./magic.py config -f build-config.toml
;- Get
PyTorch c++
, check from PyTorch c++; cmake -DCMAKE_PREFIX_PATH=deps=/path/to/pytorch/share/cmake/Torch/ .
;- Get
Intel Math Kernel Library
, check from MKL. Using the default steps for install. Usually, MKL is installed at/opt/intel/mkl
. If that's not the case, please modify themkl = ...
entry inbuild-config.toml
and rerun./magic.py config -f build-config.toml; cmake -DCMAKE_PREFIX_PATH=deps=/path/to/pytorch/share/cmake/Torch/ .
; - Build
boost
:make boost
(May take sometime depends on network conditional and computation power); - Build
jemalloc
:cd ./deps/jemalloc; autoconf; cd ../../; make jemalloc
; cmake .
Most unittests are in separate modules (e.g., x_ml). For building and running unittests in each module, you can directly goto each module.
For instance, the following steps will run unittests in xkv_core
,
which implements the core storage of XStore.
cd xkv_core
;./test.sh
make -j12