BansheeEngine/CamelotRenderer/Include/CmRenderSystem.h

/*
-----------------------------------------------------------------------------
This source file is part of OGRE
(Object-oriented Graphics Rendering Engine)
For the latest info, see http://www.ogre3d.org

Copyright (c) 2000-2011 Torus Knot Software Ltd

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in
all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
THE SOFTWARE.
-----------------------------------------------------------------------------
*/
#ifndef __RenderSystem_H_
#define __RenderSystem_H_

// Precompiler options
#include "CmPrerequisites.h"

#include <memory>

#include "CmString.h"

#include "CmSamplerState.h"
#include "CmCommon.h"

#include "CmRenderOperation.h"
#include "CmRenderSystemCapabilities.h"
#include "CmRenderTarget.h"
#include "CmRenderTexture.h"
#include "CmGpuProgram.h"
#include "CmPlane.h"

#include "boost/function.hpp"
#include "boost/signal.hpp"

namespace CamelotEngine
{
	/** \addtogroup Core
	*  @{
	*/
	/** \addtogroup RenderSystem
	*  @{
	*/

	typedef multimap<UINT8, RenderTarget * >::type RenderTargetPriorityMap;

	class TextureManager;

	/** Defines the functionality of a 3D API
	@remarks
	The RenderSystem class provides a base interface
	which abstracts the general functionality of the 3D API
	e.g. Direct3D or OpenGL. Whilst a few of the general
	methods have implementations, most of this class is
	abstract, requiring a subclass based on a specific API
	to be constructed to provide the full functionality.
	Note there are 2 levels to the interface - one which
	will be used often by the caller of the Ogre library,
	and one which is at a lower level and will be used by the
	other classes provided by Ogre. These lower level
	methods are prefixed with '_' to differentiate them.
	The advanced user of the library may use these lower
	level methods to access the 3D API at a more fundamental
	level (dealing direct with render states and rendering
	primitives), but still benefiting from Ogre's abstraction
	of exactly which 3D API is in use.
	@author
	Steven Streeting
	@version
	1.0
	*/
	class CM_EXPORT RenderSystem
	{
	public:
		/** Default Constructor.
		*/
		RenderSystem();

		/** Destructor.
		*/
		virtual ~RenderSystem();

		/** Returns the name of the rendering system.
		*/
		virtual const String& getName(void) const = 0;

		 /* @brief	Start up the RenderSystem. Call before doing any operations on the render system.  
		 *			Make sure all subsequent calls to the RenderSystem are done from the same thread it was started on.  
		 *
		 * @remark	If you want to access the render system from other threads, call RenderSystem::createRenderContext,
		 * 			set the active context using RenderSystem::setActiveRenderContext and call the render system normally.
		 * 			By default an automatically created primary render context is used.
		 */
		void startUp();

		// TODO - Classes below (shutdown to getErrorDescription) are not yet thread safe


		/** Shutdown the renderer and cleanup resources.
		*/
		void shutdown(void);

