_->caller_graph()->GetConstant(cid);
BranchInstr* branch;
BranchInstr* upper_limit_branch = NULL;
BlockEntryInstr* cid_test_entry_block = current_block;
if (test_is_range) {
// Double branch for testing a range of Cids.
// TODO(ajcbik): Make a special instruction that uses subtraction
// and unsigned comparison to do this with a single branch.
const Smi& cid_end = Smi::ZoneHandle(Smi::New(variant.cid_end));
ConstantInstr* cid_constant_end =
owner_->caller_graph()->GetConstant(cid_end);
RelationalOpInstr* compare_top = new RelationalOpInstr(
call_->instance_call()->token_pos(), Token::kLTE,
new Value(load_cid), new Value(cid_constant_end), kSmiCid,
call_->deopt_id());
BranchInstr* branch_top = upper_limit_branch =
new BranchInstr(compare_top, DeoptId::kNone);
branch_top->InheritDeoptTarget(zone(), call_);
cursor = AppendInstruction(cursor, branch_top);
current_block->set_last_instruction(branch_top);
TargetEntryInstr* below_target =
new TargetEntryInstr(AllocateBlockId(), try_idx, DeoptId::kNone);
below_target->InheritDeoptTarget(zone(), call_);
current_block->AddDominatedBlock(below_target);
cursor = current_block = below_target;
*branch_top->true_successor_address() = below_target;
RelationalOpInstr* compare_bottom = new RelationalOpInstr(
call_->instance_call()->token_pos(), Token::kGTE,
new Value(load_cid), new Value(cid_constant), kSmiCid,
call_->deopt_id());
branch = new BranchInstr(compare_bottom, DeoptId::kNone);
} else {
StrictCompareInstr* compare = new StrictCompareInstr(
call_->instance_call()->token_pos(), Token::kEQ_STRICT,
new Value(load_cid), new Value(cid_constant),
/* number_check = */ false, DeoptId::kNone);
branch = new BranchInstr(compare, DeoptId::kNone);
}
branch->InheritDeoptTarget(zone(), call_);
cursor = AppendInstruction(cursor, branch);
current_block->set_last_instruction(branch);
cursor = NULL;
// 2. Handle a match by linking to the inlined body. There are three
// cases (unshared, shared first predecessor, and shared subsequent
// predecessors).
BlockEntryInstr* callee_entry = inlined_entries_[i];
TargetEntryInstr* true_target = NULL;
if (callee_entry->IsGraphEntry()) {
// Unshared.
auto graph_entry = callee_entry->AsGraphEntry();
auto function_entry = graph_entry->normal_entry();
true_target = BranchSimplifier::ToTargetEntry(zone(), function_entry);
function_entry->ReplaceAsPredecessorWith(true_target);
for (intptr_t j = 0; j < function_entry->dominated_blocks().length();
++j) {
BlockEntryInstr* block = function_entry->dominated_blocks()[j];
true_target->AddDominatedBlock(block);
}
// Unuse all inputs of the graph entry. It is not in the graph anymore.
graph_entry->UnuseAllInputs();
} else if (callee_entry->IsTargetEntry()) {
ASSERT(!callee_entry->IsFunctionEntry());
// Shared inlined body and this is the first entry. We have already
// constructed a join and this target jumps to it.
true_target = callee_entry->AsTargetEntry();
BlockEntryInstr* join = true_target->last_instruction()->SuccessorAt(0);
current_block->AddDominatedBlock(join);
} else {
// Shared inlined body and this is a subsequent entry. We have
// already constructed a join. We need a fresh target that jumps to
// the join.
JoinEntryInstr* join = callee_entry->AsJoinEntry();
ASSERT(join != NULL);
ASSERT(join->dominator() != NULL);
true_target =
new TargetEntryInstr(AllocateBlockId(), try_idx, DeoptId::kNone);
true_target->InheritDeoptTarget(zone(), join);
GotoInstr* goto_join = new GotoInstr(join, DeoptId::kNone);
goto_join->InheritDeoptTarget(zone(), join);
true_target->LinkTo(goto_join);
true_target->set_last_instruction(goto_join);
}
*branch->true_successor_address() = true_target;
current_block->AddDominatedBlock(true_target);
// 3. Prepare to handle a match failure on the next iteration or the
// fall-through code below for non-inlined variants.
