bitmask-based optimizations

This commit is contained in:
David Rose 2009-02-10 20:39:26 +00:00
parent 2677e77fb7
commit 66a120ea8e

View File

@ -33,6 +33,65 @@ class TexMemWatcher(DirectObject):
self.cleanedUp = False
self.top = 1.0
# The textures managed by the TexMemWatcher are packed
# arbitrarily into the canvas, which is the viewable region
# that represents texture memory allocation. The packing
# arrangement has no relation to actual layout within texture
# memory (which we have no way to determine).
# The visual size of each texture is chosen in proportion to
# the total number of bytes of texture memory the texture
# consumes. This includes mipmaps, and accounts for texture
# compression. Visually, a texture with mipmaps will be
# represented by a rectangle 33% larger than an
# equivalent-sized texture without mipmaps. Of course, this
# once again has little bearing to the way the textures are
# actually arranged in memory; but it serves to give a visual
# indication of how much texture memory each texture consumes.
# There is an arbitrary limit, self.limit, which may have been
# passed to the constructor, or which may be arbitrarily
# determined. This represents the intended limit to texture
# memory utilization. We (generously) assume that the
# graphics card will implement a perfect texture packing
# algorithm, so that as long as our total utilization <=
# self.limit, it must fit within texture memory. We represent
# this visually by aggressively packing textures within the
# self.limit block so that they are guaranteed to fit, as long
# as we do not exceed the total utilization. This may
# sometimes mean distorting a texture block or even breaking
# it into multiple pieces to get it to fit, clearly
# fictionalizing whatever the graphics driver is actually
# doing.
# Internally, textures are packed into an integer grid of
# Q-units. Q-units are in proportion to texture bytes.
# Specifically, each Q-unit corresponds to a block of
# self.quantize * self.quantize texture bytes in the Texture
# Memory window. The Q-units are the smallest packable unit;
# increasing self.quantize therefore reduces the visual
# packing resolution correspondingly.
# This number defines the size of a Q-unit square, in texture
# bytes. It is automatically adjusted in repack().
self.quantize = 1
# This is the maximum number of bitmask rows (within
# self.limit) to allocate for packing. This controls the
# value assigned to self.quantize in repack().
self.maxHeight = base.config.GetInt('tex-mem-max-height', 300)
# The total number of texture bytes tracked, including overflow.
self.totalSize = 0
# The total number of texture bytes placed, not including
# overflow (that is, within self.limit).
self.placedSize = 0
# The total number of Q-units placed, not including overflow.
self.placedQSize = 0
# If no GSG is specified, use the main GSG.
if gsg is None:
@ -87,6 +146,11 @@ class TexMemWatcher(DirectObject):
self.win.setWindowEvent(eventName)
self.accept(eventName, self.windowEvent)
# Listen for this event so we can update appropriately, if
# anyone changes the window's graphics memory limit,
self.accept('graphics_memory_limit_changed',
self.graphicsMemoryLimitChanged)
# We'll need a mouse object to get mouse events.
self.mouse = base.dataRoot.attachNewNode(MouseAndKeyboard(self.win, 0, '%s-mouse' % (self.name)))
bt = ButtonThrower('%s-thrower' % (self.name))
@ -106,8 +170,11 @@ class TexMemWatcher(DirectObject):
self.isolate = None
self.isolated = None
self.needsRepack = False
self.task = taskMgr.doMethodLater(0.5, self.updateTextures, 'TexMemWatcher')
# How frequently should the texture memory window check for
# state changes?
updateInterval = base.config.GetDouble("tex-mem-update-interval", 0.5)
self.task = taskMgr.doMethodLater(updateInterval, self.updateTextures, 'TexMemWatcher')
self.setLimit(limit)
@ -224,45 +291,58 @@ class TexMemWatcher(DirectObject):
self.canvasBackground.setTexture(self.checkTex)
def setLimit(self, limit):
def setLimit(self, limit = None):
""" Indicates the texture memory limit. If limit is None or
unspecified, the limit is taken from the GSG, if any; or there
is no limit. """
self.__doSetLimit(limit)
self.reconfigureWindow()
def __doSetLimit(self, limit):
""" Internal implementation of setLimit(). """
self.limit = limit
self.lruLimit = False
self.dynamicLimit = False
if limit is None:
if not limit:
