diffcore-rename, merge-ort: use handle_early_known_dir_renames()

Put the work of the last several patches to work by calling the new
handle_early_known_dir_renames() function just after doing
basename-guided rename detection.

For the testcases mentioned in commit 557ac03 ("merge-ort: begin
performance work; instrument with trace2_region_* calls", 2020-10-28),
this change improves the performance as follows:

                            Before                  After
    no-renames:        5.671 s ±  0.052 s     5.714 s ±  0.087 s
    mega-renames:     13.653 s ±  0.104 s    11.439 s ±  0.192 s
    just-one-mega:   499.4  ms ±  3.6  ms   501.1  ms ±  3.8  ms

While this improvement looks rather modest for these testcases (because
all the previous optimizations were sufficient to nearly remove all time
spent in rename detection already), there was a cherry-pick in a real
world (but private) repository at $DAYJOB that showed a speedup factor
of ~7x from this optimization.

An important side note for future further optimization:

There is a possible improvement to this optimization that I have not yet
attempted: we could first check whether exact renames provide enough
information for us to determine directory renames, and avoid doing
basename-guided rename detection on some or all of the RELEVANT_LOCATION
files within those directories.  In effect, this variant would mean
doing the handle_early_known_dir_renames() both after exact rename
detection and again after basename-guided rename detection, though I
remember thinking at one point that there was some extra cleanup needed
between the steps if we were to try that.  Checking for skippable
renames an extra time might be a valuable optimization based on an
understanding of the difference in performance of exact rename detection
and basename-guided rename detection:

    While basename-guided rename detection is faster than full inexact
    rename detection with the latter's NxM file content comparisons,
    basename-guided rename detection is vastly slower than exact rename
    detection.  The reason for this is that the first time you need to
    compare a file's contents to another is the most expensive; after
    diffcore-rename has generated a sequence of hashes for a file's
    contents, any subsequent comparisons of that file (via the sequence
    of hashes) to another file is much cheaper.

So, removing files from rename detection before doing some
basename-guided rename detection could potentially improve this
optimization further.  However, this particular optimization was
actually the last one I did in original implementation order, and by the
time I implemented this idea, every testcase I had was sufficiently fast
that further optimization was unwarranted.  If future testcases arise
that tax rename detection more heavily, it may be worth implementing
this more involved variant.

Signed-off-by: Elijah Newren <[email protected]>

Author: newren

diffcore-rename.c

@@ -1077,7 +1077,6 @@ static void remove_unneeded_paths_from_src(int detecting_copies,
 	rename_src_nr = new_num_src;
 }
 
-MAYBE_UNUSED
 static void handle_early_known_dir_renames(struct dir_rename_info *info,
 					   struct strintmap *relevant_sources,
 					   struct strintmap *dirs_removed)
@@ -1306,9 +1305,15 @@ void diffcore_rename_extended(struct diff_options *options,
 		 * Cull sources, again:
 		 *   - remove ones involved in renames (found via basenames)
 		 *   - remove ones not found in relevant_sources
+		 * and
+		 *   - remove ones in relevant_sources which are needed only
+		 *     for directory renames, if no ancestory directory actually
+		 *     needs to know any more individual path renames under them
 		 */
 		trace2_region_enter("diff", "cull basename", options->repo);
 		remove_unneeded_paths_from_src(want_copies, relevant_sources);
+		handle_early_known_dir_renames(&info, relevant_sources,
+					       dirs_removed);
 		trace2_region_leave("diff", "cull basename", options->repo);
 	}