		/** Creates a new rendering window.
		@remarks
		This method creates a new rendering window as specified
		by the paramteters. The rendering system could be
		responible for only a single window (e.g. in the case
		of a game), or could be in charge of multiple ones (in the
		case of a level editor). The option to create the window
		as a child of another is therefore given.
		This method will create an appropriate subclass of
		RenderWindow depending on the API and platform implementation.
		@par
		After creation, this window can be retrieved using getRenderTarget().
		@param
		name The name of the window. Used in other methods
		later like setRenderTarget and getRenderTarget.
		@param
		width The width of the new window.
		@param
		height The height of the new window.
		@param
		fullScreen Specify true to make the window full screen
		without borders, title bar or menu bar.
		@param
		miscParams A NameValuePairList describing the other parameters for the new rendering window. 
		Options are case sensitive. Unrecognised parameters will be ignored silently.
		These values might be platform dependent, but these are present for all platforms unless
		indicated otherwise:
		<table>
		<tr>
			<td><b>Key</b></td>
			<td><b>Type/Values</b></td>
			<td><b>Default</b></td>
			<td><b>Description</b></td>
			<td><b>Notes</b></td>
		</tr>
		<tr>
			<td>title</td>
			<td>Any string</td>
			<td>RenderTarget name</td>
			<td>The title of the window that will appear in the title bar</td>
			<td>&nbsp;</td>
		</tr>
		<tr>
			<td>colourDepth</td>
			<td>16, 32</td>
			<td>Desktop depth</td>
			<td>Colour depth of the resulting rendering window; only applies if fullScreen</td>
			<td>Win32 Specific</td>
		</tr>
		<tr>
			<td>left</td>
			<td>Positive integers</td>
			<td>Centred</td>
			<td>Screen x coordinate from left</td>
			<td>&nbsp;</td>
		</tr>
		<tr>
			<td>top</td>
			<td>Positive integers</td>
			<td>Centred</td>
			<td>Screen y coordinate from left</td>
			<td>&nbsp;</td>
		</tr>
		<tr>
			<td>depthBuffer</td>
			<td>true, false</td>
			<td>true</td>
			<td>Use depth buffer</td>
			<td>DirectX9 specific</td>
		</tr>
		<tr>
			<td>externalWindowHandle</td>
			<td>Win32: HWND as integer<br/>
			    GLX: poslong:posint:poslong (display*:screen:windowHandle) or poslong:posint:poslong:poslong (display*:screen:windowHandle:XVisualInfo*)</td>
			<td>0 (none)</td>
			<td>External window handle, for embedding the OGRE render in an existing window</td>
			<td>&nbsp;</td>
		</tr>
		<tr>
			<td>externalGLControl</td>
			<td>true, false</td>
			<td>false</td>
			<td>Let the external window control OpenGL i.e. don't select a pixel format for the window,
			do not change v-sync and do not swap buffer. When set to true, the calling application
			is responsible of OpenGL initialization and buffer swapping. It should also create an
			OpenGL context for its own rendering, Ogre will create one for its use. Then the calling
			application must also enable Ogre OpenGL context before calling any Ogre function and
			restore its OpenGL context after these calls.</td>
			<td>OpenGL specific</td>
		</tr>
		<tr>
			<td>externalGLContext</td>
			<td>Context as Unsigned Long</td>
			<td>0 (create own context)</td>
			<td>Use an externally created GL context</td>
			<td>OpenGL Specific</td>
		</tr>
		<tr>
			<td>parentWindowHandle</td>
			<td>Win32: HWND as integer<br/>
			    GLX: poslong:posint:poslong (display*:screen:windowHandle) or poslong:posint:poslong:poslong (display*:screen:windowHandle:XVisualInfo*)</td>
			<td>0 (none)</td>
			<td>Parent window handle, for embedding the engine in a child of an external window</td>
			<td>&nbsp;</td>
		</tr>
		<tr>
			<td>macAPI</td>
			<td>String: "cocoa" or "carbon"</td>
			<td>"carbon"</td>
			<td>Specifies the type of rendering window on the Mac Platform.</td>
			<td>&nbsp;</td>
		 </tr>
		 <tr>
			<td>macAPICocoaUseNSView</td>
			<td>bool "true" or "false"</td>
			<td>"false"</td>
			<td>On the Mac platform the most diffused method to embed engine in a custom application is to use Interface Builder
				and add to the interface an instance of OgreView.
				The pointer to this instance is then used as "externalWindowHandle".
				However, there are cases where you are NOT using Interface Builder and you get the Cocoa NSView* of an existing interface.
				For example, this is happens when you want to render into a Java/AWT interface.
				In short, by setting this flag to "true" the Ogre::Root::createRenderWindow interprets the "externalWindowHandle" as a NSView*
				instead of an OgreView*. See OgreOSXCocoaView.h/mm.
			</td>
			<td>&nbsp;</td>
		 </tr>
         <tr>
			<td>FSAA</td>
			<td>Positive integer (usually 0, 2, 4, 8, 16)</td>
			<td>0</td>
			<td>Full screen antialiasing factor</td>
			<td>&nbsp;</td>
		</tr>
		<tr>
			<td>FSAAHint</td>
			<td>Depends on RenderSystem and hardware. Currently supports:<br/>
			"Quality": on systems that have an option to prefer higher AA quality over speed, use it</td>
			<td>Blank</td>
			<td>Full screen antialiasing hint</td>
			<td>&nbsp;</td>
		</tr>
		<tr>
			<td>displayFrequency</td>
			<td>Refresh rate in Hertz (e.g. 60, 75, 100)</td>
			<td>Desktop vsync rate</td>
			<td>Display frequency rate, for fullscreen mode</td>
			<td>&nbsp;</td>
		</tr>
		<tr>
			<td>vsync</td>
			<td>true, false</td>
			<td>false</td>
			<td>Synchronize buffer swaps to monitor vsync, eliminating tearing at the expense of a fixed frame rate</td>
			<td>&nbsp;</td>
		</tr>
		<tr>
			<td>vsyncInterval</td>
			<td>1, 2, 3, 4</td>
			<td>1</td>
			<td>If vsync is enabled, the minimum number of vertical blanks that should occur between renders. 
			For example if vsync is enabled, the refresh rate is 60 and this is set to 2, then the
			frame rate will be locked at 30.</td>
			<td>&nbsp;</td>
		</tr>
		<tr>
			<td>border</td>
			<td>none, fixed, resize</td>
			<td>resize</td>
			<td>The type of window border (in windowed mode)</td>
			<td>&nbsp;</td>
		</tr>
		<tr>
			<td>outerDimensions</td>
			<td>true, false</td>
			<td>false</td>
			<td>Whether the width/height is expressed as the size of the 
			outer window, rather than the content area</td>
			<td>&nbsp;</td>
		</tr>
		<tr>
			<td>useNVPerfHUD</td>
			<td>true, false</td>
			<td>false</td>
			<td>Enable the use of nVidia NVPerfHUD</td>
			<td>&nbsp;</td>
		</tr>
		<tr>
			<td>gamma</td>
			<td>true, false</td>
			<td>false</td>
			<td>Enable hardware conversion from linear colour space to gamma
			colour space on rendering to the window.</td>
			<td>&nbsp;</td>
		</tr>
		*/
		RenderWindow* createRenderWindow(const String &name, unsigned int width, unsigned int height, 
			bool fullScreen, const NameValuePairList *miscParams = 0);
		virtual void createRenderWindow_internal(const String &name, unsigned int width, unsigned int height, 
			bool fullScreen, const NameValuePairList& miscParams, AsyncOp& asyncOp) = 0;