TargetEntryInstr* false_target =
new TargetEntryInstr(AllocateBlockId(), try_idx, DeoptId::kNone);
false_target->InheritDeoptTarget(zone(), call_);
*branch->false_successor_address() = false_target;
cid_test_entry_block->AddDominatedBlock(false_target);
cursor = current_block = false_target;
if (test_is_range) {
// If we tested against a range of Cids there are two different tests
// that can go to the no-cid-match target.
JoinEntryInstr* join =
new JoinEntryInstr(AllocateBlockId(), try_idx, DeoptId::kNone);
TargetEntryInstr* false_target2 =
new TargetEntryInstr(AllocateBlockId(), try_idx, DeoptId::kNone);
*upper_limit_branch->false_successor_address() = false_target2;
cid_test_entry_block->AddDominatedBlock(false_target2);
cid_test_entry_block->AddDominatedBlock(join);
GotoInstr* goto_1 = new GotoInstr(join, DeoptId::kNone);
GotoInstr* goto_2 = new GotoInstr(join, DeoptId::kNone);
false_target->LinkTo(goto_1);
false_target2->LinkTo(goto_2);
false_target->set_last_instruction(goto_1);
false_target2->set_last_instruction(goto_2);
join->InheritDeoptTarget(zone(), call_);
false_target2->InheritDeoptTarget(zone(), call_);
goto_1->InheritDeoptTarget(zone(), call_);
goto_2->InheritDeoptTarget(zone(), call_);
cursor = current_block = join;
}
}
}
// Handle any non-inlined variants.
if (!non_inlined_variants_->is_empty()) {
// Move push arguments of the call.
for (intptr_t i = 0; i < call_->ArgumentCount(); ++i) {
PushArgumentInstr* push = call_->PushArgumentAt(i);
push->ReplaceUsesWith(push->value()->definition());
push->previous()->LinkTo(push->next());
cursor->LinkTo(push);
cursor = push;
}
PolymorphicInstanceCallInstr* fallback_call =
new PolymorphicInstanceCallInstr(
call_->instance_call(), *non_inlined_variants_, call_->complete());
fallback_call->set_ssa_temp_index(
owner_->caller_graph()->alloc_ssa_temp_index());
fallback_call->InheritDeoptTarget(zone(), call_);
fallback_call->set_total_call_count(call_->CallCount());
ReturnInstr* fallback_return =
new ReturnInstr(call_->instance_call()->token_pos(),
new Value(fallback_call), DeoptId::kNone);
fallback_return->InheritDeoptTargetAfter(owner_->caller_graph(), call_,
fallback_call);
AppendInstruction(AppendInstruction(cursor, fallback_call),
fallback_return);
exit_collector_->AddExit(fallback_return);
cursor = NULL;
} else {
// Remove push arguments of the call.
for (intptr_t i = 0; i < call_->ArgumentCount(); ++i) {
PushArgumentInstr* push = call_->PushArgumentAt(i);
push->ReplaceUsesWith(push->value()->definition());
push->RemoveFromGraph();
}
}
return entry;
}
static void TracePolyInlining(const CallTargets& targets,
intptr_t idx,
intptr_t total,
const char* message) {
String& name =
String::Handle(targets.TargetAt(idx)->target->QualifiedUserVisibleName());
int percent = total == 0 ? 0 : (100 * targets.TargetAt(idx)->count) / total;
THR_Print("%s cid %" Pd "-%" Pd ": %" Pd "/%" Pd " %d%% %s\n",
name.ToCString(), targets[idx].cid_start, targets[idx].cid_end,
targets.TargetAt(idx)->count, total, percent, message);
}
bool PolymorphicInliner::trace_inlining() const {
return owner_->trace_inlining();
}
bool PolymorphicInliner::Inline() {
ASSERT(&variants_ == &call_->targets_);
intptr_t total = call_->total_call_count();
for (intptr_t var_idx = 0; var_idx < variants_.length(); ++var_idx) {
TargetInfo* info = variants_.TargetAt(var_idx);
if (variants_.length() > FLAG_max_polymorphic_checks) {
non_inlined_variants_->Add(info);
continue;
}
const Function& target = *variants_.TargetAt(var_idx)->target;
const intptr_t count = variants_.TargetAt(var_idx)->count;
// We we almost inlined all the cases then try a little harder to inline
// the last two, because it's a big win if we inline all of them (compiler
// can see all side effects).