# If no limit was specified, use the specified graphics
# memory limit, if any.
lruLimit = self.gsg.getPreparedObjects().getGraphicsMemoryLimit()
if lruLimit < 2**32 - 1:
# Got a real lruLimit. Use it.
self.limit = lruLimit
lruSize = self.gsg.getPreparedObjects().getGraphicsMemoryLimit()
if lruSize and lruSize < 2**32 - 1:
# Got a real lruSize. Use it.
self.limit = lruSize
self.lruLimit = True
else:
# No LRU limit either, so there won't be a practical
# limit to the TexMemWatcher. We'll determine our
# limit on-the-fly instead.
self.dynamicLimit = True
if not self.dynamicLimit:
# Set our GSG to limit itself to no more textures than we
# expect to display onscreen, so we don't go crazy with
# texture memory.
self.win.getGsg().getPreparedObjects().setGraphicsMemoryLimit(self.limit)
if self.dynamicLimit:
# Choose a suitable limit by rounding to the next power of two.
self.limit = Texture.upToPower2(self.totalSize)
# Set our GSG to limit itself to no more textures than we
# expect to display onscreen, so we don't go crazy with
# texture memory.
self.win.getGsg().getPreparedObjects().setGraphicsMemoryLimit(self.limit)
# The actual height of the canvas, including the overflow
# area. The texture memory itself is restricted to (0..1)
# vertically; anything higher than 1 is overflow.
self.top = 1.25
top = 1.25
if self.dynamicLimit:
# Actually, we'll never exceed texture memory, so never mind.
self.top = 1
self.makeCanvasBackground()
top = 1
if top != self.top:
self.top = top
self.makeCanvasBackground()
self.canvasLens.setFilmSize(1, self.top)
self.canvasLens.setFilmOffset(0.5, self.top / 2.0) # lens covers 0..1 in x and y
self.reconfigureWindow()
def cleanup(self):
if not self.cleanedUp:
self.cleanedUp = True
@ -284,6 +364,10 @@ class TexMemWatcher(DirectObject):
self.texRecordsByKey = {}
self.texPlacements = {}
def graphicsMemoryLimitChanged(self):
if self.dynamicLimit or self.lruLimit:
self.__doSetLimit(None)
self.reconfigureWindow()
def windowEvent(self, win):
if win == self.win:
@ -537,7 +621,7 @@ class TexMemWatcher(DirectObject):
else:
overflowCount = sum(map(lambda tp: tp.overflowed, self.texPlacements.keys()))
if overflowCount:
if totalSize <= self.limit and overflowCount:
# Shouldn't be overflowing any more. Better repack.
self.repack()
@ -546,7 +630,7 @@ class TexMemWatcher(DirectObject):
# Sort the regions from largest to smallest to maximize
# packing effectiveness.
texRecords.sort(key = lambda tr: (-tr.w, -tr.h))
texRecords.sort(key = lambda tr: (tr.tw, tr.th), reverse = True)
self.overflowing = False
for tr in texRecords:
@ -564,10 +648,13 @@ class TexMemWatcher(DirectObject):
self.texRecordsByTex = {}
self.texRecordsByKey = {}
self.texPlacements = {}
self.bitmasks = []
self.mw.clearRegions()
self.setRollover(None, None)
self.w = 1
self.h = 1
self.placedSize = 0
self.placedQSize = 0
pgo = self.gsg.getPreparedObjects()
totalSize = 0
@ -589,14 +676,9 @@ class TexMemWatcher(DirectObject):
if not self.totalSize:
return
if self.dynamicLimit:
# Choose a suitable limit by rounding to the next power of two.
self.limit = Texture.upToPower2(self.totalSize)
# Set our GSG to limit itself to no more textures than we
# expect to display onscreen, so we don't go crazy with
# texture memory.
self.win.getGsg().getPreparedObjects().setGraphicsMemoryLimit(self.limit)
if self.dynamicLimit or self.lruLimit:
# Adjust the limit to ensure we keep tracking the lru size.
self.__doSetLimit(None)