		/** Attaches the passed render target to the render system.
		*/
		virtual void attachRenderTarget(RenderTarget &target);

		/** Detaches the render target from the render system.
		@note
		If the render target cannot be found, NULL is returned.
		*/
		virtual void detachRenderTarget(RenderTarget& renderTarget);

		/** Destroys a render window */
		virtual void destroyRenderWindow(RenderWindow* renderWindow);
		/** Destroys a render texture */
		virtual void destroyRenderTexture(RenderTexture* renderTexture);
		/** Destroys a render target of any sort */
		virtual void destroyRenderTarget(RenderTarget* renderTarget);

		/** Defines whether or now fullscreen render windows wait for the vertical blank before flipping buffers.
		@remarks
		By default, all rendering windows wait for a vertical blank (when the CRT beam turns off briefly to move
		from the bottom right of the screen back to the top left) before flipping the screen buffers. This ensures
		that the image you see on the screen is steady. However it restricts the frame rate to the refresh rate of
		the monitor, and can slow the frame rate down. You can speed this up by not waiting for the blank, but
		this has the downside of introducing 'tearing' artefacts where part of the previous frame is still displayed
		as the buffers are switched. Speed vs quality, you choose.
		@note
		Has NO effect on windowed mode render targets. Only affects fullscreen mode.
		@param
		enabled If true, the system waits for vertical blanks - quality over speed. If false it doesn't - speed over quality.
		*/
		void setWaitForVerticalBlank(bool enabled);

		/** Returns true if the system is synchronising frames with the monitor vertical blank.
		*/
		bool getWaitForVerticalBlank(void) const;

		// ------------------------------------------------------------------------
		//                     Internal Rendering Access
		// All methods below here are normally only called by other Camelot classes
		// They can be called by library user if required
		// ------------------------------------------------------------------------

		/**
		 * @brief	Sets a sampler state for the specified texture unit.
		 */
		virtual void setSamplerState(UINT16 texUnit, const SamplerState& samplerState) = 0;
		/** Turns off a texture unit. */
		virtual void disableTextureUnit(UINT16 texUnit);
		/** Disables all texture units from the given unit upwards */
		virtual void disableTextureUnitsFrom(UINT16 texUnit);

		/** Sets the size of points and how they are attenuated with distance.
		@remarks
		When performing point rendering or point sprite rendering,
		point size can be attenuated with distance. The equation for
		doing this is attenuation = 1 / (constant + linear * dist + quadratic * d^2) .
		@par
		For example, to disable distance attenuation (constant screensize) 
		you would set constant to 1, and linear and quadratic to 0. A
		standard perspective attenuation would be 0, 1, 0 respectively.
		*/
		virtual void setPointParameters(float size, bool attenuationEnabled, 
			float constant, float linear, float quadratic, float minSize, float maxSize) = 0;


		/**
		Sets the texture to bind to a given texture unit.

		User processes would not normally call this direct unless rendering
		primitives themselves.

