def _createRT(self, name, width, height, attachAuxTexture=False, shaderName="", layers=1):
""" Internal shortcut to create a new render target. The name should be
a unique identifier. When attachAuxTexture is True, an aux texture will
be attached to the buffer, additionally to the color texture.
The shader name determines the shader to load for the target, see below.
When layers is > 1, a layered render target will be created to render to
a 3D texture."""
# Setup the render target
target = RenderTarget("Scattering" + name)
target.setSize(width, height)
target.addColorTexture()
target.setColorBits(32)
# Adds aux textures if specified
if attachAuxTexture:
target.addAuxTextures(1)
target.setAuxBits(32)
# Add render layers if specified
if layers > 1:
target.setLayers(layers)
target.prepareOffscreenBuffer()
# Load the appropriate shader
sArgs = ["Shader/Scattering/DefaultVertex.vertex",
"Shader/Scattering/" + shaderName + ".fragment"]
# When using layered rendering, a geometry shader is required
if layers > 1:
sArgs.append("Shader/Scattering/DefaultGeometry.geometry")
shader = Shader.load(Shader.SLGLSL, *sArgs)
target.setShader(shader)
# Make the scattering options available
self._setInputs(target, "options")
# Lowercase the first letter
lowerCaseFirst = lambda x: x[0].lower() + x[1:]
# Make all rendered textures so far available to the target
for key, tex in self.textures.iteritems():
target.setShaderInput(key, tex)
# Register the created textures
self.textures[lowerCaseFirst(name) + "Color"] = target.getColorTexture()
if attachAuxTexture:
self.textures[lowerCaseFirst(name) + "Aux"] = target.getAuxTexture(0)
return target
def _createRT(self, name, w, h, aux=False, shaderName="", layers=1):
""" Internal shortcut to create a new render target """
rt = RenderTarget("Scattering" + name)
rt.setSize(w, h)
rt.addColorTexture()
rt.setColorBits(16)
# rt.setEngine(self.engine)s
if aux:
rt.addAuxTextures(1)
rt.setAuxBits(16)
if layers > 1:
rt.setLayers(layers)
rt.prepareOffscreenBuffer()
# self._engine.openWindows()
sArgs = [
"Shader/Scattering/DefaultVertex.vertex",
"Shader/Scattering/" + shaderName + ".fragment"
]
if layers > 1:
sArgs.append("Shader/Scattering/DefaultGeometry.geometry")
shader = BetterShader.load(*sArgs)
rt.setShader(shader)
self._setInputs(rt, "options")
lc = lambda x: x[0].lower() + x[1:]
for key, tex in self.textures.items():
rt.setShaderInput(lc(key), tex)
self.textures[lc(name) + "Color"] = rt.getColorTexture()
if aux:
self.textures[lc(name) + "Aux"] = rt.getAuxTexture(0)
return rt
class RenderingPipeline(DebugObject):
""" This is the core class, driving all other classes. To use this
pipeline, your code has to call this *after* the initialization of ShowBase:
renderPipeline = RenderingPipeline()
renderPipeline.loadSettings("pipeline.ini")
renderPipeline.create()
The pipeline will setup all required buffers, tasks and shaders itself. To
add lights, see the documentation of LightManager.
How it works:
You can see an example buffer view at http://i.imgur.com/mZK6TVj.png
The pipeline first renders all normal objects (parented to render) into a
buffer, using multiple render targets. These buffers store normals,
position and material properties. Your shaders have to output these
values, but there is a handy api, just look at
Shaders/DefaultObjectShader.fragment.
After that, the pipeline splits the screen into tiles, typically of the
size 32x32. For each tile, it computes which lights affect which tile,
called Tiled Deferred Shading. This is written to a buffer.
The next step is applying the lighting. This is done at half window
resolution only, using Temporal Reprojection. I don't aim to explain
Temporal Reprojection here, but basically, I render only each second
pixel each frame. This is simply for performance.
The lighting pass iterates through the list of lights per tile, and
applies both lighting and shadows to each pixel, using the material
information from the previous rendered buffers.
After the lighting pass, a combiner pass combines both the current
frame and the last frame, this is required because of
Temporal Reprojection.
At this step, we already have a frame we could display. In the next passes,
only anti-aliasing and post-processing effects like motion blur are added.
In the meantime, the LightManager builds a list of ShadowSources which
need an update. It creates a scene render and renders the scene from the
view of the shadow sources to the global shadow atlas. There are a limited
amount of shadow updates per frame available, and the updates are stored
in a queue. So when displaying many shadow-lights, not each shadowmap is
update each frame. The reason is, again, performance. When you need a
custom shadow caster shader, e.g. for alpha blending, you should use the
Shader/DefaultShaowCaster.* as prefab.
"""
def __init__(self, showbase):
""" Creates a new pipeline """
DebugObject.__init__(self, "RenderingPipeline")
self.showbase = showbase
self.settings = None
self.mountManager = MountManager()
def getMountManager(self):
""" Returns the mount manager. You can use this to set the
write directory and base path """
return self.mountManager
def loadSettings(self, filename):
""" Loads the pipeline settings from an ini file """
self.settings = PipelineSettingsManager()
self.settings.loadFromFile(filename)
def getSettings(self):
""" Returns the current pipeline settings """
return self.settings
def create(self):
""" Creates this pipeline """
self.debug("Setting up render pipeline")
if self.settings is None:
self.error("You have to call loadSettings first!")
return
self.debug("Analyzing system ..")
SystemAnalyzer.analyze()
self.debug("Checking required Panda3D version ..")
SystemAnalyzer.checkPandaVersionOutOfDate(01, 12, 2014)
# Mount everything first
self.mountManager.mount()
# Store globals, as cython can't handle them
self.debug("Setting up globals")
Globals.load(self.showbase)
Globals.font = loader.loadFont("Data/Font/SourceSansPro-Semibold.otf")
Globals.font.setPixelsPerUnit(25)
# Setting up shader loading
BetterShader._DumpShaders = self.settings.dumpGeneratedShaders
# We use PTA's for shader inputs, because that's faster than
# using setShaderInput
self.temporalProjXOffs = PTAInt.emptyArray(1)
self.cameraPosition = PTAVecBase3f.emptyArray(1)
self.motionBlurFactor = PTAFloat.emptyArray(1)
self.lastMVP = PTALMatrix4f.emptyArray(1)
self.currentMVP = PTALMatrix4f.emptyArray(1)
self.currentShiftIndex = PTAInt.emptyArray(1)
# Initialize variables
self.camera = self.showbase.cam
self.size = self._getSize()
self.cullBounds = None
# For the temporal reprojection it is important that the window width
# is a multiple of 2
if self.settings.enableTemporalReprojection and self.size.x % 2 == 1:
self.error(
"The window has to have a width which is a multiple of 2 "
"(Current: ", self.showbase.win.getXSize(), ")")
self.error(
"I'll correct that for you, but next time pass the correct "
"window size!")
wp = WindowProperties()
wp.setSize(self.showbase.win.getXSize() + 1,
self.showbase.win.getYSize())
self.showbase.win.requestProperties(wp)
self.showbase.graphicsEngine.openWindows()
# Get new size
self.size = self._getSize()
# Debug variables to disable specific features
self.haveLightingPass = True
# haveCombiner can only be true when haveLightingPass is enabled
self.haveCombiner = True
self.haveMRT = True
# Not as good as I want it, so disabled. I'll work on it.
self.blurEnabled = False
self.debug("Window size is", self.size.x, "x", self.size.y)
self.showbase.camLens.setNearFar(0.1, 50000)
self.showbase.camLens.setFov(90)
self.showbase.win.setClearColor(Vec4(1.0, 0.0, 1.0, 1.0))
# Create GI handler
if self.settings.enableGlobalIllumination:
self._setupGlobalIllumination()
# Create occlusion handler
self._setupOcclusion()
if self.settings.displayOnscreenDebugger:
self.guiManager = PipelineGuiManager(self)
self.guiManager.setup()
# Generate auto-configuration for shaders
self._generateShaderConfiguration()
# Create light manager, which handles lighting + shadows
if self.haveLightingPass:
self.lightManager = LightManager(self)
self.patchSize = LVecBase2i(self.settings.computePatchSizeX,
self.settings.computePatchSizeY)
# Create separate scene graphs. The deferred graph is render
self.forwardScene = NodePath("Forward-Rendering")
self.transparencyScene = NodePath("Transparency-Rendering")
self.transparencyScene.setBin("transparent", 30)
# We need no transparency (we store other information in the alpha
# channel)
self.showbase.render.setAttrib(
TransparencyAttrib.make(TransparencyAttrib.MNone), 100)
# Now create deferred render buffers
self._makeDeferredTargets()
# Create the target which constructs the view-space normals and
# position from world-space position
if self.occlusion.requiresViewSpacePosNrm():
self._createNormalPrecomputeBuffer()
if self.settings.enableGlobalIllumination:
self._creatGIPrecomputeBuffer()
# Setup the buffers for lighting
self._createLightingPipeline()
# Setup combiner for temporal reprojetion
if self.haveCombiner and self.settings.enableTemporalReprojection:
self._createCombiner()
if self.occlusion.requiresBlurring():
self._createOcclusionBlurBuffer()
self._setupAntialiasing()
if self.blurEnabled:
self._createDofStorage()
self._createBlurBuffer()
# Not sure why it has to be 0.25. But that leads to the best result
aspect = float(self.size.y) / self.size.x
self.onePixelShift = Vec2(0.125 / self.size.x, 0.125 / self.size.y /
aspect) * self.settings.jitterAmount
# Annoying that Vec2 has no multliply-operator for non-floats
multiplyVec2 = lambda a, b: Vec2(a.x * b.x, a.y * b.y)
if self.antialias.requiresJittering():
self.pixelShifts = [
multiplyVec2(self.onePixelShift, Vec2(-0.25, 0.25)),
multiplyVec2(self.onePixelShift, Vec2(0.25, -0.25))
]
else:
self.pixelShifts = [Vec2(0), Vec2(0)]
self.currentPixelShift = PTAVecBase2f.emptyArray(1)
self.lastPixelShift = PTAVecBase2f.emptyArray(1)
self._setupFinalPass()
self._setShaderInputs()
# Give the gui a hint when the pipeline is done loading
if self.settings.displayOnscreenDebugger:
self.guiManager.onPipelineLoaded()
# add update task
self._attachUpdateTask()
def getForwardScene(self):
""" Reparent objects to this scene to use forward rendering.
