v8/test/cctest/test-feedback-vector.cc

778 lines
24 KiB
C++
Raw Normal View History

// Copyright 2014 the V8 project authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
#include "src/init/v8.h"
#include "test/cctest/cctest.h"
#include "src/api/api-inl.h"
#include "src/codegen/macro-assembler.h"
#include "src/debug/debug.h"
#include "src/execution/execution.h"
#include "src/handles/global-handles.h"
#include "src/heap/factory.h"
#include "src/objects/feedback-cell-inl.h"
#include "src/objects/objects-inl.h"
#include "test/cctest/test-feedback-vector.h"
namespace v8 {
namespace internal {
namespace {
#define CHECK_SLOT_KIND(helper, index, expected_kind) \
CHECK_EQ(expected_kind, helper.vector()->GetKind(helper.slot(index)));
static Handle<JSFunction> GetFunction(const char* name) {
v8::MaybeLocal<v8::Value> v8_f = CcTest::global()->Get(
CcTest::isolate()->GetCurrentContext(), v8_str(name));
Handle<JSFunction> f =
Handle<JSFunction>::cast(v8::Utils::OpenHandle(*v8_f.ToLocalChecked()));
return f;
}
TEST(VectorStructure) {
LocalContext context;
v8::HandleScope scope(context->GetIsolate());
Isolate* isolate = CcTest::i_isolate();
Factory* factory = isolate->factory();
Zone zone(isolate->allocator(), ZONE_NAME);
Handle<FeedbackVector> vector;
{
FeedbackVectorSpec one_slot(&zone);
[turbofan] Optimize fast enum cache driven for..in. This CL adds support to optimize for..in in fast enum-cache mode to the same degree that it was optimized in Crankshaft, without adding the same deoptimization loop that Crankshaft had with missing enum cache indices. That means code like for (var k in o) { var v = o[k]; // ... } and code like for (var k in o) { if (Object.prototype.hasOwnProperty.call(o, k)) { var v = o[k]; // ... } } which follows the https://eslint.org/docs/rules/guard-for-in linter rule, can now utilize the enum cache indices if o has only fast properties on the receiver, which speeds up the access o[k] significantly and reduces the pollution of the global megamorphic stub cache. For example the micro-benchmark in the tracking bug v8:6702 now runs faster than ever before: forIn: 1516 ms. forInHasOwnProperty: 1674 ms. forInHasOwnPropertySafe: 1595 ms. forInSum: 2051 ms. forInSumSafe: 2215 ms. Compared to numbers from V8 5.8 which is the last version running with Crankshaft forIn: 1641 ms. forInHasOwnProperty: 1719 ms. forInHasOwnPropertySafe: 1802 ms. forInSum: 2226 ms. forInSumSafe: 2409 ms. and V8 6.0 which is the current stable version with TurboFan: forIn: 1713 ms. forInHasOwnProperty: 5417 ms. forInHasOwnPropertySafe: 5324 ms. forInSum: 7556 ms. forInSumSafe: 11067 ms. It also improves the throughput on the string-fasta benchmark by around 7-10%, and there seems to be a ~5% improvement on the Speedometer/React benchmark locally. For this to work, the ForInPrepare bytecode was split into ForInEnumerate and ForInPrepare, which is very similar to how it was handled in Fullcodegen initially. In TurboFan we introduce a new operator LoadFieldByIndex that does the dynamic property load. This also removes the CheckMapValue operator again in favor of just using LoadField, ReferenceEqual and CheckIf, which work automatically with the EscapeAnalysis and the BranchConditionElimination. Bug: v8:6702 Change-Id: I91235413eea478ba77ace7bd14bb2f62e155dd9a Reviewed-on: https://chromium-review.googlesource.com/645949 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Leszek Swirski <leszeks@chromium.org> Cr-Commit-Position: refs/heads/master@{#47768}
2017-09-01 10:49:06 +00:00
one_slot.AddForInSlot();
vector = NewFeedbackVector(isolate, &one_slot);
FeedbackVectorHelper helper(vector);
CHECK_EQ(1, helper.slot_count());
}
{
FeedbackVectorSpec one_icslot(&zone);
one_icslot.AddCallICSlot();
vector = NewFeedbackVector(isolate, &one_icslot);
FeedbackVectorHelper helper(vector);
CHECK_EQ(1, helper.slot_count());
}
{
FeedbackVectorSpec spec(&zone);
for (int i = 0; i < 3; i++) {
[turbofan] Optimize fast enum cache driven for..in. This CL adds support to optimize for..in in fast enum-cache mode to the same degree that it was optimized in Crankshaft, without adding the same deoptimization loop that Crankshaft had with missing enum cache indices. That means code like for (var k in o) { var v = o[k]; // ... } and code like for (var k in o) { if (Object.prototype.hasOwnProperty.call(o, k)) { var v = o[k]; // ... } } which follows the https://eslint.org/docs/rules/guard-for-in linter rule, can now utilize the enum cache indices if o has only fast properties on the receiver, which speeds up the access o[k] significantly and reduces the pollution of the global megamorphic stub cache. For example the micro-benchmark in the tracking bug v8:6702 now runs faster than ever before: forIn: 1516 ms. forInHasOwnProperty: 1674 ms. forInHasOwnPropertySafe: 1595 ms. forInSum: 2051 ms. forInSumSafe: 2215 ms. Compared to numbers from V8 5.8 which is the last version running with Crankshaft forIn: 1641 ms. forInHasOwnProperty: 1719 ms. forInHasOwnPropertySafe: 1802 ms. forInSum: 2226 ms. forInSumSafe: 2409 ms. and V8 6.0 which is the current stable version with TurboFan: forIn: 1713 ms. forInHasOwnProperty: 5417 ms. forInHasOwnPropertySafe: 5324 ms. forInSum: 7556 ms. forInSumSafe: 11067 ms. It also improves the throughput on the string-fasta benchmark by around 7-10%, and there seems to be a ~5% improvement on the Speedometer/React benchmark locally. For this to work, the ForInPrepare bytecode was split into ForInEnumerate and ForInPrepare, which is very similar to how it was handled in Fullcodegen initially. In TurboFan we introduce a new operator LoadFieldByIndex that does the dynamic property load. This also removes the CheckMapValue operator again in favor of just using LoadField, ReferenceEqual and CheckIf, which work automatically with the EscapeAnalysis and the BranchConditionElimination. Bug: v8:6702 Change-Id: I91235413eea478ba77ace7bd14bb2f62e155dd9a Reviewed-on: https://chromium-review.googlesource.com/645949 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Leszek Swirski <leszeks@chromium.org> Cr-Commit-Position: refs/heads/master@{#47768}