# Now make that into a 2-D rectangle of the appropriate shape,
# such that w * h == limit.
@ -613,8 +695,28 @@ class TexMemWatcher(DirectObject):
# Region size
w = math.sqrt(self.limit) / math.sqrt(r)
h = w * r
# Now choose self.quantize so that we don't exceed
# self.maxHeight.
if h > self.maxHeight:
self.quantize = int(math.ceil(h / self.maxHeight))
else:
self.quantize = 1
w = max(int(w / self.quantize + 0.5), 1)
h = max(int(h / self.quantize + 0.5), 1)
self.w = w
self.h = h
self.area = self.w * self.h
# We store a bitarray for each row, for fast lookup for
# unallocated space on the canvas. Each pixel on the row
# corresponds to a bit in the bitarray, where bit 0 is pixel
# 0, bit 1 is pixel 1, and so on. If the bit is set, the
# space is occupied.
self.bitmasks = []
for i in range(self.h):
self.bitmasks.append(BitArray())
self.canvas.setScale(1.0 / w, 1.0, 1.0 / h)
self.mw.setFrame(0, w, 0, h * self.top)
@ -622,7 +724,7 @@ class TexMemWatcher(DirectObject):
# Sort the regions from largest to smallest to maximize
# packing effectiveness.
texRecords = self.texRecordsByTex.values()
texRecords.sort(key = lambda tr: (tr.w, tr.h), reverse = True)
texRecords.sort(key = lambda tr: (tr.tw, tr.th), reverse = True)
self.overflowing = False
for tr in texRecords:
@ -644,34 +746,35 @@ class TexMemWatcher(DirectObject):
def unplaceTexture(self, tr):
""" Removes the texture from its place on the canvas. """
for tp in tr.placements:
tp.clearBitmasks(self.bitmasks)
if not tp.overflowed:
self.placedQSize -= tp.area
del self.texPlacements[tp]
tr.placements = []
tr.clearCard(self)
if not tp.overflowed:
self.placedSize -= tr.size
def placeTexture(self, tr):
""" Places the texture somewhere on the canvas where it will
fit. """
if not self.overflowing:
tp = self.findHole(tr.w, tr.h)
if tp:
tr.placements = [tp]
tr.makeCard(self)
self.texPlacements[tp] = tr
return
tr.computePlacementSize(self)
# Couldn't find a hole; can we fit it if we rotate?
tp = self.findHole(tr.h, tr.w)
if tp:
tp.rotated = True
tr.placements = [tp]
tr.makeCard(self)
self.texPlacements[tp] = tr
return
# Couldn't find a hole of the right shape; can we find a
# single rectangular hole of the right area, but of any shape?
tp = self.findArea(tr.w, tr.h)
shouldFit = False
availableSize = self.limit - self.placedSize
if availableSize >= tr.size:
shouldFit = True
availableQSize = self.area - self.placedQSize
if availableQSize < tr.area:
# The texture should fit, but won't, due to roundoff
# error. Make it correspondingly smaller, so we can
# place it anyway.
tr.area = availableQSize
if shouldFit:
# Look for a single rectangular hole to hold this piece.
tp = self.findHole(tr.area, tr.w, tr.h)
if tp:
texCmp = cmp(tr.w, tr.h)
holeCmp = cmp(tp.p[1] - tp.p[0], tp.p[3] - tp.p[2])
@ -679,13 +782,16 @@ class TexMemWatcher(DirectObject):
tp.rotated = True
tr.placements = [tp]
tr.makeCard(self)
tp.setBitmasks(self.bitmasks)
self.placedQSize += tp.area
self.texPlacements[tp] = tr
self.placedSize += tr.size
return
# Couldn't find a single rectangular hole. We'll have to
# divide the texture up into several smaller pieces to cram it
# in.
tpList = self.findHolePieces(tr.h * tr.w)
tpList = self.findHolePieces(tr.area)
if tpList:
texCmp = cmp(tr.w, tr.h)
tr.placements = tpList
@ -693,79 +799,50 @@ class TexMemWatcher(DirectObject):
holeCmp = cmp(tp.p[1] - tp.p[0], tp.p[3] - tp.p[2])
if texCmp != 0 and holeCmp != 0 and texCmp != holeCmp:
tp.rotated = True
tp.setBitmasks(self.bitmasks)
self.placedQSize += tp.area
self.texPlacements[tp] = tr
self.placedSize += tr.size
tr.makeCard(self)