Objects in this scene will directly get rendered, with no
lighting etc. applied. """
return self.forwardScene
def getTransparentScene(self):
""" Reparent objects to this scene to allow this objects to
have transparency. Objects in this scene will get directly
rendered and no lighting will get applied. """
return self.transparencyScene
def _createCombiner(self):
""" Creates the target which combines the result from the
lighting computation and last frame together
(Temporal Reprojection) """
self.combiner = RenderTarget("Combine-Temporal")
self.combiner.addColorTexture()
self.combiner.setColorBits(16)
self.combiner.prepareOffscreenBuffer()
self._setCombinerShader()
def _setupGlobalIllumination(self):
""" Creates the GI handler """
self.globalIllum = GlobalIllumination(self)
self.globalIllum.setup()
def _setupAntialiasing(self):
""" Creates the antialiasing technique """
technique = self.settings.antialiasingTechnique
self.debug("Creating antialiasing handler for", technique)
if technique == "None":
self.antialias = AntialiasingTechniqueNone()
elif technique == "SMAA":
self.antialias = AntialiasingTechniqueSMAA()
elif technique == "FXAA":
self.antialias = AntialiasingTechniqueFXAA()
else:
self.error("Unkown antialiasing technique", technique,
"-> using None:")
self.antialias = AntialiasingTechniqueNone()
if self.occlusion.requiresBlurring():
self.antialias.setColorTexture(
self.blurOcclusionH.getColorTexture())
else:
if self.haveCombiner and self.settings.enableTemporalReprojection:
self.antialias.setColorTexture(self.combiner.getColorTexture())
else:
self.antialias.setColorTexture(
self.lightingComputeContainer.getColorTexture())
self.antialias.setDepthTexture(self.deferredTarget.getDepthTexture())
self.antialias.setVelocityTexture(self.deferredTarget.getAuxTexture(1))
self.antialias.setup()
def _setupOcclusion(self):
""" Creates the occlusion technique """
technique = self.settings.occlusionTechnique
self.debug("Creating occlusion handle for", technique)
if technique == "None":
self.occlusion = AmbientOcclusionTechniqueNone()
elif technique == "SAO":
self.occlusion = AmbientOcclusionTechniqueSAO()
else:
self.error("Unkown occlusion technique:", technique)
self.occlusion = AmbientOcclusionTechniqueNone()
def _makeDeferredTargets(self):
""" Creates the multi-render-target """
self.debug("Creating deferred targets")
self.deferredTarget = RenderTarget("DeferredTarget")
self.deferredTarget.addColorAndDepth()
if self.haveMRT:
self.deferredTarget.addAuxTextures(3)
self.deferredTarget.setAuxBits(16)
self.deferredTarget.setColorBits(32)
self.deferredTarget.setDepthBits(32)
self.deferredTarget.prepareSceneRender()
def _setupFinalPass(self):
""" Setups the final pass which applies motion blur and so on """
# Set wrap for motion blur
colorTex = self.antialias.getResultTexture()
colorTex.setWrapU(Texture.WMClamp)
colorTex.setWrapV(Texture.WMClamp)
self._setFinalPassShader()
def _makeLightPerTileStorage(self):
""" Creates a texture to store the lights per tile into. Should
get replaced with ssbos later """
storageSizeX = self.precomputeSize.x * 8
storageSizeY = self.precomputeSize.y * 8
self.debug("Creating per tile storage of size", storageSizeX, "x",
storageSizeY)
self.lightPerTileStorage = Texture("LightsPerTile")
self.lightPerTileStorage.setup2dTexture(storageSizeX, storageSizeY,
Texture.TUnsignedShort,
Texture.FR32i)
self.lightPerTileStorage.setMinfilter(Texture.FTNearest)
self.lightPerTileStorage.setMagfilter(Texture.FTNearest)
def _creatGIPrecomputeBuffer(self):
""" Creates the half-resolution buffer which computes gi and gi
reflections. We use half-res for performance """
self.giPrecomputeBuffer = RenderTarget("GICompute")
self.giPrecomputeBuffer.setSize(self.size.x / 2, self.size.y / 2)
self.giPrecomputeBuffer.addColorTexture()
self.giPrecomputeBuffer.addAuxTextures(1)
self.giPrecomputeBuffer.setColorBits(16)
self.giPrecomputeBuffer.prepareOffscreenBuffer()
def _createLightingPipeline(self):
""" Creates the lighting pipeline, including shadow handling """
if not self.haveLightingPass:
self.debug("Skipping lighting pipeline")
return
self.debug("Creating lighting pipeline ..")
# size has to be a multiple of the compute unit size
# but still has to cover the whole screen
sizeX = int(math.ceil(float(self.size.x) / self.patchSize.x))
sizeY = int(math.ceil(float(self.size.y) / self.patchSize.y))
self.precomputeSize = LVecBase2i(sizeX, sizeY)
self.debug("Batch size =", sizeX, "x", sizeY, "Actual Buffer size=",
int(sizeX * self.patchSize.x), "x",
int(sizeY * self.patchSize.y))
self._makeLightPerTileStorage()
# Create a buffer which computes which light affects which tile
self._makeLightBoundsComputationBuffer(sizeX, sizeY)
# Create a buffer which applies the lighting
self._makeLightingComputeBuffer()
# Register for light manager
self.lightManager.setLightingComputator(self.lightingComputeContainer)
self.lightManager.setLightingCuller(self.lightBoundsComputeBuff)
self._loadFallbackCubemap()
self._loadLookupCubemap()
def _setShaderInputs(self):
""" Sets most of the required shader inputs to the targets """
# Shader inputs for the light-culling pass
if self.haveLightingPass:
self.lightBoundsComputeBuff.setShaderInput(
"destination", self.lightPerTileStorage)
self.lightBoundsComputeBuff.setShaderInput(
"depth", self.deferredTarget.getDepthTexture())
self.lightBoundsComputeBuff.setShaderInput("mainCam",
self.showbase.cam)
self.lightBoundsComputeBuff.setShaderInput("mainRender",
self.showbase.render)
# Shader inputs for the light-applying pass
self.lightingComputeContainer.setShaderInput(
"data0", self.deferredTarget.getColorTexture())
self.lightingComputeContainer.setShaderInput(
"data1", self.deferredTarget.getAuxTexture(0))
self.lightingComputeContainer.setShaderInput(
"data2", self.deferredTarget.getAuxTexture(1))
self.lightingComputeContainer.setShaderInput(
"data3", self.deferredTarget.getAuxTexture(2))
self.lightingComputeContainer.setShaderInput(
"depth", self.deferredTarget.getDepthTexture())
self.lightingComputeContainer.setShaderInput(
"mainCam", self.showbase.cam)
self.lightingComputeContainer.setShaderInput(
"mainRender", self.showbase.render)
if self.occlusion.requiresViewSpacePosNrm():
self.lightingComputeContainer.setShaderInput(
"viewSpaceNormals",
self.normalPrecompute.getColorTexture())
self.lightingComputeContainer.setShaderInput(
"viewSpacePosition",
self.normalPrecompute.getAuxTexture(0))
self.lightingComputeContainer.setShaderInput(
"shadowAtlas", self.lightManager.getAtlasTex())
if self.settings.useHardwarePCF:
self.lightingComputeContainer.setShaderInput(
"shadowAtlasPCF", self.lightManager.getAtlasTex(),
self.lightManager.getPCFSampleState())
self.lightingComputeContainer.setShaderInput(
"destination", self.lightingComputeCombinedTex)
self.lightingComputeContainer.setShaderInput(
"temporalProjXOffs", self.temporalProjXOffs)
self.lightingComputeContainer.setShaderInput(
"cameraPosition", self.cameraPosition)
self.lightingComputeContainer.setShaderInput(
"noiseTexture",
self.showbase.loader.loadTexture(
"Data/Occlusion/noise4x4.png"))
self.lightingComputeContainer.setShaderInput(
"lightsPerTile", self.lightPerTileStorage)
if self.settings.enableGlobalIllumination:
self.lightingComputeContainer.setShaderInput(
"giDiffuseTex", self.giPrecomputeBuffer.getColorTexture())
self.lightingComputeContainer.setShaderInput(
"giReflectionTex",
self.giPrecomputeBuffer.getAuxTexture(0))
# Shader inputs for the occlusion blur passes
if self.occlusion.requiresBlurring() and self.haveCombiner:
self.blurOcclusionH.setShaderInput(
"colorTex", self.blurOcclusionV.getColorTexture())
if self.settings.enableTemporalReprojection:
self.blurOcclusionV.setShaderInput(
"colorTex", self.combiner.getColorTexture())
else:
self.blurOcclusionV.setShaderInput(
"colorTex",
self.lightingComputeContainer.getColorTexture())
self.blurOcclusionH.setShaderInput(
"normalTex", self.deferredTarget.getAuxTexture(0))
self.blurOcclusionV.setShaderInput(
"normalTex", self.deferredTarget.getAuxTexture(0))
self.blurOcclusionH.setShaderInput(
"normalsView", self.normalPrecompute.getAuxTexture(0))
self.blurOcclusionV.setShaderInput(
"normalsView", self.normalPrecompute.getAuxTexture(0))
# Shader inputs for the blur passes
if self.blurEnabled:
self.blurColorH.setShaderInput("dofStorage", self.dofStorage)
self.blurColorV.setShaderInput("dofStorage", self.dofStorage)
self.blurColorH.setShaderInput("colorTex",
self.antialias.getResultTexture())
self.blurColorH.setShaderInput(
"depthTex", self.deferredTarget.getDepthTexture())
self.blurColorV.setShaderInput("colorTex",
self.blurColorH.getColorTexture())
# Shader inputs for the temporal reprojection
if self.haveCombiner and self.settings.enableTemporalReprojection:
self.combiner.setShaderInput(
"currentComputation",
self.lightingComputeContainer.getColorTexture())
self.combiner.setShaderInput("lastFrame",
self.lightingComputeCombinedTex)
self.combiner.setShaderInput("positionBuffer",
self.deferredTarget.getColorTexture())
self.combiner.setShaderInput("velocityBuffer",
self.deferredTarget.getAuxTexture(1))
self.combiner.setShaderInput("currentPixelShift",
self.currentPixelShift)
self.combiner.setShaderInput("lastPixelShift", self.lastPixelShift)
if self.blurEnabled:
self.combiner.setShaderInput("dofStorage", self.dofStorage)
self.combiner.setShaderInput("depthTex",
self.deferredTarget.getDepthTexture())
self.combiner.setShaderInput("lastPosition",
self.lastPositionBuffer)
self.combiner.setShaderInput("temporalProjXOffs",
self.temporalProjXOffs)
self.combiner.setShaderInput("lastMVP", self.lastMVP)
self.combiner.setShaderInput("cameraPosition", self.cameraPosition)
self.combiner.setShaderInput("currentMVP", self.lastMVP)
# Shader inputs for the final pass
if self.blurEnabled:
self.deferredTarget.setShaderInput(
"colorTex", self.blurColorV.getColorTexture())
else:
self.deferredTarget.setShaderInput(
"colorTex", self.antialias.getResultTexture())
if self.occlusion.requiresBlurring():
self.normalPrecompute.setShaderInput(
"positionTex", self.deferredTarget.getColorTexture())
self.normalPrecompute.setShaderInput("mainCam", self.showbase.cam)
self.normalPrecompute.setShaderInput("mainRender",
self.showbase.render)
self.normalPrecompute.setShaderInput(
"depthTex", self.deferredTarget.getDepthTexture())
if self.haveMRT:
self.deferredTarget.setShaderInput(
"velocityTex", self.deferredTarget.getAuxTexture(1))
self.deferredTarget.setShaderInput(
"depthTex", self.deferredTarget.getDepthTexture())
self.deferredTarget.setShaderInput("motionBlurFactor",
self.motionBlurFactor)
if self.haveLightingPass:
self.deferredTarget.setShaderInput("lastFrame",
self.lightingComputeCombinedTex)
if self.haveCombiner and self.settings.enableTemporalReprojection:
self.deferredTarget.setShaderInput("newFrame",
self.combiner.getColorTexture())
self.deferredTarget.setShaderInput("lastPosition",
self.lastPositionBuffer)
self.deferredTarget.setShaderInput("debugTex",
self.combiner.getColorTexture())
else:
self.deferredTarget.setShaderInput(
"debugTex", self.antialias.getResultTexture())
self.deferredTarget.setShaderInput(
"currentPosition", self.deferredTarget.getColorTexture())
# Set last / current mvp handles
self.showbase.render.setShaderInput("lastMVP", self.lastMVP)
# Set GI inputs
if self.settings.enableGlobalIllumination:
self.globalIllum.bindTo(self.giPrecomputeBuffer, "giData")
self.giPrecomputeBuffer.setShaderInput(
"data0", self.deferredTarget.getColorTexture())
self.giPrecomputeBuffer.setShaderInput(
"data1", self.deferredTarget.getAuxTexture(0))
self.giPrecomputeBuffer.setShaderInput(
"data2", self.deferredTarget.getAuxTexture(1))
self.giPrecomputeBuffer.setShaderInput(
"data3", self.deferredTarget.getAuxTexture(2))
self.giPrecomputeBuffer.setShaderInput("cameraPosition",
self.cameraPosition)
# Finally, set shaders
self.reloadShaders()
def _loadFallbackCubemap(self):
""" Loads the cubemap for image based lighting """
print self.settings.defaultReflectionCubemap
cubemap = self.showbase.loader.loadCubeMap(
self.settings.defaultReflectionCubemap)
cubemap.setMinfilter(Texture.FTLinearMipmapLinear)
cubemap.setMagfilter(Texture.FTLinearMipmapLinear)
cubemap.setFormat(Texture.F_srgb)
print math.log(cubemap.getXSize(), 2)
self.lightingComputeContainer.setShaderInput("fallbackCubemap",
cubemap)
self.lightingComputeContainer.setShaderInput(
"fallbackCubemapMipmaps", math.log(cubemap.getXSize(), 2))
def _loadLookupCubemap(self):
self.debug("Loading lookup cubemap")
cubemap = self.showbase.loader.loadCubeMap(
"Data/Cubemaps/DirectionLookup/#.png")
cubemap.setMinfilter(Texture.FTNearest)
cubemap.setMagfilter(Texture.FTNearest)
cubemap.setFormat(Texture.F_rgb8)
self.lightingComputeContainer.setShaderInput("directionToFace",
cubemap)
def _makeLightBoundsComputationBuffer(self, w, h):
""" Creates the buffer which precomputes the lights per tile """
self.debug("Creating light precomputation buffer of size", w, "x", h)
self.lightBoundsComputeBuff = RenderTarget("ComputeLightTileBounds")
self.lightBoundsComputeBuff.setSize(w, h)
self.lightBoundsComputeBuff.setColorWrite(False)
self.lightBoundsComputeBuff.prepareOffscreenBuffer()
def _makeLightingComputeBuffer(self):
""" Creates the buffer which applies the lighting """
self.lightingComputeContainer = RenderTarget("ComputeLighting")
if self.settings.enableTemporalReprojection:
self.lightingComputeContainer.setSize(self.size.x / 2, self.size.y)
else:
self.lightingComputeContainer.setSize(self.size.x, self.size.y)
self.lightingComputeContainer.addColorTexture()
self.lightingComputeContainer.setColorBits(16)
self.lightingComputeContainer.prepareOffscreenBuffer()
self.lightingComputeCombinedTex = Texture("Lighting-Compute-Combined")
self.lightingComputeCombinedTex.setup2dTexture(self.size.x,
self.size.y,
Texture.TFloat,
Texture.FRgba8)
self.lightingComputeCombinedTex.setMinfilter(Texture.FTLinear)
self.lightingComputeCombinedTex.setMagfilter(Texture.FTLinear)
self.lastPositionBuffer = Texture("Last-Position-Buffer")
self.lastPositionBuffer.setup2dTexture(self.size.x, self.size.y,
Texture.TFloat, Texture.FRgba16)
self.lastPositionBuffer.setMinfilter(Texture.FTNearest)
self.lastPositionBuffer.setMagfilter(Texture.FTNearest)
def _createOcclusionBlurBuffer(self):
""" Creates the buffers needed to blur the occlusion """
self.blurOcclusionV = RenderTarget("blurOcclusionVertical")
self.blurOcclusionV.addColorTexture()
self.blurOcclusionV.prepareOffscreenBuffer()
self.blurOcclusionH = RenderTarget("blurOcclusionHorizontal")
self.blurOcclusionH.addColorTexture()
self.blurOcclusionH.prepareOffscreenBuffer()
