khanat-code-old/code/nel/include/nel/3d/driver.h

1277 lines
50 KiB
C
Raw Normal View History

// NeL - MMORPG Framework <http://dev.ryzom.com/projects/nel/>
// Copyright (C) 2010 Winch Gate Property Limited
//
// This program is free software: you can redistribute it and/or modify
// it under the terms of the GNU Affero General Public License as
// published by the Free Software Foundation, either version 3 of the
// License, or (at your option) any later version.
//
// This program is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU Affero General Public License for more details.
//
// You should have received a copy of the GNU Affero General Public License
// along with this program. If not, see <http://www.gnu.org/licenses/>.
#ifndef NL_DRV_H
#define NL_DRV_H
#include "nel/misc/types_nl.h"
#include "nel/misc/common.h"
#include "nel/misc/smart_ptr.h"
#include "nel/misc/rgba.h"
#include "nel/misc/matrix.h"
#include "nel/misc/stream.h"
#include "nel/misc/uv.h"
#include "nel/misc/hierarchical_timer.h"
#include "nel/3d/texture.h"
#include "nel/3d/shader.h"
#include "nel/3d/vertex_buffer.h"
#include "nel/3d/index_buffer.h"
#include "nel/3d/vertex_program.h"
#include "nel/3d/material.h"
#include "nel/misc/mutex.h"
#include "nel/3d/primitive_profile.h"
#include <vector>
#include <list>
namespace NLMISC
{
class IEventEmitter;
struct IMouseDevice;
struct IKeyboardDevice;
struct IInputDeviceManager;
class CRect;
class CLog;
}
namespace NL3D
{
using NLMISC::CRefPtr;
using NLMISC::CRefCount;
using NLMISC::CSmartPtr;
using NLMISC::CRGBA;
using NLMISC::CVector;
using NLMISC::CMatrix;
using NLMISC::CSynchronized;
class CMaterial;
class CIndexBuffer;
class CLight;
class CScissor;
class CViewport;
struct CMonitorColorProperties;
struct IOcclusionQuery;
// ****************************************************************************
/// A Graphic Mode descriptor.
struct GfxMode
{
bool OffScreen;
bool Windowed;
uint16 Width;
uint16 Height;
uint8 Depth;
uint Frequency; // In hz. Default is Windows selection
sint8 AntiAlias; // -1 = no AA, 0 = max, 2 = 2x sample, 4 = 4x sample, ...
GfxMode()
{
OffScreen=false;
Windowed=false;
Width = 0;
Height = 0;
Depth = 0;
Frequency = 0;
AntiAlias = -1;
}
GfxMode(uint16 w, uint16 h, uint8 d, bool windowed = true, bool offscreen = false, uint frequency = 0, sint8 aa = -1);
};
// ****************************************************************************
// Exceptions.
struct EBadDisplay : public NLMISC::Exception
{
EBadDisplay(const std::string &reason) : Exception(reason) { }
};
// ****************************************************************************
typedef void (*emptyProc)(void);
// ****************************************************************************
// *** IMPORTANT ********************
// *** IF YOU MODIFY THE STRUCTURE OF THIS CLASS, PLEASE INCREMENT IDriver::InterfaceVersion TO INVALIDATE OLD DRIVER DLL
// **********************************
//
// * Driver implementation notes:
// *
// * Driver implementation must save monitor color parameters at initialization and restore it at release.
class IDriver : public NLMISC::CRefCount
{
public:
/// Version of the driver interface. To increment when the interface change.
static const uint32 InterfaceVersion;
public:
enum TMessageBoxId { okId=0, yesId, noId, abortId, retryId, cancelId, ignoreId, idCount };
enum TMessageBoxType { okType=0, okCancelType, yesNoType, abortRetryIgnoreType, yesNoCancelType, retryCancelType, typeCount };
enum TMessageBoxIcon { noIcon=0, handIcon, questionIcon, exclamationIcon, asteriskIcon, warningIcon, errorIcon, informationIcon, stopIcon, iconCount };
enum TCullMode { CCW = 0, CW };
enum TStencilOp { keep = 0, zero, replace, incr, decr, invert };
enum TStencilFunc { never = 0, less, lessequal, equal, notequal, greaterequal, greater, always};
/**
* Driver's polygon modes.
*
* \see setPolygonMode, getPolygonMode
*/
enum TPolygonMode { Filled=0, Line, Point };
/**
* Driver Max matrix count.