const bool try_harder = (var_idx >= variants_.length() - 2) &&
non_inlined_variants_->length() == 0;
intptr_t size = target.optimized_instruction_count();
bool small = (size != 0 && size < FLAG_inlining_size_threshold);
// If it's less than 3% of the dispatches, we won't even consider
// checking for the class ID and branching to another already-inlined
// version.
if (!try_harder && count < (total >> 5)) {
TRACE_INLINING(
TracePolyInlining(variants_, var_idx, total, "way too infrequent"));
non_inlined_variants_->Add(info);
continue;
}
// First check if this is the same target as an earlier inlined variant.
if (CheckInlinedDuplicate(target)) {
TRACE_INLINING(TracePolyInlining(variants_, var_idx, total,
"duplicate already inlined"));
inlined_variants_.Add(info);
continue;
}
// If it's less than 12% of the dispatches and it's not already inlined, we
// don't consider inlining. For very small functions we are willing to
// consider inlining for 6% of the cases.
if (!try_harder && count < (total >> (small ? 4 : 3))) {
TRACE_INLINING(
TracePolyInlining(variants_, var_idx, total, "too infrequent"));
non_inlined_variants_->Add(&variants_[var_idx]);
continue;
}
// Also check if this is the same target as an earlier non-inlined
// variant. If so and since inlining decisions are costly, do not try
// to inline this variant.
if (CheckNonInlinedDuplicate(target)) {
TRACE_INLINING(
TracePolyInlining(variants_, var_idx, total, "already not inlined"));
non_inlined_variants_->Add(&variants_[var_idx]);
continue;
}
// Make an inlining decision.
if (TryInliningPoly(*info)) {
TRACE_INLINING(TracePolyInlining(variants_, var_idx, total, "inlined"));
inlined_variants_.Add(&variants_[var_idx]);
} else {
TRACE_INLINING(
TracePolyInlining(variants_, var_idx, total, "not inlined"));
non_inlined_variants_->Add(&variants_[var_idx]);
}
}
// If there are no inlined variants, leave the call in place.
if (inlined_variants_.is_empty()) return false;
// Now build a decision tree (a DAG because of shared inline variants) and
// inline it at the call site.
TargetEntryInstr* entry = BuildDecisionGraph();
exit_collector_->ReplaceCall(entry);
return true;
}
FlowGraphInliner::FlowGraphInliner(
FlowGraph* flow_graph,
GrowableArray<const Function*>* inline_id_to_function,
GrowableArray<TokenPosition>* inline_id_to_token_pos,
GrowableArray<intptr_t>* caller_inline_id,
SpeculativeInliningPolicy* speculative_policy,
Precompiler* precompiler)
: flow_graph_(flow_graph),
inline_id_to_function_(inline_id_to_function),
inline_id_to_token_pos_(inline_id_to_token_pos),
caller_inline_id_(caller_inline_id),
trace_inlining_(FLAG_trace_inlining && flow_graph->should_print()),
speculative_policy_(speculative_policy),
precompiler_(precompiler) {}
void FlowGraphInliner::CollectGraphInfo(FlowGraph* flow_graph, bool force) {
const Function& function = flow_graph->function();
if (force || (function.optimized_instruction_count() == 0)) {
GraphInfoCollector info;
info.Collect(*flow_graph);
function.SetOptimizedInstructionCountClamped(info.instruction_count());
function.SetOptimizedCallSiteCountClamped(info.call_site_count());
}
}
// TODO(srdjan): This is only needed when disassembling and/or profiling.
// Sets inlining id for all instructions of this flow-graph, as well for the
// FlowGraph itself.
void FlowGraphInliner::SetInliningId(FlowGraph* flow_graph,
intptr_t inlining_id) {
ASSERT(flow_graph->inlining_id() < 0);
flow_graph->set_inlining_id(inlining_id);
for (BlockIterator block_it = flow_graph->postorder_iterator();
!block_it.Done(); block_it.Advance()) {
for (ForwardInstructionIterator it(block_it.Current()); !it.Done();
it.Advance()) {
Instruction* current = it.Current();
// Do not overwrite owner function.
