v8/test/cctest/test-serialize.cc

647 lines
21 KiB
C++
Raw Normal View History

// Copyright 2007-2010 the V8 project authors. All rights reserved.
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
//
// * Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
// * Redistributions in binary form must reproduce the above
// copyright notice, this list of conditions and the following
// disclaimer in the documentation and/or other materials provided
// with the distribution.
// * Neither the name of Google Inc. nor the names of its
// contributors may be used to endorse or promote products derived
// from this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#include <signal.h>
#include "sys/stat.h"
#include "v8.h"
#include "debug.h"
#include "ic-inl.h"
#include "runtime.h"
#include "serialize.h"
#include "scopeinfo.h"
#include "snapshot.h"
#include "cctest.h"
#include "spaces.h"
#include "objects.h"
#include "natives.h"
#include "bootstrapper.h"
using namespace v8::internal;
static const unsigned kCounters = 256;
static int local_counters[kCounters];
static const char* local_counter_names[kCounters];
static unsigned CounterHash(const char* s) {
unsigned hash = 0;
while (*++s) {
hash |= hash << 5;
hash += *s;
}
return hash;
}
// Callback receiver to track counters in test.
static int* counter_function(const char* name) {
unsigned hash = CounterHash(name) % kCounters;
unsigned original_hash = hash;
USE(original_hash);
while (true) {
if (local_counter_names[hash] == name) {
return &local_counters[hash];
}
if (local_counter_names[hash] == 0) {
local_counter_names[hash] = name;
return &local_counters[hash];
}
if (strcmp(local_counter_names[hash], name) == 0) {
return &local_counters[hash];
}
hash = (hash + 1) % kCounters;
ASSERT(hash != original_hash); // Hash table has been filled up.
}
}
template <class T>
static Address AddressOf(T id) {
return ExternalReference(id, i::Isolate::Current()).address();
}
template <class T>
static uint32_t Encode(const ExternalReferenceEncoder& encoder, T id) {
return encoder.Encode(AddressOf(id));
}
static int make_code(TypeCode type, int id) {
return static_cast<uint32_t>(type) << kReferenceTypeShift | id;
}
TEST(ExternalReferenceEncoder) {
Isolate* isolate = i::Isolate::Current();
isolate->stats_table()->SetCounterFunction(counter_function);
v8::V8::Initialize();
ExternalReferenceEncoder encoder;
CHECK_EQ(make_code(BUILTIN, Builtins::kArrayCode),
Encode(encoder, Builtins::kArrayCode));
CHECK_EQ(make_code(v8::internal::RUNTIME_FUNCTION, Runtime::kAbort),
Encode(encoder, Runtime::kAbort));
ExternalReference total_compile_size =
ExternalReference(isolate->counters()->total_compile_size());
CHECK_EQ(make_code(STATS_COUNTER, Counters::k_total_compile_size),
encoder.Encode(total_compile_size.address()));
ExternalReference stack_limit_address =
ExternalReference::address_of_stack_limit(isolate);
CHECK_EQ(make_code(UNCLASSIFIED, 4),
encoder.Encode(stack_limit_address.address()));
ExternalReference real_stack_limit_address =
ExternalReference::address_of_real_stack_limit(isolate);
CHECK_EQ(make_code(UNCLASSIFIED, 5),
encoder.Encode(real_stack_limit_address.address()));
#ifdef ENABLE_DEBUGGER_SUPPORT
CHECK_EQ(make_code(UNCLASSIFIED, 16),
encoder.Encode(ExternalReference::debug_break(isolate).address()));
#endif // ENABLE_DEBUGGER_SUPPORT
CHECK_EQ(make_code(UNCLASSIFIED, 10),
encoder.Encode(
ExternalReference::new_space_start(isolate).address()));
CHECK_EQ(make_code(UNCLASSIFIED, 3),
encoder.Encode(
ExternalReference::roots_array_start(isolate).address()));
CHECK_EQ(make_code(UNCLASSIFIED, 52),
encoder.Encode(ExternalReference::cpu_features().address()));
}
TEST(ExternalReferenceDecoder) {
Isolate* isolate = i::Isolate::Current();
isolate->stats_table()->SetCounterFunction(counter_function);
v8::V8::Initialize();
ExternalReferenceDecoder decoder;
CHECK_EQ(AddressOf(Builtins::kArrayCode),
decoder.Decode(make_code(BUILTIN, Builtins::kArrayCode)));
CHECK_EQ(AddressOf(Runtime::kAbort),
decoder.Decode(make_code(v8::internal::RUNTIME_FUNCTION,
Runtime::kAbort)));
ExternalReference total_compile_size =
ExternalReference(isolate->counters()->total_compile_size());
CHECK_EQ(total_compile_size.address(),
decoder.Decode(
make_code(STATS_COUNTER,
Counters::k_total_compile_size)));
CHECK_EQ(ExternalReference::address_of_stack_limit(isolate).address(),
decoder.Decode(make_code(UNCLASSIFIED, 4)));
CHECK_EQ(ExternalReference::address_of_real_stack_limit(isolate).address(),
decoder.Decode(make_code(UNCLASSIFIED, 5)));
#ifdef ENABLE_DEBUGGER_SUPPORT
CHECK_EQ(ExternalReference::debug_break(isolate).address(),
decoder.Decode(make_code(UNCLASSIFIED, 16)));
#endif // ENABLE_DEBUGGER_SUPPORT
CHECK_EQ(ExternalReference::new_space_start(isolate).address(),
decoder.Decode(make_code(UNCLASSIFIED, 10)));
}
class FileByteSink : public SnapshotByteSink {
public:
explicit FileByteSink(const char* snapshot_file) {
fp_ = OS::FOpen(snapshot_file, "wb");
file_name_ = snapshot_file;
if (fp_ == NULL) {
PrintF("Unable to write to snapshot file \"%s\"\n", snapshot_file);
exit(1);
}
}
virtual ~FileByteSink() {
if (fp_ != NULL) {
fclose(fp_);
}
}
virtual void Put(int byte, const char* description) {
if (fp_ != NULL) {
fputc(byte, fp_);
}
}
virtual int Position() {
return ftell(fp_);
}
void WriteSpaceUsed(
int new_space_used,
int pointer_space_used,
int data_space_used,
int code_space_used,
int map_space_used,
int cell_space_used,
int property_cell_space_used);
private:
FILE* fp_;
const char* file_name_;
};
void FileByteSink::WriteSpaceUsed(
int new_space_used,
