I have spotted the biggest bottleneck in "bdiff.c". Actually it was pretty easy to find after I recompiled the python interpreter and mercurial for profiling. In "bdiff.c" function "equatelines" allocates the minimum hash table size, which can lead to tons of collisions. I introduced an "overcommit" factor of 16, this is, I allocate 16 times more memory than the minimum value. Overcommiting 128 times does not improve the performance over the 16-times case.

% init
% commit
adding base
% qnew mqbase
% qrefresh
% qdiff
diff -r 67e992f2c4f3 base
--- a/base
+++ b/base
@@ -1,1 +1,1 @@ base
-base
+patched
% qdiff dirname
diff -r 67e992f2c4f3 base
--- a/base
+++ b/base
@@ -1,1 +1,1 @@ base
-base
+patched