		@param unit The index of the texture unit to modify. Multitexturing 
		hardware can support multiple units (see 
		RenderSystemCapabilites::getNumTextureUnits)
		@param enabled Boolean to turn the unit on/off
		@param texPtr Pointer to the texture to use.
		*/
		virtual void setTexture(UINT16 unit, bool enabled, 
			const TexturePtr &texPtr) = 0;

		/** Sets the global blending factors for combining subsequent renders with the existing frame contents.
		The result of the blending operation is:</p>
		<p align="center">final = (texture * sourceFactor) + (pixel * destFactor)</p>
		Each of the factors is specified as one of a number of options, as specified in the SceneBlendFactor
		enumerated type.
		By changing the operation you can change addition between the source and destination pixels to a different operator.
		@param sourceFactor The source factor in the above calculation, i.e. multiplied by the texture colour components.
		@param destFactor The destination factor in the above calculation, i.e. multiplied by the pixel colour components.
		@param op The blend operation mode for combining pixels
		*/
		virtual void setSceneBlending(SceneBlendFactor sourceFactor, SceneBlendFactor destFactor, SceneBlendOperation op = SBO_ADD) = 0;

		/** Sets the global blending factors for combining subsequent renders with the existing frame contents.
		The result of the blending operation is:</p>
		<p align="center">final = (texture * sourceFactor) + (pixel * destFactor)</p>
		Each of the factors is specified as one of a number of options, as specified in the SceneBlendFactor
		enumerated type.
		@param sourceFactor The source factor in the above calculation, i.e. multiplied by the texture colour components.
		@param destFactor The destination factor in the above calculation, i.e. multiplied by the pixel colour components.
		@param sourceFactorAlpha The source factor in the above calculation for the alpha channel, i.e. multiplied by the texture alpha components.
		@param destFactorAlpha The destination factor in the above calculation for the alpha channel, i.e. multiplied by the pixel alpha components.
		@param op The blend operation mode for combining pixels
		@param alphaOp The blend operation mode for combining pixel alpha values
		*/
		virtual void setSeparateSceneBlending(SceneBlendFactor sourceFactor, SceneBlendFactor destFactor, SceneBlendFactor sourceFactorAlpha, 
			SceneBlendFactor destFactorAlpha, SceneBlendOperation op = SBO_ADD, SceneBlendOperation alphaOp = SBO_ADD) = 0;

		/** Sets the global alpha rejection approach for future renders.
		By default images are rendered regardless of texture alpha. This method lets you change that.
		@param func The comparison function which must pass for a pixel to be written.
		@param val The value to compare each pixels alpha value to (0-255)
		@param alphaToCoverage Whether to enable alpha to coverage, if supported
		*/
		virtual void setAlphaRejectSettings(CompareFunction func, unsigned char value, bool alphaToCoverage) = 0;

		/**
		* Signifies the beginning of a frame, i.e. the start of rendering on a single viewport. Will occur
		* several times per complete frame if multiple viewports exist.
		*/
		virtual void beginFrame(void) = 0;
		
		/**
		* Ends rendering of a frame to the current viewport.
		*/
		virtual void endFrame(void) = 0;
		/**
		Sets the provided viewport as the active one for future
		rendering operations. This viewport is aware of it's own
		camera and render target. Must be implemented by subclass.

		@param target Viewport to render to.
		*/
		virtual void setViewport(const Viewport& vp) = 0;
		/** Get the current active viewport for rendering. */
		virtual Viewport getViewport(void);

		/** Sets the culling mode for the render system based on the 'vertex winding'.
		A typical way for the rendering engine to cull triangles is based on the
		'vertex winding' of triangles. Vertex winding refers to the direction in
		which the vertices are passed or indexed to in the rendering operation as viewed
		from the camera, and will wither be clockwise or anticlockwise (that's 'counterclockwise' for
		you Americans out there ;) The default is CULL_CLOCKWISE i.e. that only triangles whose vertices
		are passed/indexed in anticlockwise order are rendered - this is a common approach and is used in 3D studio models
		for example. You can alter this culling mode if you wish but it is not advised unless you know what you are doing.
		You may wish to use the CULL_NONE option for mesh data that you cull yourself where the vertex
		winding is uncertain.
		*/
		virtual void setCullingMode(CullingMode mode) = 0;

		virtual CullingMode getCullingMode(void) const;

		/** Sets the mode of operation for depth buffer tests from this point onwards.
		Sometimes you may wish to alter the behaviour of the depth buffer to achieve
		special effects. Because it's unlikely that you'll set these options for an entire frame,
		but rather use them to tweak settings between rendering objects, this is an internal
		method (indicated by the '_' prefix) which will be used by a SceneManager implementation
		rather than directly from the client application.
		If this method is never called the settings are automatically the same as the default parameters.
		@param depthTest If true, the depth buffer is tested for each pixel and the frame buffer is only updated
		if the depth function test succeeds. If false, no test is performed and pixels are always written.
		@param depthWrite If true, the depth buffer is updated with the depth of the new pixel if the depth test succeeds.
		If false, the depth buffer is left unchanged even if a new pixel is written.
		@param depthFunction Sets the function required for the depth test.
		*/
		virtual void setDepthBufferParams(bool depthTest = true, bool depthWrite = true, CompareFunction depthFunction = CMPF_LESS_EQUAL) = 0;

