Hi Werner, Werner LEMBERG wrote on Sat, Mar 07, 2015 at 07:38:44PM +0100:
> Well, the ms macros are *much* simpler than mdoc's highly nested macro > parsing. You should really try unprocessed vs. processed mdoc, > applied to a very long man page, and you will see a significant > difference in processing time. Actually, the relative gain in speed is larger for *smaller* manual pages. In my last mail, i have shown that the stripping only shaves 10-15% off of the time to format the 137kB, 5671 line ksh(1) manual. For the 4.9kB, 188 line cat(1) manual, the speedup is twice as large, about 25-30% (81 vs. 111 Milliseconds; mandoc(1) for comparison: 6.3 Milliseconds, which is 13 to 18 times faster; by the way, running just wc(1) on the manual takes 3 Milliseconds, so mandoc(1) only consumes twice the time already needed for merely reading the file from the buffer cache). I guess this indicates that the speedup is mostly due to the economy *reading in* the macro file, not so much the economy *expanding* the macros. In any case, saving time is more relevant for large documents and less so for small ones that take only fractions of a second in the first place, which makes the optimization even more questionable. Yours, Ingo P.S. Note that the stripping reduces the size of the mdoc macro set from 240kB to 137kB (-43%). Comparing that to the speedup of just below 30% confirms the suspicion that the effect from reading the macroset probably dominates the speedup.
