Improve hardware feature detection & consolidate duplicated settings in CMake files #824
+76
−158
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Previously, the CMake configuration was such that it would always build the AVX2, AVX512 & SSE2 Pybind11 modules without testing whether the current system supported those options. The changes here use CMake features to detect whether the architectural features are in fact available, and only attempt to build the appropriate modules. In addition, there is a minor bit of refactoring to group some of the code more logically, and to add some more informational message printouts.
Each of 8 or so `pybind_interface/` subdirectories had `CMakeLists.txt` files that contained the same text for setting certain compilation flags. This commit removes the duplication in favor of putting the settings into the top-level CMake file.
On Ubuntu, one sees warnings like this: ``` lto-wrapper: warning: using serial compilation of 13 LTRANS jobs lto-wrapper: note: see the ‘-flto’ option documentation for more information lto-wrapper: warning: using serial compilation of 15 LTRANS jobs lto-wrapper: note: see the ‘-flto’ option documentation for more information lto-wrapper: warning: using serial compilation of 16 LTRANS jobs lto-wrapper: note: see the ‘-flto’ option documentation for more information lto-wrapper: warning: using serial compilation of 16 LTRANS jobs lto-wrapper: note: see the ‘-flto’ option documentation for more information ``` This seems to be the default behavior if the option `-flto` is not given a value (c.f. https://stackoverflow.com/a/72222512/28972686). Giving it a value of "auto" makes the warning go away and lets the compilation toolchain decide how much parallism it can use.
mhucka
changed the title
This can enable additional optimizations without requiring specific avx/sse/etc. instructions.
mhucka
force-pushed
the
mh-consolidate-cmake-configs
branch
from
1b753ba to
3941a3f
Compare
The common wisdom is wrong, apparently: `-march=native` does not work for AVX on MacOS.
mhucka
force-pushed
the
mh-consolidate-cmake-configs
branch
from
1756d01 to
20c9dd8
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Changes:
Previously, the CMake configuration was such that it would always build the AVX2, AVX512 & SSE2 Pybind11 modules without testing whether the current system supported those options. The changes here use CMake features to detect whether the architectural features are in fact available on the host computer, and only attempt to build the appropriate modules.
Previously, each of 8 or so
pybind_interface/subdirectories hadCMakeLists.txtfiles that contained the same text for setting certain compilation flags. This PR removes the duplication in favor of putting the settings into the top-level CMake file.Finally, there is a minor bit of refactoring to the top level
CMakeLists.txtfile to group some of the code more logically, and to add some more informational message printouts.