int pointer_space_used,
int data_space_used,
int code_space_used,
int map_space_used,
int cell_space_used,
int property_cell_space_used) {
int file_name_length = StrLength(file_name_) + 10;
Vector<char> name = Vector<char>::New(file_name_length + 1);
OS::SNPrintF(name, "%s.size", file_name_);
FILE* fp = OS::FOpen(name.start(), "w");
name.Dispose();
fprintf(fp, "new %d\n", new_space_used);
fprintf(fp, "pointer %d\n", pointer_space_used);
fprintf(fp, "data %d\n", data_space_used);
fprintf(fp, "code %d\n", code_space_used);
fprintf(fp, "map %d\n", map_space_used);
fprintf(fp, "cell %d\n", cell_space_used);
fprintf(fp, "property cell %d\n", property_cell_space_used);
fclose(fp);
}
static bool WriteToFile(const char* snapshot_file) {
FileByteSink file(snapshot_file);
StartupSerializer ser(&file);
ser.Serialize();
Refactoring of snapshots. This simplifies and improves the speed of deserializing code. The current startup time improvement for V8 is around 6%, but code deserialization is speeded up disproportionately, and we will soon have more code in the snapshot. * Removed support for deserializing into large object space. The regular pages are 1Mbyte now and that is plenty. This is a big simplification. * Instead of reserving space for the snapshot we actually allocate it now. This removes some special casing from the memory management and simplifies deserialization since we are just bumping a pointer rather than calling the normal allocation routines during deserialization. * Record in the snapshot how much we need to boot up and allocate it instead of just assuming that allocations in a new VM will always be linear. * In the snapshot we always address an object as a negative offset from the current allocation point. We used to sometimes address from the start of the deserialized data, but this is less useful now that we have good support for roots and repetitions in the deserialization data. * Code objects were previously deserialized (like other objects) by alternating raw data (deserialized with memcpy) and pointers (to external references, other objects, etc.). Now we deserialize code objects with a single memcpy, followed by a series of skips and pointers that partially overwrite the code we memcopied out of the snapshot. The skips are sometimes merged into the following instruction in the deserialization data to reduce dispatch time. * Integers in the snapshot were stored in a variable length format that gives a compact representation for small positive integers. This is still the case, but the new encoding can be decoded without branches or conditional instructions, which is faster on a modern CPU. Review URL: https://chromiumcodereview.appspot.com/10918067 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12505 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2012-09-14 11:16:56 +00:00
file.WriteSpaceUsed(
ser.CurrentAllocationAddress(NEW_SPACE),
ser.CurrentAllocationAddress(OLD_POINTER_SPACE),
ser.CurrentAllocationAddress(OLD_DATA_SPACE),
ser.CurrentAllocationAddress(CODE_SPACE),
ser.CurrentAllocationAddress(MAP_SPACE),
ser.CurrentAllocationAddress(CELL_SPACE),
ser.CurrentAllocationAddress(PROPERTY_CELL_SPACE));
Refactoring of snapshots. This simplifies and improves the speed of deserializing code. The current startup time improvement for V8 is around 6%, but code deserialization is speeded up disproportionately, and we will soon have more code in the snapshot. * Removed support for deserializing into large object space. The regular pages are 1Mbyte now and that is plenty. This is a big simplification. * Instead of reserving space for the snapshot we actually allocate it now. This removes some special casing from the memory management and simplifies deserialization since we are just bumping a pointer rather than calling the normal allocation routines during deserialization. * Record in the snapshot how much we need to boot up and allocate it instead of just assuming that allocations in a new VM will always be linear. * In the snapshot we always address an object as a negative offset from the current allocation point. We used to sometimes address from the start of the deserialized data, but this is less useful now that we have good support for roots and repetitions in the deserialization data. * Code objects were previously deserialized (like other objects) by alternating raw data (deserialized with memcpy) and pointers (to external references, other objects, etc.). Now we deserialize code objects with a single memcpy, followed by a series of skips and pointers that partially overwrite the code we memcopied out of the snapshot. The skips are sometimes merged into the following instruction in the deserialization data to reduce dispatch time. * Integers in the snapshot were stored in a variable length format that gives a compact representation for small positive integers. This is still the case, but the new encoding can be decoded without branches or conditional instructions, which is faster on a modern CPU. Review URL: https://chromiumcodereview.appspot.com/10918067 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12505 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2012-09-14 11:16:56 +00:00
return true;
}
static void Serialize() {
// We have to create one context. One reason for this is so that the builtins
// can be loaded from v8natives.js and their addresses can be processed. This
// will clear the pending fixups array, which would otherwise contain GC roots
// that would confuse the serialization/deserialization process.