		/** Sets whether or not the depth buffer check is performed before a pixel write.
		@param enabled If true, the depth buffer is tested for each pixel and the frame buffer is only updated
		if the depth function test succeeds. If false, no test is performed and pixels are always written.
		*/
		virtual void setDepthBufferCheckEnabled(bool enabled = true) = 0;

		/** Sets whether or not the depth buffer is updated after a pixel write.
		@param enabled If true, the depth buffer is updated with the depth of the new pixel if the depth test succeeds.
		If false, the depth buffer is left unchanged even if a new pixel is written.
		*/
		virtual void setDepthBufferWriteEnabled(bool enabled = true) = 0;

		/** Sets the comparison function for the depth buffer check.
		Advanced use only - allows you to choose the function applied to compare the depth values of
		new and existing pixels in the depth buffer. Only an issue if the deoth buffer check is enabled
		(see _setDepthBufferCheckEnabled)
		@param  func The comparison between the new depth and the existing depth which must return true
		for the new pixel to be written.
		*/
		virtual void setDepthBufferFunction(CompareFunction func = CMPF_LESS_EQUAL) = 0;

		/** Sets whether or not colour buffer writing is enabled, and for which channels. 
		@remarks
		For some advanced effects, you may wish to turn off the writing of certain colour
		channels, or even all of the colour channels so that only the depth buffer is updated
		in a rendering pass. However, the chances are that you really want to use this option
		through the Material class.
		@param red, green, blue, alpha Whether writing is enabled for each of the 4 colour channels. */
		virtual void setColorBufferWriteEnabled(bool red, bool green, bool blue, bool alpha) = 0;

		/** Sets the depth bias, NB you should use the Material version of this. 
		@remarks
		When polygons are coplanar, you can get problems with 'depth fighting' where
		the pixels from the two polys compete for the same screen pixel. This is particularly
		a problem for decals (polys attached to another surface to represent details such as
		bulletholes etc.).
		@par
		A way to combat this problem is to use a depth bias to adjust the depth buffer value
		used for the decal such that it is slightly higher than the true value, ensuring that
		the decal appears on top.
		@note
		The final bias value is a combination of a constant bias and a bias proportional
		to the maximum depth slope of the polygon being rendered. The final bias
		is constantBias + slopeScaleBias * maxslope. Slope scale biasing is
		generally preferable but is not available on older hardware.
		@param constantBias The constant bias value, expressed as a value in 
		homogeneous depth coordinates.
		@param slopeScaleBias The bias value which is factored by the maximum slope
		of the polygon, see the description above. This is not supported by all
		cards.

		*/
		virtual void setDepthBias(float constantBias, float slopeScaleBias = 0.0f) = 0;

		/** Sets how to rasterise triangles, as points, wireframe or solid polys. */
		virtual void setPolygonMode(PolygonMode level) = 0;

		/** Turns stencil buffer checking on or off. 
		@remarks
		Stencilling (masking off areas of the rendering target based on the stencil 
		buffer) can be turned on or off using this method. By default, stencilling is
		disabled.
		*/
		virtual void setStencilCheckEnabled(bool enabled) = 0;

		/** This method allows you to set all the stencil buffer parameters in one call.
		@remarks
		The stencil buffer is used to mask out pixels in the render target, allowing
		you to do effects like mirrors, cut-outs, stencil shadows and more. Each of
		your batches of rendering is likely to ignore the stencil buffer, 
		update it with new values, or apply it to mask the output of the render.
		The stencil test is:<PRE>
		(Reference Value & Mask) CompareFunction (Stencil Buffer Value & Mask)</PRE>
		The result of this will cause one of 3 actions depending on whether the test fails,
		succeeds but with the depth buffer check still failing, or succeeds with the
		depth buffer check passing too.
		@par
		Unlike other render states, stencilling is left for the application to turn
		on and off when it requires. This is because you are likely to want to change
		parameters between batches of arbitrary objects and control the ordering yourself.
		In order to batch things this way, you'll want to use OGRE's separate render queue
		groups (see RenderQueue) and register a RenderQueueListener to get notifications
		between batches.
		@par
		There are individual state change methods for each of the parameters set using 
		this method. 
		Note that the default values in this method represent the defaults at system 
		start up too.
		@param func The comparison function applied.
		@param refValue The reference value used in the comparison
		@param mask The bitmask applied to both the stencil value and the reference value 
		before comparison
		@param stencilFailOp The action to perform when the stencil check fails
		@param depthFailOp The action to perform when the stencil check passes, but the
		depth buffer check still fails
		@param passOp The action to take when both the stencil and depth check pass.
		@param twoSidedOperation If set to true, then if you render both back and front faces 
		(you'll have to turn off culling) then these parameters will apply for front faces, 
		and the inverse of them will happen for back faces (keep remains the same).
		*/
		virtual void setStencilBufferParams(CompareFunction func = CMPF_ALWAYS_PASS, 
			UINT32 refValue = 0, UINT32 mask = 0xFFFFFFFF, 
			StencilOperation stencilFailOp = SOP_KEEP, 
			StencilOperation depthFailOp = SOP_KEEP,
			StencilOperation passOp = SOP_KEEP, 
			bool twoSidedOperation = false) = 0;