2017-09-01 10:49:06 +00:00
spec.AddForInSlot();
}
for (int i = 0; i < 5; i++) {
spec.AddCallICSlot();
}
vector = NewFeedbackVector(isolate, &spec);
FeedbackVectorHelper helper(vector);
CHECK_EQ(8, helper.slot_count());
int index = vector->GetIndex(helper.slot(0));
CHECK_EQ(helper.slot(0), vector->ToSlot(index));
index = vector->GetIndex(helper.slot(3));
CHECK_EQ(helper.slot(3), vector->ToSlot(index));
index = vector->GetIndex(helper.slot(7));
CHECK_EQ(3 + 4 * FeedbackMetadata::GetSlotSize(FeedbackSlotKind::kCall),
index);
CHECK_EQ(helper.slot(7), vector->ToSlot(index));
CHECK_EQ(3 + 5 * FeedbackMetadata::GetSlotSize(FeedbackSlotKind::kCall),
vector->length());
}
{
FeedbackVectorSpec spec(&zone);
[turbofan] Optimize fast enum cache driven for..in. This CL adds support to optimize for..in in fast enum-cache mode to the same degree that it was optimized in Crankshaft, without adding the same deoptimization loop that Crankshaft had with missing enum cache indices. That means code like for (var k in o) { var v = o[k]; // ... } and code like for (var k in o) { if (Object.prototype.hasOwnProperty.call(o, k)) { var v = o[k]; // ... } } which follows the https://eslint.org/docs/rules/guard-for-in linter rule, can now utilize the enum cache indices if o has only fast properties on the receiver, which speeds up the access o[k] significantly and reduces the pollution of the global megamorphic stub cache. For example the micro-benchmark in the tracking bug v8:6702 now runs faster than ever before: forIn: 1516 ms. forInHasOwnProperty: 1674 ms. forInHasOwnPropertySafe: 1595 ms. forInSum: 2051 ms. forInSumSafe: 2215 ms. Compared to numbers from V8 5.8 which is the last version running with Crankshaft forIn: 1641 ms. forInHasOwnProperty: 1719 ms. forInHasOwnPropertySafe: 1802 ms. forInSum: 2226 ms. forInSumSafe: 2409 ms. and V8 6.0 which is the current stable version with TurboFan: forIn: 1713 ms. forInHasOwnProperty: 5417 ms. forInHasOwnPropertySafe: 5324 ms. forInSum: 7556 ms. forInSumSafe: 11067 ms. It also improves the throughput on the string-fasta benchmark by around 7-10%, and there seems to be a ~5% improvement on the Speedometer/React benchmark locally. For this to work, the ForInPrepare bytecode was split into ForInEnumerate and ForInPrepare, which is very similar to how it was handled in Fullcodegen initially. In TurboFan we introduce a new operator LoadFieldByIndex that does the dynamic property load. This also removes the CheckMapValue operator again in favor of just using LoadField, ReferenceEqual and CheckIf, which work automatically with the EscapeAnalysis and the BranchConditionElimination. Bug: v8:6702 Change-Id: I91235413eea478ba77ace7bd14bb2f62e155dd9a Reviewed-on: https://chromium-review.googlesource.com/645949 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Leszek Swirski <leszeks@chromium.org> Cr-Commit-Position: refs/heads/master@{#47768}
2017-09-01 10:49:06 +00:00
spec.AddForInSlot();
spec.AddFeedbackCellForCreateClosure();
[turbofan] Optimize fast enum cache driven for..in. This CL adds support to optimize for..in in fast enum-cache mode to the same degree that it was optimized in Crankshaft, without adding the same deoptimization loop that Crankshaft had with missing enum cache indices. That means code like for (var k in o) { var v = o[k]; // ... } and code like for (var k in o) { if (Object.prototype.hasOwnProperty.call(o, k)) { var v = o[k]; // ... } } which follows the https://eslint.org/docs/rules/guard-for-in linter rule, can now utilize the enum cache indices if o has only fast properties on the receiver, which speeds up the access o[k] significantly and reduces the pollution of the global megamorphic stub cache. For example the micro-benchmark in the tracking bug v8:6702 now runs faster than ever before: forIn: 1516 ms. forInHasOwnProperty: 1674 ms. forInHasOwnPropertySafe: 1595 ms. forInSum: 2051 ms. forInSumSafe: 2215 ms. Compared to numbers from V8 5.8 which is the last version running with Crankshaft forIn: 1641 ms. forInHasOwnProperty: 1719 ms. forInHasOwnPropertySafe: 1802 ms. forInSum: 2226 ms. forInSumSafe: 2409 ms. and V8 6.0 which is the current stable version with TurboFan: forIn: 1713 ms. forInHasOwnProperty: 5417 ms. forInHasOwnPropertySafe: 5324 ms. forInSum: 7556 ms. forInSumSafe: 11067 ms. It also improves the throughput on the string-fasta benchmark by around 7-10%, and there seems to be a ~5% improvement on the Speedometer/React benchmark locally. For this to work, the ForInPrepare bytecode was split into ForInEnumerate and ForInPrepare, which is very similar to how it was handled in Fullcodegen initially. In TurboFan we introduce a new operator LoadFieldByIndex that does the dynamic property load. This also removes the CheckMapValue operator again in favor of just using LoadField, ReferenceEqual and CheckIf, which work automatically with the EscapeAnalysis and the BranchConditionElimination. Bug: v8:6702 Change-Id: I91235413eea478ba77ace7bd14bb2f62e155dd9a Reviewed-on: https://chromium-review.googlesource.com/645949 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Leszek Swirski <leszeks@chromium.org> Cr-Commit-Position: refs/heads/master@{#47768}
2017-09-01 10:49:06 +00:00
spec.AddForInSlot();
vector = NewFeedbackVector(isolate, &spec);
FeedbackVectorHelper helper(vector);
FeedbackCell cell = *vector->GetClosureFeedbackCell(0);
CHECK_EQ(cell.value(), *factory->undefined_value());
}
}
// IC slots need an encoding to recognize what is in there.
