v8/src/mksnapshot.cc

404 lines
13 KiB
C++
Raw Normal View History

// Copyright 2006-2008 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 <errno.h>
#include <stdio.h>
#ifdef COMPRESS_STARTUP_DATA_BZ2
#include <bzlib.h>
#endif
#include <signal.h>
#include "v8.h"
#include "bootstrapper.h"
#include "flags.h"
#include "natives.h"
#include "platform.h"
#include "serialize.h"
#include "list.h"
using namespace v8;
class Compressor {
public:
virtual ~Compressor() {}
virtual bool Compress(i::Vector<char> input) = 0;
virtual i::Vector<char>* output() = 0;
};
class PartialSnapshotSink : public i::SnapshotByteSink {
public:
PartialSnapshotSink() : data_(), raw_size_(-1) { }
virtual ~PartialSnapshotSink() { data_.Free(); }
virtual void Put(int byte, const char* description) {
data_.Add(byte);
}
virtual int Position() { return data_.length(); }
void Print(FILE* fp) {
int length = Position();
for (int j = 0; j < length; j++) {
if ((j & 0x1f) == 0x1f) {
fprintf(fp, "\n");
}
if (j != 0) {
fprintf(fp, ",");
}
fprintf(fp, "%u", static_cast<unsigned char>(at(j)));
}
}
char at(int i) { return data_[i]; }
bool Compress(Compressor* compressor) {
ASSERT_EQ(-1, raw_size_);
raw_size_ = data_.length();
if (!compressor->Compress(data_.ToVector())) return false;
data_.Clear();
data_.AddAll(*compressor->output());
return true;
}
int raw_size() { return raw_size_; }
private:
i::List<char> data_;
int raw_size_;
};
class CppByteSink : public PartialSnapshotSink {
public:
explicit CppByteSink(const char* snapshot_file) {
fp_ = i::OS::FOpen(snapshot_file, "wb");
if (fp_ == NULL) {
i::PrintF("Unable to write to snapshot file \"%s\"\n", snapshot_file);
exit(1);
}
fprintf(fp_, "// Autogenerated snapshot file. Do not edit.\n\n");
fprintf(fp_, "#include \"v8.h\"\n");
fprintf(fp_, "#include \"platform.h\"\n\n");
fprintf(fp_, "#include \"snapshot.h\"\n\n");
fprintf(fp_, "namespace v8 {\nnamespace internal {\n\n");
fprintf(fp_, "const byte Snapshot::data_[] = {");
}
virtual ~CppByteSink() {
fprintf(fp_, "const int Snapshot::size_ = %d;\n", Position());
#ifdef COMPRESS_STARTUP_DATA_BZ2
fprintf(fp_, "const byte* Snapshot::raw_data_ = NULL;\n");
fprintf(fp_,
"const int Snapshot::raw_size_ = %d;\n\n",
raw_size());
#else
fprintf(fp_,
"const byte* Snapshot::raw_data_ = Snapshot::data_;\n");
fprintf(fp_,
"const int Snapshot::raw_size_ = Snapshot::size_;\n\n");
#endif
fprintf(fp_, "} } // namespace v8::internal\n");
fclose(fp_);
}
void WriteSpaceUsed(
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
const char* prefix,
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) {
fprintf(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
"const int Snapshot::%snew_space_used_ = %d;\n",
prefix,
new_space_used);
fprintf(fp_,
"const int Snapshot::%spointer_space_used_ = %d;\n",
prefix,
pointer_space_used);
fprintf(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
"const int Snapshot::%sdata_space_used_ = %d;\n",
prefix,
data_space_used);
fprintf(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
"const int Snapshot::%scode_space_used_ = %d;\n",
prefix,
code_space_used);
fprintf(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
"const int Snapshot::%smap_space_used_ = %d;\n",
prefix,
map_space_used);
fprintf(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
"const int Snapshot::%scell_space_used_ = %d;\n",
prefix,
cell_space_used);
fprintf(fp_,
"const int Snapshot::%sproperty_cell_space_used_ = %d;\n",
prefix,
property_cell_space_used);
}
void WritePartialSnapshot() {
int length = partial_sink_.Position();
fprintf(fp_, "};\n\n");
fprintf(fp_, "const int Snapshot::context_size_ = %d;\n", length);
