build: speedup compilation of mksnapshot output

Incremental compilation of Node.js is slow. Currently on a powerful
Linux machine, it takes about 5.6 seconds to compile
`gen/node_snapshot.cc` with g++.

As in the previous PR which dealt with `node_js2c`, we add a new flag
`--use-string-literals` to `node_mksnapshot`. When this flag is set, we
emit string literals instead of array literals for the snapshot blob and
for the code cache, i.e.:

```c++
// old: static const uint8_t X[] = { ... };
static const uint8_t *X = "...";
```

I only enabled the new flag on Linux/macOS, since those are systems that
I have available for testing. On my Linux system with gcc, it speeds up
compilation of this file by 3.7s (5.8s -> 2.1s). On my Mac system with
clang, it speeds up compilation by 1.7s (3.4s -> 1.7s).

Again, the right thing here is probably to generate separate files for
the snapshot blob and for each code cache output, but this is a nice
intermediate speedup.

The thing I'm most unsure about in this PR is how to actually thread the
argument through. I considered adding it to the general argument parser,
but that felt strange, since this flag only makes sense during the build
process. So I kind of hacked it in, which also feels weird. Suggestions
are very welcome.

Refs: nodejs#47984
Refs: nodejs#48160

node.gyp

@@ -650,8 +650,16 @@
                   ],
                   'action': [
                     '<@(_inputs)',
+                    '<@(node_mksnapshot_use_string_literals_flag)',
                     '<@(_outputs)',
                   ],
+                  'conditions': [
+                    ['OS=="linux" or OS=="mac"', {
+                      'variables': {'node_mksnapshot_use_string_literals_flag': ['--use-string-literals']},
+                    }, {
+                      'variables': {'node_mksnapshot_use_string_literals_flag': []},
+                    }],
+                  ],
                 },
               ],
             }],

src/node_snapshot_builder.h

@@ -18,7 +18,8 @@ class NODE_EXTERN_PRIVATE SnapshotBuilder {
  public:
   static ExitCode Generate(std::ostream& out,
                            const std::vector<std::string> args,
-                           const std::vector<std::string> exec_args);
+                           const std::vector<std::string> exec_args,
+                           bool use_string_literals);
 
   // Generate the snapshot into out.
   static ExitCode Generate(SnapshotData* out,

src/node_snapshotable.cc

@@ -742,23 +742,61 @@ static std::string FormatSize(size_t size) {
   return buf;
 }
 
-static void WriteStaticCodeCacheData(std::ostream* ss,
-                                     const builtins::CodeCacheInfo& info) {
+static void WriteDataAsCharString(std::ostream* ss,
+                                  const uint8_t* data,
+                                  size_t length) {
+  for (size_t i = 0; i < length; i++) {
+    const uint8_t ch = data[i];
+    // We can print most printable characters directly. The exceptions are '\'
+    // (escape characters), " (would end the string), and ? (trigraphs). The
+    // latter may be overly conservative: we compile with C++17 which doesn't
+    // support trigraphs.
+    if (ch >= ' ' && ch <= '~' && ch != '\\' && ch != '"' && ch != '?') {
+      *ss << ch;
+    } else {
+      // All other characters are blindly output as octal.
+      const char c0 = '0' + ((ch & 0700) >> 6);
+      const char c1 = '0' + ((ch & 0070) >> 3);
+      const char c2 = '0' + (ch & 7);
+      *ss << "\\" << c0 << c1 << c2;
+    }
+    if (i % 64 == 63) {
+      // Go to a newline every 64 bytes since many text editors have
+      // problems with very long lines.
+      *ss << "\"\n\"";
+    }
+  }
+}
+
+static void WriteStaticCodeCacheDataAsArray(
+    std::ostream* ss, const builtins::CodeCacheInfo& info) {
   *ss << "static const uint8_t " << GetCodeCacheDefName(info.id) << "[] = {\n";
   WriteVector(ss, info.data.data, info.data.length);
-  *ss << "};";
+  *ss << "};\n";
+}
+
+static void WriteStaticCodeCacheDataAsStringLiteral(
+    std::ostream* ss, const builtins::CodeCacheInfo& info) {
+  *ss << "static const uint8_t *" << GetCodeCacheDefName(info.id)
+      << "= reinterpret_cast<const uint8_t *>(\"";
+  WriteDataAsCharString(ss, info.data.data, info.data.length);
+  *ss << "\");\n";
 }
 
-static void WriteCodeCacheInitializer(std::ostream* ss, const std::string& id) {
+static void WriteCodeCacheInitializer(std::ostream* ss,
+                                      const std::string& id,
+                                      size_t size) {
   std::string def_name = GetCodeCacheDefName(id);
   *ss << "    { \"" << id << "\",\n";
   *ss << "      {" << def_name << ",\n";
-  *ss << "       arraysize(" << def_name << "),\n";
+  *ss << "       " << size << ",\n";
   *ss << "      }\n";
   *ss << "    },\n";
 }
 
