Simplify nvtx3 CMake handling, always use nvtx3 #153784
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/153784
Note: Links to docs will display an error until the docs builds have been completed. ❌ 6 New Failures, 1 Unrelated FailureAs of commit a705ea2 with merge base 4613081 ( 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. |
be25ee9 to
81a9132
Compare
a767cf2 to
b26bbcc
Compare
|
@pytorchbot label "topic: not user facing" |
22de8de to
2b21267
Compare
| @@ -127,36 +124,12 @@ endif() | ||
|
|
||
| # ---[ CUDA libraries wrapper | ||
|
|
||
| # find lbnvrtc.so |
There was a problem hiding this comment.
Would it be better to separate nvtx and nvrtc changes?
| @@ -961,20 +961,15 @@ if(USE_CUDNN) | ||
| endif() | ||
|
|
||
| # ---[ nvtx | ||
| if(USE_SYSTEM_NVTX) |
There was a problem hiding this comment.
Although this might not be a common use case since we have shipped nvtx header files, deleting this would eliminate the possibility of using system nvtx?
Would it always find the nvtx shipped with the src code? Or how can it find system nvtx?
| @@ -107,7 +107,6 @@ function(caffe2_print_configuration_summary) | ||
| get_target_property(__tmp torch::cudss INTERFACE_LINK_LIBRARIES) | ||
| message(STATUS " cuDSS library : ${__tmp}") | ||
| endif() | ||
| message(STATUS " nvrtc : ${CUDA_nvrtc_LIBRARY}") |
There was a problem hiding this comment.
Where would nvrtc info be printed? Again, this PR title seems to be specific to nvtx, suggest nvrtc in a separate PR.
There was a problem hiding this comment.
Such changes need to be tested with downstream tasks, is there a way to test this before merge? e.g. build with torchvision, torchaudio, https://github.com/NVIDIA/Fuser etc.
Fixes #ISSUE_NUMBER