[FBGEMM] Modularize FBGEMM CMake Target Definitions, pt 3b #4524
Conversation
✅ Deploy Preview for pytorch-fbgemm-docs ready!
To edit notification comments on pull requests, go to your Netlify project configuration. |
q10
changed the title
q10
force-pushed
the
bm/fbgemm-arm-build-3
branch
2 times, most recently
from
2d2de86 to
a6184c2
Compare
|
In PT we use the Sanitizer::address and other similar targets for sanitizers, which are more accurate. Therefore should we check the existence of these targets and prefer them? |
q10
force-pushed
the
bm/fbgemm-arm-build-3
branch
2 times, most recently
from
f54d31a to
752294e
Compare
| -Wno-infinite-recursion | ||
| -Wno-sign-compare | ||
| -Wno-gnu-zero-variadic-macro-arguments | ||
| -Wno-unused-parameter |
There was a problem hiding this comment.
We can enable -Wunused-parameter, the code is clean enough to support it.
cmake/modules/CppLibrary.cmake
Outdated
| if(CMAKE_COMPILER_IS_GNUCXX AND (CMAKE_CXX_COMPILER_VERSION VERSION_LESS 8.3.0)) | ||
| # Workaround for https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80947 | ||
| list(APPEND lib_cc_flags | ||
| -Wno-attributes) |
There was a problem hiding this comment.
Should we really support old g++? Is it for META builds? but how can they build C++20 code?
|
Some disabled |
|
One more concern: we use cpuinfo for runtime CPU architecture detection; meanwhile we also use conditional preprocessing inclusion for build time CPU architecture, are the actions conflicting? |
q10
force-pushed
the
bm/fbgemm-arm-build-3
branch
3 times, most recently
from
32d9fc3 to
b83aa34
Compare
q10
changed the title
q10
force-pushed
the
bm/fbgemm-arm-build-3
branch
from
b83aa34 to
47fb27a
Compare
q10
force-pushed
the
bm/fbgemm-arm-build-3
branch
from
47fb27a to
d89366e
Compare
No description provided.