Re: State of the overlay tree branch?

[Stefan Monnier <monnier@iro.umontreal.ca>]

> this is the profiler report I get for the slow case (BTW, is there a way to
> have the profiler resolve functions within line-number-at-pos?

It should do that without you asking.  I mean, it won't show you
`goto-char` and `point-min` kind of things since these are "inlined"
(actually turned into their own byte-code), but `count-lines` should
definitely be there.

>              - let                                              18368  86%
>                 line-number-at-pos                              18348  86%

It's very odd that there's no `count-lines` down here (and `count-lines`
is a perfectly normal Elisp function that's not inlined or otherwise
treated specially), since it should be where most of the time is spent!

If we assume the code works as intended, it would imply that the time is
not spent in `count-lines` but elsewhere, e.g. in `goto-char`.

> with perf, the ("self") time taken by buf_charpos_to_bytepos increases from
> ~60% (fast case) to >98%. This is the diff generated by perf diff
> <fast.perf> <slow.perf>:

So I guess that could be it: the (goto-char opoint) spends an inordinate
amount of time in buf_charpos_to_bytepos.

Could you try the patch below, to see if it makes a difference?


        Stefan


diff --git a/src/marker.c b/src/marker.c
index 7773c4fce0..3d808fd6fa 100644
--- a/src/marker.c
+++ b/src/marker.c
@@ -141,6 +141,7 @@ buf_charpos_to_bytepos (struct buffer *b, ptrdiff_t charpos)
   struct Lisp_Marker *tail;
   ptrdiff_t best_above, best_above_byte;
   ptrdiff_t best_below, best_below_byte;
+  ptrdiff_t distance = 50;
 
   eassert (BUF_BEG (b) <= charpos && charpos <= BUF_Z (b));
 
@@ -180,8 +181,10 @@ buf_charpos_to_bytepos (struct buffer *b, ptrdiff_t charpos)
       /* If we are down to a range of 50 chars,
 	 don't bother checking any other markers;
 	 scan the intervening chars directly now.  */
-      if (best_above - best_below < 50)
+      if (best_above - best_below < distance)
 	break;
+      else
+        distance = distance + 10;
     }
 
   /* We get here if we did not exactly hit one of the known places.