Revision tags: llvmorg-18.1.8, llvmorg-18.1.7, llvmorg-18.1.6, llvmorg-18.1.5, llvmorg-18.1.4, llvmorg-18.1.3, llvmorg-18.1.2, llvmorg-18.1.1, llvmorg-18.1.0, llvmorg-18.1.0-rc4, llvmorg-18.1.0-rc3, llvmorg-18.1.0-rc2, llvmorg-18.1.0-rc1, llvmorg-19-init |
|
#
9584f583 |
| 01-Dec-2023 |
Amir Ayupov <aaupov@fb.com> |
[BOLT][utils] Bump default time threshold to 2s in nfc-stat-parser
|
#
202dda8e |
| 28-Nov-2023 |
Amir Ayupov <aaupov@fb.com> |
[BOLT][utils] Bump default time threshold to 1s in nfc-stat-parser
|
Revision tags: llvmorg-17.0.6, llvmorg-17.0.5 |
|
#
12c66251 |
| 11-Nov-2023 |
Amir Ayupov <aaupov@fb.com> |
[BOLT][Utils] Add nfc-stat-parser.py (#71979)
Add a utility to parse output from llvm-bolt-wrapper script and detect individual and aggregate time/memory swings. The wrapper reports wall time and pe
[BOLT][Utils] Add nfc-stat-parser.py (#71979)
Add a utility to parse output from llvm-bolt-wrapper script and detect individual and aggregate time/memory swings. The wrapper reports wall time and peak RSS for each BOLT invocation.
Exit code: The utility exits with non-zero exit code if any individual test has time or memory swing larger than `threshold_single` (default 10%), or the aggregate (geometric mean) swing larger than `threshold_agg` (default 5%). Short tests where BOLT wall time is less than `check_longer_than` seconds (default 0.5s) are excluded from threshold calculation.
Output: The script prints test results exceeding the individual threshold, and geomean values if it exceeds aggregate results. In `--verbose` mode all individual results are printed (short time results are marked with '?').
Example usage: ``` $ cd ~/llvm-build # build folder where NFC testing was invoked $ python3 ~/llvm-project/bolt/utils/nfc-stat-parser.py \ --check_longer_than 0.1 `find -name timing.log` ./tools/bolt/test/runtime/X86/exceptions-pic.test/ -88.46% -0.13% Geomean -19.78% +0.37% $ echo $? 1 ```
show more ...
|