Query on Non-Open-Source Components in Self-Compiled SCIP (Replacing pyscipopt Without Parallel Support)

[cjh668]

I originally used pyscipopt, but it does not support parallel computing. Therefore, I have tried compiling SCIP myself. However, I have an additional requirement: the program must be fully open-source—since I may need to develop my own software later, I want to know if the current compilation configuration will install any non-open-source components.

I performed the relevant operations in Developer PowerShell for VS 2022

cmake .. -G "Visual Studio 17 2022" -A x64 ^
-DCMAKE_BUILD_TYPE=Release ^
-DENABLE_PARALLEL=ON ^
-DCMAKE_INSTALL_PREFIX="D:\SCIP-9.2.3-install" ^
-DDISABLE_ZIMPL=ON ^
-DDISABLE_GCG=ON ^
-DDISABLE_PAPILO=ON ^
-DDISABLE_UG=ON ^
-DREADLINE=Off ^
-DAUTOBUILD=ON

[mmghannam]

If you use AUTOBUILD=On we cannot know which software is linked to your SCIP installation without looking at the log. You could check these in the log and check their licenses. Looking at the log you will also see that some of the parameters you passed are not supported by the cmake system, for example to turn off PaPiLO you should pass -DPAPILO=Off.
Anyhow, I think you can get a build that makes SCIPconcurrentSolve work with fully open-source software but I'm not very sure. Still as I said in your other issue, I wouldn't expect a big (if any) speedup out of the box.

[cjh668]

Got it, thank you for your reply. cjh ***@***.***

[cjh668]

cd D:\tool\scip\scip-9.2.3 mkdir -Force build cd build cmake .. -G "Visual Studio 17 2022" -A x64 `  -DCMAKE_BUILD_TYPE=Release `  -DENABLE_PARALLEL=ON `  -DCMAKE_INSTALL_PREFIX="D:\tool\scip\scip-9.2.3-install" `  -DZIMPL=Off `  -DGCG=Off `  -DPAPILO=Off `  -DUG=Off `  -DREADLINE=Off `  -DAUTOBUILD=ON I have tried to compile using the above code in Developer PowerShell for VS 2022, but errors keep occurring. Where could the problem be? I have also tried downloading SoPlex separately, yet I can't seem to find it as the picture show. CMake Error at _deps/soplex-build/src/cmake_install.cmake:213 (file): file INSTALL cannot find "D:/tool/scip/scip-9.2.3/build/_deps/soplex-build/bin/Release/soplex.exe": File exists.   Call Stack (most recent call first):   _deps/soplex-build/cmake_install.cmake:37 (include)   Call Stack (most recent call first): cmake/Dependencies.cmake:50 (execute_process_handle_output) cmake/Dependencies.cmake:87 (build_package) CMakeLists.txt:420 (find_or_download_package)   -- Configuring incomplete, errors occurred! cjh ***@***.*** 原始邮件 发件人：Mohammed Ghannam ***@***.***> 发件时间：2025年10月8日 21:13 收件人：scipopt/PySCIPOpt ***@***.***> 抄送：cjh668 ***@***.***>, Author ***@***.***> 主题：Re: [scipopt/PySCIPOpt] Query on Non-Open-Source Components inSelf-Compiled SCIP (Replacing pyscipopt Without Parallel Support) (Issue#1077) mmghannam left a comment (scipopt/PySCIPOpt#1077) If you use AUTOBUILD=On we cannot know which software is linked to your SCIP installation without looking at the log. You could check these in the log and check their licenses. Looking at the log you will also see that some of the parameters you passed are not supported by the cmake system, for example to turn off PaPiLO you should pass -DPAPILO=Off. Anyhow, I think you can get a build that makes SCIPconcurrentSolve work with fully open-source software but I'm not very sure. Still as I said in your other issue, I wouldn't expect a big (if any) speedup out of the box. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.

[mmghannam]

How did you compile and install SoPlex?

[cjh668]

I tried to handle the SoPlex dependency through SCIP's automatic build feature (setting -DAUTOBUILD=ON). However, from the error message, there was an issue during the installation phase where soplex.exe could not be found. This is the sequence of steps I followed, but an error occurred at this step. cd D:\tool\scip\scip-9.2.3   mkdir -Force build  cd build   cmake .. -G "Visual Studio 17 2022" -A x64 `   -DCMAKE_BUILD_TYPE=Release `   -DENABLE_PARALLEL=ON `   -DCMAKE_INSTALL_PREFIX="D:\tool\scip\scip-9.2.3-install" `   -DZIMPL=Off `   -DGCG=Off `   -DPAPILO=Off `   -DUG=Off `   -DREADLINE=Off `   -DAUTOBUILD=ON After that, I also tried to download SoPlex manually from the official website. But when I attempted to obtain soplex-6.0.3.tgz, the website displayed a "File not found!" error. I'm not sure if it's an improper version selection or if there's a problem with the specific version link on the official website. cjh ***@***.***

[mmghannam]

You can also get the latest soplex version from Github https://github.com/scipopt/soplex/releases/tag/release-715

Try to compile that, and set -DSOPLEX_DIR for the cmake configuration of SCIP to the directory you installed soplex in.

[cjh668]

I've tried deploying SCIP locally, but as you mentioned, the solving performance hasn't changed significantly. During program execution, I also observed that the CPU usage remained around 20% throughout; in contrast, when using other optimization solvers, the CPU utilization can sometimes reach as high as 100%. I'm wondering if this is an issue with my usage, or if SCIP's performance is inherently limited to this level? From my understanding, with parallel computing enabled, even if it doesn't help at the root node stage, it should at least speed up the computation during the branching phase. Additionally, is there any relevant log output that can confirm whether SCIP has successfully invoked multi-threading? cjh ***@***.***