-void FormatBlob(std::ostream& ss, const SnapshotData* data) {
+void FormatBlob(std::ostream& ss,
+                const SnapshotData* data,
+                const bool use_string_literals) {
   ss << R"(#include <cstddef>
 #include "env.h"
 #include "node_snapshot_builder.h"
@@ -767,21 +805,32 @@ void FormatBlob(std::ostream& ss, const SnapshotData* data) {
 // This file is generated by tools/snapshot. Do not edit.
 
 namespace node {
-
-static const char v8_snapshot_blob_data[] = {
 )";
-  WriteVector(&ss,
-              data->v8_snapshot_blob_data.data,
-              data->v8_snapshot_blob_data.raw_size);
-  ss << R"(};
-
-static const int v8_snapshot_blob_size = )"
+  if (use_string_literals) {
+    ss << R"(static const char *v8_snapshot_blob_data = ")";
+    WriteDataAsCharString(
+        &ss,
+        reinterpret_cast<const uint8_t*>(data->v8_snapshot_blob_data.data),
+        data->v8_snapshot_blob_data.raw_size);
+    ss << R"(";)";
+  } else {
+    ss << R"(static const char v8_snapshot_blob_data[] = {)";
+    WriteVector(&ss,
+                data->v8_snapshot_blob_data.data,
+                data->v8_snapshot_blob_data.raw_size);
+    ss << R"(};)";
+  }
+  ss << R"(static const int v8_snapshot_blob_size = )"
      << data->v8_snapshot_blob_data.raw_size << ";";
 
-  // Windows can't deal with too many large vector initializers.
-  // Store the data into static arrays first.
   for (const auto& item : data->code_cache) {
-    WriteStaticCodeCacheData(&ss, item);
+    if (use_string_literals) {
+      WriteStaticCodeCacheDataAsStringLiteral(&ss, item);
+    } else {
+      // Windows can't deal with too many large vector initializers.
+      // Store the data into static arrays first.
+      WriteStaticCodeCacheDataAsArray(&ss, item);
+    }
   }
 
   ss << R"(const SnapshotData snapshot_data {
@@ -808,7 +857,7 @@ static const int v8_snapshot_blob_size = )"
   // -- code_cache begins --
   {)";
   for (const auto& item : data->code_cache) {
-    WriteCodeCacheInitializer(&ss, item.id);
+    WriteCodeCacheInitializer(&ss, item.id, item.data.length);
   }
   ss << R"(
   }
@@ -1022,13 +1071,14 @@ ExitCode SnapshotBuilder::CreateSnapshot(SnapshotData* out,
 
 ExitCode SnapshotBuilder::Generate(std::ostream& out,
                                    const std::vector<std::string> args,
-                                   const std::vector<std::string> exec_args) {
+                                   const std::vector<std::string> exec_args,
+                                   const bool use_string_literals) {
   SnapshotData data;
   ExitCode exit_code = Generate(&data, args, exec_args);
   if (exit_code != ExitCode::kNoFailure) {
     return exit_code;
   }
-  FormatBlob(out, &data);
+  FormatBlob(out, &data, use_string_literals);
   return exit_code;
 }
 

tools/snapshot/node_mksnapshot.cc

@@ -58,15 +58,23 @@ int main(int argc, char* argv[]) {
 
 int BuildSnapshot(int argc, char* argv[]) {
   if (argc < 2) {
-    std::cerr << "Usage: " << argv[0] << " <path/to/output.cc>\n";
+    std::cerr << "Usage: " << argv[0]
+              << " [--use-string-literals] <path/to/output.cc>\n";
     std::cerr << "       " << argv[0] << " --build-snapshot "
               << "<path/to/script.js> <path/to/output.cc>\n";
     return 1;
   }
 
+  std::vector<std::string> args{argv, argv + argc};
+  size_t size_before_remove = args.size();
+  args.erase(std::remove(args.begin(), args.end(), "--use-string-literals"),
+             args.end());
+  // If the size of args changed, we must have removed a
+  // "--use-string-literals".
+  const bool use_string_literals = args.size() != size_before_remove;
+
   std::unique_ptr<node::InitializationResult> result =
-      node::InitializeOncePerProcess(
-          std::vector<std::string>(argv, argv + argc));
+      node::InitializeOncePerProcess(args);
 
   CHECK(!result->early_return());
   CHECK_EQ(result->exit_code(), 0);
@@ -87,7 +95,7 @@ int BuildSnapshot(int argc, char* argv[]) {
   node::ExitCode exit_code = node::ExitCode::kNoFailure;
   {
     exit_code = node::SnapshotBuilder::Generate(
-        out, result->args(), result->exec_args());
+        out, result->args(), result->exec_args(), use_string_literals);
     if (exit_code == node::ExitCode::kNoFailure) {
       if (!out) {
         std::cerr << "Failed to write " << out_path << "\n";