		/** Sets the current vertex declaration, ie the source of vertex data. */
		virtual void setVertexDeclaration(VertexDeclarationPtr decl) = 0;
		/** Sets the current vertex buffer binding state. */
		virtual void setVertexBufferBinding(VertexBufferBinding* binding) = 0;

		/**
		Render something to the active viewport.

		Low-level rendering interface to perform rendering
		operations. Unlikely to be used directly by client
		applications, since the SceneManager and various support
		classes will be responsible for calling this method.
		Can only be called between _beginScene and _endScene

		@param op A rendering operation instance, which contains
		details of the operation to be performed.
		*/
		virtual void render(const RenderOperation& op);

		/** Gets the capabilities of the render system. */
		const RenderSystemCapabilities* getCapabilities(void) const;

		/** Returns the driver version.
		*/
		virtual const DriverVersion& getDriverVersion(void) const;

		/** Binds a given GpuProgram (but not the parameters). 
		@remarks Only one GpuProgram of each type can be bound at once, binding another
		one will simply replace the existing one.
		*/
		virtual void bindGpuProgram(GpuProgramHandle prg);

		/** Bind Gpu program parameters.
		@param gptype The type of program to bind the parameters to
		@param params The parameters to bind
		@param variabilityMask A mask of GpuParamVariability identifying which params need binding
		*/
		virtual void bindGpuProgramParameters(GpuProgramType gptype, 
			GpuProgramParametersSharedPtr params, UINT16 variabilityMask) = 0;

		/** Unbinds GpuPrograms of a given GpuProgramType.
		@remarks
		This returns the pipeline to fixed-function processing for this type.
		*/
		virtual void unbindGpuProgram(GpuProgramType gptype);

		/** Returns whether or not a Gpu program of the given type is currently bound. */
		virtual bool isGpuProgramBound(GpuProgramType gptype);

		/** Sets the user clipping region.
		*/
		virtual void setClipPlanes(const PlaneList& clipPlanes);

		/** Add a user clipping plane. */
		virtual void addClipPlane (const Plane& p);
		/** Add a user clipping plane. */
		virtual void addClipPlane (float A, float B, float C, float D);

		/** Clears the user clipping region.
		*/
		virtual void resetClipPlanes();

		/** Internal method for swapping all the buffers on all render targets,
		if _updateAllRenderTargets was called with a 'false' parameter. */
		virtual void swapAllRenderTargetBuffers(bool waitForVsync = true);

		/** Sets whether or not vertex windings set should be inverted; this can be important
		for rendering reflections. */
		virtual void setInvertVertexWinding_(bool invert);

		/** Indicates whether or not the vertex windings set will be inverted for the current render (e.g. reflections)
		@see RenderSystem::setInvertVertexWinding
		*/
		virtual bool getInvertVertexWinding(void) const;

		/** Sets the 'scissor region' ie the region of the target in which rendering can take place.
		@remarks
		This method allows you to 'mask off' rendering in all but a given rectangular area
		as identified by the parameters to this method.
		@note
		Not all systems support this method. Check the RenderSystemCapabilities for the
		RSC_SCISSOR_TEST capability to see if it is supported.
		@param enabled True to enable the scissor test, false to disable it.
		@param left, top, right, bottom The location of the corners of the rectangle, expressed in
		<i>pixels</i>.
		*/
		virtual void setScissorTest(bool enabled, UINT32 left = 0, UINT32 top = 0, 
			UINT32 right = 800, UINT32 bottom = 600) = 0;