TEST(VectorICMetadata) {
LocalContext context;
v8::HandleScope scope(context->GetIsolate());
Isolate* isolate = CcTest::i_isolate();
Zone zone(isolate->allocator(), ZONE_NAME);
FeedbackVectorSpec spec(&zone);
// Set metadata.
for (int i = 0; i < 40; i++) {
switch (i % 4) {
case 0:
[turbofan] Optimize fast enum cache driven for..in. This CL adds support to optimize for..in in fast enum-cache mode to the same degree that it was optimized in Crankshaft, without adding the same deoptimization loop that Crankshaft had with missing enum cache indices. That means code like for (var k in o) { var v = o[k]; // ... } and code like for (var k in o) { if (Object.prototype.hasOwnProperty.call(o, k)) { var v = o[k]; // ... } } which follows the https://eslint.org/docs/rules/guard-for-in linter rule, can now utilize the enum cache indices if o has only fast properties on the receiver, which speeds up the access o[k] significantly and reduces the pollution of the global megamorphic stub cache. For example the micro-benchmark in the tracking bug v8:6702 now runs faster than ever before: forIn: 1516 ms. forInHasOwnProperty: 1674 ms. forInHasOwnPropertySafe: 1595 ms. forInSum: 2051 ms. forInSumSafe: 2215 ms. Compared to numbers from V8 5.8 which is the last version running with Crankshaft forIn: 1641 ms. forInHasOwnProperty: 1719 ms. forInHasOwnPropertySafe: 1802 ms. forInSum: 2226 ms. forInSumSafe: 2409 ms. and V8 6.0 which is the current stable version with TurboFan: forIn: 1713 ms. forInHasOwnProperty: 5417 ms. forInHasOwnPropertySafe: 5324 ms. forInSum: 7556 ms. forInSumSafe: 11067 ms. It also improves the throughput on the string-fasta benchmark by around 7-10%, and there seems to be a ~5% improvement on the Speedometer/React benchmark locally. For this to work, the ForInPrepare bytecode was split into ForInEnumerate and ForInPrepare, which is very similar to how it was handled in Fullcodegen initially. In TurboFan we introduce a new operator LoadFieldByIndex that does the dynamic property load. This also removes the CheckMapValue operator again in favor of just using LoadField, ReferenceEqual and CheckIf, which work automatically with the EscapeAnalysis and the BranchConditionElimination. Bug: v8:6702 Change-Id: I91235413eea478ba77ace7bd14bb2f62e155dd9a Reviewed-on: https://chromium-review.googlesource.com/645949 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Leszek Swirski <leszeks@chromium.org> Cr-Commit-Position: refs/heads/master@{#47768}
2017-09-01 10:49:06 +00:00
spec.AddForInSlot();
break;
case 1:
spec.AddCallICSlot();
break;
case 2:
spec.AddLoadICSlot();
break;
case 3:
spec.AddKeyedLoadICSlot();
break;
}
}
Handle<FeedbackVector> vector = NewFeedbackVector(isolate, &spec);
FeedbackVectorHelper helper(vector);
CHECK_EQ(40, helper.slot_count());
// Meanwhile set some feedback values and type feedback values to
// verify the data structure remains intact.
vector->Set(FeedbackSlot(0), MaybeObject::FromObject(*vector));
// Verify the metadata is correctly set up from the spec.