{
v8::Isolate* isolate = v8::Isolate::GetCurrent();
v8::HandleScope scope(isolate);
v8::Context::New(isolate);
}
WriteToFile(FLAG_testing_serialization_file);
}
// Test that the whole heap can be serialized.
TEST(Serialize) {
if (!Snapshot::HaveASnapshotToStartFrom()) {
Serializer::Enable();
v8::V8::Initialize();
Serialize();
}
}
// Test that heap serialization is non-destructive.
TEST(SerializeTwice) {
if (!Snapshot::HaveASnapshotToStartFrom()) {
Serializer::Enable();
v8::V8::Initialize();
Serialize();
Serialize();
}
}
//----------------------------------------------------------------------------
// Tests that the heap can be deserialized.
static void Deserialize() {
CHECK(Snapshot::Initialize(FLAG_testing_serialization_file));
}
static void SanityCheck() {
Isolate* isolate = Isolate::Current();
v8::HandleScope scope(v8::Isolate::GetCurrent());
#ifdef VERIFY_HEAP
HEAP->Verify();
#endif
CHECK(isolate->global_object()->IsJSObject());
CHECK(isolate->native_context()->IsContext());
CHECK(HEAP->string_table()->IsStringTable());
CHECK(!isolate->factory()->InternalizeOneByteString(
STATIC_ASCII_VECTOR("Empty"))->IsFailure());
}
DEPENDENT_TEST(Deserialize, Serialize) {
// The serialize-deserialize tests only work if the VM is built without
// serialization. That doesn't matter. We don't need to be able to
// serialize a snapshot in a VM that is booted from a snapshot.
if (!Snapshot::HaveASnapshotToStartFrom()) {
v8::Isolate* isolate = v8::Isolate::GetCurrent();
v8::HandleScope scope(isolate);
Deserialize();
v8::Local<v8::Context> env = v8::Context::New(isolate);
env->Enter();
SanityCheck();
}
}
DEPENDENT_TEST(DeserializeFromSecondSerialization, SerializeTwice) {
if (!Snapshot::HaveASnapshotToStartFrom()) {
v8::Isolate* isolate = v8::Isolate::GetCurrent();
v8::HandleScope scope(isolate);
Deserialize();
v8::Local<v8::Context> env = v8::Context::New(isolate);
env->Enter();
SanityCheck();
}
}
DEPENDENT_TEST(DeserializeAndRunScript2, Serialize) {
if (!Snapshot::HaveASnapshotToStartFrom()) {
v8::Isolate* isolate = v8::Isolate::GetCurrent();
v8::HandleScope scope(isolate);
Deserialize();
v8::Local<v8::Context> env = v8::Context::New(isolate);
env->Enter();
const char* c_source = "\"1234\".length";
v8::Local<v8::String> source = v8::String::New(c_source);
v8::Local<v8::Script> script = v8::Script::Compile(source);
CHECK_EQ(4, script->Run()->Int32Value());
}
}
DEPENDENT_TEST(DeserializeFromSecondSerializationAndRunScript2,
SerializeTwice) {
if (!Snapshot::HaveASnapshotToStartFrom()) {
v8::Isolate* isolate = v8::Isolate::GetCurrent();
v8::HandleScope scope(isolate);
Deserialize();
v8::Local<v8::Context> env = v8::Context::New(isolate);
env->Enter();
const char* c_source = "\"1234\".length";
v8::Local<v8::String> source = v8::String::New(c_source);
v8::Local<v8::Script> script = v8::Script::Compile(source);
CHECK_EQ(4, script->Run()->Int32Value());
}
}
TEST(PartialSerialization) {
if (!Snapshot::HaveASnapshotToStartFrom()) {
Serializer::Enable();
v8::V8::Initialize();
Isolate* isolate = Isolate::Current();
v8::Isolate* v8_isolate = reinterpret_cast<v8::Isolate*>(isolate);
Heap* heap = isolate->heap();
v8::Persistent<v8::Context> env;
{
HandleScope scope(isolate);
env.Reset(v8_isolate, v8::Context::New(v8_isolate));
}
ASSERT(!env.IsEmpty());
{
v8::HandleScope handle_scope(v8_isolate);
v8::Local<v8::Context>::New(v8_isolate, env)->Enter();
}
// Make sure all builtin scripts are cached.