		/** Clears one or more frame buffers on the active render target. 
		@param buffers Combination of one or more elements of FrameBufferType
		denoting which buffers are to be cleared
		@param colour The colour to clear the colour buffer with, if enabled
		@param depth The value to initialise the depth buffer with, if enabled
		@param stencil The value to initialise the stencil buffer with, if enabled.
		*/
		virtual void clearFrameBuffer(unsigned int buffers, 
			const Color& color = Color::Black, 
			float depth = 1.0f, unsigned short stencil = 0) = 0;

		/**
         * Set current render target to target, enabling its device context if needed
         */
        virtual void setRenderTarget(RenderTarget *target) = 0;

		/************************************************************************/
		/* 								UTILITY METHODS                    		*/
		/************************************************************************/

		/** Get the native VertexElementType for a compact 32-bit colour value
		for this rendersystem.
		*/
		virtual VertexElementType getColorVertexElementType(void) const = 0;

		/** Converts a uniform projection matrix to suitable for this render system.
		@remarks
		Because different APIs have different requirements (some incompatible) for the
		projection matrix, this method allows each to implement their own correctly and pass
		back a generic Camelot matrix for storage in the engine.
		*/
		virtual void convertProjectionMatrix(const Matrix4& matrix,
			Matrix4& dest, bool forGpuProgram = false) = 0;

		/** Returns the horizontal texel offset value required for mapping 
		texel origins to pixel origins in this rendersystem.
		@remarks
		Since rendersystems sometimes disagree on the origin of a texel, 
		mapping from texels to pixels can sometimes be problematic to 
		implement generically. This method allows you to retrieve the offset
		required to map the origin of a texel to the origin of a pixel in
		the horizontal direction.
		*/
		virtual float getHorizontalTexelOffset(void) = 0;

		/** Returns the vertical texel offset value required for mapping 
		texel origins to pixel origins in this rendersystem.
		@remarks
		Since rendersystems sometimes disagree on the origin of a texel, 
		mapping from texels to pixels can sometimes be problematic to 
		implement generically. This method allows you to retrieve the offset
		required to map the origin of a texel to the origin of a pixel in
		the vertical direction.
		*/
		virtual float getVerticalTexelOffset(void) = 0;

		/** Gets the minimum (closest) depth value to be used when rendering
		using identity transforms.
		@remarks
		When using identity transforms you can manually set the depth
		of a vertex; however the input values required differ per
		rendersystem. This method lets you retrieve the correct value.
		@see Renderable::getUseIdentityView, Renderable::getUseIdentityProjection
		*/
		virtual float getMinimumDepthInputValue(void) = 0;

		/** Gets the maximum (farthest) depth value to be used when rendering
		using identity transforms.
		@remarks
		When using identity transforms you can manually set the depth
		of a vertex; however the input values required differ per
		rendersystem. This method lets you retrieve the correct value.
		@see Renderable::getUseIdentityView, Renderable::getUseIdentityProjection
		*/
		virtual float getMaximumDepthInputValue(void) = 0;
		/************************************************************************/
		/* 						INTERNAL DATA & METHODS                      	*/
		/************************************************************************/
	protected:
		/** The render targets. */
		vector<RenderTarget*>::type mRenderTargets;
		/** The render targets, ordered by priority. */
		RenderTargetPriorityMap mPrioritisedRenderTargets;
		/** The Active render target. */
		RenderTarget * mActiveRenderTarget;
		/** The Active GPU programs and gpu program parameters*/
		GpuProgramParametersSharedPtr mActiveVertexGpuProgramParameters;
		GpuProgramParametersSharedPtr mActiveGeometryGpuProgramParameters;
		GpuProgramParametersSharedPtr mActiveFragmentGpuProgramParameters;

		// Active viewport (dest for future rendering operations)
		Viewport mActiveViewport;

		CullingMode mCullingMode;

		bool mVsync;
		unsigned int mVSyncInterval;

		bool mInvertVertexWinding;

		/// Texture units from this upwards are disabled
		UINT16 mDisabledTexUnitsFrom;

		bool mVertexProgramBound;
		bool mGeometryProgramBound;
		bool mFragmentProgramBound;

		// Recording user clip planes
		PlaneList mClipPlanes;
		// Indicator that we need to re-set the clip planes on next render call
		bool mClipPlanesDirty;

		/// Used to store the capabilities of the graphics card
		RenderSystemCapabilities* mCurrentCapabilities;

		virtual void startUp_internal();
		virtual void shutdown_internal();

		/// Internal method used to set the underlying clip planes when needed
		virtual void setClipPlanesImpl(const PlaneList& clipPlanes) = 0;

		/** Query the real capabilities of the GPU and driver in the RenderSystem*/
		virtual RenderSystemCapabilities* createRenderSystemCapabilities() const = 0;

		/** Initialize the render system from the capabilities*/
		virtual void initialiseFromRenderSystemCapabilities(RenderSystemCapabilities* caps, RenderTarget* primary) = 0;

