For years I wondered why performance of the code generated by the lhs2TeX styling template I use (due to Richard Eisenberg) is so abyssmal when using acmart. It turns out this is due to an interaction between microtype and newtxmath. I could speed up the pipeline greatly by getting rid of any entering of math mode from within \text, e.g., \text{$x$} is "bad". (Measurably so only with acmart.)
For years I wondered why performance of the code generated by the
lhs2TeXstyling template I use (due to Richard Eisenberg) is so abyssmal when usingacmart. It turns out this is due to an interaction betweenmicrotypeandnewtxmath. I could speed up the pipeline greatly by getting rid of any entering of math mode from within\text, e.g.,\text{$x$}is "bad". (Measurably so only withacmart.)I added a FAQ entry.