Only tested in 18.04. In principle it should work on higher versions too, but no guarantees
Currently the build system does not automatically pull the LLVM dependencies so we need to manually install them
# Download precompiled LLVM 10 binaries
wget -c https://github.com/llvm/llvm-project/releases/download/llvmorg-10.0.0/clang+llvm-10.0.0-x86_64-linux-gnu-ubuntu-18.04.tar.xz
tar xf clang+llvm-10.0.0-x86_64-linux-gnu-ubuntu-18.04.tar.xz -C /my/llvm/install/dir --strip-components=1
# (Optional) Create `$LLVM_HOME` env var pointing to the llvm installation
echo -e export LLVM_HOME=$(realpath /my/llvm/install/dir) >> ~/.bashrc
source ~/.bashrc(we should really have CMake configure this but for now its ok)
sudo apt update
sudo apt install libcxxtools-dev
- dot
- Doxygen
sudo apt install graphviz
sudo apt install doxygen
cd kerma
mkdir build
cd build && cmake ..
make -j4
If all went well there should be a directory out under build with the following structure:
out/
├── bin/
├── docs/
├── include/
├── lib/
└── rt/
This is the directory to use as KERMA_HOME in KermaView, meaning that in KermaView you should set KERMA_HOME=path/to/kerma/build/out in the .env file.
+---------------+
| Device IR |
|Instrumentation|
+----+----+-----+
^ |
| |
| |
| |
clang + v llc ptxas fatbinary
+---------------> device.bc +-------> device.ptx +-------> device.o +---------> device.fatbin
| +
+ |
program.cu |
+ |
| clang clang clang |
+---------------> host.cu.inc +-------> host.bc +-------> host.o +---------------->+ clang (link)
(host src w. + ^ |
Cuda wrappers) | | |
| | v
| | program.exe
v |
+----+---+------+
| Host IR |
|Instrumentation|
+----+---+------+
TODO: Add flags for each step
-
Step 0
clang++ -std=c++11 -c -S -g -O0 -emit-llvm -fno-discard-value-names \ -Xclang -disable-O0-optnone \ <file> -cuda-gpu-arch=<arch> -
Step 1
opt -S -mem2reg -instnamer <file>.ll > <file>.1.ll -
Step 2
opt -S -load <KermaTransformPlugin> --kerma-md --kerma-mi [<kerma-md options>] [<kerma-mi-options>] <file>.1.ll > <file>.2.ll