[SymmMem] find_path does not search /usr/local/lib#157765
[SymmMem] find_path does not search /usr/local/lib#157765pytorchbot wants to merge 1 commit intocherry-pick-157513-by-pytorch_bot_bot_from
Conversation
This PR uses `find_library` to replace `find_path`. It also searches for NVSHMEM host lib and device lib separately. Tested against system install location: /usr/local/lib and /usr/local/include. Pull Request resolved: #157695 Approved by: https://github.com/Skylion007 ghstack dependencies: #157513 (cherry picked from commit c558907)
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/157765
Note: Links to docs will display an error until the docs builds have been completed. ❌ 2 New Failures, 1 Unrelated FailureAs of commit 57e29b8 with merge base 3a7ff82 ( NEW FAILURES - The following jobs have failed:
UNSTABLE - The following job is marked as unstable, possibly due to flakiness on trunk:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
We still need to merge the build_cuda changes, but yeah. |
|
@Skylion007 Yeah, I cherry-picked your PR onto 2.8 release branch: #156745 (comment) |
|
This stack has no effect until this merged: #157774 and vice versa |
| find_path(NVSHMEM_INCLUDE_DIR | ||
| NAMES nvshmem.h | ||
| PATHS $ENV{NVSHMEM_HOME}/include ${NVSHMEM_PY_DIR}/include | ||
| HINTS $ENV{NVSHMEM_HOME}/include ${NVSHMEM_PY_DIR}/include |
There was a problem hiding this comment.
I just noticed the default /usr/local/include is excluded. I hope it will be able to find it.
There was a problem hiding this comment.
@Skylion007 I cleared all the hints on my machine, and the build log show me the following:
-- NVSHMEM_HOST_LIB: '/usr/local/lib/libnvshmem_host.so'
-- NVSHMEM_DEVICE_LIB: '/usr/local/lib/libnvshmem_device.a'
-- NVSHMEM_INCLUDE_DIR: '/usr/local/include'
-- NVSHMEM found, building with NVSHMEM support
so I guess it is able to find the header and libs?
I would like to download a 2.8 build to verify this eventually tho.
@atalman can you please let us know after all these 3 PRs are landed in 2.8 release branch?
#157774
#157755
#157765
Then I would like install the wheel and do the verification.
|
Looks like this PR hasn't been updated in a while so we're going to go ahead and mark this as |
Stack from ghstack (oldest at bottom):
This PR uses
find_libraryto replacefind_path.It also searches for NVSHMEM host lib and device lib separately.
Tested against system install location: /usr/local/lib and /usr/local/include.
cc @H-Huang @awgu @wanchaol @fegin @fduwjj @wz337 @wconstab @d4l3k @pragupta @ezyang @msaroufim