two changes: 1) a python script to combine multiple training corpus into one; 2) push Colibrow's PR (https://github.com/google/ml-compiler-opt/pull/8). Thanks, Colibrow!
diff --git a/compiler_opt/tools/combine_training_corpus.py b/compiler_opt/tools/combine_training_corpus.py
new file mode 100644
index 0000000..e3bace5
--- /dev/null
+++ b/compiler_opt/tools/combine_training_corpus.py
@@ -0,0 +1,74 @@
+# coding=utf-8
+# Copyright 2020 Google LLC
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+r"""Combine multiple training corpus into a single training corpus.
+
+Usage: we'd like to combine training corpus corpus1 and corpus2 into
+combinedcorpus; we first structure the files as follows:
+
+combinedcorpus
+combinedcorpus/corpus1
+combinedcorpus/corpus2
+
+Running this script with
+
+python3 compiler_opt/tools/combine_training_corpus.py \
+ --root_dir=$PATH_TO_combinedcorpus
+
+generates combinedcorpus/module_path file. In this way corpus1 and
+corpus2 are combined into combinedcorpus.
+"""
+
+import os
+
+from absl import app
+from absl import flags
+from absl import logging
+
+import tensorflow as tf
+
+flags.DEFINE_string('root_dir', '', 'root dir of module paths to combine.')
+
+FLAGS = flags.FLAGS
+
+_FILE_NAME = 'module_paths'
+
+
+def main(argv):
+ if len(argv) > 1:
+ raise app.UsageError('Too many command-line arguments.')
+
+ module_names = []
+
+ for sub_dir in tf.io.gfile.listdir(FLAGS.root_dir):
+ path = os.path.join(FLAGS.root_dir, sub_dir, _FILE_NAME)
+
+ logging.info('processing %s', path)
+
+ if not tf.io.gfile.exists(path):
+ logging.error('%s does not exist.', path)
+ continue
+
+ with tf.io.gfile.GFile(path, 'r') as f:
+ module_names.extend(
+ [os.path.join(sub_dir, name.rstrip('\n')) for name in f])
+
+ with tf.io.gfile.GFile(os.path.join(FLAGS.root_dir, _FILE_NAME), 'w') as f:
+ for module in module_names:
+ f.write(module + '\n')
+
+
+if __name__ == '__main__':
+ app.run(main)
diff --git a/docs/demo/demo.md b/docs/demo/demo.md
index 99e842e..db95183 100644
--- a/docs/demo/demo.md
+++ b/docs/demo/demo.md
@@ -145,7 +145,9 @@
Fuchsia build conveniently generates a size report. Let's copy it for reference.
-**Note** The `--args=clang_embed_bitcode=true` option above adds the compilation
+**Note**
+The `clang_prefix` is the absolute path of $LLVM_INSTALLDIR/bin(replace it by
+yours). The `--args=clang_embed_bitcode=true` option above adds the compilation
flag `-Xclang=-fembed-bitcode=all`. This can be seen in the compilation database.
The effect of this is that the object files have the llvm bytecode produced by
clang, before the optimization passes, and the clang command line, captured in