# Mipmaps for blur?
# self.blurOcclusionV.getColorTexture().setMinfilter(
# Texture.FTLinearMipmapLinear)
# self.combiner.getColorTexture().setMinfilter(
# Texture.FTLinearMipmapLinear)
def _createBlurBuffer(self):
""" Creates the buffers for the dof """
self.blurColorV = RenderTarget("blurColorVertical")
self.blurColorV.addColorTexture()
self.blurColorV.prepareOffscreenBuffer()
self.blurColorH = RenderTarget("blurColorHorizontal")
self.blurColorH.addColorTexture()
self.blurColorH.prepareOffscreenBuffer()
# self.blurColorH.getColorTexture().setMinfilter(
# Texture.FTLinearMipmapLinear)
# self.antialias.getResultTexture().setMinfilter(
# Texture.FTLinearMipmapLinear)
def _createNormalPrecomputeBuffer(self):
""" Creates a buffer which reconstructs the normals and position
from view-space """
self.normalPrecompute = RenderTarget("PrecomputeNormals")
self.normalPrecompute.addColorTexture()
self.normalPrecompute.addAuxTextures(1)
self.normalPrecompute.setColorBits(16)
self.normalPrecompute.setAuxBits(16)
self.normalPrecompute.prepareOffscreenBuffer()
def _createDofStorage(self):
""" Creates the texture where the dof factor is stored in, so we
don't recompute it each pass """
self.dofStorage = Texture("DOFStorage")
self.dofStorage.setup2dTexture(self.size.x, self.size.y,
Texture.TFloat, Texture.FRg16)
def _setOcclusionBlurShader(self):
""" Sets the shaders which blur the occlusion """
blurVShader = BetterShader.load(
"Shader/DefaultPostProcess.vertex",
"Shader/BlurOcclusionVertical.fragment")
blurHShader = BetterShader.load(
"Shader/DefaultPostProcess.vertex",
"Shader/BlurOcclusionHorizontal.fragment")
self.blurOcclusionV.setShader(blurVShader)
self.blurOcclusionH.setShader(blurHShader)
def _setGIComputeShader(self):
""" Sets the shader which computes the GI """
giShader = BetterShader.load("Shader/DefaultPostProcess.vertex",
"Shader/ComputeGI.fragment")
self.giPrecomputeBuffer.setShader(giShader)
def _setBlurShader(self):
""" Sets the shaders which blur the color """
blurVShader = BetterShader.load("Shader/DefaultPostProcess.vertex",
"Shader/BlurVertical.fragment")
blurHShader = BetterShader.load("Shader/DefaultPostProcess.vertex",
"Shader/BlurHorizontal.fragment")
self.blurColorV.setShader(blurVShader)
self.blurColorH.setShader(blurHShader)
def _setLightingShader(self):
""" Sets the shader which applies the light """
lightShader = BetterShader.load("Shader/DefaultPostProcess.vertex",
"Shader/ApplyLighting.fragment")
self.lightingComputeContainer.setShader(lightShader)
def _setCombinerShader(self):
""" Sets the shader which combines the lighting with the previous frame
(temporal reprojection) """
cShader = BetterShader.load("Shader/DefaultPostProcess.vertex",
"Shader/Combiner.fragment")
self.combiner.setShader(cShader)
def _setPositionComputationShader(self):
""" Sets the shader which computes the lights per tile """
pcShader = BetterShader.load("Shader/DefaultPostProcess.vertex",
"Shader/PrecomputeLights.fragment")
self.lightBoundsComputeBuff.setShader(pcShader)
def _setFinalPassShader(self):
""" Sets the shader which computes the final frame,
with motion blur and so on """
fShader = BetterShader.load("Shader/DefaultPostProcess.vertex",
"Shader/Final.fragment")
self.deferredTarget.setShader(fShader)
def _getSize(self):
""" Returns the window size. """
return LVecBase2i(self.showbase.win.getXSize(),
self.showbase.win.getYSize())
def reloadShaders(self):
""" Reloads all shaders """
if self.haveLightingPass:
self.lightManager.debugReloadShader()
self._setPositionComputationShader()
self._setLightingShader()
if self.haveCombiner and self.settings.enableTemporalReprojection:
self._setCombinerShader()
self._setFinalPassShader()
if self.settings.enableGlobalIllumination:
self._setGIComputeShader()
if self.occlusion.requiresBlurring():
self._setOcclusionBlurShader()
if self.blurEnabled:
self._setBlurShader()
if self.occlusion.requiresViewSpacePosNrm():
self._setNormalExtractShader()
self.antialias.reloadShader()
if self.settings.enableGlobalIllumination:
self.globalIllum.reloadShader()
def _setNormalExtractShader(self):
""" Sets the shader which constructs the normals from position """
npShader = BetterShader.load("Shader/DefaultPostProcess.vertex",
"Shader/ExtractNormals.fragment")
self.normalPrecompute.setShader(npShader)
def _attachUpdateTask(self):
""" Attaches the update tasks to the showbase """
self.showbase.addTask(self._preRenderCallback,
"RP_BeforeRender",
sort=-5000)
self.showbase.addTask(self._update, "RP_Update", sort=-10)
if self.haveLightingPass:
self.showbase.addTask(self._updateLights,
"RP_UpdateLights",
sort=-9)
self.showbase.addTask(self._updateShadows,
"RP_UpdateShadows",
sort=-8)
self.showbase.addTask(self._processShadowCallbacks,
"RP_ShadowCallbacks",
sort=-5)
if self.settings.displayOnscreenDebugger:
self.showbase.addTask(self._updateGUI, "RP_UpdateGUI", sort=7)
self.showbase.addTask(self._postRenderCallback,
"RP_AfterRender",
sort=5000)
def _preRenderCallback(self, task=None):
""" Called before rendering """
if self.settings.enableGlobalIllumination:
self.globalIllum.process()
self.antialias.preRenderUpdate()
if task is not None:
return task.cont
def _postRenderCallback(self, task=None):
""" Called after rendering """
self.antialias.postRenderUpdate()
if task is not None:
return task.cont
def _computeCameraBounds(self):
""" Computes the current camera bounds, i.e. for light culling """
cameraBounds = self.camera.node().getLens().makeBounds()
cameraBounds.xform(self.camera.getMat(self.showbase.render))
return cameraBounds
def _updateLights(self, task=None):
""" Task which updates/culls the lights """
self.lightManager.updateLights()
if task is not None:
return task.cont
def _processShadowCallbacks(self, task=None):
self.lightManager.processCallbacks()
if task is not None:
return task.cont
def _updateShadows(self, task=None):
""" Task which updates the shadow maps """
self.lightManager.updateShadows()
if task is not None:
return task.cont
def _updateGUI(self, task=None):
""" Task which updates the onscreen gui debugger """
self.guiManager.update()
if task is not None:
return task.cont
def _update(self, task=None):
""" Main update task """
self.currentShiftIndex[0] = 1 - self.currentShiftIndex[0]
currentFPS = 1.0 / self.showbase.taskMgr.globalClock.getDt()
self.temporalProjXOffs[0] = 1 - self.temporalProjXOffs[0]
self.cameraPosition[0] = self.showbase.cam.getPos(self.showbase.render)
self.motionBlurFactor[0] = min(
1.5, currentFPS / 60.0) * self.settings.motionBlurFactor
self.cullBounds = self._computeCameraBounds()
if self.haveLightingPass:
self.lightManager.setCullBounds(self.cullBounds)
self.lastMVP[0] = self.currentMVP[0]
self.currentMVP[0] = self._computeMVP()
shift = self.pixelShifts[self.currentShiftIndex[0]]
self.lastPixelShift[0] = self.currentPixelShift[0]
self.currentPixelShift[0] = shift
Globals.base.camLens.setFilmOffset(shift.x, shift.y)
if task is not None:
return task.cont
def _computeMVP(self):
""" Computes the current mvp. Actually, this is the
worldViewProjectionMatrix, but for convience it's called mvp. """
camLens = self.showbase.camLens
projMat = Mat4.convertMat(
CSYupRight,
camLens.getCoordinateSystem()) * camLens.getProjectionMat()
transformMat = TransformState.makeMat(
Mat4.convertMat(
self.showbase.win.getGsg().getInternalCoordinateSystem(),
CSZupRight))
modelViewMat = transformMat.invertCompose(
self.showbase.render.getTransform(self.showbase.cam)).getMat()
return UnalignedLMatrix4f(modelViewMat * projMat)
def getLightManager(self):
""" Returns a handle to the light manager """
return self.lightManager
def getDefaultObjectShader(self, tesselated=False):
""" Returns the default shader for objects """
if not tesselated:
shader = BetterShader.load(
"Shader/DefaultObjectShader/vertex.glsl",
"Shader/DefaultObjectShader/fragment.glsl")
else:
self.warn("Tesselation is only experimental! Remember "
"to convert the geometry to patches first!")
