Build projects in wamr-test-suites with multiple cores #811
Conversation
Merge bytecodealliance:main into wenyongh:main
Apply clang-format for core/shared and product-mini files (#785)
Merge bytecodealliance:main into wenyongh:main
Enhance spec test workflow to support x86_32 target (#787)
Merge bytecodealliance:main into wenyongh:main
Merge wasm-micro-runtime:main into wenyongh:main
Apply clang-format for more source files (#795)
And fix host_tool help show issue.
Disable source debugging by default (#804)
Merge bytecodealliance:main into wenyongh:main
Merge bytecodealliance:main into wenyongh:main
| # keep going and do not care if it is success or not | ||
| make -ki clean | true | ||
| cmake ${compile_flag} ${WORK_DIR}/../../unit && make | ||
| cmake ${compile_flag} ${WORK_DIR}/../../unit && make -j 4 |
There was a problem hiding this comment.
had better use $(nproc)
| cmake ${compile_flag} ${WORK_DIR}/../../unit && make -j 4 | |
| cmake ${compile_flag} ${WORK_DIR}/../../unit && make -j $(nproc) |
There was a problem hiding this comment.
nproc may not be installed. 4 might be a safe option
There was a problem hiding this comment.
Yes, so let's use -j 4.
Another place to reduce the workloads compilation time is building iwasm in compilation_on_android_ubuntu_macos.yml, currently we get llvm libraries
before building iwasm for each mode, in fact we only need to get llvm libraries
for the mode jit and lazyjit. For fast/classic interpreter mode and aot mode,
we don't need to get the llvm libraires, is it easy to add condition check
in the scripts?
- name: Get LLVM libraries
id: cache_llvm
if: ${{ matrix.light == 'green' }}
uses: actions/cache@v2
with:
path: |
./core/deps/llvm/build/bin
./core/deps/llvm/build/include
./core/deps/llvm/build/lib
./core/deps/llvm/build/libexec
./core/deps/llvm/build/share
key: ${{ matrix.os }}-${{ env.LLVM_CACHE_SUFFIX }}e.g. modify this line: if: ${{ matrix.light == 'green' }} to something like:
if: ${{ matrix.light == 'green'
&& (make_options_run_mode == $JIT_BUILD_OPTIONS
|| make_options_run_mode == $LAZY_JIT_BUILD_OPTIONS)
}}There was a problem hiding this comment.
Time consumption is defined by the longest job in parallel mode. In our workflow, the followings will not start until all five build_iwasm jobs are finished.
There was a problem hiding this comment.
There should be total 243 build_iwasm jobs but not five in compilation_on_android_ubuntu_macos.yml:
https://github.com/bytecodealliance/wasm-micro-runtime/actions/runs/1402357515
It is the longest phase in all the workflows.
Currently each build_iwasm job will get LLVM libraries firstly which costs some time, but there
are a lot of interpreter/aot mode jobs which don't need llvm libraries, if we can skip this step for them,
we should be able to reduce the total time of 243 build_iwasm jobs, so as to reduce the total time of
all workflows.
| && cd build \ | ||
| && cmake .. \ | ||
| && make | ||
| && make -j 4 |
|
LGTM |
Build projects in wamr-test-suites with multiple cores (#811)
…nce#811) Build several projects in wamr-test-suites with multiple cores to reduce the compilation time.
3 participants
Build several projects in wamr-test-suites with multiple cores to reduce
the compilation time.