Added framework for generating/consuming benchmarking data sets.

This takes the code that was sitting in benchmarks/
already and makes it easier for language-specific
benchmarks to consume.  Future PRs will enhance this
so that the language-specific benchmarks can report
metrics back that will be tracked over time in PerfKit.

1 files changed, 75 insertions, 0 deletions

diff --git a/benchmarks/Makefile.am b/benchmarks/Makefile.am
new file mode 100644
index 00000000..79581ee9
--- /dev/null
+++ b/benchmarks/Makefile.am
@@ -0,0 +1,75 @@
+
+benchmarks_protoc_inputs =                                     \
+  benchmarks.proto                                             \
+  benchmark_messages_proto3.proto
+
+benchmarks_protoc_inputs_proto2 =                              \
+  benchmark_messages_proto2.proto
+
+benchmarks_protoc_outputs =                                    \
+  benchmarks.pb.cc                                             \
+  benchmarks.pb.h                                              \
+  benchmark_messages_proto3.pb.cc                              \
+  benchmark_messages_proto3.pb.h
+
+benchmarks_protoc_outputs_proto2 =                             \
+  benchmark_messages_proto2.pb.cc                              \
+  benchmark_messages_proto2.pb.h
+
+bin_PROGRAMS = generate-datasets
+
+generate_datasets_LDADD = $(top_srcdir)/src/libprotobuf.la
+generate_datasets_SOURCES = generate_datasets.cc
+generate_datasets_CPPFLAGS = -I$(top_srcdir)/src -I$(srcdir)
+nodist_generate_datasets_SOURCES =                             \
+  google_message1.h                                            \
+  google_message2.h                                            \
+  $(benchmarks_protoc_outputs)                                 \
+  $(benchmarks_protoc_outputs_proto2)
+
+# Explicit deps beacuse BUILT_SOURCES are only done before a "make all/check"
+# so a direct "make test_cpp" could fail if parallel enough.
+generate_datasets-generate_datasets.$(OBJEXT): benchmarks.pb.h google_message1.h google_message2.h
+
+$(benchmarks_protoc_outputs): protoc_middleman
+$(benchmarks_protoc_outputs_proto2): protoc_middleman2
+
+google_message1.h: google_message1.dat
+	xxd -i $< $@
+
+google_message2.h: google_message2.dat
+	xxd -i $< $@
+
+CLEANFILES =                                                   \
+  $(benchmarks_protoc_outputs)                                 \
+  $(benchmarks_protoc_outputs_proto2)                          \
+  google_message1.h                                            \
+  google_message2.h                                            \
+  protoc_middleman                                             \
+  protoc_middleman2                                            \
+  dataset.*
+
+if USE_EXTERNAL_PROTOC
+
+protoc_middleman: $(benchmarks_protoc_inputs)
+	$(PROTOC) -I$(srcdir) -I$(top_srcdir) --cpp_out=. $(benchmarks_protoc_inputs)
+	touch protoc_middleman
+
+protoc_middleman2: $(benchmarks_protoc_inputs_proto2)
+	$(PROTOC) -I$(srcdir) -I$(top_srcdir) --cpp_out=. $(benchmarks_protoc_inputs_proto2)
+	touch protoc_middleman2
+
+else
+
+# We have to cd to $(srcdir) before executing protoc because $(protoc_inputs) is
+# relative to srcdir, which may not be the same as the current directory when
+# building out-of-tree.
+protoc_middleman: $(top_srcdir)/src/protoc$(EXEEXT) $(benchmarks_protoc_inputs) $(well_known_type_protoc_inputs)
+	oldpwd=`pwd` && ( cd $(srcdir) && $$oldpwd/../src/protoc$(EXEEXT) -I. -I$(top_srcdir)/src --cpp_out=$$oldpwd $(benchmarks_protoc_inputs) )
+	touch protoc_middleman
+
+protoc_middleman2: $(top_srcdir)/src/protoc$(EXEEXT) $(benchmarks_protoc_inputs_proto2) $(well_known_type_protoc_inputs)
+	oldpwd=`pwd` && ( cd $(srcdir) && $$oldpwd/../src/protoc$(EXEEXT) -I. -I$(top_srcdir)/src --cpp_out=$$oldpwd $(benchmarks_protoc_inputs_proto2) )
+	touch protoc_middleman
+
+endif