diff --git a/TODO b/TODO
index 76edded6..f8ea7a4a 100644
--- a/TODO
+++ b/TODO
@@ -1,8 +1,31 @@
+- vipsprofile reports a leak, strangely
+
+- vipsprofile performance is very poor for large data sets, eg. 
+
+	time vips sharpen wtc.jpg x.jpg --radius 20  --vips-profile
+	recording profile in vips-profile.txt
+	real	0m14.728s
+	user	0m55.515s
+	sys	0m0.200s
+	john@bambam ~/pics $ vipsprofile 
+	reading from vips-profile.txt
+	loaded 157716 events
+	last time = 14.584175
+	name		alive	wait%	work%	unkn%	memory	peakm
+	worker 20	    14	2.22	95.8	2	22.7	22.7	
+	worker 21	    14	2.67	95.4	1.93	8.5	8.5	
+	worker 22	    14	2.95	95.2	1.84	17.8	17.8	
+	worker 23	    14	2.44	95.5	2.1	11.4	11.4	
+	wbuffer 24	    15	96	4.02	0.000654	0 0	
+	wbuffer 25	    15	95.4	4.62	0.000696 0	0	
+	main 26	    15	99.1	0	0.923	-37.4 6.87	
+	peak memory = 67.3 MB
+	leak! final memory = 23 MB
+	positioning work/wait/mem ...
+	0% complete   
 
 - vipsprofile needs a man page for Debian, I guess
 
-- vipsprofile reports a leak, strangely
-
 - new_heart.ws fails with libvips master
 
 	has the sharing fix resolved this?
diff --git a/tools/vipsprofile b/tools/vipsprofile
index f58e730e..08ad62a3 100644
--- a/tools/vipsprofile
+++ b/tools/vipsprofile
@@ -239,13 +239,24 @@ for thread in threads:
 
     thread.total_y = total_y
 
+    n_thread_events = len(thread.events)
+    if n_thread_events == 0:
+        continue
+
     # first pass .. move work and wait events to y == 0
+    print 'positioning work/wait/mem ...'
+    i = 0
     gate_positions = {}
     for event in thread.events:
+        i += 1
+        if i % (1 + n_thread_events / 100) == 0:
+            print '%d%% complete   \r' %  (100 * i / n_thread_events),
+        print '%d%% complete   \r' %  (100 * i / n_thread_events),
+
         if not event.work and not event.wait and not event.memory:
             continue
 
-        # no works and waits must overlap
+        # works and waits must not overlap
         if event.work or event.wait:
             if not event.gate_name in gate_positions:
                 for gate_name in gate_positions:
@@ -261,8 +272,14 @@ for thread in threads:
         event.total_y = total_y
 
     # second pass: move all other events to non-overlapping ys
+    print 'finding maximal sets of non-overlapping gates ...'
     y = 1
+    i = 0
     for event in thread.events:
+        i += 1
+        if i % (1 + n_thread_events / 100) == 0:
+            print '%d%% complete   \r' %  (100 * i / n_thread_events),
+
         if event.work or event.wait or event.memory:
             continue
 
@@ -292,6 +309,7 @@ for thread in threads:
 
     # third pass: flip the order of the ys to get the lowest-level ones at the
     # top, next to the wait/work line
+    print 'ordering timelines by granularity ...'
     for event in thread.events:
         if event.work or event.wait or event.memory:
             continue