{ HandleScope scope(isolate);
for (int i = 0; i < Natives::GetBuiltinsCount(); i++) {
isolate->bootstrapper()->NativesSourceLookup(i);
}
}
heap->CollectAllGarbage(Heap::kNoGCFlags);
heap->CollectAllGarbage(Heap::kNoGCFlags);
Object* raw_foo;
{
v8::HandleScope handle_scope(v8_isolate);
v8::Local<v8::String> foo = v8::String::New("foo");
ASSERT(!foo.IsEmpty());
raw_foo = *(v8::Utils::OpenHandle(*foo));
}
int file_name_length = StrLength(FLAG_testing_serialization_file) + 10;
Vector<char> startup_name = Vector<char>::New(file_name_length + 1);
OS::SNPrintF(startup_name, "%s.startup", FLAG_testing_serialization_file);
{
v8::HandleScope handle_scope(v8_isolate);
v8::Local<v8::Context>::New(v8_isolate, env)->Exit();
}
env.Dispose(v8_isolate);
FileByteSink startup_sink(startup_name.start());
StartupSerializer startup_serializer(&startup_sink);
startup_serializer.SerializeStrongReferences();
FileByteSink partial_sink(FLAG_testing_serialization_file);
PartialSerializer p_ser(&startup_serializer, &partial_sink);
p_ser.Serialize(&raw_foo);
startup_serializer.SerializeWeakReferences();
Refactoring of snapshots. This simplifies and improves the speed of deserializing code. The current startup time improvement for V8 is around 6%, but code deserialization is speeded up disproportionately, and we will soon have more code in the snapshot. * Removed support for deserializing into large object space. The regular pages are 1Mbyte now and that is plenty. This is a big simplification. * Instead of reserving space for the snapshot we actually allocate it now. This removes some special casing from the memory management and simplifies deserialization since we are just bumping a pointer rather than calling the normal allocation routines during deserialization. * Record in the snapshot how much we need to boot up and allocate it instead of just assuming that allocations in a new VM will always be linear. * In the snapshot we always address an object as a negative offset from the current allocation point. We used to sometimes address from the start of the deserialized data, but this is less useful now that we have good support for roots and repetitions in the deserialization data. * Code objects were previously deserialized (like other objects) by alternating raw data (deserialized with memcpy) and pointers (to external references, other objects, etc.). Now we deserialize code objects with a single memcpy, followed by a series of skips and pointers that partially overwrite the code we memcopied out of the snapshot. The skips are sometimes merged into the following instruction in the deserialization data to reduce dispatch time. * Integers in the snapshot were stored in a variable length format that gives a compact representation for small positive integers. This is still the case, but the new encoding can be decoded without branches or conditional instructions, which is faster on a modern CPU. Review URL: https://chromiumcodereview.appspot.com/10918067 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12505 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2012-09-14 11:16:56 +00:00
partial_sink.WriteSpaceUsed(
p_ser.CurrentAllocationAddress(NEW_SPACE),
p_ser.CurrentAllocationAddress(OLD_POINTER_SPACE),
p_ser.CurrentAllocationAddress(OLD_DATA_SPACE),
p_ser.CurrentAllocationAddress(CODE_SPACE),
p_ser.CurrentAllocationAddress(MAP_SPACE),
p_ser.CurrentAllocationAddress(CELL_SPACE),
p_ser.CurrentAllocationAddress(PROPERTY_CELL_SPACE));
Refactoring of snapshots. This simplifies and improves the speed of deserializing code. The current startup time improvement for V8 is around 6%, but code deserialization is speeded up disproportionately, and we will soon have more code in the snapshot. * Removed support for deserializing into large object space. The regular pages are 1Mbyte now and that is plenty. This is a big simplification. * Instead of reserving space for the snapshot we actually allocate it now. This removes some special casing from the memory management and simplifies deserialization since we are just bumping a pointer rather than calling the normal allocation routines during deserialization. * Record in the snapshot how much we need to boot up and allocate it instead of just assuming that allocations in a new VM will always be linear. * In the snapshot we always address an object as a negative offset from the current allocation point. We used to sometimes address from the start of the deserialized data, but this is less useful now that we have good support for roots and repetitions in the deserialization data. * Code objects were previously deserialized (like other objects) by alternating raw data (deserialized with memcpy) and pointers (to external references, other objects, etc.). Now we deserialize code objects with a single memcpy, followed by a series of skips and pointers that partially overwrite the code we memcopied out of the snapshot. The skips are sometimes merged into the following instruction in the deserialization data to reduce dispatch time. * Integers in the snapshot were stored in a variable length format that gives a compact representation for small positive integers. This is still the case, but the new encoding can be decoded without branches or conditional instructions, which is faster on a modern CPU. Review URL: https://chromiumcodereview.appspot.com/10918067 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12505 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2012-09-14 11:16:56 +00:00
startup_sink.WriteSpaceUsed(
startup_serializer.CurrentAllocationAddress(NEW_SPACE),
startup_serializer.CurrentAllocationAddress(OLD_POINTER_SPACE),
startup_serializer.CurrentAllocationAddress(OLD_DATA_SPACE),
startup_serializer.CurrentAllocationAddress(CODE_SPACE),
startup_serializer.CurrentAllocationAddress(MAP_SPACE),
startup_serializer.CurrentAllocationAddress(CELL_SPACE),
startup_serializer.CurrentAllocationAddress(PROPERTY_CELL_SPACE));