for (int i = 0; i < 40; i++) {
FeedbackSlotKind kind = vector->GetKind(helper.slot(i));
switch (i % 4) {
case 0:
[turbofan] Optimize fast enum cache driven for..in. This CL adds support to optimize for..in in fast enum-cache mode to the same degree that it was optimized in Crankshaft, without adding the same deoptimization loop that Crankshaft had with missing enum cache indices. That means code like for (var k in o) { var v = o[k]; // ... } and code like for (var k in o) { if (Object.prototype.hasOwnProperty.call(o, k)) { var v = o[k]; // ... } } which follows the https://eslint.org/docs/rules/guard-for-in linter rule, can now utilize the enum cache indices if o has only fast properties on the receiver, which speeds up the access o[k] significantly and reduces the pollution of the global megamorphic stub cache. For example the micro-benchmark in the tracking bug v8:6702 now runs faster than ever before: forIn: 1516 ms. forInHasOwnProperty: 1674 ms. forInHasOwnPropertySafe: 1595 ms. forInSum: 2051 ms. forInSumSafe: 2215 ms. Compared to numbers from V8 5.8 which is the last version running with Crankshaft forIn: 1641 ms. forInHasOwnProperty: 1719 ms. forInHasOwnPropertySafe: 1802 ms. forInSum: 2226 ms. forInSumSafe: 2409 ms. and V8 6.0 which is the current stable version with TurboFan: forIn: 1713 ms. forInHasOwnProperty: 5417 ms. forInHasOwnPropertySafe: 5324 ms. forInSum: 7556 ms. forInSumSafe: 11067 ms. It also improves the throughput on the string-fasta benchmark by around 7-10%, and there seems to be a ~5% improvement on the Speedometer/React benchmark locally. For this to work, the ForInPrepare bytecode was split into ForInEnumerate and ForInPrepare, which is very similar to how it was handled in Fullcodegen initially. In TurboFan we introduce a new operator LoadFieldByIndex that does the dynamic property load. This also removes the CheckMapValue operator again in favor of just using LoadField, ReferenceEqual and CheckIf, which work automatically with the EscapeAnalysis and the BranchConditionElimination. Bug: v8:6702 Change-Id: I91235413eea478ba77ace7bd14bb2f62e155dd9a Reviewed-on: https://chromium-review.googlesource.com/645949 Commit-Queue: Benedikt Meurer <bmeurer@chromium.org> Reviewed-by: Yang Guo <yangguo@chromium.org> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org> Reviewed-by: Leszek Swirski <leszeks@chromium.org> Cr-Commit-Position: refs/heads/master@{#47768}
2017-09-01 10:49:06 +00:00
CHECK_EQ(FeedbackSlotKind::kForIn, kind);
break;
case 1:
CHECK_EQ(FeedbackSlotKind::kCall, kind);
break;
case 2:
CHECK_EQ(FeedbackSlotKind::kLoadProperty, kind);
break;
case 3:
CHECK_EQ(FeedbackSlotKind::kLoadKeyed, kind);
break;
}
}
}
TEST(VectorCallICStates) {
if (!i::FLAG_use_ic) return;
if (i::FLAG_always_opt) return;
FLAG_allow_natives_syntax = true;
CcTest::InitializeVM();
LocalContext context;
v8::HandleScope scope(context->GetIsolate());
Isolate* isolate = CcTest::i_isolate();
// Make sure function f has a call that uses a type feedback slot.
CompileRun(
"function foo() { return 17; };"
"%EnsureFeedbackVectorForFunction(f);"
"function f(a) { a(); } f(foo);");
Handle<JSFunction> f = GetFunction("f");
// There should be one IC.
Handle<FeedbackVector> feedback_vector =
Handle<FeedbackVector>(f->feedback_vector(), isolate);
FeedbackSlot slot(0);
FeedbackNexus nexus(feedback_vector, slot);
CHECK_EQ(MONOMORPHIC, nexus.ic_state());
CompileRun("f(function() { return 16; })");
CHECK_EQ(GENERIC, nexus.ic_state());
// After a collection, state should remain GENERIC.
CcTest::CollectAllGarbage();
CHECK_EQ(GENERIC, nexus.ic_state());
}
TEST(VectorCallFeedback) {
if (!i::FLAG_use_ic) return;
if (i::FLAG_always_opt) return;
FLAG_allow_natives_syntax = true;
CcTest::InitializeVM();
LocalContext context;
v8::HandleScope scope(context->GetIsolate());
Isolate* isolate = CcTest::i_isolate();
// Make sure function f has a call that uses a type feedback slot.
CompileRun(
"function foo() { return 17; }"
"%EnsureFeedbackVectorForFunction(f);"
"function f(a) { a(); } f(foo);");
Handle<JSFunction> f = GetFunction("f");
Handle<JSFunction> foo = GetFunction("foo");
// There should be one IC.
Handle<FeedbackVector> feedback_vector =
Handle<FeedbackVector>(f->feedback_vector(), isolate);
FeedbackSlot slot(0);
FeedbackNexus nexus(feedback_vector, slot);
CHECK_EQ(MONOMORPHIC, nexus.ic_state());
HeapObject heap_object;
CHECK(nexus.GetFeedback()->GetHeapObjectIfWeak(&heap_object));
CHECK_EQ(*foo, heap_object);
CcTest::CollectAllGarbage();
// It should stay monomorphic even after a GC.
CHECK_EQ(MONOMORPHIC, nexus.ic_state());
}
TEST(VectorCallFeedbackForArray) {
if (!i::FLAG_use_ic) return;
if (i::FLAG_always_opt) return;
FLAG_allow_natives_syntax = true;
CcTest::InitializeVM();
LocalContext context;
v8::HandleScope scope(context->GetIsolate());
Isolate* isolate = CcTest::i_isolate();
// Make sure function f has a call that uses a type feedback slot.
CompileRun(
"function f(a) { a(); };"
"%EnsureFeedbackVectorForFunction(f);"
"f(Array);");
Handle<JSFunction> f = GetFunction("f");
// There should be one IC.
