Skip to content

Latest commit

 

History

History
84 lines (65 loc) · 2.61 KB

README.cmake.md

File metadata and controls

84 lines (65 loc) · 2.61 KB
Raw

Building/Using Gdev with CMake

Follow the instruction below to use Gdev with OS runtime support. You may be required to install additional software packages depending on your environment. $(TOPDIR) represents your top working directory.

1. Download

Check out the Gdev and envytools source code via git.

cd $(TOPDIR)
git clone git://github.com/shinpei0208/gdev.git  # OR git://github.com/CPFL/gdev.git
git clone git://github.com/envytools/envytools.git

2. envytools

envytools is a rich set of open-source tools to compile or decompile NVIDIA GPU program code, firmware code, macro code, and so on. It is also used to generate C header files with GPU command definitions. In addition, envytools document the NVIDIA GPU architecture details, while are not disclosed to the public so far. If you are interested in GPU system software development, this is what you should read! Please follow the instruction below to install envytools.

cd $(TOPDIR)/envytools
mkdir build
cd build
cmake .. # may require some packages on your distro
make
sudo make install # will install tools to /usr/local/{bin,lib}

3. Building Gdev

Gdev provides the build instruction with CMake. Leveraging CMake, you can build/install Gdev and CUDA runtime. The following options are supported.

option description values
driver select GPU driver. must set driver name in the user-space mode nouveau,pscnv,nvrm,barra
user user mode (default off) ON/OFF
runtime enable CUDA runtime API (default on) ON/OFF
usched enable user mode scheduler (default off) ON/OFF
use_as use assembler (default off) ON/OFF

For example, you can pass the above options to CMake like below.

cmake -H. -Brelease \
    -Ddriver=nouveau \
    -Duser=ON \
    -Druntime=ON \
    -Dusched=OFF \
    -Duse_as=OFF \
    -DCMAKE_BUILD_TYPE=Release
make -C release
make -C release install

This will mainly produce

  1. user-space mode Gdev libgdev.so and
  2. a limited set of CUDA Driver/Runtime API libucuda.so

make builds them with the specified options and they are installed under the directory /usr/local/gdev.

CAUTION: Especially, the libraries are installed under the directory /usr/local/gdev/lib64. So please remember to add /usr/local/gdev/lib64 to $LD_LIBRARY_PATH.

4. CUDA Driver API test (user-space programs)

You can build / test examples using CUDA Driver API. They are located under the directory $(TOPDIR)/test/cuda/user/. Follow the instruction to build and use the examples.

cd $(TOPDIR)/test/cuda/user/madd
make
./user_test 256 # a[256] + b[256] = c[256]