ASSERT(!current->has_inlining_id());
current->set_inlining_id(inlining_id);
}
}
}
// Use function name to determine if inlineable operator.
// Add names as necessary.
static bool IsInlineableOperator(const Function& function) {
return (function.name() == Symbols::IndexToken().raw()) ||
(function.name() == Symbols::AssignIndexToken().raw()) ||
(function.name() == Symbols::Plus().raw()) ||
(function.name() == Symbols::Minus().raw());
}
bool FlowGraphInliner::AlwaysInline(const Function& function) {
const char* kAlwaysInlineAnnotation = "AlwaysInline";
if (FLAG_enable_inlining_annotations &&
HasAnnotation(function, kAlwaysInlineAnnotation)) {
TRACE_INLINING(
THR_Print("AlwaysInline annotation for %s\n", function.ToCString()));
return true;
}
// We don't want to inline DIFs for recognized methods because we would rather
// replace them with inline FG before inlining introduces any superfluous
// AssertAssignable instructions.
if (function.IsDispatcherOrImplicitAccessor() &&
!(function.kind() == RawFunction::kDynamicInvocationForwarder &&
function.IsRecognized())) {
// Smaller or same size as the call.
return true;
}
if (function.is_const()) {
// Inlined const fields are smaller than a call.
return true;
}
if (function.IsGetterFunction() || function.IsSetterFunction() ||
IsInlineableOperator(function) ||
(function.kind() == RawFunction::kConstructor)) {
const intptr_t count = function.optimized_instruction_count();
if ((count != 0) && (count < FLAG_inline_getters_setters_smaller_than)) {
return true;
}
}
return MethodRecognizer::AlwaysInline(function);
}
int FlowGraphInliner::Inline() {
// Collect graph info and store it on the function.
// We might later use it for an early bailout from the inlining.
CollectGraphInfo(flow_graph_);
const Function& top = flow_graph_->function();
if ((FLAG_inlining_filter != NULL) &&
(strstr(top.ToFullyQualifiedCString(), FLAG_inlining_filter) == NULL)) {
return 0;
}
if (trace_inlining()) {
String& name = String::Handle(top.QualifiedUserVisibleName());
THR_Print("Inlining calls in %s\n", name.ToCString());
}
if (FLAG_support_il_printer && trace_inlining() &&
(FLAG_print_flow_graph || FLAG_print_flow_graph_optimized)) {
THR_Print("Before Inlining of %s\n",
flow_graph_->function().ToFullyQualifiedCString());
FlowGraphPrinter printer(*flow_graph_);
printer.PrintBlocks();
}
intptr_t inlining_depth_threshold = FLAG_inlining_depth_threshold;
CallSiteInliner inliner(this, inlining_depth_threshold);
inliner.InlineCalls();
if (FLAG_print_inlining_tree) {
inliner.PrintInlinedInfo(top);
}
if (inliner.inlined()) {
flow_graph_->DiscoverBlocks();
if (trace_inlining()) {
THR_Print("Inlining growth factor: %f\n", inliner.GrowthFactor());
if (FLAG_support_il_printer &&
(FLAG_print_flow_graph || FLAG_print_flow_graph_optimized)) {
THR_Print("After Inlining of %s\n",
flow_graph_->function().ToFullyQualifiedCString());
FlowGraphPrinter printer(*flow_graph_);
printer.PrintBlocks();
}
}
}
return inliner.inlining_depth();
}
intptr_t FlowGraphInliner::NextInlineId(const Function& function,
TokenPosition tp,
intptr_t parent_id) {
const intptr_t id = inline_id_to_function_->length();
// TODO(johnmccutchan): Do not allow IsNoSource once all nodes have proper
// source positions.
ASSERT(tp.IsReal() || tp.IsSynthetic() || tp.IsNoSource());
RELEASE_ASSERT(!function.IsNull());
inline_id_to_function_->Add(&function);
inline_id_to_token_pos_->Add(tp);
caller_inline_id_->Add(parent_id);
// We always have one less token position than functions.
