[wasm] Make StreamingDecoder abstract, introduce AsyncStreamingDecoder

The implementation of the StreamingDecoder depends on async compilation.
However, when the --single-threaded flag is set, async compilation is
not available. Therefore V8 does not support streaming compilation at
the moment if the --single-threaded flag is set.

This CL is the first step to support streaming compilation in
--single-threaded mode. This CL makes the StreamingDecoder an abstract
class, and the current implementation a sub-class called
AsyncStreamingDecoder. A follow-up CL will provided a second sub-class
implementation for streaming compilation in --single-threaded mode.

Bug: v8:10548
Change-Id: Ice5c01340d3df18f836a4a05d30571207ca8ccf6
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2208869
Commit-Queue: Andreas Haas <ahaas@chromium.org>
Reviewed-by: Thibaud Michaud <thibaudm@chromium.org>
Cr-Commit-Position: refs/heads/master@{#67931}

src/wasm/module-compiler.cc

@@ -1591,8 +1591,9 @@ class AsyncStreamingProcessor final : public StreamingProcessor {
 
 std::shared_ptr<StreamingDecoder> AsyncCompileJob::CreateStreamingDecoder() {
   DCHECK_NULL(stream_);
-  stream_.reset(new StreamingDecoder(std::make_unique<AsyncStreamingProcessor>(
-      this, isolate_->async_counters(), isolate_->allocator())));
+  stream_ = StreamingDecoder::CreateAsyncStreamingDecoder(
+      std::make_unique<AsyncStreamingProcessor>(
+          this, isolate_->async_counters(), isolate_->allocator()));
   return stream_;
 }
 

src/wasm/streaming-decoder.h

@@ -66,220 +66,35 @@ class V8_EXPORT_PRIVATE StreamingProcessor {
 // and function bodies.
 class V8_EXPORT_PRIVATE StreamingDecoder {
  public:
-  explicit StreamingDecoder(std::unique_ptr<StreamingProcessor> processor);
+  virtual ~StreamingDecoder() = default;
 
   // The buffer passed into OnBytesReceived is owned by the caller.
-  void OnBytesReceived(Vector<const uint8_t> bytes);
+  virtual void OnBytesReceived(Vector<const uint8_t> bytes) = 0;
 
-  void Finish();
+  virtual void Finish() = 0;
 
-  void Abort();
+  virtual void Abort() = 0;
 
   // Notify the StreamingDecoder that compilation ended and the
   // StreamingProcessor should not be called anymore.
-  void NotifyCompilationEnded() { Fail(); }
+  virtual void NotifyCompilationEnded() = 0;
 
   // Caching support.
   // Sets the callback that is called after the module is fully compiled.
   using ModuleCompiledCallback =
       std::function<void(const std::shared_ptr<NativeModule>&)>;
-  void SetModuleCompiledCallback(ModuleCompiledCallback callback);
+  virtual void SetModuleCompiledCallback(ModuleCompiledCallback callback) = 0;
   // Passes previously compiled module bytes from the embedder's cache.
-  bool SetCompiledModuleBytes(Vector<const uint8_t> compiled_module_bytes);
+  virtual bool SetCompiledModuleBytes(
+      Vector<const uint8_t> compiled_module_bytes) = 0;
 
-  void NotifyNativeModuleCreated(
-      const std::shared_ptr<NativeModule>& native_module);
+  virtual void NotifyNativeModuleCreated(
+      const std::shared_ptr<NativeModule>& native_module) = 0;
 
