[ci] Add build script for win #3410
Conversation
|
✔️ Deploy Preview for jovial-fermat-aa59dc ready! 🔨 Explore the source changes: 1fbd7d6 🔍 Inspect the deploy log: https://app.netlify.com/sites/jovial-fermat-aa59dc/deploys/6188f1d07ec7490008cf8698 😎 Browse the preview: https://deploy-preview-3410--jovial-fermat-aa59dc.netlify.app |
frostming
force-pushed
the
ci/win-build-script
branch
from
ee16cea
to
1d0cb0a
Compare
| New-Item -ItemType Directory -Path $build | ||
| } | ||
| Push-Location $build | ||
| WriteInfo("Download and extract LLVM") |
There was a problem hiding this comment.
Note that the LLVM/clang/Vulkan libs usually are usually downloaded once and placed somewhere outside the repo dir/build dir, (i.e., a more stable place, e.g. C:\ThirdParty\Libs). I'd suggest that we add one additional flag to hold all these required dependent libs, WDYT?
In addition, let's make the default behavior not to download these libs?
There was a problem hiding this comment.
the lib location can be specified with -build option, and if the library exists, won't download again.
So -build C:\ThirdParty\Libs can do the trick.
There was a problem hiding this comment.
I see, then let's give it a more obvious name, e.g. libsdir, depsdir, ...
There was a problem hiding this comment.
The name is changed to libdir
There was a problem hiding this comment.
Huh, I guess the naming convention here is $libsDir? Also, the default value should be empty (or at least not _build)
There was a problem hiding this comment.
I've updated the name and default to the current location (.) since empty isn't a valid path parameter
This would be a great step forward! We used to think about using conda to automate/simplify the process. Any suggestions on this :-) ? |
Related issue = #3391 #2715
a. Usage:
win_build.ps1 [[-libdir] <string>] [-clone] [-installVulkan] [-develop] [-install]b. Support rerun, won't download again if already exists
c. build and install python library in a virtual environment to avoid contaminating the system site-packages
d. support install in develop mode(with
-developflag), likepython setup.py develope. CMake args and environments are populated automatically, users don't bother to set them.
I've tested on Windows 10, CPU build, both build and install are successful. If you have a Windows PC, please help to test the script.
Other thoughts
Now setting up a taichi development environment on Windows is as easy as downloading a script from the internet and run it. We may also do the same for linux and mac. And development installation doc can be updated accordingly.