return
# Just let it overflow.
self.overflowing = True
tp = self.findHole(tr.w, tr.h, allowOverflow = True)
tp = self.findOverflowHole(tr.area, tr.w, tr.h)
if tp:
if tp.p[1] > self.w or tp.p[3] > self.h:
tp.overflowed = 1
while len(self.bitmasks) <= tp.p[3]:
self.bitmasks.append(BitArray())
tr.placements = [tp]
tr.makeCard(self)
tp.setBitmasks(self.bitmasks)
if not tp.overflowed:
self.placedQSize += tp.area
self.placedSize += tr.size
self.texPlacements[tp] = tr
return
# Something went wrong.
assert False
def findHole(self, w, h, allowOverflow = False):
""" Searches for a hole large enough for (w, h). If one is
found, returns an appropriate TexPlacement; otherwise, returns
None. """
if w > self.w:
# It won't fit within the row at all.
if not allowOverflow:
return None
# Just stack it on the top.
y = 0
if self.texPlacements:
y = max(map(lambda tp: tp.p[3], self.texPlacements.keys()))
tp = TexPlacement(0, w, y, y + h)
return tp
y = 0
while y + h <= self.h or allowOverflow:
nextY = None
# Scan along the row at 'y'.
x = 0
while x + w <= self.w:
# Consider the spot at x, y.
tp = TexPlacement(x, x + w, y, y + h)
overlap = self.findOverlap(tp)
if not overlap:
# Hooray!
return tp
nextX = overlap.p[1]
if nextY is None:
nextY = overlap.p[3]
else:
nextY = min(nextY, overlap.p[3])
assert nextX > x
x = nextX
assert nextY > y
y = nextY
# Nope, wouldn't fit anywhere.
return None
def findArea(self, w, h):
def findHole(self, area, w, h):
""" Searches for a rectangular hole that is at least area
square units big, regardless of its shape, but attempt to find
one that comes close to the right shape, at least. If one is
found, returns an appropriate TexPlacement; otherwise, returns
None. """
aspect = float(min(w, h)) / float(max(w, h))
area = w * h
holes = self.findAvailableHoles(area)
if area == 0:
tp = TexPlacement(0, 0, 0, 0)
return tp
# Rotate the hole to horizontal first.
w, h = max(w, h), min(w, h)
aspect = float(w) / float(h)
holes = self.findAvailableHoles(area, w, h)
# Walk through the list and find the one with the best aspect
# match.
@ -779,22 +856,24 @@ class TexMemWatcher(DirectObject):
# we have to squish it?
if tw < w:
# We'd have to make it taller.
nh = area / tw
assert nh <= th
nh = min(area / tw, th)
th = nh
elif th < h:
# We'd have to make it narrower.
nw = area / th
assert nw <= tw
nw = min(area / th, tw)
tw = nw
else:
# We don't have to squish it? Shouldn't have gotten
# here.
assert False
# Hey, we don't have to squish it after all! Just
# return this hole.
tw = w
th = h
# Make a new tp that has the right area.
tp = TexPlacement(l, l + tw, b, b + th)
ta = float(min(tw, th)) / float(max(tw, th))
ta = float(max(tw, th)) / float(min(tw, th))
if ta == aspect:
return tp
match = min(ta, aspect) / max(ta, aspect)
matches.append((match, tp))
@ -803,58 +882,6 @@ class TexMemWatcher(DirectObject):
return max(matches)[1]
return None
def findAllArea(self, area):
""" Searches for a rectangular hole that is at least area
square units big, regardless of its shape. Returns a list of
all such holes found. """
result = []
y = 0
while y < self.h:
nextY = self.h
# Scan along the row at 'y'.
x = 0
while x < self.w:
nextX = self.w
# Consider the spot at x, y.
# How wide can we go? Start by trying to go all the
# way to the edge of the region.
tpw = self.w - x
# Now, given this particular width, how tall do we
# need to go?
tph = area / tpw
while y + tph < self.h:
tp = TexPlacement(x, x + tpw, y, y + tph)
overlap = self.findOverlap(tp)
if not overlap:
result.append(tp)
nextX = min(nextX, overlap.p[1])
nextY = min(nextY, overlap.p[3])
# Shorten the available region.
tpw0 = overlap.p[0] - x
if tpw0 <= 0.0:
break
if x + tpw0 == x + tpw:
tpw0 *= 0.999 # imprecision hack
tpw = tpw0
tph = area / tpw
assert nextX > x
x = nextX
assert nextY > y
y = nextY
return result
def findHolePieces(self, area):
""" Returns a list of holes whose net area sums to the given
area, or None if there are not enough holes. """
@ -862,10 +889,19 @@ class TexMemWatcher(DirectObject):