Refactoring of snapshots. This simplifies and improves the speed of deserializing code. The current startup time improvement for V8 is around 6%, but code deserialization is speeded up disproportionately, and we will soon have more code in the snapshot. * Removed support for deserializing into large object space. The regular pages are 1Mbyte now and that is plenty. This is a big simplification. * Instead of reserving space for the snapshot we actually allocate it now. This removes some special casing from the memory management and simplifies deserialization since we are just bumping a pointer rather than calling the normal allocation routines during deserialization. * Record in the snapshot how much we need to boot up and allocate it instead of just assuming that allocations in a new VM will always be linear. * In the snapshot we always address an object as a negative offset from the current allocation point. We used to sometimes address from the start of the deserialized data, but this is less useful now that we have good support for roots and repetitions in the deserialization data. * Code objects were previously deserialized (like other objects) by alternating raw data (deserialized with memcpy) and pointers (to external references, other objects, etc.). Now we deserialize code objects with a single memcpy, followed by a series of skips and pointers that partially overwrite the code we memcopied out of the snapshot. The skips are sometimes merged into the following instruction in the deserialization data to reduce dispatch time. * Integers in the snapshot were stored in a variable length format that gives a compact representation for small positive integers. This is still the case, but the new encoding can be decoded without branches or conditional instructions, which is faster on a modern CPU. Review URL: https://chromiumcodereview.appspot.com/10918067 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12505 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2012-09-14 11:16:56 +00:00
startup_name.Dispose();
}
}
Refactoring of snapshots. This simplifies and improves the speed of deserializing code. The current startup time improvement for V8 is around 6%, but code deserialization is speeded up disproportionately, and we will soon have more code in the snapshot. * Removed support for deserializing into large object space. The regular pages are 1Mbyte now and that is plenty. This is a big simplification. * Instead of reserving space for the snapshot we actually allocate it now. This removes some special casing from the memory management and simplifies deserialization since we are just bumping a pointer rather than calling the normal allocation routines during deserialization. * Record in the snapshot how much we need to boot up and allocate it instead of just assuming that allocations in a new VM will always be linear. * In the snapshot we always address an object as a negative offset from the current allocation point. We used to sometimes address from the start of the deserialized data, but this is less useful now that we have good support for roots and repetitions in the deserialization data. * Code objects were previously deserialized (like other objects) by alternating raw data (deserialized with memcpy) and pointers (to external references, other objects, etc.). Now we deserialize code objects with a single memcpy, followed by a series of skips and pointers that partially overwrite the code we memcopied out of the snapshot. The skips are sometimes merged into the following instruction in the deserialization data to reduce dispatch time. * Integers in the snapshot were stored in a variable length format that gives a compact representation for small positive integers. This is still the case, but the new encoding can be decoded without branches or conditional instructions, which is faster on a modern CPU. Review URL: https://chromiumcodereview.appspot.com/10918067 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12505 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2012-09-14 11:16:56 +00:00
static void ReserveSpaceForSnapshot(Deserializer* deserializer,
const char* file_name) {
int file_name_length = StrLength(file_name) + 10;
Vector<char> name = Vector<char>::New(file_name_length + 1);
OS::SNPrintF(name, "%s.size", file_name);
FILE* fp = OS::FOpen(name.start(), "r");
name.Dispose();
int new_size, pointer_size, data_size, code_size, map_size, cell_size,
property_cell_size;
#ifdef _MSC_VER
// Avoid warning about unsafe fscanf from MSVC.
// Please note that this is only fine if %c and %s are not being used.
#define fscanf fscanf_s
#endif
CHECK_EQ(1, fscanf(fp, "new %d\n", &new_size));
CHECK_EQ(1, fscanf(fp, "pointer %d\n", &pointer_size));
CHECK_EQ(1, fscanf(fp, "data %d\n", &data_size));
CHECK_EQ(1, fscanf(fp, "code %d\n", &code_size));
CHECK_EQ(1, fscanf(fp, "map %d\n", &map_size));
CHECK_EQ(1, fscanf(fp, "cell %d\n", &cell_size));
CHECK_EQ(1, fscanf(fp, "property cell %d\n", &property_cell_size));
#ifdef _MSC_VER
#undef fscanf
#endif
fclose(fp);
Refactoring of snapshots. This simplifies and improves the speed of deserializing code. The current startup time improvement for V8 is around 6%, but code deserialization is speeded up disproportionately, and we will soon have more code in the snapshot. * Removed support for deserializing into large object space. The regular pages are 1Mbyte now and that is plenty. This is a big simplification. * Instead of reserving space for the snapshot we actually allocate it now. This removes some special casing from the memory management and simplifies deserialization since we are just bumping a pointer rather than calling the normal allocation routines during deserialization. * Record in the snapshot how much we need to boot up and allocate it instead of just assuming that allocations in a new VM will always be linear. * In the snapshot we always address an object as a negative offset from the current allocation point. We used to sometimes address from the start of the deserialized data, but this is less useful now that we have good support for roots and repetitions in the deserialization data. * Code objects were previously deserialized (like other objects) by alternating raw data (deserialized with memcpy) and pointers (to external references, other objects, etc.). Now we deserialize code objects with a single memcpy, followed by a series of skips and pointers that partially overwrite the code we memcopied out of the snapshot. The skips are sometimes merged into the following instruction in the deserialization data to reduce dispatch time. * Integers in the snapshot were stored in a variable length format that gives a compact representation for small positive integers. This is still the case, but the new encoding can be decoded without branches or conditional instructions, which is faster on a modern CPU. Review URL: https://chromiumcodereview.appspot.com/10918067 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12505 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2012-09-14 11:16:56 +00:00
