data-flow-analyzer

Analyze data flow graph for C/C++/Fortran (maybe) programs.

Environment

Please make sure you have gcc, python, cmake, and LLVM==10.0.0

Build app with LLVM

you can use provided Makefile at app_Makefile/ directory, and please modify the target executable name at line 191 you can find the default as follow:

	$(LD) $(OBJFILES) -o Geodynamics_cpu.mpi $(FFTLIB)  $(LIB)

Analysis app and example/test

there are four supported modules: generate CFG/DFG, inter loop analysis and intra loop analysis, thus there are four scripts you can use: cfg.sh, dfg.sh, inter.sh and intra.sh.

you can use all scripts with correct configures of above envs to run and analyze.

To be specific: the LLVM compiler location and the compiled app executable file(GS_EXE) should be changed correctly for inter/intra anlysis, the python env should be checked

Analysis App

CFG

./cfg.sh

and you will find a pdf file named cfg.pdf in this directory

DFG

./dfg.sh

and you will find a pdf file named dfg.pdf in this directory

inter analysis

./inter.sh

and you will get the result at the standard output

intra analysis

./intra.sh

and you will get the generated dependence graphs named loop1.png as so on.

Analysis example/test

for cfg, dfg, and inter analysis, you only need to add the source code location without the file postfix, such as:

./cfg.sh ~/data-flow-analyzer/example/test/00test

for intra analysis, you need an extra parameter for the generated png name without the postfix, such as:

./intra.sh ~/data-flow-analyzer/example/test/00test 00test

and you can find 00test.png in this directory