shader = BetterShader.load(
"Shader/DefaultObjectShader/vertex.glsl",
"Shader/DefaultObjectShader/fragment.glsl", "",
"Shader/DefaultObjectShader/tesscontrol.glsl",
"Shader/DefaultObjectShader/tesseval.glsl")
return shader
def _getDeferredBuffer(self):
""" Returns a handle to the internal deferred target """
return self.deferredTarget.getInternalBuffer()
def addLight(self, light):
""" Adds a light to the list of rendered lights """
if self.haveLightingPass:
self.lightManager.addLight(light)
else:
self.warn("Lighting is disabled, so addLight has no effect")
def setScattering(self, scatteringModel):
""" Sets a scattering model to use. Only has an effect if enableScattering
is enabled """
self.debug("Loading scattering model ..")
if not self.settings.enableScattering:
self.error("You cannot set a scattering model as scattering is not"
" enabled in your pipeline.ini!")
return
self.lightingComputeContainer.setShaderInput(
"transmittanceSampler", scatteringModel.getTransmittanceResult())
self.lightingComputeContainer.setShaderInput(
"inscatterSampler", scatteringModel.getInscatterTexture())
scatteringModel.bindTo(self.lightingComputeContainer,
"scatteringOptions")
def enableDefaultEarthScattering(self):
""" Adds a standard scattering model, representing the atmosphere of
the earth. This is a shortcut for creating a Scattering instance and
precomputing it """
earthScattering = Scattering()
scale = 1000000000
earthScattering.setSettings({
"atmosphereOffset":
Vec3(0, 0, -(6360.0 + 9.5) * scale),
"atmosphereScale":
Vec3(scale)
})
earthScattering.precompute()
self.setScattering(earthScattering)
def setGILightSource(self, light):
""" Sets the light source for the global illumination. The GI uses this
light to shade the voxels, so this light is the only light which "casts"
global illumination. When GI is disabled, this has no effect """
if self.settings.enableGlobalIllumination:
self.globalIllum.setTargetLight(light)
def _generateShaderConfiguration(self):
""" Genrates the global shader include which defines
most values used in the shaders. """
self.debug("(Re)Generating shader configuration")
# Generate list of defines
defines = []
if self.settings.antialiasingTechnique == "SMAA":
quality = self.settings.smaaQuality.upper()
if quality in ["LOW", "MEDIUM", "HIGH", "ULTRA"]:
defines.append(("SMAA_PRESET_" + quality, ""))
else:
self.error("Unrecognized SMAA quality:", quality)
return
defines.append(
("LIGHTING_COMPUTE_PATCH_SIZE_X", self.settings.computePatchSizeX))
defines.append(
("LIGHTING_COMPUTE_PATCH_SIZE_Y", self.settings.computePatchSizeY))
defines.append(("LIGHTING_MIN_MAX_DEPTH_ACCURACY",
self.settings.minMaxDepthAccuracy))
if self.blurEnabled:
defines.append(("USE_DOF", 1))
if self.settings.useSimpleLighting:
defines.append(("USE_SIMPLE_LIGHTING", 1))
if self.settings.anyLightBoundCheck:
defines.append(("LIGHTING_ANY_BOUND_CHECK", 1))
if self.settings.accurateLightBoundCheck:
defines.append(("LIGHTING_ACCURATE_BOUND_CHECK", 1))
if self.settings.renderShadows:
defines.append(("USE_SHADOWS", 1))
defines.append(
("AMBIENT_CUBEMAP_SAMPLES", self.settings.ambientCubemapSamples))
defines.append(
("SHADOW_MAP_ATLAS_SIZE", self.settings.shadowAtlasSize))
defines.append(("SHADOW_MAX_UPDATES_PER_FRAME",
self.settings.maxShadowUpdatesPerFrame))
defines.append(("SHADOW_GEOMETRY_MAX_VERTICES",
self.settings.maxShadowUpdatesPerFrame * 3))
defines.append(("SHADOW_NUM_PCF_SAMPLES", self.settings.numPCFSamples))
defines.append(("SHADOW_NUM_PCSS_SEARCH_SAMPLES",
self.settings.numPCSSSearchSamples))
defines.append(("SHADOW_NUM_PCSS_FILTER_SAMPLES",
self.settings.numPCSSFilterSamples))
defines.append(("SHADOW_PSSM_BORDER_PERCENTAGE",
self.settings.shadowCascadeBorderPercentage))
if self.settings.useHardwarePCF:
defines.append(("USE_HARDWARE_PCF", 1))
defines.append(("WINDOW_WIDTH", self.size.x))
defines.append(("WINDOW_HEIGHT", self.size.y))
if self.settings.motionBlurEnabled:
defines.append(("USE_MOTION_BLUR", 1))
defines.append(
("MOTION_BLUR_SAMPLES", self.settings.motionBlurSamples))
# Occlusion
defines.append(
("OCCLUSION_TECHNIQUE_" + self.occlusion.getIncludeName(), 1))
defines.append(("OCCLUSION_RADIUS", self.settings.occlusionRadius))
defines.append(("OCCLUSION_STRENGTH", self.settings.occlusionStrength))
defines.append(
("OCCLUSION_SAMPLES", self.settings.occlusionSampleCount))
if self.settings.displayOnscreenDebugger:
defines.append(("DEBUGGER_ACTIVE", 1))
extraSettings = self.guiManager.getDefines()
defines += extraSettings
if self.settings.enableTemporalReprojection:
defines.append(("USE_TEMPORAL_REPROJECTION", 1))
if self.settings.enableGlobalIllumination:
defines.append(("USE_GLOBAL_ILLUMINATION", 1))
if self.settings.enableScattering:
defines.append(("USE_SCATTERING", 1))
# Pass near far
defines.append(("CAMERA_NEAR", Globals.base.camLens.getNear()))
defines.append(("CAMERA_FAR", Globals.base.camLens.getFar()))
# Generate
output = "// Autogenerated by RenderingPipeline.py\n"
output += "// Do not edit! Your changes will be lost.\n\n"
for key, value in defines:
output += "#define " + key + " " + str(value) + "\n"
# Try to write the file
try:
with open("PipelineTemp/ShaderAutoConfig.include", "w") as handle:
handle.write(output)
except Exception, msg:
self.fatal(
"Error writing shader autoconfig. Maybe no write-access?")
return
class LightManager(DebugObject):
""" This class is internally used by the RenderingPipeline to handle
Lights and their Shadows. It stores a list of lights, and updates the
required ShadowSources per frame. There are two main update methods:
updateLights processes each light and does a basic frustum check.
If the light is in the frustum, its ID is passed to the light precompute
container (set with setLightingCuller). Also, each shadowSource of
the light is checked, and if it reports to be invalid, it's queued to
the list of queued shadow updates.
updateShadows processes the queued shadow updates and setups everything
to render the shadow depth textures to the shadow atlas.
Lights can be added with addLight. Notice you cannot change the shadow
resolution or even wether the light casts shadows after you called addLight.
This is because it might already have a position in the atlas, and so
the atlas would have to delete it's map, which is not supported (yet).
This shouldn't be an issue, as you usually always know before if a
light will cast shadows or not.
"""
def __init__(self, pipeline):
""" Creates a new LightManager. It expects a RenderPipeline as parameter. """
DebugObject.__init__(self, "LightManager")
self._initArrays()
self.pipeline = pipeline
self.settings = pipeline.getSettings()
# Create arrays to store lights & shadow sources
self.lights = []
self.shadowSources = []
self.queuedShadowUpdates = []
self.allLightsArray = ShaderStructArray(Light, self.maxTotalLights)
self.updateCallbacks = []
self.cullBounds = None
self.shadowScene = Globals.render
# Create atlas
self.shadowAtlas = ShadowAtlas()
self.shadowAtlas.setSize(self.settings.shadowAtlasSize)
self.shadowAtlas.create()
self.maxShadowMaps = 24
self.maxShadowUpdatesPerFrame = self.settings.maxShadowUpdatesPerFrame
self.numShadowUpdatesPTA = PTAInt.emptyArray(1)
self.updateShadowsArray = ShaderStructArray(
ShadowSource, self.maxShadowUpdatesPerFrame)
self.allShadowsArray = ShaderStructArray(
ShadowSource, self.maxShadowMaps)
# Create shadow compute buffer
self._createShadowComputationBuffer()
# Create the initial shadow state
self.shadowComputeCamera.setTagStateKey("ShadowPassShader")
# self.shadowComputeCamera.setInitialState(RenderState.make(
# ColorWriteAttrib.make(ColorWriteAttrib.C_off),
# ColorWriteAttrib.make(ColorWriteAttrib.C_rgb),
# DepthWriteAttrib.make(DepthWriteAttrib.M_on),
# CullFaceAttrib.make(CullFaceAttrib.MCullNone),
# 100))
self._createTagStates()
self.shadowScene.setTag("ShadowPassShader", "Default")
# Create debug overlay
self._createDebugTexts()
# Disable buffer on start
self.shadowComputeTarget.setActive(False)
# Bind arrays
self.updateShadowsArray.bindTo(self.shadowScene, "updateSources")
self.updateShadowsArray.bindTo(
self.shadowComputeTarget, "updateSources")
# Set initial inputs
for target in [self.shadowComputeTarget, self.shadowScene]:
target.setShaderInput("numUpdates", self.numShadowUpdatesPTA)
self.lightingComputator = None
self.lightCuller = None
self.skip = 0
self.skipRate = 0
def _createTagStates(self):
# Create shadow caster shader
self.shadowCasterShader = BetterShader.load(
"Shader/DefaultShadowCaster/vertex.glsl",
"Shader/DefaultShadowCaster/fragment.glsl",
"Shader/DefaultShadowCaster/geometry.glsl")
initialState = NodePath("ShadowCasterState")
initialState.setShader(self.shadowCasterShader, 30)
self.shadowComputeCamera.setTagState(
"Default", initialState.getState())
def _createShadowComputationBuffer(self):
""" This creates the internal shadow buffer which also is the
shadow atlas. Shadow maps are rendered to this using Viewports
(thank you rdb for adding this!). It also setups the base camera
which renders the shadow objects, although a custom mvp is passed
to the shaders, so the camera is mainly a dummy """
# Create camera showing the whole scene
self.shadowComputeCamera = Camera("ShadowComputeCamera")
self.shadowComputeCameraNode = self.shadowScene.attachNewNode(
self.shadowComputeCamera)
self.shadowComputeCamera.getLens().setFov(90, 90)
self.shadowComputeCamera.getLens().setNearFar(10.0, 100000.0)
# Disable culling
self.shadowComputeCamera.setBounds(OmniBoundingVolume())
self.shadowComputeCameraNode.setPos(0, 0, 150)
self.shadowComputeCameraNode.lookAt(0, 0, 0)
self.shadowComputeTarget = RenderTarget("ShadowAtlas")
self.shadowComputeTarget.setSize(self.shadowAtlas.getSize())
self.shadowComputeTarget.addDepthTexture()
self.shadowComputeTarget.setDepthBits(32)
if self.settings.enableGlobalIllumination:
self.shadowComputeTarget.addColorTexture()
self.shadowComputeTarget.setColorBits(16)
self.shadowComputeTarget.setSource(
self.shadowComputeCameraNode, Globals.base.win)
self.shadowComputeTarget.prepareSceneRender()
# This took me a long time to figure out. If not removing the quad
# children, the color and aux buffers will be overridden each frame.
# Quite annoying!
self.shadowComputeTarget.getQuad().node().removeAllChildren()
self.shadowComputeTarget.getInternalRegion().setSort(-200)
self.shadowComputeTarget.getInternalRegion().setNumRegions(
self.maxShadowUpdatesPerFrame + 1)
self.shadowComputeTarget.getInternalRegion().setDimensions(0,
(0, 0, 0, 0))
self.shadowComputeTarget.getInternalBuffer().setSort(-300)