-  Vector<const char> url() { return VectorOf(url_); }
-  void SetUrl(Vector<const char> url) {
-    url_.assign(url.begin(), url.length());
-  }
-
- private:
-  // TODO(ahaas): Put the whole private state of the StreamingDecoder into the
-  // cc file (PIMPL design pattern).
-
-  // The SectionBuffer is the data object for the content of a single section.
-  // It stores all bytes of the section (including section id and section
-  // length), and the offset where the actual payload starts.
-  class SectionBuffer : public WireBytesStorage {
-   public:
-    // id: The section id.
-    // payload_length: The length of the payload.
-    // length_bytes: The section length, as it is encoded in the module bytes.
-    SectionBuffer(uint32_t module_offset, uint8_t id, size_t payload_length,
-                  Vector<const uint8_t> length_bytes)
-        :  // ID + length + payload
-          module_offset_(module_offset),
-          bytes_(OwnedVector<uint8_t>::New(1 + length_bytes.length() +
-                                           payload_length)),
-          payload_offset_(1 + length_bytes.length()) {
-      bytes_.start()[0] = id;
-      memcpy(bytes_.start() + 1, &length_bytes.first(), length_bytes.length());
-    }
-
-    SectionCode section_code() const {
-      return static_cast<SectionCode>(bytes_.start()[0]);
-    }
-
-    Vector<const uint8_t> GetCode(WireBytesRef ref) const final {
-      DCHECK_LE(module_offset_, ref.offset());
-      uint32_t offset_in_code_buffer = ref.offset() - module_offset_;
-      return bytes().SubVector(offset_in_code_buffer,
-                               offset_in_code_buffer + ref.length());
-    }
-
-    uint32_t module_offset() const { return module_offset_; }
-    Vector<uint8_t> bytes() const { return bytes_.as_vector(); }
-    Vector<uint8_t> payload() const { return bytes() + payload_offset_; }
-    size_t length() const { return bytes_.size(); }
-    size_t payload_offset() const { return payload_offset_; }
-
-   private:
-    const uint32_t module_offset_;
-    const OwnedVector<uint8_t> bytes_;
-    const size_t payload_offset_;
-  };
-
-  // The decoding of a stream of wasm module bytes is organized in states. Each
-  // state provides a buffer to store the bytes required for the current state,
-  // information on how many bytes have already been received, how many bytes
-  // are needed, and a {Next} function which starts the next state once all
-  // bytes of the current state were received.
-  //
-  // The states change according to the following state diagram:
-  //
-  //       Start
-  //         |
-  //         |
-  //         v
-  // DecodeModuleHeader
-  //         |   _________________________________________
-  //         |   |                                        |
-  //         v   v                                        |
-  //  DecodeSectionID --> DecodeSectionLength --> DecodeSectionPayload
-  //         A                  |
-  //         |                  | (if the section id == code)
-  //         |                  v
-  //         |      DecodeNumberOfFunctions -- > DecodeFunctionLength
-  //         |                                          A    |
-  //         |                                          |    |
-  //         |  (after all functions were read)         |    v
-  //         ------------------------------------- DecodeFunctionBody
-  //
-  class DecodingState {
-   public:
-    virtual ~DecodingState() = default;
-
-    // Reads the bytes for the current state and returns the number of read
-    // bytes.
-    virtual size_t ReadBytes(StreamingDecoder* streaming,
-                             Vector<const uint8_t> bytes);
-
-    // Returns the next state of the streaming decoding.
-    virtual std::unique_ptr<DecodingState> Next(
-        StreamingDecoder* streaming) = 0;
-    // The buffer to store the received bytes.
-    virtual Vector<uint8_t> buffer() = 0;
-    // The number of bytes which were already received.
-    size_t offset() const { return offset_; }
-    void set_offset(size_t value) { offset_ = value; }
-    // A flag to indicate if finishing the streaming decoder is allowed without
-    // error.
-    virtual bool is_finishing_allowed() const { return false; }
-
-   private:
-    size_t offset_ = 0;
-  };
-
-  // Forward declarations of the concrete states. This is needed so that they
-  // can access private members of the StreamingDecoder.
-  class DecodeVarInt32;
-  class DecodeModuleHeader;
-  class DecodeSectionID;
-  class DecodeSectionLength;
-  class DecodeSectionPayload;
-  class DecodeNumberOfFunctions;
-  class DecodeFunctionLength;
-  class DecodeFunctionBody;
-
-  // Creates a buffer for the next section of the module.
-  SectionBuffer* CreateNewBuffer(uint32_t module_offset, uint8_t section_id,
-                                 size_t length,
-                                 Vector<const uint8_t> length_bytes);
-
-  std::unique_ptr<DecodingState> Error(const WasmError& error) {
-    if (ok()) processor_->OnError(error);
-    Fail();
-    return std::unique_ptr<DecodingState>(nullptr);
-  }
-
-  std::unique_ptr<DecodingState> Error(std::string message) {
-    return Error(WasmError{module_offset_ - 1, std::move(message)});
-  }
-
-  void ProcessModuleHeader() {
-    if (!ok()) return;
-    if (!processor_->ProcessModuleHeader(state_->buffer(), 0)) Fail();
-  }
-
-  void ProcessSection(SectionBuffer* buffer) {
-    if (!ok()) return;
-    if (!processor_->ProcessSection(
-            buffer->section_code(), buffer->payload(),
-            buffer->module_offset() +
-                static_cast<uint32_t>(buffer->payload_offset()))) {
-      Fail();
-    }
-  }
-
-  void StartCodeSection(int num_functions,
-                        std::shared_ptr<WireBytesStorage> wire_bytes_storage,
-                        int code_section_length) {
-    if (!ok()) return;
-    // The offset passed to {ProcessCodeSectionHeader} is an error offset and
-    // not the start offset of a buffer. Therefore we need the -1 here.
-    if (!processor_->ProcessCodeSectionHeader(
-            num_functions, module_offset() - 1, std::move(wire_bytes_storage),
-            code_section_length)) {
-      Fail();
-    }
-  }
-
-  void ProcessFunctionBody(Vector<const uint8_t> bytes,
-                           uint32_t module_offset) {
-    if (!ok()) return;
-    if (!processor_->ProcessFunctionBody(bytes, module_offset)) Fail();
-  }
-
-  void Fail() {
-    // We reset the {processor_} field to represent failure. This also ensures
-    // that we do not accidentally call further methods on the processor after
-    // failure.
-    processor_.reset();
-  }
-
-  bool ok() const { return processor_ != nullptr; }
-
-  uint32_t module_offset() const { return module_offset_; }
-
-  bool deserializing() const { return !compiled_module_bytes_.empty(); }
-
-  std::unique_ptr<StreamingProcessor> processor_;
-  std::unique_ptr<DecodingState> state_;
-  std::vector<std::shared_ptr<SectionBuffer>> section_buffers_;
-  bool code_section_processed_ = false;
-  uint32_t module_offset_ = 0;
-  size_t total_size_ = 0;
-  std::string url_;
-
-  // Caching support.
-  ModuleCompiledCallback module_compiled_callback_ = nullptr;
-  // We need wire bytes in an array for deserializing cached modules.
-  std::vector<uint8_t> wire_bytes_for_deserializing_;
-  Vector<const uint8_t> compiled_module_bytes_;
-
-  DISALLOW_COPY_AND_ASSIGN(StreamingDecoder);
+  virtual Vector<const char> url() = 0;
+  virtual void SetUrl(Vector<const char> url) = 0;
+  static std::unique_ptr<StreamingDecoder> CreateAsyncStreamingDecoder(
+      std::unique_ptr<StreamingProcessor> processor);
 };
 
