Fix too big to optimize in test, actually use O0 when aot_inductor.compile_wrapper_with_O0 is set #148714
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/148714
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit b20b51c with merge base 1e37e5b ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
This pull request was exported from Phabricator. Differential Revision: D70670957 |
|
This pull request was exported from Phabricator. Differential Revision: D70670957 |
yushangdi
force-pushed
the
export-D70670957
branch
from
8e4ed93 to
6b0506c
Compare
…mpile_wrapper_with_O0 is set (pytorch#148714) Summary: 1. the '\0' char cannot be properly recognized. Change to use '0' instead. 2. We got the following error when using anything other than O0 flag: `error: Function ZN5torch12aot_inductorL22__check_inputs_outputsEPP16AtenTensorOpaqueS3 is too big to optimize [-Werror,-Wignored-optimization-argument]` So we use O0 flag in wrapper code when `aot_inductor.compile_wrapper_with_O0` is set Test Plan: ``` buck run 'fbcode//mode/opt' fbcode//deeplearning/aot_inductor/cpu/test:ads_second_stage_dsnn_models_aoti_lowering_test -- -r AdsSecondStageDSNNModelsAOTILoweringTest ``` Differential Revision: D70670957
yushangdi
force-pushed
the
export-D70670957
branch
from
6b0506c to
f56e5d6
Compare
|
This pull request was exported from Phabricator. Differential Revision: D70670957 |
| """ | ||
| bool _check_aoti_runtime_check_inputs_env() { | ||
| const static char* env_var_value = getenv("AOTI_RUNTIME_CHECK_INPUTS"); | ||
| const static bool result = env_var_value != nullptr && env_var_value[0] != '\0'; |
There was a problem hiding this comment.
Is just escaping the backslash with a double backslash the proper fix?
There was a problem hiding this comment.
@Skylion007 I intend to check if it's a 0, not checking against the null character. The original code was wrong. I guess the PR summary was misleading, I updated it now.
|
The second bullet of your PR description is also stale. Please update. |
…mpile_wrapper_with_O0 is set (pytorch#148714) Summary: 1. Check against the "0" char instead 2. We got the following error when using anything other than O0 flag: `error: Function ZN5torch12aot_inductorL22__check_inputs_outputsEPP16AtenTensorOpaqueS3 is too big to optimize [-Werror,-Wignored-optimization-argument]` So we use O0 flag in wrapper code when `aot_inductor.compile_wrapper_opt_level` is set to `O0`. Test Plan: ``` buck run 'fbcode//mode/opt' fbcode//deeplearning/aot_inductor/cpu/test:ads_second_stage_dsnn_models_aoti_lowering_test -- -r AdsSecondStageDSNNModelsAOTILoweringTest ``` Reviewed By: desertfire Differential Revision: D70670957
yushangdi
force-pushed
the
export-D70670957
branch
from
f56e5d6 to
b20b51c
Compare
|
This pull request was exported from Phabricator. Differential Revision: D70670957 |
|
@pytorchbot merge (Initiating merge automatically since Phabricator Diff has merged) |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
5 participants
Summary:
Check against the "0" char instead
We got the following error when using anything other than O0 flag:
error: Function ZN5torch12aot_inductorL22__check_inputs_outputsEPP16AtenTensorOpaqueS3 is too big to optimize [-Werror,-Wignored-optimization-argument]So we use O0 flag in wrapper code whenaot_inductor.compile_wrapper_opt_levelis set toO0.Test Plan:
Differential Revision: D70670957
cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @yf225 @chenyang78 @kadeng @muchulee8 @amjames @chauhang @aakhundov