# First, save the original value of self.texPlacements, since
# we will be modifying that during this search.
savedTexPlacements = copy.copy(self.texPlacements)
savedBitmasks = []
for ba in self.bitmasks:
savedBitmasks.append(BitArray(ba))
result = []
while area > 0:
# We have to call findLargestHole() each time through this
# loop, instead of just walking through
# findAvailableHoles() in order, because
# findAvailableHoles() might return a list of overlapping
# holes.
tp = self.findLargestHole()
if not tp:
break
@ -875,18 +911,23 @@ class TexMemWatcher(DirectObject):
if tpArea >= area:
# we're done.
shorten = (tpArea - area) / (r - l)
tp.p = (l, r, b, t - shorten)
t -= shorten
tp.p = (l, r, b, t)
tp.area = (r - l) * (t - b)
result.append(tp)
self.texPlacements = savedTexPlacements
self.bitmasks = savedBitmasks
return result
# Keep going.
area -= tpArea
result.append(tp)
tp.setBitmasks(self.bitmasks)
self.texPlacements[tp] = None
# Huh, not enough room, or no more holes.
self.texPlacements = savedTexPlacements
self.bitmasks = savedBitmasks
return None
def findLargestHole(self):
@ -895,83 +936,149 @@ class TexMemWatcher(DirectObject):
return max(holes)[1]
return None
def findAvailableHoles(self, area):
def findAvailableHoles(self, area, w = None, h = None):
""" Finds a list of available holes, of at least the indicated
area. Returns a list of tuples, where each tuple is of the
form (area, tp)."""
form (area, tp).
If w and h are non-None, this will short-circuit on the first
hole it finds that fits w x h, and return just that hole in a
singleton list.
"""
holes = []
y = 0
while y < self.h:
nextY = self.h
lastTuples = set()
lastBitmask = None
b = 0
while b < self.h:
# Separate this row into (l, r) tuples.
bm = self.bitmasks[b]
if bm == lastBitmask:
# This row is exactly the same as the row below; no
# need to reexamine.
b += 1
continue
# Scan along the row at 'y'.
x = 0
while x < self.w:
nextX = self.w
# Consider the spot at x, y.
# How wide can we go? Start by trying to go all the
# way to the edge of the region.
tpw = self.w - x
# And how tall can we go? Start by trying to go to
# the top of the region.
tph = self.h - y
while tpw > 0.0 and tph > 0.0:
tp = TexPlacement(x, x + tpw, y, y + tph)
overlap = self.findOverlap(tp)
if not overlap:
# Here's a hole.
tarea = tpw * tph
if tarea >= area:
holes.append((tarea, tp))
break
nextX = min(nextX, overlap.p[1])
nextY = min(nextY, overlap.p[3])
# We've been intersected either on the top or the
# right. We need to shorten either width or
# height. Which way results in the largest
# remaining area?
tpw0 = overlap.p[0] - x
tph0 = overlap.p[2] - y
if tpw0 * tph > tpw * tph0:
# Shortening width results in larger.
if x + tpw == x + tpw0:
tpw0 *= 0.999 # imprecision hack
tpw = tpw0
else:
# Shortening height results in larger.
if y + tph == y + tph0:
tph0 *= 0.999 # imprecision hack
tph = tph0
#print "x = %s, y = %s, tpw = %s, tph = %s" % (x, y, tpw, tph)
lastBitmask = bm
assert nextX > x
x = nextX
tuples = self.findEmptyRuns(bm)
newTuples = tuples.difference(lastTuples)
assert nextY > y
y = nextY
for l, r in newTuples:
# Find out how high we can go with this bitmask.
mask = BitArray.range(l, r - l)
t = b + 1
while t < self.h and (self.bitmasks[t] & mask).isZero():
t += 1
tpw = (r - l)
tph = (t - b)
tarea = tpw * tph
assert tarea > 0
if tarea >= area:
tp = TexPlacement(l, r, b, t)
if w and h and \
((tpw >= w and tph >= h) or \
(tph >= w and tpw >= h)):
