diff --git a/tools/vipsprofile.py b/tools/vipsprofile.py
index 533a115d..d64c06f1 100755
--- a/tools/vipsprofile.py
+++ b/tools/vipsprofile.py
@@ -4,8 +4,6 @@ import re
 import math
 import cairo
 
-WIDTH, HEIGHT = 256, 256
-
 class ReadFile:
     def __init__(self, filename):
         self.filename = filename
@@ -40,20 +38,34 @@ def read_times(rf):
 
     return times[::-1]
 
-class Event:
-    def __init__(self, thread_name, gate_name, start, stop):
+class Thread:
+    thread_number = 0
+
+    def __init__(self, thread_name):
         self.thread_name = thread_name
+        self.thread_number = Thread.thread_number
+        self.events = []
+        Thread.thread_number += 1
+
+class Event:
+    def __init__(self, thread, gate_name, start, stop):
+        self.thread = thread
         self.gate_name = gate_name
         self.start = start
         self.stop = stop
 
-        if re.match(': work', gate_name):
+        self.work = False
+        self.wait = False
+        if re.match('.*: .*work.*', gate_name):
             self.work = True
-        if re.match(': wait', gate_name):
+        if re.match('.*: .*wait.*', gate_name):
             self.wait = True
 
-events = []
+        thread.events.append(self)
+
 thread_id = 0
+threads = []
+n_events = 0
 with ReadFile('vips-profile.txt') as rf:
     while rf:
         match = re.match('thread: (.*)', rf.line)
@@ -61,6 +73,8 @@ with ReadFile('vips-profile.txt') as rf:
             print 'parse error line %d, expected "thread"' % rf.lineno
         thread_name = match.group(1) + " " + str(thread_id)
         thread_id += 1
+        thread = Thread(thread_name)
+        threads.append(thread)
         rf.getnext()
 
         while True:
@@ -88,54 +102,95 @@ with ReadFile('vips-profile.txt') as rf:
                 print 'start and stop length mismatch'
 
             for a, b in zip(start, stop):
-                event = Event(thread_name, gate_name, a, b)
-                events.append(event)
+                Event(thread, gate_name, a, b)
+                n_events += 1
 
-events.sort(lambda x, y: cmp(x.start, y.start))
+for thread in threads:
+    thread.events.sort(lambda x, y: cmp(x.start, y.start))
 
-print 'loaded %d events' % len(events)
+print 'loaded %d events' % n_events
 
 # normalise time axis to secs of computation
 ticks_per_sec = 1000000.0
-start_time = events[0].start
-for event in events:
-    event.start = (event.start - start_time) / ticks_per_sec
-    event.stop = (event.stop - start_time) / ticks_per_sec
-last_time = events[-1].stop
+start_time = threads[0].events[0].start
+last_time = 0
+for thread in threads:
+    for event in thread.events:
+        event.start = (event.start - start_time) / ticks_per_sec
+        event.stop = (event.stop - start_time) / ticks_per_sec
+
+        if event.stop > last_time:
+            last_time = event.stop
+
 print 'last time =', last_time
 
-# within each thread, allocate a Y position
-threads = []
-for event in events:
-    if not event.thread_name in threads:
-        threads.append(event.thread_name)
+# do two gates overlap?
+def is_overlap(events, gate_name1, gate_name2):
+    for event1 in events:
+        if event1.gate_name != gate_name1:
+            continue
 
+        for event2 in events:
+            if event2.gate_name != gate_name2:
+                continue
 
+            # if either endpoint of 1 is within 2
+            if event1.start > event2.start and event1.stop < event2.stop:
+                return True
+            if event1.stop > event2.start and event1.stop < event2.stop:
+                return True
 
+    return False
+
+# allocate a y position for each gate
+total_y = 0
+for thread in threads:
+    y = 1
+    gate_positions = {}
+    for event in thread.events:
+        if event.work or event.wait:
+            gate_positions[event.gate_name] = 0
+        elif not event.gate_name in gate_positions:
+            overlap = True
+            for gate_name in gate_positions:
+                if not is_overlap(thread.events, gate_name, event.gate_name):
+                    gate_positions[event.gate_name] = gate_positions[gate_name]
+                    overlap = False
+                    break
+
+            if overlap:
+                gate_positions[event.gate_name] = y
+                y += 1
+
+        event.y = gate_positions[event.gate_name]
+        event.total_y = total_y + y
+
+    total_y += y
+
+PIXELS_PER_SECOND = 1000
+PIXELS_PER_GATE = 20
+WIDTH = int(last_time * PIXELS_PER_SECOND)
+HEIGHT = int(total_y * PIXELS_PER_GATE)
 
 surface = cairo.ImageSurface (cairo.FORMAT_ARGB32, WIDTH, HEIGHT)
 ctx = cairo.Context (surface)
 
-ctx.scale (WIDTH, HEIGHT) # Normalizing the canvas
+ctx.scale (PIXELS_PER_SECOND, PIXELS_PER_GATE) 
 
-pat = cairo.LinearGradient (0.0, 0.0, 0.0, 1.0)
-pat.add_color_stop_rgba (1, 0.7, 0, 0, 0.5) # First stop, 50% opacity
-pat.add_color_stop_rgba (0, 0.9, 0.7, 0.2, 1) # Last stop, 100% opacity
+for thread in threads:
+    for event in thread.events:
+        ctx.move_to (event.start, event.total_y)
+        ctx.line_to (event.stop, event.total_y)
+        ctx.close_path ()
+        ctx.set_line_width (0.5)
 
-ctx.rectangle (0, 0, 1, 1) # Rectangle(x0, y0, x1, y1)
-ctx.set_source (pat)
-ctx.fill ()
+        if event.wait:
+            ctx.set_source_rgb (0.9, 0.1, 0.1)
+        elif event.work:
+            ctx.set_source_rgb (0.1, 0.9, 0.1)
+        else:
+            ctx.set_source_rgb (0.1, 0.1, 0.9)
 
-ctx.translate (0.1, 0.1) # Changing the current transformation matrix
-
-ctx.move_to (0, 0)
-ctx.arc (0.2, 0.1, 0.1, -math.pi/2, 0) # Arc(cx, cy, radius, start_angle, stop_angle)
-ctx.line_to (0.5, 0.1) # Line to (x,y)
-ctx.curve_to (0.5, 0.2, 0.5, 0.4, 0.2, 0.8) # Curve(x1, y1, x2, y2, x3, y3)
-ctx.close_path ()
-
-ctx.set_source_rgb (0.3, 0.2, 0.5) # Solid color
-ctx.set_line_width (0.02)
-ctx.stroke ()
+        ctx.stroke ()
 
 surface.write_to_png ("example.png") # Output to PNG