		/**	Create a MultiRenderTarget, which is a render target that renders to multiple RenderTextures
		at once. Surfaces can be bound and unbound at will.
		This fails if mCapabilities->getNumMultiRenderTargets() is smaller than 2.
		*/
		virtual MultiRenderTarget * createMultiRenderTarget(const String & name) = 0; 

		/** Returns a description of an error code.
		*/
		virtual String getErrorDescription(long errorNumber) const = 0;

		DriverVersion mDriverVersion;

		/************************************************************************/
		/* 								THREADING	                     		*/
		/************************************************************************/

		class RenderWorkerFunc CM_THREAD_WORKER_INHERIT
		{
		public:
			RenderWorkerFunc(RenderSystem* rs);

			void operator()();

		private:
			RenderSystem* mRS;
		};

		RenderWorkerFunc* mRenderThreadFunc;
		bool mRenderThreadShutdown;

		CM_THREAD_ID_TYPE mRenderThreadId;
		CM_THREAD_SYNCHRONISER(mRenderThreadStartCondition)
		CM_MUTEX(mRenderThreadStartMutex)
		CM_MUTEX(mCommandQueueMutex)
		CM_THREAD_SYNCHRONISER(mCommandReadyCondition)
		CM_MUTEX(mCommandNotifyMutex)
		CM_THREAD_SYNCHRONISER(mCommandCompleteCondition)

#if CM_THREAD_SUPPORT
		CM_THREAD_TYPE* mRenderThread;
#endif

		CommandQueue* mCommandQueue;

		UINT32 mMaxCommandNotifyId; // ID that will be assigned to the next command with a notifier callback
		vector<UINT32>::type mCommandsCompleted; // Completed commands that have notifier callbacks set up

		/**
		 * @brief	Initializes a separate render thread. Should only be called once.
		 */
		void initRenderThread();

		/**
		 * @brief	Main function of the render thread. Called once thread is started.
		 */
		void runRenderThread();

		/**
		 * @brief	Shutdowns the render thread. It will complete all ready commands
		 * 			before shutdown.
		 */
		void shutdownRenderThread();

		/**
		 * @brief	Throws an exception if current thread isn't the render thread;
		 */
		void throwIfNotRenderThread() const;

		/**
		 * @brief	Blocks the calling thread until the command with the specified ID completes.
		 * 			Make sure that the specified ID actually exists, otherwise this will block forever.
		 */
		void blockUntilCommandCompleted(UINT32 commandId);

		/**
		 * @brief	Callback called by the command list when a specific command finishes executing.
		 * 			This is only called on commands that have a special notify on complete flag set.
		 *
		 * @param	commandId	Identifier for the command.
		 */
		void commandCompletedNotify(UINT32 commandId);

	public:
		/**
		 * @brief	Returns the id of the render thread. If a separate render thread
		 * 			is not used, then it returns the id of the thread RenderSystem
		 * 			was initialized on.
		 */
		CM_THREAD_ID_TYPE getRenderThreadId() const { return mRenderThreadId; }

		/**
		 * @brief	Creates a new render system context that you can use for rendering on 
		 * 			a non-render thread. You can have as many of these as you wish, the only limitation
		 * 			is that you do not use a single instance on more than one thread. Each thread
		 * 			requires its own context. The context will be bound to the thread you call this method on.
		 */
		DeferredRenderContextPtr createDeferredContext();

		/**
		 * @brief	Queues a new command that will be added to the global command queue. You are allowed to call this from any thread,
		 * 			however be aware that it involves possibly slow synchronization primitives, so limit your usage.
		 * 			
		 * @param	blockUntilComplete If true the thread will be blocked until the command executes. Be aware that there be many commands queued before it
		 * 							   and they all need to be executed in order before the current command is reached, which might take a long time.
		 * 	
		 * @see		CommandQueue::queueReturn
		 */
		AsyncOp queueReturnCommand(boost::function<void(AsyncOp&)> commandCallback, bool blockUntilComplete = false);

		/**
		* @brief	Queues a new command that will be added to the global command queue.You are allowed to call this from any thread,
		 * 			however be aware that it involves possibly slow synchronization primitives, so limit your usage.
		 * 	
		 * @param	blockUntilComplete If true the thread will be blocked until the command executes. Be aware that there be many commands queued before it
		 * 							   and they all need to be executed in order before the current command is reached, which might take a long time.
		 * @see		CommandQueue::queue
		 */
		void queueCommand(boost::function<void()> commandCallback, bool blockUntilComplete = false);
	};
	/** @} */
	/** @} */
}

#endif