It seems the overhead is due to there being one asciidoctor process spawned per file. Running asciidoctor in batch mode is fast.
My benchmarking attempts with
-
hyperfine (benchmarking tool)
-
laptop: 2020 arm m1 macbook air, 8gb ram, macos ventura 13.5.1
-
hugo version: v0.111.3+extended
-
site Summary
type count Pages 121 Non-page files 29 Static files 9 -
bench results
bench name command time (sec) stock hugo8.9 shim alias asciidoctor=cat0.7 standalone fd -e adoc-xasciidoctor -o -6.2 batch fd -e adoc-Xasciidoctor -o -0.38