deserializer->set_reservation(NEW_SPACE, new_size);
deserializer->set_reservation(OLD_POINTER_SPACE, pointer_size);
deserializer->set_reservation(OLD_DATA_SPACE, data_size);
deserializer->set_reservation(CODE_SPACE, code_size);
deserializer->set_reservation(MAP_SPACE, map_size);
deserializer->set_reservation(CELL_SPACE, cell_size);
deserializer->set_reservation(PROPERTY_CELL_SPACE, property_cell_size);
}
DEPENDENT_TEST(PartialDeserialization, PartialSerialization) {
if (!Snapshot::IsEnabled()) {
int file_name_length = StrLength(FLAG_testing_serialization_file) + 10;
Vector<char> startup_name = Vector<char>::New(file_name_length + 1);
OS::SNPrintF(startup_name, "%s.startup", FLAG_testing_serialization_file);
CHECK(Snapshot::Initialize(startup_name.start()));
startup_name.Dispose();
const char* file_name = FLAG_testing_serialization_file;
int snapshot_size = 0;
byte* snapshot = ReadBytes(file_name, &snapshot_size);
Object* root;
{
SnapshotByteSource source(snapshot, snapshot_size);
Deserializer deserializer(&source);
Refactoring of snapshots. This simplifies and improves the speed of deserializing code. The current startup time improvement for V8 is around 6%, but code deserialization is speeded up disproportionately, and we will soon have more code in the snapshot. * Removed support for deserializing into large object space. The regular pages are 1Mbyte now and that is plenty. This is a big simplification. * Instead of reserving space for the snapshot we actually allocate it now. This removes some special casing from the memory management and simplifies deserialization since we are just bumping a pointer rather than calling the normal allocation routines during deserialization. * Record in the snapshot how much we need to boot up and allocate it instead of just assuming that allocations in a new VM will always be linear. * In the snapshot we always address an object as a negative offset from the current allocation point. We used to sometimes address from the start of the deserialized data, but this is less useful now that we have good support for roots and repetitions in the deserialization data. * Code objects were previously deserialized (like other objects) by alternating raw data (deserialized with memcpy) and pointers (to external references, other objects, etc.). Now we deserialize code objects with a single memcpy, followed by a series of skips and pointers that partially overwrite the code we memcopied out of the snapshot. The skips are sometimes merged into the following instruction in the deserialization data to reduce dispatch time. * Integers in the snapshot were stored in a variable length format that gives a compact representation for small positive integers. This is still the case, but the new encoding can be decoded without branches or conditional instructions, which is faster on a modern CPU. Review URL: https://chromiumcodereview.appspot.com/10918067 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12505 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2012-09-14 11:16:56 +00:00
ReserveSpaceForSnapshot(&deserializer, file_name);
deserializer.DeserializePartial(&root);
CHECK(root->IsString());
}
v8::HandleScope handle_scope(v8::Isolate::GetCurrent());
Handle<Object> root_handle(root, Isolate::Current());
Object* root2;
{
SnapshotByteSource source(snapshot, snapshot_size);
Deserializer deserializer(&source);
Refactoring of snapshots. This simplifies and improves the speed of deserializing code. The current startup time improvement for V8 is around 6%, but code deserialization is speeded up disproportionately, and we will soon have more code in the snapshot. * Removed support for deserializing into large object space. The regular pages are 1Mbyte now and that is plenty. This is a big simplification. * Instead of reserving space for the snapshot we actually allocate it now. This removes some special casing from the memory management and simplifies deserialization since we are just bumping a pointer rather than calling the normal allocation routines during deserialization. * Record in the snapshot how much we need to boot up and allocate it instead of just assuming that allocations in a new VM will always be linear. * In the snapshot we always address an object as a negative offset from the current allocation point. We used to sometimes address from the start of the deserialized data, but this is less useful now that we have good support for roots and repetitions in the deserialization data. * Code objects were previously deserialized (like other objects) by alternating raw data (deserialized with memcpy) and pointers (to external references, other objects, etc.). Now we deserialize code objects with a single memcpy, followed by a series of skips and pointers that partially overwrite the code we memcopied out of the snapshot. The skips are sometimes merged into the following instruction in the deserialization data to reduce dispatch time. * Integers in the snapshot were stored in a variable length format that gives a compact representation for small positive integers. This is still the case, but the new encoding can be decoded without branches or conditional instructions, which is faster on a modern CPU. Review URL: https://chromiumcodereview.appspot.com/10918067 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12505 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2012-09-14 11:16:56 +00:00
ReserveSpaceForSnapshot(&deserializer, file_name);
deserializer.DeserializePartial(&root2);
CHECK(root2->IsString());
CHECK(*root_handle == root2);
}
}
}
TEST(ContextSerialization) {
if (!Snapshot::HaveASnapshotToStartFrom()) {
Serializer::Enable();
v8::V8::Initialize();
Isolate* isolate = Isolate::Current();
v8::Isolate* v8_isolate = reinterpret_cast<v8::Isolate*>(isolate);
Heap* heap = isolate->heap();
v8::Persistent<v8::Context> env;
{
HandleScope scope(isolate);
env.Reset(v8_isolate, v8::Context::New(v8_isolate));
}
ASSERT(!env.IsEmpty());
{
v8::HandleScope handle_scope(v8_isolate);
v8::Local<v8::Context>::New(v8_isolate, env)->Enter();
}
// Make sure all builtin scripts are cached.