#ifdef COMPRESS_STARTUP_DATA_BZ2
fprintf(fp_,
"const int Snapshot::context_raw_size_ = %d;\n",
partial_sink_.raw_size());
#else
fprintf(fp_,
"const int Snapshot::context_raw_size_ = "
"Snapshot::context_size_;\n");
#endif
fprintf(fp_, "const byte Snapshot::context_data_[] = {\n");
partial_sink_.Print(fp_);
fprintf(fp_, "};\n\n");
#ifdef COMPRESS_STARTUP_DATA_BZ2
fprintf(fp_, "const byte* Snapshot::context_raw_data_ = NULL;\n");
#else
fprintf(fp_, "const byte* Snapshot::context_raw_data_ ="
" Snapshot::context_data_;\n");
#endif
}
void WriteSnapshot() {
Print(fp_);
}
PartialSnapshotSink* partial_sink() { return &partial_sink_; }
private:
FILE* fp_;
PartialSnapshotSink partial_sink_;
};
#ifdef COMPRESS_STARTUP_DATA_BZ2
class BZip2Compressor : public Compressor {
public:
BZip2Compressor() : output_(NULL) {}
virtual ~BZip2Compressor() {
delete output_;
}
virtual bool Compress(i::Vector<char> input) {
delete output_;
output_ = new i::ScopedVector<char>((input.length() * 101) / 100 + 1000);
unsigned int output_length_ = output_->length();
int result = BZ2_bzBuffToBuffCompress(output_->start(), &output_length_,
input.start(), input.length(),
9, 1, 0);
if (result == BZ_OK) {
output_->Truncate(output_length_);
return true;
} else {
fprintf(stderr, "bzlib error code: %d\n", result);
return false;
}
}
virtual i::Vector<char>* output() { return output_; }
private:
i::ScopedVector<char>* output_;
};
class BZip2Decompressor : public StartupDataDecompressor {
public:
virtual ~BZip2Decompressor() { }
protected:
virtual int DecompressData(char* raw_data,
int* raw_data_size,
const char* compressed_data,
int compressed_data_size) {
ASSERT_EQ(StartupData::kBZip2,
V8::GetCompressedStartupDataAlgorithm());
unsigned int decompressed_size = *raw_data_size;
int result =
BZ2_bzBuffToBuffDecompress(raw_data,
&decompressed_size,
const_cast<char*>(compressed_data),
compressed_data_size,
0, 1);
if (result == BZ_OK) {
*raw_data_size = decompressed_size;
}
return result;
}
};
#endif
void DumpException(Handle<Message> message) {
String::Utf8Value message_string(message->Get());
String::Utf8Value message_line(message->GetSourceLine());
fprintf(stderr, "%s at line %d\n", *message_string, message->GetLineNumber());
fprintf(stderr, "%s\n", *message_line);
for (int i = 0; i <= message->GetEndColumn(); ++i) {
fprintf(stderr, "%c", i < message->GetStartColumn() ? ' ' : '^');
}
fprintf(stderr, "\n");
}
int main(int argc, char** argv) {
V8::InitializeICU();
i::Isolate::SetCrashIfDefaultIsolateInitialized();
// By default, log code create information in the snapshot.
i::FLAG_log_code = true;
// Print the usage if an error occurs when parsing the command line
// flags or if the help flag is set.
int result = i::FlagList::SetFlagsFromCommandLine(&argc, argv, true);
if (result > 0 || argc != 2 || i::FLAG_help) {
::printf("Usage: %s [flag] ... outfile\n", argv[0]);
i::FlagList::PrintHelp();
return !i::FLAG_help;
}
#ifdef COMPRESS_STARTUP_DATA_BZ2
BZip2Decompressor natives_decompressor;
int bz2_result = natives_decompressor.Decompress();
if (bz2_result != BZ_OK) {
fprintf(stderr, "bzip error code: %d\n", bz2_result);
exit(1);
}
#endif
i::FLAG_logfile_per_isolate = false;
Isolate* isolate = v8::Isolate::New();
isolate->Enter();
i::Isolate* internal_isolate = reinterpret_cast<i::Isolate*>(isolate);
i::Serializer::Enable(internal_isolate);
Persistent<Context> context;
{
HandleScope handle_scope(isolate);
context.Reset(isolate, Context::New(isolate));
}
if (context.IsEmpty()) {
fprintf(stderr,
"\nException thrown while compiling natives - see above.\n\n");
exit(1);
}
if (i::FLAG_extra_code != NULL) {
// Capture 100 frames if anything happens.