Handle<FeedbackVector> feedback_vector =
Handle<FeedbackVector>(f->feedback_vector(), isolate);
FeedbackSlot slot(0);
FeedbackNexus nexus(feedback_vector, slot);
CHECK_EQ(MONOMORPHIC, nexus.ic_state());
HeapObject heap_object;
CHECK(nexus.GetFeedback()->GetHeapObjectIfWeak(&heap_object));
CHECK_EQ(*isolate->array_function(), heap_object);
CcTest::CollectAllGarbage();
// It should stay monomorphic even after a GC.
CHECK_EQ(MONOMORPHIC, nexus.ic_state());
}
size_t GetFeedbackVectorLength(Isolate* isolate, const char* src,
bool with_oneshot_opt) {
i::FLAG_enable_one_shot_optimization = with_oneshot_opt;
i::Handle<i::Object> i_object = v8::Utils::OpenHandle(*CompileRun(src));
i::Handle<i::JSFunction> f = i::Handle<i::JSFunction>::cast(i_object);
Handle<FeedbackVector> feedback_vector =
Handle<FeedbackVector>(f->feedback_vector(), isolate);
return feedback_vector->length();
}
TEST(OneShotCallICSlotCount) {
if (!i::FLAG_use_ic) return;
if (i::FLAG_always_opt) return;
if (i::FLAG_lazy_feedback_allocation) return;
FLAG_allow_natives_syntax = true;
CcTest::InitializeVM();
LocalContext context;
v8::HandleScope scope(context->GetIsolate());
Isolate* isolate = CcTest::i_isolate();
i::FLAG_compilation_cache = false;
const char* no_call = R"(
function f1() {};
function f2() {};
(function() {
return arguments.callee;
})();
)";
// len = 2 * 1 ldaNamed property
CHECK_EQ(GetFeedbackVectorLength(isolate, no_call, false), 2);
// no slots of named property loads/stores in one shot
CHECK_EQ(GetFeedbackVectorLength(isolate, no_call, true), 0);
const char* single_call = R"(
function f1() {};
function f2() {};
(function() {
f1();
return arguments.callee;
})();
)";
// len = 2 * 1 ldaNamed Slot + 2 * 1 CachedGlobalSlot + 2 * 1 CallICSlot
CHECK_EQ(GetFeedbackVectorLength(isolate, single_call, false), 6);
// len = 2 * 1 CachedGlobalSlot
CHECK_EQ(GetFeedbackVectorLength(isolate, single_call, true), 2);
const char* multiple_calls = R"(
function f1() {};
function f2() {};
(function() {
f1();
f2();
f1();
f2();
return arguments.callee;
})();
)";
// len = 2 * 1 ldaNamedSlot + 2 * 2 CachedGlobalSlot (one for each unique
// function) + 2 * 4 CallICSlot (one for each function call)
CHECK_EQ(GetFeedbackVectorLength(isolate, multiple_calls, false), 14);
// CachedGlobalSlot (one for each unique function)
// len = 2 * 2 CachedGlobalSlot (one for each unique function)
CHECK_EQ(GetFeedbackVectorLength(isolate, multiple_calls, true), 4);
}
TEST(VectorCallCounts) {
if (!i::FLAG_use_ic) return;
if (i::FLAG_always_opt) return;
FLAG_allow_natives_syntax = true;
CcTest::InitializeVM();
LocalContext context;
v8::HandleScope scope(context->GetIsolate());
Isolate* isolate = CcTest::i_isolate();
// Make sure function f has a call that uses a type feedback slot.
CompileRun(
"function foo() { return 17; }"
"%EnsureFeedbackVectorForFunction(f);"
"function f(a) { a(); } f(foo);");
Handle<JSFunction> f = GetFunction("f");
// There should be one IC.
Handle<FeedbackVector> feedback_vector =
Handle<FeedbackVector>(f->feedback_vector(), isolate);
FeedbackSlot slot(0);
FeedbackNexus nexus(feedback_vector, slot);
CHECK_EQ(MONOMORPHIC, nexus.ic_state());
CompileRun("f(foo); f(foo);");
CHECK_EQ(MONOMORPHIC, nexus.ic_state());
CHECK_EQ(3, nexus.GetCallCount());
// Send the IC megamorphic, but we should still have incrementing counts.
CompileRun("f(function() { return 12; });");
CHECK_EQ(GENERIC, nexus.ic_state());
CHECK_EQ(4, nexus.GetCallCount());
}
TEST(VectorConstructCounts) {
if (!i::FLAG_use_ic) return;
if (i::FLAG_always_opt) return;
FLAG_allow_natives_syntax = true;
CcTest::InitializeVM();
LocalContext context;
v8::HandleScope scope(context->GetIsolate());
Isolate* isolate = CcTest::i_isolate();
// Make sure function f has a call that uses a type feedback slot.
CompileRun(
"function Foo() {}"
"%EnsureFeedbackVectorForFunction(f);"
"function f(a) { new a(); } f(Foo);");
Handle<JSFunction> f = GetFunction("f");
Handle<FeedbackVector> feedback_vector =
Handle<FeedbackVector>(f->feedback_vector(), isolate);
FeedbackSlot slot(0);
FeedbackNexus nexus(feedback_vector, slot);
CHECK_EQ(MONOMORPHIC, nexus.ic_state());
CHECK(feedback_vector->Get(slot)->IsWeak());
CompileRun("f(Foo); f(Foo);");
CHECK_EQ(MONOMORPHIC, nexus.ic_state());
CHECK_EQ(3, nexus.GetCallCount());
// Send the IC megamorphic, but we should still have incrementing counts.