# This hole is big enough; short circuit.
return [(tarea, tp)]
holes.append((tarea, tp))
lastTuples = tuples
b += 1
return holes
def findOverlap(self, tp):
""" If there is another placement that overlaps the indicated
TexPlacement, returns it. Otherwise, returns None. """
def findOverflowHole(self, area, w, h):
""" Searches for a hole large enough for (w, h), in the
overflow space. Since the overflow space is infinite, this
will always succeed. """
for other in self.texPlacements.keys():
if other.intersects(tp):
return other
if w > self.w:
# It won't fit within the margins at all; just stack it on
# the top.
return None
# Scan down past all of the empty bitmasks that may be
# stacked on top.
b = len(self.bitmasks)
while b > self.h and self.bitmasks[b - 1].isZero():
b -= 1
tp = TexPlacement(0, w, b, b + h)
return tp
# It fits within the margins; find the first row with enough
# space for it.
lastTuples = set()
lastBitmask = None
b = self.h
while True:
if b >= len(self.bitmasks):
# Off the top. Just leave it here.
tp = TexPlacement(0, w, b, b + h)
return tp
# Separate this row into (l, r) tuples.
bm = self.bitmasks[b]
if bm == lastBitmask:
# This row is exactly the same as the row below; no
# need to reexamine.
b += 1
continue
lastBitmask = bm
tuples = self.findEmptyRuns(bm)
newTuples = tuples.difference(lastTuples)
for l, r in newTuples:
# Is this region wide enough?
if r - l < w:
continue
# Is it tall enough?
r = l + w
mask = BitArray.range(l, r - l)
t = b + 1
while t < b + h and \
(t > len(self.bitmasks) or (self.bitmasks[t] & mask).isZero()):
t += 1
if t < b + h:
# Not tall enough.
continue
tp = TexPlacement(l, r, b, t)
return tp
lastTuples = tuples
b += 1
def findEmptyRuns(self, bm):
""" Separates a bitmask into a list of (l, r) tuples,
corresponding to the empty regions in the row between 0 and
self.w. """
tuples = set()
l = bm.getLowestOffBit()
assert l != -1
if l < self.w:
r = bm.getNextHigherDifferentBit(l)
if r == l or r >= self.w:
r = self.w
tuples.add((l, r))
l = bm.getNextHigherDifferentBit(r)
while l != r and l < self.w:
r = bm.getNextHigherDifferentBit(l)
if r == l or r >= self.w:
r = self.w
tuples.add((l, r))
l = bm.getNextHigherDifferentBit(r)
return tuples
class TexRecord:
@ -991,12 +1098,14 @@ class TexRecord:
y = self.tex.getYSize()
r = float(y) / float(x)
# Card size
w = math.sqrt(self.size) / math.sqrt(r)
h = w * r
# Card size, in unscaled texel units.
self.tw = math.sqrt(self.size) / math.sqrt(r)
self.th = self.tw * r
self.w = w
self.h = h
def computePlacementSize(self, tmw):
self.w = max(int(self.tw / tmw.quantize + 0.5), 1)
self.h = max(int(self.th / tmw.quantize + 0.5), 1)
self.area = self.w * self.h
def setActive(self, flag):
@ -1100,6 +1209,7 @@ class TexRecord:
class TexPlacement:
def __init__(self, l, r, b, t):
self.p = (l, r, b, t)
self.area = (r - l) * (t - b)
self.rotated = False
self.overflowed = 0
@ -1113,3 +1223,36 @@ class TexPlacement:
return (tl < mr and tr > ml and
tb < mt and tt > mb)
def setBitmasks(self, bitmasks):
""" Sets all of the appropriate bits to indicate this region
is taken. """
l, r, b, t = self.p
mask = BitArray.range(l, r - l)
for yi in range(b, t):
assert (bitmasks[yi] & mask).isZero()
bitmasks[yi] |= mask
def clearBitmasks(self, bitmasks):
""" Clears all of the appropriate bits to indicate this region
is available. """
l, r, b, t = self.p
mask = ~BitArray.range(l, r - l)
for yi in range(b, t):
assert (bitmasks[yi] | mask).isAllOn()
bitmasks[yi] &= mask
def hasOverlap(self, bitmasks):
""" Returns true if there is an overlap with this region and
any other region, false otherwise. """
l, r, b, t = self.p
mask = BitArray.range(l, r - l)
for yi in range(b, t):
if not (bitmasks[yi] & mask).isZero():
return True
return False