# We can't clear the depth per viewport.
# But we need to clear it in any way, as we still want
# z-testing in the buffers. So well, we create a
# display region *below* (smaller sort value) each viewport
# which has a depth-clear assigned. This is hacky, I know.
self.depthClearer = []
for i in range(self.maxShadowUpdatesPerFrame):
buff = self.shadowComputeTarget.getInternalBuffer()
dr = buff.makeDisplayRegion()
dr.setSort(-250)
for k in xrange(16):
dr.setClearActive(k, True)
dr.setClearValue(k, Vec4(0.5,0.5,0.5,1))
dr.setClearDepthActive(True)
dr.setClearDepth(1.0)
dr.setDimensions(0,0,0,0)
dr.setActive(False)
self.depthClearer.append(dr)
# When using hardware pcf, set the correct filter types
dTex = self.shadowComputeTarget.getDepthTexture()
if self.settings.useHardwarePCF:
dTex.setMinfilter(Texture.FTShadow)
dTex.setMagfilter(Texture.FTShadow)
dTex.setWrapU(Texture.WMClamp)
dTex.setWrapV(Texture.WMClamp)
def getAllLights(self):
""" Returns all attached lights """
return self.lights
def _createDebugTexts(self):
""" Creates a debug overlay if specified in the pipeline settings """
self.lightsVisibleDebugText = None
self.lightsUpdatedDebugText = None
if self.settings.displayDebugStats:
try:
from Code.GUI.FastText import FastText
self.lightsVisibleDebugText = FastText(pos=Vec2(
Globals.base.getAspectRatio() - 0.1, 0.84), rightAligned=True, color=Vec3(1, 1, 0), size=0.036)
self.lightsUpdatedDebugText = FastText(pos=Vec2(
Globals.base.getAspectRatio() - 0.1, 0.8), rightAligned=True, color=Vec3(1, 1, 0), size=0.036)
except Exception, msg:
self.debug(
"Overlay is disabled because FastText wasn't loaded")
class RenderingPipeline(DebugObject):
def __init__(self, showbase):
DebugObject.__init__(self, "RenderingPipeline")
self.showbase = showbase
self.lightManager = LightManager()
self.size = self._getSize()
self.precomputeSize = Vec2(0)
self.camera = base.cam
self.cullBounds = None
self.patchSize = Vec2(32, 32)
self.temporalProjXOffs = 0
self.temporalProjFactor = 2
self.forwardScene = NodePath("Forward Rendering")
self.lastMVP = None
self._setup()
def _setup(self):
self.debug("Setting up render pipeline")
# First, we need no transparency
render.setAttrib(TransparencyAttrib.make(TransparencyAttrib.MNone),
100)
# Now create deferred render buffers
self._makeDeferredTargets()
# Setup compute shader for lighting
self._createLightingPipeline()
# Setup combiner
self._createCombiner()
self.deferredTarget.setShader(
BetterShader.load("Shader/DefaultPostProcess.vertex",
"Shader/TextureDisplay.fragment"))
self._setupAntialiasing()
self._createFinalPass()
self.antialias.getFirstBuffer().setShaderInput(
"lastFrame", self.lightingComputeCombinedTex)
self.antialias.getFirstBuffer().setShaderInput("lastPosition",
self.lastPositionBuffer)
self.antialias.getFirstBuffer().setShaderInput(
"currentPosition", self.deferredTarget.getColorTexture())
# self.deferredTarget.setShaderInput("sampler", self.lightingComputeCombinedTex)
# self.deferredTarget.setShaderInput("sampler", self.antialias.getResultTexture())
self.deferredTarget.setShaderInput("sampler",
self.finalPass.getColorTexture())
# self.deferredTarget.setShaderInput("sampler", self.combiner.getColorTexture())
# self.deferredTarget.setShaderInput("sampler", self.lightingComputeCombinedTex)
# self.deferredTarget.setShaderInput("sampler", self.antialias._neighborBuffer.getColorTexture())
# self.deferredTarget.setShaderInput("sampler", self.antialias._blendBuffer.getColorTexture())
# self.deferredTarget.setShaderInput("sampler", self.lightingComputeCombinedTex)
# add update task
self._attachUpdateTask()
# compute first mvp
self._computeMVP()
self.lastLastMVP = self.lastMVP
# DirectFrame(frameColor=(1, 1, 1, 0.2), frameSize=(-0.28, 0.28, -0.27, 0.4), pos=(base.getAspectRatio() - 0.35, 0.0, 0.49))
self.atlasDisplayImage = OnscreenImage(
image=self.lightManager.getAtlasTex(),
pos=(base.getAspectRatio() - 0.35, 0, 0.5),
scale=(0.25, 0, 0.25))
self.lastPosImage = OnscreenImage(
image=self.lightingComputeCombinedTex,
pos=(base.getAspectRatio() - 0.35, 0, -0.05),
scale=(0.25, 0, 0.25))
# self.atlasDisplayImage = OnscreenImage(image = self.lightManager.getAtlasTex(), pos = (0,0,0), scale=(0.8,1,0.8))
# self.atlasDisplayImage = OnscreenImage(image = self.lightPerTileStorage, pos = (base.getAspectRatio() - 0.35, 0, 0.5), scale=(0.25,0,0.25))
def _createCombiner(self):
self.combiner = RenderTarget("Combine-Temporal")
self.combiner.setColorBits(8)
self.combiner.addRenderTexture(RenderTargetType.Color)
self.combiner.prepareOffscreenBuffer()
self.combiner.setShaderInput(
"currentComputation",
self.lightingComputeContainer.getColorTexture())
self.combiner.setShaderInput("lastFrame",
self.lightingComputeCombinedTex)
self.combiner.setShaderInput("positionBuffer",
self.deferredTarget.getColorTexture())
self.combiner.setShaderInput("velocityBuffer",
self.deferredTarget.getAuxTexture(1))
self.combiner.setShaderInput("lastPosition", self.lastPositionBuffer)
self._setCombinerShader()
def _setupAntialiasing(self):
self.debug("Creating antialiasing handler ..")
self.antialias = Antialiasing()
# self.antialias.setColorTexture(self.lightingComputeContainer.getColorTexture())
self.antialias.setColorTexture(self.combiner.getColorTexture())
self.antialias.setDepthTexture(self.deferredTarget.getDepthTexture())
self.antialias.setup()
# Creates all the render targets
def _makeDeferredTargets(self):
self.debug("Creating deferred targets")
self.deferredTarget = RenderTarget("DeferredTarget")
self.deferredTarget.addRenderTexture(RenderTargetType.Color)
self.deferredTarget.addRenderTexture(RenderTargetType.Depth)
self.deferredTarget.addRenderTexture(RenderTargetType.Aux0)
self.deferredTarget.addRenderTexture(RenderTargetType.Aux1)
self.deferredTarget.setAuxBits(16)
self.deferredTarget.setColorBits(16)
self.deferredTarget.setDepthBits(32)
# self.deferredTarget.setSize(400, 240) # check for overdraw
self.deferredTarget.prepareSceneRender()
def _createFinalPass(self):
self.debug("Creating final pass")
self.finalPass = RenderTarget("FinalPass")
self.finalPass.addRenderTexture(RenderTargetType.Color)
self.finalPass.prepareOffscreenBuffer()
colorTex = self.antialias.getResultTexture()
# Set wrap for motion blur
colorTex.setWrapU(Texture.WMMirror)
colorTex.setWrapV(Texture.WMMirror)
self.finalPass.setShaderInput("colorTex", colorTex)
self.finalPass.setShaderInput("velocityTex",
self.deferredTarget.getAuxTexture(1))
self.finalPass.setShaderInput("depthTex",
self.deferredTarget.getDepthTexture())
self._setFinalPassShader()
# Creates the storage to store the list of visible lights per tile
def _makeLightPerTileStorage(self):
storageSizeX = int(self.precomputeSize.x * 8)
storageSizeY = int(self.precomputeSize.y * 8)
self.debug("Creating per tile storage of size", storageSizeX, "x",
storageSizeY)
self.lightPerTileStorage = Texture("LightsPerTile")
self.lightPerTileStorage.setup2dTexture(storageSizeX, storageSizeY,
Texture.TUnsignedShort,
Texture.FR32i)
self.lightPerTileStorage.setMinfilter(Texture.FTNearest)
self.lightPerTileStorage.setMagfilter(Texture.FTNearest)
# Inits the lighting pipeline
def _createLightingPipeline(self):
self.debug("Creating lighting pipeline ..")
# size has to be a multiple of the compute unit size
# but still has to cover the whole screen
sizeX = int(math.ceil(self.size.x / self.patchSize.x))
sizeY = int(math.ceil(self.size.y / self.patchSize.y))
self.precomputeSize = Vec2(sizeX, sizeY)
self.debug("Batch size =", sizeX, "x", sizeY, "Actual Buffer size=",
int(sizeX * self.patchSize.x), "x",
int(sizeY * self.patchSize.y))
self._makeLightPerTileStorage()
# Create a buffer which computes which light affects which tile
self._makeLightBoundsComputationBuffer(sizeX, sizeY)
# Create a buffer which applies the lighting
self._makeLightingComputeBuffer()
# Register for light manager
self.lightManager.setLightingComputator(self.lightingComputeContainer)
self.lightManager.setLightingCuller(self.lightBoundsComputeBuff)
self.lightingComputeContainer.setShaderInput("lightsPerTile",
self.lightPerTileStorage)
self.lightingComputeContainer.setShaderInput("cameraPosition",
base.cam.getPos(render))
# Ensure the images have the correct filter mode
for bmode in [RenderTargetType.Color]:
tex = self.lightBoundsComputeBuff.getTexture(bmode)
tex.setMinfilter(Texture.FTNearest)
tex.setMagfilter(Texture.FTNearest)
self._loadFallbackCubemap()
# Create storage for the bounds computation
# Set inputs
self.lightBoundsComputeBuff.setShaderInput("destination",
self.lightPerTileStorage)
self.lightBoundsComputeBuff.setShaderInput(
"depth", self.deferredTarget.getDepthTexture())
self.lightingComputeContainer.setShaderInput(
"data0", self.deferredTarget.getColorTexture())
self.lightingComputeContainer.setShaderInput(
"data1", self.deferredTarget.getAuxTexture(0))
self.lightingComputeContainer.setShaderInput(
"data2", self.deferredTarget.getAuxTexture(1))
self.lightingComputeContainer.setShaderInput(
"shadowAtlas", self.lightManager.getAtlasTex())
self.lightingComputeContainer.setShaderInput(
"destination", self.lightingComputeCombinedTex)
# self.lightingComputeContainer.setShaderInput("sampleTex", loader.loadTexture("Data/Antialiasing/Unigine01.png"))
def _loadFallbackCubemap(self):
cubemap = loader.loadCubeMap("Cubemap/#.png")
cubemap.setMinfilter(Texture.FTLinearMipmapLinear)
cubemap.setMagfilter(Texture.FTLinearMipmapLinear)
cubemap.setFormat(Texture.F_srgb_alpha)
self.lightingComputeContainer.setShaderInput("fallbackCubemap",
cubemap)
def _makeLightBoundsComputationBuffer(self, w, h):
self.debug("Creating light precomputation buffer of size", w, "x", h)
self.lightBoundsComputeBuff = RenderTarget("ComputeLightTileBounds")
self.lightBoundsComputeBuff.setSize(w, h)
self.lightBoundsComputeBuff.addRenderTexture(RenderTargetType.Color)
self.lightBoundsComputeBuff.setColorBits(16)
self.lightBoundsComputeBuff.prepareOffscreenBuffer()
self.lightBoundsComputeBuff.setShaderInput("mainCam", base.cam)
self.lightBoundsComputeBuff.setShaderInput("mainRender", base.render)
self._setPositionComputationShader()
def _makeLightingComputeBuffer(self):
self.lightingComputeContainer = RenderTarget("ComputeLighting")
