Merge pull request #16 from falseywinchnet/feature-branch

locality

Author: falseywinchnet

CMakeLists.txt

@@ -87,6 +87,7 @@ if(BFFT_BUILD_EXAMPLES)
     target_link_libraries(benchmark PRIVATE ${CMAKE_DL_LIBS})
   endif()
 
+  bfft_add_cpp_executable(locality_probe examples/locality_probe.cpp)
   bfft_add_cpp_executable(cpp_api_demo examples/cpp_api_demo.cpp)
   bfft_add_c_executable(c_api_demo examples/c_api_demo.c)
 

Makefile

@@ -40,6 +40,7 @@ STATIC_LIB := $(BUILD_DIR)/lib$(LIB_NAME).a
 SHARED_LIB := $(BUILD_DIR)/lib$(LIB_NAME).so
 PC_FILE := $(BUILD_DIR)/$(LIB_NAME).pc
 BENCH := $(BUILD_DIR)/examples/benchmark
+LOCALITY_PROBE := $(BUILD_DIR)/examples/locality_probe
 C_DEMO := $(BUILD_DIR)/examples/c_api_demo
 CPP_DEMO := $(BUILD_DIR)/examples/cpp_api_demo
 CORRECTNESS_TEST := $(BUILD_DIR)/tests/correctness
@@ -73,11 +74,14 @@ $(PC_FILE): pkgconfig/bfft.pc.in | $(BUILD_DIR)
 		-e 's|@PROJECT_VERSION@|$(VERSION)|g' \
 		$< > $@
 
-examples: $(BENCH) $(C_DEMO) $(CPP_DEMO)
+examples: $(BENCH) $(LOCALITY_PROBE) $(C_DEMO) $(CPP_DEMO)
 
 $(BENCH): examples/benchmark.cpp include/bfft/bfft.hpp $(STATIC_LIB) | $(BUILD_DIR)
 	$(CXX) $(CPPFLAGS) $(INCLUDES) $(CXXFLAGS) $< $(STATIC_LIB) $(LDLIBS) $(DL_LIBS) -o $@
 
+$(LOCALITY_PROBE): examples/locality_probe.cpp include/bfft/bfft.hpp $(STATIC_LIB) | $(BUILD_DIR)
+	$(CXX) $(CPPFLAGS) $(INCLUDES) $(CXXFLAGS) $< $(STATIC_LIB) $(LDLIBS) -o $@
+
 $(C_DEMO): examples/c_api_demo.c include/bfft/bfft.h $(STATIC_LIB) | $(BUILD_DIR)
 	$(CC) $(CPPFLAGS) $(INCLUDES) $(CFLAGS) $< $(STATIC_LIB) $(LDLIBS) -lstdc++ -o $@
 

examples/benchmark_compile_instructions.txt

@@ -55,3 +55,11 @@ Then:
 
 build/examples/benchmark --intel-mkl
 build/examples/benchmark --intel-mkl 1024 1000
+
+Locality probe:
+
+The ordinary make examples target also builds build/examples/locality_probe.
+It compares hot-cache timing with a rotating working set sized by ring_mib.
+
+build/examples/locality_probe
+build/examples/locality_probe 65536 240 64