ASSERT(inline_id_to_token_pos_->length() ==
(inline_id_to_function_->length() - 1));
return id;
}
static bool ShouldInlineSimd() {
return FlowGraphCompiler::SupportsUnboxedSimd128();
}
static bool CanUnboxDouble() {
return FlowGraphCompiler::SupportsUnboxedDoubles();
}
static bool ShouldInlineInt64ArrayOps() {
return FlowGraphCompiler::SupportsUnboxedInt64();
}
static bool CanUnboxInt32() {
// Int32/Uint32 can be unboxed if it fits into a smi or the platform
// supports unboxed mints.
return (compiler::target::kSmiBits >= 32) ||
FlowGraphCompiler::SupportsUnboxedInt64();
}
// Quick access to the current one.
#undef Z
#define Z (flow_graph->zone())
static intptr_t PrepareInlineIndexedOp(FlowGraph* flow_graph,
Instruction* call,
intptr_t array_cid,
Definition** array,
Definition** index,
Instruction** cursor) {
// Insert array length load and bounds check.
LoadFieldInstr* length = new (Z) LoadFieldInstr(
new (Z) Value(*array), Slot::GetLengthFieldForArrayCid(array_cid),
call->token_pos());
*cursor = flow_graph->AppendTo(*cursor, length, NULL, FlowGraph::kValue);
*index = flow_graph->CreateCheckBound(length, *index, call->deopt_id());
*cursor =
flow_graph->AppendTo(*cursor, *index, call->env(), FlowGraph::kValue);
if (array_cid == kGrowableObjectArrayCid) {
// Insert data elements load.
LoadFieldInstr* elements = new (Z)
LoadFieldInstr(new (Z) Value(*array), Slot::GrowableObjectArray_data(),
call->token_pos());
*cursor = flow_graph->AppendTo(*cursor, elements, NULL, FlowGraph::kValue);
// Load from the data from backing store which is a fixed-length array.
*array = elements;
array_cid = kArrayCid;
} else if (RawObject::IsExternalTypedDataClassId(array_cid)) {
LoadUntaggedInstr* elements = new (Z)
LoadUntaggedInstr(new (Z) Value(*array),
compiler::target::TypedDataBase::data_field_offset());
*cursor = flow_graph->AppendTo(*cursor, elements, NULL, FlowGraph::kValue);
*array = elements;
}
return array_cid;
}
static bool InlineGetIndexed(FlowGraph* flow_graph,
MethodRecognizer::Kind kind,
Instruction* call,
Definition* receiver,
GraphEntryInstr* graph_entry,
FunctionEntryInstr** entry,
Instruction** last,
Definition** result) {
intptr_t array_cid = MethodRecognizer::MethodKindToReceiverCid(kind);
Definition* array = receiver;
Definition* index = call->ArgumentAt(1);
*entry =
new (Z) FunctionEntryInstr(graph_entry, flow_graph->allocate_block_id(),
call->GetBlock()->try_index(), DeoptId::kNone);
(*entry)->InheritDeoptTarget(Z, call);
Instruction* cursor = *entry;
array_cid = PrepareInlineIndexedOp(flow_graph, call, array_cid, &array,
&index, &cursor);
intptr_t deopt_id = DeoptId::kNone;
if ((array_cid == kTypedDataInt32ArrayCid) ||
(array_cid == kTypedDataUint32ArrayCid)) {
// Deoptimization may be needed if result does not always fit in a Smi.
deopt_id =
(compiler::target::kSmiBits >= 32) ? DeoptId::kNone : call->deopt_id();
}
// Array load and return.
intptr_t index_scale = compiler::target::Instance::ElementSizeFor(array_cid);
LoadIndexedInstr* load = new (Z)
LoadIndexedInstr(new (Z) Value(array), new (Z) Value(index), index_scale,
array_cid, kAlignedAccess, deopt_id, call->token_pos());
*last = load;
cursor = flow_graph->AppendTo(cursor, load,
deopt_id != DeoptId::kNone ? call->env() : NULL,
FlowGraph::kValue);
const bool value_needs_boxing =
array_cid == kTypedDataInt8ArrayCid ||
array_cid == kTypedDataInt16ArrayCid ||
array_cid == kTypedDataUint8ArrayCid ||
array_cid == kTypedDataUint8ClampedArrayCid ||
array_cid == kTypedDataUint16ArrayCid ||
array_cid == kExternalTypedDataUint8ArrayCid ||
array_cid == kExternalTypedDataUint8ClampedArrayCid;