self.lightingComputeContainer.setSize(
base.win.getXSize() / self.temporalProjFactor, base.win.getYSize())
self.lightingComputeContainer.addRenderTexture(RenderTargetType.Color)
self.lightingComputeContainer.setColorBits(16)
self.lightingComputeContainer.prepareOffscreenBuffer()
self.lightingComputeCombinedTex = Texture("Lighting-Compute-Combined")
self.lightingComputeCombinedTex.setup2dTexture(base.win.getXSize(),
base.win.getYSize(),
Texture.TFloat,
Texture.FRgba16)
self.lightingComputeCombinedTex.setMinfilter(Texture.FTLinear)
self.lightingComputeCombinedTex.setMagfilter(Texture.FTLinear)
self.lastPositionBuffer = Texture("Last-Position-Buffer")
self.lastPositionBuffer.setup2dTexture(base.win.getXSize(),
base.win.getYSize(),
Texture.TFloat, Texture.FRgba16)
self.lastPositionBuffer.setMinfilter(Texture.FTNearest)
self.lastPositionBuffer.setMagfilter(Texture.FTNearest)
def _setLightingShader(self):
lightShader = BetterShader.load("Shader/DefaultPostProcess.vertex",
"Shader/ApplyLighting.fragment")
self.lightingComputeContainer.setShader(lightShader)
def _setCombinerShader(self):
cShader = BetterShader.load("Shader/DefaultPostProcess.vertex",
"Shader/Combiner.fragment")
self.combiner.setShader(cShader)
def _setPositionComputationShader(self):
pcShader = BetterShader.load("Shader/DefaultPostProcess.vertex",
"Shader/PrecomputeLights.fragment")
self.lightBoundsComputeBuff.setShader(pcShader)
def _setFinalPassShader(self):
fShader = BetterShader.load("Shader/DefaultPostProcess.vertex",
"Shader/Final.fragment")
self.finalPass.setShader(fShader)
def _getSize(self):
return Vec2(int(self.showbase.win.getXSize()),
int(self.showbase.win.getYSize()))
def debugReloadShader(self):
self.lightManager.debugReloadShader()
self._setPositionComputationShader()
self._setCombinerShader()
self._setLightingShader()
self._setFinalPassShader()
self.antialias.reloadShader()
def _attachUpdateTask(self):
self.showbase.addTask(self._update,
"UpdateRenderingPipeline",
sort=-10000)
def _computeCameraBounds(self):
# compute camera bounds in render space
cameraBounds = self.camera.node().getLens().makeBounds()
cameraBounds.xform(self.camera.getMat(render))
return cameraBounds
def _update(self, task=None):
self.temporalProjXOffs += 1
self.temporalProjXOffs = self.temporalProjXOffs % self.temporalProjFactor
self.cullBounds = self._computeCameraBounds()
self.lightManager.setCullBounds(self.cullBounds)
self.lightManager.update()
self.lightingComputeContainer.setShaderInput("cameraPosition",
base.cam.getPos(render))
self.lightingComputeContainer.setShaderInput(
"temporalProjXOffs", LVecBase2i(self.temporalProjXOffs))
self.combiner.setShaderInput("lastMVP", self.lastMVP)
render.setShaderInput("lastMVP", self.lastMVP)
self.combiner.setShaderInput("temporalProjXOffs",
LVecBase2i(self.temporalProjXOffs))
self._computeMVP()
self.combiner.setShaderInput("currentMVP", self.lastMVP)
self.combiner.setShaderInput("cameraPosition", base.cam.getPos(render))
if task is not None:
return task.cont
def _computeMVP(self):
projMat = Mat4.convertMat(CSYupRight, base.camLens.getCoordinateSystem(
)) * base.camLens.getProjectionMat()
transformMat = TransformState.makeMat(
Mat4.convertMat(base.win.getGsg().getInternalCoordinateSystem(),
CSZupRight))
modelViewMat = transformMat.invertCompose(render.getTransform(
base.cam)).getMat()
self.lastMVP = modelViewMat * projMat
# print "Self.lastMVP is now from frame",globalClock.getFrameTime()
def getLightManager(self):
return self.lightManager
def getDefaultObjectShader(self):
shader = BetterShader.load("Shader/DefaultObjectShader.vertex",
"Shader/DefaultObjectShader.fragment")
return shader
class RenderingPipeline(DebugObject):
""" This is the core class, driving all other classes. To use this
pipeline, your code has to call *after* the initialization of ShowBase:
renderPipeline = RenderingPipeline()
renderPipeline.loadSettings("pipeline.ini")
renderPipeline.create()
The pipeline will setup all required buffers, tasks and shaders itself. To
add lights, see the documentation of LightManager.
How it works:
You can see an example buffer view at http://i.imgur.com/mZK6TVj.png
The pipeline first renders all normal objects (parented to render) into a
buffer, using multiple render targets. These buffers store normals,
position and material properties. Your shaders have to output these
values, but there is a handy api, just look at
Shaders/DefaultObjectShader.fragment.
After that, the pipeline splits the screen into tiles, typically of the
size 32x32. For each tile, it computes which lights affect which tile,
called Tiled Deferred Shading. This is written to a buffer.
The next step is applying the lighting. This is done at half window
resolution only, using Temporal Reprojection. I don't aim to explain
Temporal Reprojection here, but basically, I render only each second
pixel each frame. This is simply for performance.
The lighting pass iterates through the list of lights per tile, and
applies both lighting and shadows to each pixel, using the material
information from the previous rendered buffers.
After the lighting pass, a combiner pass combines both the current
frame and the last frame, this is required because of
Temporal Reprojection.
At this step, we already have a frame we could display. In the next passes,
only anti-aliasing and post-processing effects like motion blur are added.
In the meantime, the LightManager builds a list of ShadowSources which
need an update. It creates a scene render and renders the scene from the
view of the shadow sources to the global shadow atlas. There are a limited
amount of shadow updates per frame available, and the updates are stored
in a queue. So when displaying many shadow-lights, not each shadowmap is
update each frame. The reason is, again, performance. When you need a
custom shadow caster shader, e.g. for alpha blending, you should use the
Shader/DefaultShaowCaster.* as prefab.
"""
def __init__(self, showbase):
""" Creates a new pipeline """
DebugObject.__init__(self, "RenderingPipeline")
self.showbase = showbase
self.settings = None
self.mountManager = MountManager()
def getMountManager(self):
""" Returns the mount manager. You can use this to set the
write directory and base path """
return self.mountManager
def loadSettings(self, filename):
""" Loads the pipeline settings from an ini file """
self.settings = PipelineSettingsManager()
self.settings.loadFromFile(filename)
def getSettings(self):
""" Returns the current pipeline settings """
return self.settings
def create(self):
""" Creates this pipeline """
self.debug("Setting up render pipeline")
if self.settings is None:
self.error("You have to call loadSettings first!")
return
# Mount everything first
self.mountManager.mount()
# Store globals, as cython can't handle them
self.debug("Setting up globals")
Globals.load(self.showbase)
# Setting up shader loading
BetterShader._DumpShaders = self.settings.dumpGeneratedShaders
# We use PTA's for shader inputs, because that's faster than
# using setShaderInput
self.temporalProjXOffs = PTAInt.emptyArray(1)
self.cameraPosition = PTAVecBase3f.emptyArray(1)
self.motionBlurFactor = PTAFloat.emptyArray(1)
self.lastMVP = PTALMatrix4f.emptyArray(1)
self.currentMVP = PTALMatrix4f.emptyArray(1)
# Create onscreen gui
# For the temporal reprojection it is important that the window width
# is a multiple of 2
if self.showbase.win.getXSize() % 2 == 1:
self.error(
"The window has to have a width which is a multiple of 2 "
"(Current: ", self.showbase.win.getXSize(), ")")
self.error(
"I'll correct that for you, but next time pass the correct "
"window size!")
wp = WindowProperties()
wp.setSize(
self.showbase.win.getXSize() + 1, self.showbase.win.getYSize())
self.showbase.win.requestProperties(wp)
self.showbase.graphicsEngine.openWindows()
self.camera = self.showbase.cam
self.size = self._getSize()
self.cullBounds = None
# Debug variables to disable specific features
self.haveLightingPass = True
# haveCombiner can only be true when haveLightingPass is enabled
self.haveCombiner = True
self.haveMRT = True
# Not as good as I want it, so disabled. I'll work on it.
self.blurEnabled = False
self.debug("Window size is", self.size.x, "x", self.size.y)
self.showbase.camLens.setNearFar(0.1, 50000)
self.showbase.camLens.setFov(90)
self.showbase.win.setClearColor(Vec4(1.0,0.0,1.0,1.0))
# Create occlusion handler
self._setupOcclusion()
if self.settings.displayOnscreenDebugger:
self.guiManager = PipelineGuiManager(self)
self.guiManager.setup()
# Generate auto-configuration for shaders
self._generateShaderConfiguration()
# Create light manager, which handles lighting + shadows
if self.haveLightingPass:
self.lightManager = LightManager(self)
self.patchSize = LVecBase2i(
self.settings.computePatchSizeX,
self.settings.computePatchSizeY)
# Create separate scene graphs. The deferred graph is render
self.forwardScene = NodePath("Forward-Rendering")
self.transparencyScene = NodePath("Transparency-Rendering")
# We need no transparency (we store other information in the alpha
# channel)
self.showbase.render.setAttrib(
TransparencyAttrib.make(TransparencyAttrib.MNone), 100)
# Now create deferred render buffers
self._makeDeferredTargets()
# Create the target which constructs the view-space normals and
# position from world-space position
if self.occlusion.requiresViewSpacePosNrm():
self._createNormalPrecomputeBuffer()
# Setup the buffers for lighting
self._createLightingPipeline()
# Setup combiner for temporal reprojetion
if self.haveCombiner:
self._createCombiner()
if self.occlusion.requiresBlurring():
self._createOcclusionBlurBuffer()
self._setupAntialiasing()
if self.blurEnabled:
self._createDofStorage()
self._createBlurBuffer()
self._setupFinalPass()
self._setShaderInputs()
# add update task
self._attachUpdateTask()
# display shadow atlas is defined
# todo: move this to the gui manager
# if self.settings.displayShadowAtlas and self.haveLightingPass:
# self.atlasDisplayImage = OnscreenImage(
# image=self.lightManager.getAtlasTex(), pos=(
# self.showbase.getAspectRatio() - 0.55, 0, 0.2),
# scale=(0.5, 0, 0.5))
def getForwardScene(self):
""" Reparent objects to this scene to use forward rendering.