V8::SetCaptureStackTraceForUncaughtExceptions(true, 100);
HandleScope scope(isolate);
v8::Context::Scope cscope(v8::Local<v8::Context>::New(isolate, context));
const char* name = i::FLAG_extra_code;
FILE* file = i::OS::FOpen(name, "rb");
if (file == NULL) {
fprintf(stderr, "Failed to open '%s': errno %d\n", name, errno);
exit(1);
}
fseek(file, 0, SEEK_END);
int size = ftell(file);
rewind(file);
char* chars = new char[size + 1];
chars[size] = '\0';
for (int i = 0; i < size;) {
int read = static_cast<int>(fread(&chars[i], 1, size - i, file));
if (read < 0) {
fprintf(stderr, "Failed to read '%s': errno %d\n", name, errno);
exit(1);
}
i += read;
}
fclose(file);
Local<String> source = String::NewFromUtf8(isolate, chars);
TryCatch try_catch;
Local<Script> script = Script::Compile(source);
if (try_catch.HasCaught()) {
fprintf(stderr, "Failure compiling '%s'\n", name);
DumpException(try_catch.Message());
exit(1);
}
script->Run();
if (try_catch.HasCaught()) {
fprintf(stderr, "Failure running '%s'\n", name);
DumpException(try_catch.Message());
exit(1);
}
}
// Make sure all builtin scripts are cached.
{ HandleScope scope(isolate);
for (int i = 0; i < i::Natives::GetBuiltinsCount(); i++) {
internal_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 the context.
internal_isolate->heap()->CollectAllGarbage(
i::Heap::kNoGCFlags, "mksnapshot");
i::Object* raw_context = *v8::Utils::OpenPersistent(context);
context.Reset();
CppByteSink sink(argv[1]);
// This results in a somewhat smaller snapshot, probably because it gets rid
// of some things that are cached between garbage collections.
i::StartupSerializer ser(internal_isolate, &sink);
ser.SerializeStrongReferences();
i::PartialSerializer partial_ser(
internal_isolate, &ser, sink.partial_sink());
partial_ser.Serialize(&raw_context);
ser.SerializeWeakReferences();
#ifdef COMPRESS_STARTUP_DATA_BZ2
BZip2Compressor compressor;
if (!sink.Compress(&compressor))
return 1;
if (!sink.partial_sink()->Compress(&compressor))
return 1;
#endif
sink.WriteSnapshot();
sink.WritePartialSnapshot();
sink.WriteSpaceUsed(
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
"context_",
partial_ser.CurrentAllocationAddress(i::NEW_SPACE),
partial_ser.CurrentAllocationAddress(i::OLD_POINTER_SPACE),
partial_ser.CurrentAllocationAddress(i::OLD_DATA_SPACE),
partial_ser.CurrentAllocationAddress(i::CODE_SPACE),
partial_ser.CurrentAllocationAddress(i::MAP_SPACE),
partial_ser.CurrentAllocationAddress(i::CELL_SPACE),
partial_ser.CurrentAllocationAddress(i::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
sink.WriteSpaceUsed(
"",
ser.CurrentAllocationAddress(i::NEW_SPACE),
ser.CurrentAllocationAddress(i::OLD_POINTER_SPACE),
ser.CurrentAllocationAddress(i::OLD_DATA_SPACE),
ser.CurrentAllocationAddress(i::CODE_SPACE),
ser.CurrentAllocationAddress(i::MAP_SPACE),
ser.CurrentAllocationAddress(i::CELL_SPACE),
ser.CurrentAllocationAddress(i::PROPERTY_CELL_SPACE));
return 0;
}