CompileRun("f(function() {});");
CHECK_EQ(GENERIC, nexus.ic_state());
CHECK_EQ(4, nexus.GetCallCount());
}
TEST(VectorSpeculationMode) {
if (!i::FLAG_use_ic) return;
if (i::FLAG_always_opt) return;
FLAG_allow_natives_syntax = true;
CcTest::InitializeVM();
LocalContext context;
v8::HandleScope scope(context->GetIsolate());
Isolate* isolate = CcTest::i_isolate();
// Make sure function f has a call that uses a type feedback slot.
CompileRun(
"function Foo() {}"
"%EnsureFeedbackVectorForFunction(f);"
"function f(a) { new a(); } f(Foo);");
Handle<JSFunction> f = GetFunction("f");
Handle<FeedbackVector> feedback_vector =
Handle<FeedbackVector>(f->feedback_vector(), isolate);
FeedbackSlot slot(0);
FeedbackNexus nexus(feedback_vector, slot);
CHECK_EQ(SpeculationMode::kAllowSpeculation, nexus.GetSpeculationMode());
CompileRun("f(Foo); f(Foo);");
CHECK_EQ(3, nexus.GetCallCount());
CHECK_EQ(SpeculationMode::kAllowSpeculation, nexus.GetSpeculationMode());
nexus.SetSpeculationMode(SpeculationMode::kDisallowSpeculation);
CHECK_EQ(SpeculationMode::kDisallowSpeculation, nexus.GetSpeculationMode());
CHECK_EQ(3, nexus.GetCallCount());
nexus.SetSpeculationMode(SpeculationMode::kAllowSpeculation);
CHECK_EQ(SpeculationMode::kAllowSpeculation, nexus.GetSpeculationMode());
CHECK_EQ(3, nexus.GetCallCount());
}
TEST(VectorLoadICStates) {
if (!i::FLAG_use_ic) return;
if (i::FLAG_always_opt) return;
FLAG_allow_natives_syntax = true;
CcTest::InitializeVM();
LocalContext context;
v8::HandleScope scope(context->GetIsolate());
Isolate* isolate = CcTest::i_isolate();
// Make sure function f has a call that uses a type feedback slot.
CompileRun(
"var o = { foo: 3 };"
"%EnsureFeedbackVectorForFunction(f);"
"function f(a) { return a.foo; } f(o);");
Handle<JSFunction> f = GetFunction("f");
// There should be one IC.
Handle<FeedbackVector> feedback_vector =
Handle<FeedbackVector>(f->feedback_vector(), isolate);
FeedbackSlot slot(0);
FeedbackNexus nexus(feedback_vector, slot);
CHECK_EQ(MONOMORPHIC, nexus.ic_state());
// Verify that the monomorphic map is the one we expect.
v8::MaybeLocal<v8::Value> v8_o =
CcTest::global()->Get(context.local(), v8_str("o"));
Handle<JSObject> o =
Handle<JSObject>::cast(v8::Utils::OpenHandle(*v8_o.ToLocalChecked()));
CHECK_EQ(o->map(), nexus.GetFirstMap());
// Now go polymorphic.
CompileRun("f({ blarg: 3, foo: 2 })");
CHECK_EQ(POLYMORPHIC, nexus.ic_state());
CompileRun(
"delete o.foo;"
"f(o)");
CHECK_EQ(POLYMORPHIC, nexus.ic_state());
CompileRun("f({ blarg: 3, torino: 10, foo: 2 })");
CHECK_EQ(POLYMORPHIC, nexus.ic_state());
MapHandles maps;
nexus.ExtractMaps(&maps);
CHECK_EQ(4, maps.size());
// Finally driven megamorphic.
CompileRun("f({ blarg: 3, gran: 3, torino: 10, foo: 2 })");
CHECK_EQ(MEGAMORPHIC, nexus.ic_state());
CHECK(nexus.GetFirstMap().is_null());
// After a collection, state should not be reset to PREMONOMORPHIC.
CcTest::CollectAllGarbage();
CHECK_EQ(MEGAMORPHIC, nexus.ic_state());
}
TEST(VectorLoadGlobalICSlotSharing) {
if (!i::FLAG_use_ic) return;
if (i::FLAG_always_opt) return;
FLAG_allow_natives_syntax = true;
CcTest::InitializeVM();
LocalContext context;
v8::HandleScope scope(context->GetIsolate());
Isolate* isolate = CcTest::i_isolate();
// Function f has 5 LoadGlobalICs: 3 for {o} references outside of "typeof"
// operator and 2 for {o} references inside "typeof" operator.
CompileRun(
"o = 10;"
"function f() {"
" var x = o || 10;"
" var y = typeof o;"
" return o , typeof o, x , y, o;"
"}"
"%EnsureFeedbackVectorForFunction(f);"
"f();");
Handle<JSFunction> f = GetFunction("f");
// There should be two IC slots for {o} references outside and inside
// typeof operator respectively.
Handle<FeedbackVector> feedback_vector =
Handle<FeedbackVector>(f->feedback_vector(), isolate);
FeedbackVectorHelper helper(feedback_vector);
CHECK_EQ(2, helper.slot_count());
CHECK_SLOT_KIND(helper, 0, FeedbackSlotKind::kLoadGlobalNotInsideTypeof);
CHECK_SLOT_KIND(helper, 1, FeedbackSlotKind::kLoadGlobalInsideTypeof);
FeedbackSlot slot1 = helper.slot(0);
FeedbackSlot slot2 = helper.slot(1);
CHECK_EQ(MONOMORPHIC, FeedbackNexus(feedback_vector, slot1).ic_state());
CHECK_EQ(MONOMORPHIC, FeedbackNexus(feedback_vector, slot2).ic_state());
}
TEST(VectorLoadICOnSmi) {
if (!i::FLAG_use_ic) return;
if (i::FLAG_always_opt) return;
FLAG_allow_natives_syntax = true;
CcTest::InitializeVM();
LocalContext context;
v8::HandleScope scope(context->GetIsolate());
Isolate* isolate = CcTest::i_isolate();
Heap* heap = isolate->heap();
// Make sure function f has a call that uses a type feedback slot.
