-
Notifications
You must be signed in to change notification settings - Fork 3
lipeng28/Gklee
Folders and files
Name | Name | Last commit message | Last commit date | |
---|---|---|---|---|
Repository files navigation
GKLEE is a symbolic analyser and test generator tailored for CUDA C++ programs, it supports almost all of CUDA intrinsic functions in kernel code and majority of CUDA runtimes in host code. For more details, please visit GKLEE's homepage: www.cs.utah.edu/fv/GKLEE ======================= Build GKLEE ToolChain =========================== 1. Before you install the GKLEE toolchain, please ensure "flex" and "bison" are both installed in your Linux system. 2. In the 'Gklee' dir, and run the script 'BuildGklee'. You will be asked to build Clang/LLVM and GKLEE with the "Release" or "Debug" mode. This script is able to download and build all necessary parts automatically. ======================= Set Environment Variables ========================= 1. After you finish the instalation, please customize the environment variables in the following two ways: -- Add the following lines to your /home/[username]/.bashrc file -- Create an environment file including the following lines export KLEE_HOME_DIR=/Path/To/Gklee export PATH=$KLEE_HOME_DIR/bin:$KLEE_HOME_DIR/Gklee/Release+Asserts/bin:$KLEE_HOME_DIR/llvm/Release+Asserts/bin:$PATH NOTE: The above environment variables are present under the "Release" building mode, you should adjust "Release" to "Debug" accordingly if you build GKLEE under the "Debug" mode. ======================= Run GKLEE ========================= After building GKLEE and setting environment variables properly, you could run GKLEE. 1. Compile your CUDA programs into LLVM bytecode with the command: gklee-nvcc xxx.cu [-o executable] [other options used in nvcc ...] NOTE: if you do not specify a executable name, gklee-nvcc will use the prefix of the CUDA program as the default name 2. Run your program with GKLEE command gklee executable [Normal Mode] gklee --symbolic-config executable [Parametric Flow Mode] 3. The test cases will be generated in the same dir where you execute gklee, and appear as 'klee-out-[number]' and 'klee-last'.
About
GKLEE is a symbolic analyser and test generator tailored for CUDA C++ programs
Resources
Stars
Watchers
Forks
Packages 0
No packages published