Objects in this scene will directly get rendered, with no
lighting etc. applied. """
return self.forwardScene
def getTransparentScene(self):
""" Reparent objects to this scene to allow this objects to
have transparency. Objects in this scene will get directly
rendered and no lighting will get applied. """
return self.transparencyScene
def _createCombiner(self):
""" Creates the target which combines the result from the
lighting computation and last frame together
(Temporal Reprojection) """
self.combiner = RenderTarget("Combine-Temporal")
self.combiner.addColorTexture()
self.combiner.setColorBits(16)
self.combiner.prepareOffscreenBuffer()
self._setCombinerShader()
def _setupAntialiasing(self):
""" Creates the antialiasing technique """
technique = self.settings.antialiasingTechnique
self.debug("Creating antialiasing handler for", technique)
if technique == "None":
self.antialias = AntialiasingTechniqueNone()
elif technique == "SMAA":
self.antialias = AntialiasingTechniqueSMAA()
else:
self.error(
"Unkown antialiasing technique", technique, "-> using None:")
self.antialias = AntialiasingTechniqueNone()
if self.occlusion.requiresBlurring():
self.antialias.setColorTexture(
self.blurOcclusionH.getColorTexture())
else:
if self.haveCombiner:
self.antialias.setColorTexture(self.combiner.getColorTexture())
else:
self.antialias.setColorTexture(
self.deferredTarget.getColorTexture())
self.antialias.setDepthTexture(self.deferredTarget.getDepthTexture())
self.antialias.setup()
def _setupOcclusion(self):
""" Creates the occlusion technique """
technique = self.settings.occlusionTechnique
self.debug("Creating occlusion handle for", technique)
if technique == "None":
self.occlusion = AmbientOcclusionTechniqueNone()
elif technique == "SAO":
self.occlusion = AmbientOcclusionTechniqueSAO()
else:
self.error("Unkown occlusion technique:", technique)
self.occlusion = AmbientOcclusionTechniqueNone()
def _makeDeferredTargets(self):
""" Creates the multi-render-target """
self.debug("Creating deferred targets")
self.deferredTarget = RenderTarget("DeferredTarget")
self.deferredTarget.addColorAndDepth()
if self.haveMRT:
self.deferredTarget.addAuxTextures(3)
self.deferredTarget.setAuxBits(16)
self.deferredTarget.setColorBits(16)
self.deferredTarget.setDepthBits(32)
# GL_INVALID_OPERATION ?
# self.deferredTarget.setMultisamples(1)
self.deferredTarget.prepareSceneRender()
def _setupFinalPass(self):
""" Setups the final pass which applies motion blur and so on """
# Set wrap for motion blur
colorTex = self.antialias.getResultTexture()
colorTex.setWrapU(Texture.WMClamp)
colorTex.setWrapV(Texture.WMClamp)
self._setFinalPassShader()
def _makeLightPerTileStorage(self):
""" Creates a texture to store the lights per tile into. Should
get replaced with ssbos later """
storageSizeX = self.precomputeSize.x * 8
storageSizeY = self.precomputeSize.y * 8
self.debug(
"Creating per tile storage of size",
storageSizeX, "x", storageSizeY)
self.lightPerTileStorage = Texture("LightsPerTile")
self.lightPerTileStorage.setup2dTexture(
storageSizeX, storageSizeY, Texture.TUnsignedShort, Texture.FR32i)
self.lightPerTileStorage.setMinfilter(Texture.FTNearest)
self.lightPerTileStorage.setMagfilter(Texture.FTNearest)
def _createLightingPipeline(self):
""" Creates the lighting pipeline, including shadow handling """
if not self.haveLightingPass:
self.debug("Skipping lighting pipeline")
return
self.debug("Creating lighting pipeline ..")
# size has to be a multiple of the compute unit size
# but still has to cover the whole screen
sizeX = int(math.ceil(float(self.size.x) / self.patchSize.x))
sizeY = int(math.ceil(float(self.size.y) / self.patchSize.y))
self.precomputeSize = LVecBase2i(sizeX, sizeY)
self.debug("Batch size =", sizeX, "x", sizeY,
"Actual Buffer size=", int(sizeX * self.patchSize.x),
"x", int(sizeY * self.patchSize.y))
self._makeLightPerTileStorage()
# Create a buffer which computes which light affects which tile
self._makeLightBoundsComputationBuffer(sizeX, sizeY)
# Create a buffer which applies the lighting
self._makeLightingComputeBuffer()
# Register for light manager
self.lightManager.setLightingComputator(self.lightingComputeContainer)
self.lightManager.setLightingCuller(self.lightBoundsComputeBuff)
self._loadFallbackCubemap()
self._loadLookupCubemap()
def _setShaderInputs(self):
""" Sets most of the required shader inputs to the targets """
# Shader inputs for the light-culling pass
if self.haveLightingPass:
self.lightBoundsComputeBuff.setShaderInput(
"destination", self.lightPerTileStorage)
self.lightBoundsComputeBuff.setShaderInput(
"depth", self.deferredTarget.getDepthTexture())
self.lightBoundsComputeBuff.setShaderInput(
"mainCam", self.showbase.cam)
self.lightBoundsComputeBuff.setShaderInput(
"mainRender", self.showbase.render)
# Shader inputs for the light-applying pass
self.lightingComputeContainer.setShaderInput(
"data0", self.deferredTarget.getColorTexture())
self.lightingComputeContainer.setShaderInput(
"data1", self.deferredTarget.getAuxTexture(0))
self.lightingComputeContainer.setShaderInput(
"data2", self.deferredTarget.getAuxTexture(1))
self.lightingComputeContainer.setShaderInput(
"data3", self.deferredTarget.getAuxTexture(2))
self.lightingComputeContainer.setShaderInput(
"depth", self.deferredTarget.getDepthTexture())
if self.occlusion.requiresViewSpacePosNrm():
self.lightingComputeContainer.setShaderInput(
"viewSpaceNormals",
self.normalPrecompute.getColorTexture())
self.lightingComputeContainer.setShaderInput(
"viewSpacePosition",
self.normalPrecompute.getAuxTexture(0))
self.lightingComputeContainer.setShaderInput(
"shadowAtlas", self.lightManager.getAtlasTex())
self.lightingComputeContainer.setShaderInput(
"destination", self.lightingComputeCombinedTex)
self.lightingComputeContainer.setShaderInput(
"temporalProjXOffs", self.temporalProjXOffs)
self.lightingComputeContainer.setShaderInput(
"cameraPosition", self.cameraPosition)
self.lightingComputeContainer.setShaderInput(
"noiseTexture", self.showbase.loader.loadTexture("Data/Occlusion/noise4x4.png"))
self.lightingComputeContainer.setShaderInput(
"lightsPerTile", self.lightPerTileStorage)
# Shader inputs for the occlusion blur passes
if self.occlusion.requiresBlurring() and self.haveCombiner:
self.blurOcclusionH.setShaderInput(
"colorTex", self.blurOcclusionV.getColorTexture())
self.blurOcclusionV.setShaderInput(
"colorTex", self.combiner.getColorTexture())
self.blurOcclusionH.setShaderInput(
"normalTex", self.deferredTarget.getAuxTexture(0))
self.blurOcclusionV.setShaderInput(
"normalTex", self.deferredTarget.getAuxTexture(0))
self.blurOcclusionH.setShaderInput(
"normalsView", self.normalPrecompute.getAuxTexture(0))
self.blurOcclusionV.setShaderInput(
"normalsView", self.normalPrecompute.getAuxTexture(0))
# Shader inputs for the blur passes
if self.blurEnabled:
self.blurColorH.setShaderInput(
"dofStorage", self.dofStorage)
self.blurColorV.setShaderInput(
"dofStorage", self.dofStorage)
self.blurColorH.setShaderInput("colorTex",
self.antialias.getResultTexture())
self.blurColorH.setShaderInput("depthTex",
self.deferredTarget.getDepthTexture())
self.blurColorV.setShaderInput("colorTex",
self.blurColorH.getColorTexture())
# Shader inputs for the temporal reprojection
if self.haveCombiner:
self.combiner.setShaderInput(
"currentComputation",
self.lightingComputeContainer.getColorTexture())
self.combiner.setShaderInput(
"lastFrame", self.lightingComputeCombinedTex)
self.combiner.setShaderInput(
"positionBuffer", self.deferredTarget.getColorTexture())
self.combiner.setShaderInput(
"velocityBuffer", self.deferredTarget.getAuxTexture(1))
if self.blurEnabled:
self.combiner.setShaderInput(
"dofStorage", self.dofStorage)
self.combiner.setShaderInput(
"depthTex", self.deferredTarget.getDepthTexture())
self.combiner.setShaderInput(
"lastPosition", self.lastPositionBuffer)
self.combiner.setShaderInput(
"temporalProjXOffs", self.temporalProjXOffs)
self.combiner.setShaderInput("lastMVP", self.lastMVP)
self.combiner.setShaderInput("cameraPosition", self.cameraPosition)
self.combiner.setShaderInput("currentMVP", self.lastMVP)
# Shader inputs for the final pass
if self.blurEnabled:
self.deferredTarget.setShaderInput(
"colorTex", self.blurColorV.getColorTexture())
else:
self.deferredTarget.setShaderInput(
"colorTex", self.antialias.getResultTexture())
if self.occlusion.requiresBlurring():
self.normalPrecompute.setShaderInput(
"positionTex", self.deferredTarget.getColorTexture())
self.normalPrecompute.setShaderInput(
"mainCam", self.showbase.cam)
self.normalPrecompute.setShaderInput(
"mainRender", self.showbase.render)
self.normalPrecompute.setShaderInput(
"depthTex", self.deferredTarget.getDepthTexture())
if self.haveMRT:
self.deferredTarget.setShaderInput(
"velocityTex", self.deferredTarget.getAuxTexture(1))
self.deferredTarget.setShaderInput(
"depthTex", self.deferredTarget.getDepthTexture())
self.deferredTarget.setShaderInput(
"motionBlurFactor", self.motionBlurFactor)
if self.haveLightingPass:
self.deferredTarget.setShaderInput(
"lastFrame", self.lightingComputeCombinedTex)
if self.haveCombiner:
self.deferredTarget.setShaderInput(
"newFrame", self.combiner.getColorTexture())
self.deferredTarget.setShaderInput(
"lastPosition", self.lastPositionBuffer)
self.deferredTarget.setShaderInput(
"currentPosition", self.deferredTarget.getColorTexture())
# Set last / current mvp handles
self.showbase.render.setShaderInput("lastMVP", self.lastMVP)
# Finally, set shaders
self.reloadShaders()
def _loadFallbackCubemap(self):
""" Loads the cubemap for image based lighting """
cubemap = self.showbase.loader.loadCubeMap(
"Data/Cubemaps/Default/#.png")
cubemap.setMinfilter(Texture.FTLinearMipmapLinear)
cubemap.setMagfilter(Texture.FTLinearMipmapLinear)
cubemap.setFormat(Texture.F_srgb_alpha)
self.lightingComputeContainer.setShaderInput(
"fallbackCubemap", cubemap)
def _loadLookupCubemap(self):
self.debug("Loading lookup cubemap")
cubemap = self.showbase.loader.loadCubeMap(
"Data/Cubemaps/DirectionLookup/#.png")
cubemap.setMinfilter(Texture.FTNearest)
cubemap.setMagfilter(Texture.FTNearest)
cubemap.setFormat(Texture.F_rgb8)
self.lightingComputeContainer.setShaderInput(
"directionToFace", cubemap)
def _makeLightBoundsComputationBuffer(self, w, h):
""" Creates the buffer which precomputes the lights per tile """
self.debug("Creating light precomputation buffer of size", w, "x", h)
self.lightBoundsComputeBuff = RenderTarget("ComputeLightTileBounds")
self.lightBoundsComputeBuff.setSize(w, h)
self.lightBoundsComputeBuff.setColorWrite(False)
self.lightBoundsComputeBuff.prepareOffscreenBuffer()
def _makeLightingComputeBuffer(self):
""" Creates the buffer which applies the lighting """
self.lightingComputeContainer = RenderTarget("ComputeLighting")
self.lightingComputeContainer.setSize(
self.showbase.win.getXSize() / 2, self.showbase.win.getYSize())
self.lightingComputeContainer.addColorTexture()
self.lightingComputeContainer.setColorBits(16)
self.lightingComputeContainer.prepareOffscreenBuffer()
self.lightingComputeCombinedTex = Texture("Lighting-Compute-Combined")
self.lightingComputeCombinedTex.setup2dTexture(
self.showbase.win.getXSize(), self.showbase.win.getYSize(),
Texture.TFloat, Texture.FRgba8)
self.lightingComputeCombinedTex.setMinfilter(Texture.FTLinear)
self.lightingComputeCombinedTex.setMagfilter(Texture.FTLinear)
self.lastPositionBuffer = Texture("Last-Position-Buffer")
self.lastPositionBuffer.setup2dTexture(
self.showbase.win.getXSize(), self.showbase.win.getYSize(),
Texture.TFloat, Texture.FRgba16)
self.lastPositionBuffer.setMinfilter(Texture.FTNearest)
self.lastPositionBuffer.setMagfilter(Texture.FTNearest)
def _createOcclusionBlurBuffer(self):
""" Creates the buffers needed to blur the occlusion """
self.blurOcclusionV = RenderTarget("blurOcclusionVertical")
self.blurOcclusionV.addColorTexture()
self.blurOcclusionV.prepareOffscreenBuffer()
self.blurOcclusionH = RenderTarget("blurOcclusionHorizontal")
self.blurOcclusionH.addColorTexture()
self.blurOcclusionH.prepareOffscreenBuffer()