CompileRun(
"var o = { foo: 3 };"
"%EnsureFeedbackVectorForFunction(f);"
"function f(a) { return a.foo; } f(34);");
Handle<JSFunction> f = GetFunction("f");
// There should be one IC.
Handle<FeedbackVector> feedback_vector =
Handle<FeedbackVector>(f->feedback_vector(), isolate);
FeedbackSlot slot(0);
FeedbackNexus nexus(feedback_vector, slot);
CHECK_EQ(MONOMORPHIC, nexus.ic_state());
// Verify that the monomorphic map is the one we expect.
Map number_map = ReadOnlyRoots(heap).heap_number_map();
CHECK_EQ(number_map, nexus.GetFirstMap());
// Now go polymorphic on o.
CompileRun("f(o)");
CHECK_EQ(POLYMORPHIC, nexus.ic_state());
MapHandles maps;
nexus.ExtractMaps(&maps);
CHECK_EQ(2, maps.size());
// One of the maps should be the o map.
v8::MaybeLocal<v8::Value> v8_o =
CcTest::global()->Get(context.local(), v8_str("o"));
Handle<JSObject> o =
Handle<JSObject>::cast(v8::Utils::OpenHandle(*v8_o.ToLocalChecked()));
bool number_map_found = false;
bool o_map_found = false;
for (Handle<Map> current : maps) {
if (*current == number_map)
number_map_found = true;
else if (*current == o->map())
o_map_found = true;
}
CHECK(number_map_found && o_map_found);
// The degree of polymorphism doesn't change.
CompileRun("f(100)");
CHECK_EQ(POLYMORPHIC, nexus.ic_state());
MapHandles maps2;
nexus.ExtractMaps(&maps2);
CHECK_EQ(2, maps2.size());
}
TEST(ReferenceContextAllocatesNoSlots) {
if (!i::FLAG_use_ic) return;
if (i::FLAG_always_opt) return;
FLAG_allow_natives_syntax = true;
CcTest::InitializeVM();
LocalContext context;
v8::HandleScope scope(context->GetIsolate());
Isolate* isolate = CcTest::i_isolate();
{
CompileRun(
"function testvar(x) {"
" y = x;"
" y = a;"
" return y;"
"}"
"%EnsureFeedbackVectorForFunction(testvar);"
"a = 3;"
"testvar({});");
Handle<JSFunction> f = GetFunction("testvar");
// There should be two LOAD_ICs, one for a and one for y at the end.
Handle<FeedbackVector> feedback_vector =
handle(f->feedback_vector(), isolate);
FeedbackVectorHelper helper(feedback_vector);
CHECK_EQ(3, helper.slot_count());
CHECK_SLOT_KIND(helper, 0, FeedbackSlotKind::kStoreGlobalSloppy);
CHECK_SLOT_KIND(helper, 1, FeedbackSlotKind::kLoadGlobalNotInsideTypeof);
CHECK_SLOT_KIND(helper, 2, FeedbackSlotKind::kLoadGlobalNotInsideTypeof);
}
{
CompileRun(
"function testprop(x) {"
" 'use strict';"
" x.blue = a;"
"}"
"%EnsureFeedbackVectorForFunction(testprop);"
"testprop({ blue: 3 });");
Handle<JSFunction> f = GetFunction("testprop");
// There should be one LOAD_IC, for the load of a.
Handle<FeedbackVector> feedback_vector(f->feedback_vector(), isolate);
FeedbackVectorHelper helper(feedback_vector);
CHECK_EQ(2, helper.slot_count());
CHECK_SLOT_KIND(helper, 0, FeedbackSlotKind::kLoadGlobalNotInsideTypeof);
CHECK_SLOT_KIND(helper, 1, FeedbackSlotKind::kStoreNamedStrict);
}
{
CompileRun(
"function testpropfunc(x) {"
" x().blue = a;"
" return x().blue;"
"}"
"%EnsureFeedbackVectorForFunction(testpropfunc);"
"function makeresult() { return { blue: 3 }; }"
"testpropfunc(makeresult);");
Handle<JSFunction> f = GetFunction("testpropfunc");
// There should be 1 LOAD_GLOBAL_IC to load x (in both cases), 2 CALL_ICs
// to call x and a LOAD_IC to load blue.
Handle<FeedbackVector> feedback_vector(f->feedback_vector(), isolate);
FeedbackVectorHelper helper(feedback_vector);
CHECK_EQ(5, helper.slot_count());
CHECK_SLOT_KIND(helper, 0, FeedbackSlotKind::kCall);
CHECK_SLOT_KIND(helper, 1, FeedbackSlotKind::kLoadGlobalNotInsideTypeof);
CHECK_SLOT_KIND(helper, 2, FeedbackSlotKind::kStoreNamedSloppy);
CHECK_SLOT_KIND(helper, 3, FeedbackSlotKind::kCall);
CHECK_SLOT_KIND(helper, 4, FeedbackSlotKind::kLoadProperty);
}
{
CompileRun(
"function testkeyedprop(x) {"
" x[0] = a;"
" return x[0];"
"}"
"%EnsureFeedbackVectorForFunction(testkeyedprop);"
"testkeyedprop([0, 1, 2]);");
Handle<JSFunction> f = GetFunction("testkeyedprop");
// There should be 1 LOAD_GLOBAL_ICs for the load of a, and one
// KEYED_LOAD_IC for the load of x[0] in the return statement.
