Name Date Size #Lines LOC

..--

3D-Reconstruction/H25-Apr-2025-2,0431,766

non_greedy_mv/H25-Apr-2025-196143

README.pgo.mdH A D25-Apr-2025896 2518

cpplint.pyH A D25-Apr-2025234.5 KiB6,2454,080

diff.pyH A D25-Apr-20254.1 KiB13191

gen_authors.shH A D25-Apr-2025314 159

intersect-diffs.pyH A D25-Apr-20252.3 KiB7749

lint-hunks.pyH A D25-Apr-20255 KiB149104

set_analyzer_env.shH A D25-Apr-20253.8 KiB13097

tiny_ssim.cH A D25-Apr-202517.4 KiB570471

wrap-commit-msg.pyH A D25-Apr-20252.1 KiB7148

README.pgo.md

1# Using Profile Guided Optimizations to identify compiler optimization failures
2
3When using Clang, the `-Rpass-missed` flag enables the verbose log of failed
4compiler optimizations. However, the extensive log messages can obscure
5potential optimization opportunities.
6
7Use the following steps to generate a more transparent optimization report
8using a previously created PGO profile file. The report also includes code
9hotness diagnostics:
10
11```bash
12$ ../libvpx/configure --use-profile=perf.profdata \
13  --extra-cflags="-fsave-optimization-record -fdiagnostics-show-hotness"
14```
15
16Convert the generated YAML files into a detailed HTML report using the
17[optviewer2](https://github.com/OfekShilon/optview2) tool:
18
19```bash
20$ opt-viewer.py --output-dir=out/ --source-dir=libvpx .
21```
22
23The HTML report displays each code line's relative hotness, cross-referenced
24with the failed compiler optimizations.
25