# Mipmaps for blur?
# self.blurOcclusionV.getColorTexture().setMinfilter(
# Texture.FTLinearMipmapLinear)
# self.combiner.getColorTexture().setMinfilter(
# Texture.FTLinearMipmapLinear)
def _createBlurBuffer(self):
""" Creates the buffers for the dof """
self.blurColorV = RenderTarget("blurColorVertical")
self.blurColorV.addColorTexture()
self.blurColorV.prepareOffscreenBuffer()
self.blurColorH = RenderTarget("blurColorHorizontal")
self.blurColorH.addColorTexture()
self.blurColorH.prepareOffscreenBuffer()
# self.blurColorH.getColorTexture().setMinfilter(
# Texture.FTLinearMipmapLinear)
# self.antialias.getResultTexture().setMinfilter(
# Texture.FTLinearMipmapLinear)
def _createNormalPrecomputeBuffer(self):
""" Creates a buffer which reconstructs the normals and position
from view-space """
self.normalPrecompute = RenderTarget("PrecomputeNormals")
self.normalPrecompute.addColorTexture()
self.normalPrecompute.addAuxTextures(1)
self.normalPrecompute.setColorBits(16)
self.normalPrecompute.setAuxBits(16)
self.normalPrecompute.prepareOffscreenBuffer()
def _createDofStorage(self):
""" Creates the texture where the dof factor is stored in, so we
don't recompute it each pass """
self.dofStorage = Texture("DOFStorage")
self.dofStorage.setup2dTexture(
self.showbase.win.getXSize(), self.showbase.win.getYSize(),
Texture.TFloat, Texture.FRg16)
def _setOcclusionBlurShader(self):
""" Sets the shaders which blur the occlusion """
blurVShader = BetterShader.load(
"Shader/DefaultPostProcess.vertex",
"Shader/BlurOcclusionVertical.fragment")
blurHShader = BetterShader.load(
"Shader/DefaultPostProcess.vertex",
"Shader/BlurOcclusionHorizontal.fragment")
self.blurOcclusionV.setShader(blurVShader)
self.blurOcclusionH.setShader(blurHShader)
def _setBlurShader(self):
""" Sets the shaders which blur the color """
blurVShader = BetterShader.load(
"Shader/DefaultPostProcess.vertex",
"Shader/BlurVertical.fragment")
blurHShader = BetterShader.load(
"Shader/DefaultPostProcess.vertex",
"Shader/BlurHorizontal.fragment")
self.blurColorV.setShader(blurVShader)
self.blurColorH.setShader(blurHShader)
def _setLightingShader(self):
""" Sets the shader which applies the light """
lightShader = BetterShader.load(
"Shader/DefaultPostProcess.vertex",
"Shader/ApplyLighting.fragment")
self.lightingComputeContainer.setShader(lightShader)
def _setCombinerShader(self):
""" Sets the shader which combines the lighting with the previous frame
(temporal reprojection) """
cShader = BetterShader.load(
"Shader/DefaultPostProcess.vertex",
"Shader/Combiner.fragment")
self.combiner.setShader(cShader)
def _setPositionComputationShader(self):
""" Sets the shader which computes the lights per tile """
pcShader = BetterShader.load(
"Shader/DefaultPostProcess.vertex",
"Shader/PrecomputeLights.fragment")
self.lightBoundsComputeBuff.setShader(pcShader)
def _setFinalPassShader(self):
""" Sets the shader which computes the final frame,
with motion blur and so on """
fShader = BetterShader.load(
"Shader/DefaultPostProcess.vertex",
"Shader/Final.fragment")
self.deferredTarget.setShader(fShader)
def _getSize(self):
""" Returns the window size. """
return LVecBase2i(
self.showbase.win.getXSize(),
self.showbase.win.getYSize())
def reloadShaders(self):
""" Reloads all shaders """
if self.haveLightingPass:
self.lightManager.debugReloadShader()
self._setPositionComputationShader()
self._setLightingShader()
if self.haveCombiner:
self._setCombinerShader()
self._setFinalPassShader()
if self.occlusion.requiresBlurring():
self._setOcclusionBlurShader()
if self.blurEnabled:
self._setBlurShader()
if self.occlusion.requiresViewSpacePosNrm():
self._setNormalExtractShader()
self.antialias.reloadShader()
def _setNormalExtractShader(self):
""" Sets the shader which constructs the normals from position """
npShader = BetterShader.load(
"Shader/DefaultPostProcess.vertex",
"Shader/ExtractNormals.fragment")
self.normalPrecompute.setShader(npShader)
def _attachUpdateTask(self):
""" Attaches the update tasks to the showbase """
self.showbase.addTask(
self._update, "UpdateRenderingPipeline", sort=-10)
if self.haveLightingPass:
self.showbase.addTask(
self._updateLights, "UpdateLights", sort=-9)
self.showbase.addTask(
self._updateShadows, "updateShadows", sort=-8)
if self.settings.displayOnscreenDebugger:
self.showbase.addTask(
self._updateGUI, "UpdateGUI", sort=7)
def _computeCameraBounds(self):
""" Computes the current camera bounds, i.e. for light culling """
cameraBounds = self.camera.node().getLens().makeBounds()
cameraBounds.xform(self.camera.getMat(self.showbase.render))
return cameraBounds
def _updateLights(self, task=None):
""" Task which updates/culls the lights """
self.lightManager.updateLights()
if task is not None:
return task.cont
def _updateShadows(self, task=None):
""" Task which updates the shadow maps """
self.lightManager.updateShadows()
if task is not None:
return task.cont
def _updateGUI(self, task=None):
""" Task which updates the onscreen gui debugger """
self.guiManager.update()
if task is not None:
return task.cont
def _update(self, task=None):
""" Main update task """
currentFPS = 1.0 / self.showbase.taskMgr.globalClock.getDt()
self.temporalProjXOffs[0] = 1 - self.temporalProjXOffs[0]
self.cameraPosition[0] = self.showbase.cam.getPos(self.showbase.render)
self.motionBlurFactor[0] = min(1.5, currentFPS /
60.0) * self.settings.motionBlurFactor
self.cullBounds = self._computeCameraBounds()
if self.haveLightingPass:
self.lightManager.setCullBounds(self.cullBounds)
self.lastMVP[0] = self.currentMVP[0]
self.currentMVP[0] = self._computeMVP()
if task is not None:
return task.cont
def _computeMVP(self):
""" Computes the current mvp. Actually, this is the
worldViewProjectionMatrix, but for convience it's called mvp. """
camLens = self.showbase.camLens
projMat = Mat4.convertMat(
CSYupRight,
camLens.getCoordinateSystem()) * camLens.getProjectionMat()
transformMat = TransformState.makeMat(
Mat4.convertMat(self.showbase.win.getGsg().getInternalCoordinateSystem(),
CSZupRight))
modelViewMat = transformMat.invertCompose(
self.showbase.render.getTransform(self.showbase.cam)).getMat()
return UnalignedLMatrix4f(modelViewMat * projMat)
def getLightManager(self):
""" Returns a handle to the light manager """
return self.lightManager
def getDefaultObjectShader(self, tesselated=False):
""" Returns the default shader for objects """
if not tesselated:
shader = BetterShader.load(
"Shader/DefaultObjectShader/vertex.glsl",
"Shader/DefaultObjectShader/fragment.glsl")
else:
self.warn(
"Tesselation is only experimental! Remember "
"to convert the geometry to patches first!")
shader = BetterShader.load(
"Shader/DefaultObjectShader/vertex.glsl",
"Shader/DefaultObjectShader/fragment.glsl",
"",
"Shader/DefaultObjectShader/tesscontrol.glsl",
"Shader/DefaultObjectShader/tesseval.glsl")
return shader
def addLight(self, light):
""" Adds a light to the list of rendered lights """
if self.haveLightingPass:
self.lightManager.addLight(light)
else:
self.warn("Lighting is disabled, so addLight has no effect")
def _generateShaderConfiguration(self):
""" Genrates the global shader include which defines
most values used in the shaders. """
self.debug("(Re)Generating shader configuration")
# Generate list of defines
defines = []
if self.settings.antialiasingTechnique == "SMAA":
quality = self.settings.smaaQuality.upper()
if quality in ["LOW", "MEDIUM", "HIGH", "ULTRA"]:
defines.append(("SMAA_PRESET_" + quality, ""))
else:
self.error("Unrecognized SMAA quality:", quality)
return
defines.append(
("LIGHTING_COMPUTE_PATCH_SIZE_X", self.settings.computePatchSizeX))
defines.append(
("LIGHTING_COMPUTE_PATCH_SIZE_Y", self.settings.computePatchSizeY))
defines.append(
("LIGHTING_MIN_MAX_DEPTH_ACCURACY", self.settings.minMaxDepthAccuracy))
if self.blurEnabled:
defines.append(("USE_DOF", 1))
if self.settings.useSimpleLighting:
defines.append(("USE_SIMPLE_LIGHTING", 1))
if self.settings.anyLightBoundCheck:
defines.append(("LIGHTING_ANY_BOUND_CHECK", 1))
if self.settings.accurateLightBoundCheck:
defines.append(("LIGHTING_ACCURATE_BOUND_CHECK", 1))
if self.settings.renderShadows:
defines.append(("USE_SHADOWS", 1))
defines.append(
("SHADOW_MAP_ATLAS_SIZE", self.settings.shadowAtlasSize))
defines.append(
("SHADOW_MAX_UPDATES_PER_FRAME", self.settings.maxShadowUpdatesPerFrame))
defines.append(
("SHAODOW_GEOMETRY_MAX_VERTICES", self.settings.maxShadowUpdatesPerFrame * 3))
defines.append(("SHADOWS_NUM_SAMPLES", self.settings.numShadowSamples))
if self.settings.useHardwarePCF:
defines.append(("USE_HARDWARE_PCF", 1))
defines.append(("WINDOW_WIDTH", self.showbase.win.getXSize()))
defines.append(("WINDOW_HEIGHT", self.showbase.win.getYSize()))
if self.settings.motionBlurEnabled:
defines.append(("USE_MOTION_BLUR", 1))
defines.append(
("MOTION_BLUR_SAMPLES", self.settings.motionBlurSamples))
# Occlusion
defines.append(
("OCCLUSION_TECHNIQUE_" + self.occlusion.getIncludeName(), 1))
defines.append(
("OCCLUSION_RADIUS", self.settings.occlusionRadius))
defines.append(
("OCCLUSION_STRENGTH", self.settings.occlusionStrength))
defines.append(
("OCCLUSION_SAMPLES", self.settings.occlusionSampleCount))
if self.settings.displayOnscreenDebugger:
defines.append(("DEBUGGER_ACTIVE", 1))
extraSettings = self.guiManager.getDefines()
defines += extraSettings
# Pass near far
defines.append(("CAMERA_NEAR", Globals.base.camLens.getNear()))
defines.append(("CAMERA_FAR", Globals.base.camLens.getFar()))
# Generate
output = "// Autogenerated by RenderingPipeline.py\n"
output += "// Do not edit! Your changes will be lost.\n\n"
for key, value in defines:
output += "#define " + key + " " + str(value) + "\n"
# Try to write the file
# Todo: add error handling
with open("PipelineTemp/ShaderAutoConfig.include", "w") as handle:
handle.write(output)
def onWindowResized(self):
""" Call this whenever the window resized """
raise NotImplementedError()
def destroy(self):
""" Call this when you want to shut down the pipeline """
self.mountManager.unmount()
self.error("Destroy is not implemented yet")
def reload(self):
""" This reloads the whole pipeline, same as destroy(); create() """
self.debug("Reloading pipeline")
self.destroy()
self.create()
def setActive(self, active):
""" You can enable/disable the pipeline, for example
when the user is in the menu, the 3d scene does not have
to be rendered """
raise NotImplementedError()