{ HandleScope scope(isolate);
for (int i = 0; i < Natives::GetBuiltinsCount(); i++) {
isolate->bootstrapper()->NativesSourceLookup(i);
}
}
// If we don't do this then we end up with a stray root pointing at the
// context even after we have disposed of env.
heap->CollectAllGarbage(Heap::kNoGCFlags);
int file_name_length = StrLength(FLAG_testing_serialization_file) + 10;
Vector<char> startup_name = Vector<char>::New(file_name_length + 1);
OS::SNPrintF(startup_name, "%s.startup", FLAG_testing_serialization_file);
{
v8::HandleScope handle_scope(v8_isolate);
v8::Local<v8::Context>::New(v8_isolate, env)->Exit();
}
i::Object* raw_context = *v8::Utils::OpenPersistent(env);
env.Dispose(v8_isolate);
FileByteSink startup_sink(startup_name.start());
StartupSerializer startup_serializer(&startup_sink);
startup_serializer.SerializeStrongReferences();
FileByteSink partial_sink(FLAG_testing_serialization_file);
PartialSerializer p_ser(&startup_serializer, &partial_sink);
p_ser.Serialize(&raw_context);
startup_serializer.SerializeWeakReferences();
Refactoring of snapshots. This simplifies and improves the speed of deserializing code. The current startup time improvement for V8 is around 6%, but code deserialization is speeded up disproportionately, and we will soon have more code in the snapshot. * Removed support for deserializing into large object space. The regular pages are 1Mbyte now and that is plenty. This is a big simplification. * Instead of reserving space for the snapshot we actually allocate it now. This removes some special casing from the memory management and simplifies deserialization since we are just bumping a pointer rather than calling the normal allocation routines during deserialization. * Record in the snapshot how much we need to boot up and allocate it instead of just assuming that allocations in a new VM will always be linear. * In the snapshot we always address an object as a negative offset from the current allocation point. We used to sometimes address from the start of the deserialized data, but this is less useful now that we have good support for roots and repetitions in the deserialization data. * Code objects were previously deserialized (like other objects) by alternating raw data (deserialized with memcpy) and pointers (to external references, other objects, etc.). Now we deserialize code objects with a single memcpy, followed by a series of skips and pointers that partially overwrite the code we memcopied out of the snapshot. The skips are sometimes merged into the following instruction in the deserialization data to reduce dispatch time. * Integers in the snapshot were stored in a variable length format that gives a compact representation for small positive integers. This is still the case, but the new encoding can be decoded without branches or conditional instructions, which is faster on a modern CPU. Review URL: https://chromiumcodereview.appspot.com/10918067 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12505 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2012-09-14 11:16:56 +00:00
partial_sink.WriteSpaceUsed(
p_ser.CurrentAllocationAddress(NEW_SPACE),
p_ser.CurrentAllocationAddress(OLD_POINTER_SPACE),
p_ser.CurrentAllocationAddress(OLD_DATA_SPACE),
p_ser.CurrentAllocationAddress(CODE_SPACE),
p_ser.CurrentAllocationAddress(MAP_SPACE),
p_ser.CurrentAllocationAddress(CELL_SPACE),
p_ser.CurrentAllocationAddress(PROPERTY_CELL_SPACE));
Refactoring of snapshots. This simplifies and improves the speed of deserializing code. The current startup time improvement for V8 is around 6%, but code deserialization is speeded up disproportionately, and we will soon have more code in the snapshot. * Removed support for deserializing into large object space. The regular pages are 1Mbyte now and that is plenty. This is a big simplification. * Instead of reserving space for the snapshot we actually allocate it now. This removes some special casing from the memory management and simplifies deserialization since we are just bumping a pointer rather than calling the normal allocation routines during deserialization. * Record in the snapshot how much we need to boot up and allocate it instead of just assuming that allocations in a new VM will always be linear. * In the snapshot we always address an object as a negative offset from the current allocation point. We used to sometimes address from the start of the deserialized data, but this is less useful now that we have good support for roots and repetitions in the deserialization data. * Code objects were previously deserialized (like other objects) by alternating raw data (deserialized with memcpy) and pointers (to external references, other objects, etc.). Now we deserialize code objects with a single memcpy, followed by a series of skips and pointers that partially overwrite the code we memcopied out of the snapshot. The skips are sometimes merged into the following instruction in the deserialization data to reduce dispatch time. * Integers in the snapshot were stored in a variable length format that gives a compact representation for small positive integers. This is still the case, but the new encoding can be decoded without branches or conditional instructions, which is faster on a modern CPU. Review URL: https://chromiumcodereview.appspot.com/10918067 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12505 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2012-09-14 11:16:56 +00:00
startup_sink.WriteSpaceUsed(
startup_serializer.CurrentAllocationAddress(NEW_SPACE),
startup_serializer.CurrentAllocationAddress(OLD_POINTER_SPACE),
startup_serializer.CurrentAllocationAddress(OLD_DATA_SPACE),
startup_serializer.CurrentAllocationAddress(CODE_SPACE),
startup_serializer.CurrentAllocationAddress(MAP_SPACE),
startup_serializer.CurrentAllocationAddress(CELL_SPACE),
startup_serializer.CurrentAllocationAddress(PROPERTY_CELL_SPACE));
