Closed morrisonlevi closed 3 weeks ago
All modified and coverable lines are covered by tests :white_check_mark:
Project coverage is 79.32%. Comparing base (
e636250
) to head (76dd2bf
). Report is 2 commits behind head on master.
Difference between the commit on master and this branch:
# This branch
-rw-r--r-- 1 levi.morrison staff 425M Jun 6 21:30 dd-library-php-1.0.0+76dd2bf00d17575ba6c41682e29f419dd7795668-aarch64-linux-gnu.tar.gz
-rw-r--r-- 1 levi.morrison staff 428M Jun 6 21:30 dd-library-php-1.0.0+76dd2bf00d17575ba6c41682e29f419dd7795668-x86_64-linux-gnu.tar.gz
# master
-rw-r--r-- 1 levi.morrison staff 496M Jun 6 20:14 dd-library-php-1.0.0+e63625078edc52eb92e4c6104f0daaab3ec8fa63-aarch64-linux-gnu.tar.gz
-rw-r--r-- 1 levi.morrison staff 500M Jun 6 20:11 dd-library-php-1.0.0+e63625078edc52eb92e4c6104f0daaab3ec8fa63-x86_64-linux-gnu.tar.gz
Roughly 15% savings.
Benchmark execution time: 2024-06-07 03:46:29
Comparing candidate commit 76dd2bf00d17575ba6c41682e29f419dd7795668 in PR branch levi/shrink-panic-unwind
with baseline commit e63625078edc52eb92e4c6104f0daaab3ec8fa63 in branch master
.
Found 2 performance improvements and 1 performance regressions! Performance is the same for 26 metrics, 7 unstable metrics.
cpu_user_time
[-110.046ms; -61.233ms] or [-5.017%; -2.792%]execution_time
[-43.920ms; -27.745ms] or [-3.719%; -2.349%]wall_time
[+509.800ns; +535.407ns] or [+2.161%; +2.269%]
PROF-9904
Description
This is an attempt to shrink artifacts, but unlike #2601, this does not do
panic=abort
. Forpanic=abort
to be really effective, you also need to rebuild std from source. That means rebuilding images and also installing Rust from rustup (because the packages we are using do not include std sources). We had issues with CI last time we tried it. We played whack-a-mole for a while, but since the issues just kept going and we wanted to release 1.0.0beta1, we decide to revert back.But this PR without
panic=abort
will still shrink the size a fair bit, so it's worth splitting out. I wouldn't go so far as saying that it will fix the size issue reported in PR #2599, but it helps.Reviewer checklist