 }  // namespace wasm

test/unittests/wasm/streaming-decoder-unittest.cc

@@ -99,11 +99,11 @@ class WasmStreamingDecoderTest : public ::testing::Test {
                       size_t expected_functions) {
     for (int split = 0; split <= data.length(); ++split) {
       MockStreamingResult result;
-      StreamingDecoder stream(
+      auto stream = StreamingDecoder::CreateAsyncStreamingDecoder(
           std::make_unique<MockStreamingProcessor>(&result));
-      stream.OnBytesReceived(data.SubVector(0, split));
-      stream.OnBytesReceived(data.SubVector(split, data.length()));
-      stream.Finish();
+      stream->OnBytesReceived(data.SubVector(0, split));
+      stream->OnBytesReceived(data.SubVector(split, data.length()));
+      stream->Finish();
       EXPECT_TRUE(result.ok());
       EXPECT_EQ(expected_sections, result.num_sections);
       EXPECT_EQ(expected_functions, result.num_functions);
@@ -115,11 +115,11 @@ class WasmStreamingDecoderTest : public ::testing::Test {
                      const char* message) {
     for (int split = 0; split <= data.length(); ++split) {
       MockStreamingResult result;
-      StreamingDecoder stream(
+      auto stream = StreamingDecoder::CreateAsyncStreamingDecoder(
           std::make_unique<MockStreamingProcessor>(&result));
-      stream.OnBytesReceived(data.SubVector(0, split));
-      stream.OnBytesReceived(data.SubVector(split, data.length()));
-      stream.Finish();
+      stream->OnBytesReceived(data.SubVector(0, split));
+      stream->OnBytesReceived(data.SubVector(split, data.length()));
+      stream->Finish();
       EXPECT_FALSE(result.ok());
       EXPECT_EQ(error_offset, result.error.offset());
       EXPECT_EQ(message, result.error.message());
@@ -129,8 +129,9 @@ class WasmStreamingDecoderTest : public ::testing::Test {
 
 TEST_F(WasmStreamingDecoderTest, EmptyStream) {
   MockStreamingResult result;
-  StreamingDecoder stream(std::make_unique<MockStreamingProcessor>(&result));
-  stream.Finish();
+  auto stream = StreamingDecoder::CreateAsyncStreamingDecoder(
+      std::make_unique<MockStreamingProcessor>(&result));
+  stream->Finish();
   EXPECT_FALSE(result.ok());
 }
 
@@ -138,9 +139,10 @@ TEST_F(WasmStreamingDecoderTest, IncompleteModuleHeader) {
   const uint8_t data[] = {U32_LE(kWasmMagic), U32_LE(kWasmVersion)};
   {
     MockStreamingResult result;
-    StreamingDecoder stream(std::make_unique<MockStreamingProcessor>(&result));
-    stream.OnBytesReceived(VectorOf(data, 1));
-    stream.Finish();
+    auto stream = StreamingDecoder::CreateAsyncStreamingDecoder(
+        std::make_unique<MockStreamingProcessor>(&result));
+    stream->OnBytesReceived(VectorOf(data, 1));
+    stream->Finish();
     EXPECT_FALSE(result.ok());
   }
   for (uint32_t length = 1; length < sizeof(data); ++length) {