Refactoring of snapshots. This simplifies and improves the speed of deserializing code. The current startup time improvement for V8 is around 6%, but code deserialization is speeded up disproportionately, and we will soon have more code in the snapshot. * Removed support for deserializing into large object space. The regular pages are 1Mbyte now and that is plenty. This is a big simplification. * Instead of reserving space for the snapshot we actually allocate it now. This removes some special casing from the memory management and simplifies deserialization since we are just bumping a pointer rather than calling the normal allocation routines during deserialization. * Record in the snapshot how much we need to boot up and allocate it instead of just assuming that allocations in a new VM will always be linear. * In the snapshot we always address an object as a negative offset from the current allocation point. We used to sometimes address from the start of the deserialized data, but this is less useful now that we have good support for roots and repetitions in the deserialization data. * Code objects were previously deserialized (like other objects) by alternating raw data (deserialized with memcpy) and pointers (to external references, other objects, etc.). Now we deserialize code objects with a single memcpy, followed by a series of skips and pointers that partially overwrite the code we memcopied out of the snapshot. The skips are sometimes merged into the following instruction in the deserialization data to reduce dispatch time. * Integers in the snapshot were stored in a variable length format that gives a compact representation for small positive integers. This is still the case, but the new encoding can be decoded without branches or conditional instructions, which is faster on a modern CPU. Review URL: https://chromiumcodereview.appspot.com/10918067 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12505 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2012-09-14 11:16:56 +00:00
startup_name.Dispose();
}
}
DEPENDENT_TEST(ContextDeserialization, ContextSerialization) {
if (!Snapshot::HaveASnapshotToStartFrom()) {
int file_name_length = StrLength(FLAG_testing_serialization_file) + 10;
Vector<char> startup_name = Vector<char>::New(file_name_length + 1);
OS::SNPrintF(startup_name, "%s.startup", FLAG_testing_serialization_file);
CHECK(Snapshot::Initialize(startup_name.start()));
startup_name.Dispose();
const char* file_name = FLAG_testing_serialization_file;
int snapshot_size = 0;
byte* snapshot = ReadBytes(file_name, &snapshot_size);
Object* root;
{
SnapshotByteSource source(snapshot, snapshot_size);
Deserializer deserializer(&source);
Refactoring of snapshots. This simplifies and improves the speed of deserializing code. The current startup time improvement for V8 is around 6%, but code deserialization is speeded up disproportionately, and we will soon have more code in the snapshot. * Removed support for deserializing into large object space. The regular pages are 1Mbyte now and that is plenty. This is a big simplification. * Instead of reserving space for the snapshot we actually allocate it now. This removes some special casing from the memory management and simplifies deserialization since we are just bumping a pointer rather than calling the normal allocation routines during deserialization. * Record in the snapshot how much we need to boot up and allocate it instead of just assuming that allocations in a new VM will always be linear. * In the snapshot we always address an object as a negative offset from the current allocation point. We used to sometimes address from the start of the deserialized data, but this is less useful now that we have good support for roots and repetitions in the deserialization data. * Code objects were previously deserialized (like other objects) by alternating raw data (deserialized with memcpy) and pointers (to external references, other objects, etc.). Now we deserialize code objects with a single memcpy, followed by a series of skips and pointers that partially overwrite the code we memcopied out of the snapshot. The skips are sometimes merged into the following instruction in the deserialization data to reduce dispatch time. * Integers in the snapshot were stored in a variable length format that gives a compact representation for small positive integers. This is still the case, but the new encoding can be decoded without branches or conditional instructions, which is faster on a modern CPU. Review URL: https://chromiumcodereview.appspot.com/10918067 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12505 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2012-09-14 11:16:56 +00:00
ReserveSpaceForSnapshot(&deserializer, file_name);
deserializer.DeserializePartial(&root);
CHECK(root->IsContext());
}
v8::HandleScope handle_scope(v8::Isolate::GetCurrent());
Handle<Object> root_handle(root, Isolate::Current());
Object* root2;
{
SnapshotByteSource source(snapshot, snapshot_size);
Deserializer deserializer(&source);
Refactoring of snapshots. This simplifies and improves the speed of deserializing code. The current startup time improvement for V8 is around 6%, but code deserialization is speeded up disproportionately, and we will soon have more code in the snapshot. * Removed support for deserializing into large object space. The regular pages are 1Mbyte now and that is plenty. This is a big simplification. * Instead of reserving space for the snapshot we actually allocate it now. This removes some special casing from the memory management and simplifies deserialization since we are just bumping a pointer rather than calling the normal allocation routines during deserialization. * Record in the snapshot how much we need to boot up and allocate it instead of just assuming that allocations in a new VM will always be linear. * In the snapshot we always address an object as a negative offset from the current allocation point. We used to sometimes address from the start of the deserialized data, but this is less useful now that we have good support for roots and repetitions in the deserialization data. * Code objects were previously deserialized (like other objects) by alternating raw data (deserialized with memcpy) and pointers (to external references, other objects, etc.). Now we deserialize code objects with a single memcpy, followed by a series of skips and pointers that partially overwrite the code we memcopied out of the snapshot. The skips are sometimes merged into the following instruction in the deserialization data to reduce dispatch time. * Integers in the snapshot were stored in a variable length format that gives a compact representation for small positive integers. This is still the case, but the new encoding can be decoded without branches or conditional instructions, which is faster on a modern CPU. Review URL: https://chromiumcodereview.appspot.com/10918067 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@12505 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2012-09-14 11:16:56 +00:00
ReserveSpaceForSnapshot(&deserializer, file_name);
deserializer.DeserializePartial(&root2);
CHECK(root2->IsContext());
CHECK(*root_handle != root2);
}
}
}
TEST(TestThatAlwaysSucceeds) {
}
TEST(TestThatAlwaysFails) {
bool ArtificialFailure = false;
CHECK(ArtificialFailure);
}
DEPENDENT_TEST(DependentTestThatAlwaysFails, TestThatAlwaysSucceeds) {
bool ArtificialFailure2 = false;
CHECK(ArtificialFailure2);
}