Handle<FeedbackVector> feedback_vector(f->feedback_vector(), isolate);
FeedbackVectorHelper helper(feedback_vector);
CHECK_EQ(3, helper.slot_count());
CHECK_SLOT_KIND(helper, 0, FeedbackSlotKind::kLoadGlobalNotInsideTypeof);
CHECK_SLOT_KIND(helper, 1, FeedbackSlotKind::kStoreKeyedSloppy);
CHECK_SLOT_KIND(helper, 2, FeedbackSlotKind::kLoadKeyed);
}
{
CompileRun(
"function testkeyedprop(x) {"
" 'use strict';"
" x[0] = a;"
" return x[0];"
"}"
"%EnsureFeedbackVectorForFunction(testkeyedprop);"
"testkeyedprop([0, 1, 2]);");
Handle<JSFunction> f = GetFunction("testkeyedprop");
// There should be 1 LOAD_GLOBAL_ICs for the load of a, and one
// KEYED_LOAD_IC for the load of x[0] in the return statement.
Handle<FeedbackVector> feedback_vector(f->feedback_vector(), isolate);
FeedbackVectorHelper helper(feedback_vector);
CHECK_EQ(3, helper.slot_count());
CHECK_SLOT_KIND(helper, 0, FeedbackSlotKind::kLoadGlobalNotInsideTypeof);
CHECK_SLOT_KIND(helper, 1, FeedbackSlotKind::kStoreKeyedStrict);
CHECK_SLOT_KIND(helper, 2, FeedbackSlotKind::kLoadKeyed);
}
{
CompileRun(
"function testcompound(x) {"
" 'use strict';"
" x.old = x.young = x.in_between = a;"
" return x.old + x.young;"
"}"
"%EnsureFeedbackVectorForFunction(testcompound);"
"testcompound({ old: 3, young: 3, in_between: 3 });");
Handle<JSFunction> f = GetFunction("testcompound");
// There should be 1 LOAD_GLOBAL_IC for load of a and 2 LOAD_ICs, for load
// of x.old and x.young.
Handle<FeedbackVector> feedback_vector(f->feedback_vector(), isolate);
FeedbackVectorHelper helper(feedback_vector);
CHECK_EQ(7, helper.slot_count());
CHECK_SLOT_KIND(helper, 0, FeedbackSlotKind::kLoadGlobalNotInsideTypeof);
CHECK_SLOT_KIND(helper, 1, FeedbackSlotKind::kStoreNamedStrict);
CHECK_SLOT_KIND(helper, 2, FeedbackSlotKind::kStoreNamedStrict);
CHECK_SLOT_KIND(helper, 3, FeedbackSlotKind::kStoreNamedStrict);
CHECK_SLOT_KIND(helper, 4, FeedbackSlotKind::kBinaryOp);
CHECK_SLOT_KIND(helper, 5, FeedbackSlotKind::kLoadProperty);
CHECK_SLOT_KIND(helper, 6, FeedbackSlotKind::kLoadProperty);
}
}
TEST(VectorStoreICBasic) {
if (!i::FLAG_use_ic) return;
if (i::FLAG_always_opt) return;
FLAG_allow_natives_syntax = true;
CcTest::InitializeVM();
LocalContext context;
v8::HandleScope scope(context->GetIsolate());
CompileRun(
"function f(a) {"
" a.foo = 5;"
"};"
"%EnsureFeedbackVectorForFunction(f);"
"var a = { foo: 3 };"
"f(a);"
"f(a);"
"f(a);");
Handle<JSFunction> f = GetFunction("f");
// There should be one IC slot.
Handle<FeedbackVector> feedback_vector(f->feedback_vector(), f->GetIsolate());
FeedbackVectorHelper helper(feedback_vector);
CHECK_EQ(1, helper.slot_count());
FeedbackSlot slot(0);
FeedbackNexus nexus(feedback_vector, slot);
CHECK_EQ(MONOMORPHIC, nexus.ic_state());
}
TEST(StoreOwnIC) {
if (!i::FLAG_use_ic) return;
if (i::FLAG_always_opt) return;
FLAG_allow_natives_syntax = true;
CcTest::InitializeVM();
LocalContext context;
v8::HandleScope scope(context->GetIsolate());
CompileRun(
"function f(v) {"
" return {a: 0, b: v, c: 0};"
"}"
"%EnsureFeedbackVectorForFunction(f);"
"f(1);"
"f(2);"
"f(3);");
Handle<JSFunction> f = GetFunction("f");
// There should be one IC slot.
Handle<FeedbackVector> feedback_vector(f->feedback_vector(), f->GetIsolate());
FeedbackVectorHelper helper(feedback_vector);
CHECK_EQ(2, helper.slot_count());
CHECK_SLOT_KIND(helper, 0, FeedbackSlotKind::kLiteral);
CHECK_SLOT_KIND(helper, 1, FeedbackSlotKind::kStoreOwnNamed);
FeedbackNexus nexus(feedback_vector, helper.slot(1));
CHECK_EQ(MONOMORPHIC, nexus.ic_state());
}
} // namespace
} // namespace internal
} // namespace v8