* Kept for backward compatibility. Suppose any Hardware VertexProgram can handle only 16 matrix
*
*/
enum TMatrixCount { MaxModelMatrix= 16 };
protected:
CSynchronized<TTexDrvInfoPtrMap> _SyncTexDrvInfos;
TTexDrvSharePtrList _TexDrvShares;
TMatDrvInfoPtrList _MatDrvInfos;
TVBDrvInfoPtrList _VBDrvInfos;
TIBDrvInfoPtrList _IBDrvInfos;
TPolygonMode _PolygonMode;
TVtxPrgDrvInfoPtrList _VtxPrgDrvInfos;
TShaderDrvInfoPtrList _ShaderDrvInfos;
uint _ResetCounter;
public:
IDriver(void);
virtual ~IDriver(void);
virtual bool init (uint windowIcon = 0, emptyProc exitFunc = 0)=0;
// Test if the device is lost. Can only happen with D3D.
// The calling application may skip some part of its rendering when it is the case (this is not a requirement, but may save cpu for other applications)
virtual bool isLost() const = 0;
/// \name Disable Hardware Feature
/** Disable some Feature that may be supported by the Hardware
* Call before setDisplay() to work properly
*/
// @{
virtual void disableHardwareVertexProgram()=0;
virtual void disableHardwareVertexArrayAGP()=0;
virtual void disableHardwareTextureShader()=0;
// @}
// first param is the associated window.
// Must be a HWND for Windows (WIN32).
virtual bool setDisplay(nlWindow wnd, const GfxMode& mode, bool show = true, bool resizeable = true) throw(EBadDisplay)=0;
// Must be called after a setDisplay that initialize the mode
virtual bool setMode(const GfxMode& mode)=0;
virtual bool getModes(std::vector<GfxMode> &modes)=0;
/// Set the title of the NeL window
virtual void setWindowTitle(const ucstring &title)=0;
/// Set icon(s) of the NeL window
virtual void setWindowIcon(const std::vector<NLMISC::CBitmap> &bitmaps)=0;
/// Set the position of the NeL window
virtual void setWindowPos(sint32 x, sint32 y)=0;
/// Show or hide the NeL window
virtual void showWindow(bool show)=0;
/// return the current screen mode (if we are in windowed, return the screen mode behind the window)
virtual bool getCurrentScreenMode(GfxMode &mode)=0;
/// enter/leave the dialog mode
virtual void beginDialogMode() =0;
virtual void endDialogMode() =0;
// Return is the associated window information. (Implementation dependent)
// Must be a HWND for Windows (WIN32).
virtual nlWindow getDisplay() =0;
/**
* Setup monitor color properties.
*
* Return false if setup failed.
*/
virtual bool setMonitorColorProperties (const CMonitorColorProperties &properties) = 0;
// Return is the associated default window proc for the driver. (Implementation dependent)
// Must be a void GlWndProc(IDriver *driver, HWND hWnd, UINT message, WPARAM wParam, LPARAM lParam) for Windows (WIN32).
virtual emptyProc getWindowProc() = 0;
/// Before rendering via a driver in a thread, must activate() (per thread).
virtual bool activate(void) = 0;
/// Get the number of texture stage available, for multi texturing (Normal material shaders). Valid only after setDisplay().
virtual uint getNbTextureStages() const = 0;
/** is the texture is set up in the driver
* NB: this method is thread safe.
*/
virtual bool isTextureExist(const ITexture&tex)=0;
virtual NLMISC::IEventEmitter* getEventEmitter(void) = 0;
/* Clear the current target surface pixels. The function ignores the viewport settings but uses the scissor. */
virtual bool clear2D(CRGBA rgba) = 0;
/* Clear the current target surface zbuffer. The function ignores the viewport settings but uses the scissor. */
virtual bool clearZBuffer(float zval=1) = 0;
/* Clear the current target surface stencil buffer. The function ignores the viewport settings but uses the scissor. */
virtual bool clearStencilBuffer(float stencilval=0) = 0;
/// Set the color mask filter through where the operation done will pass
virtual void setColorMask (bool bRed, bool bGreen, bool bBlue, bool bAlpha) = 0;
/** Set depth range. Depth range specify a linear mapping from device z coordinates (in the [-1, 1] range) to window coordinates (in the [0, 1] range)
* This mapping occurs after clipping of primitives and division by w of vertices coordinates.
* Default depth range is [0, 1].
* NB : znear should be different from zfar or an assertion is raised
*/
virtual void setDepthRange(float znear, float zfar) = 0;
// Get the current depth range
virtual void getDepthRange(float &znear, float &zfar) const = 0;
/** setup a texture, generate and upload if needed. same as setupTextureEx(tex, true, dummy);
*/
virtual bool setupTexture(ITexture& tex) = 0;
/** setup a texture in the driver.
* \param bUpload if true the texture is created and uploaded to VRAM, if false the texture is only created
* it is useful for the async upload texture to only create the texture and then make invalidate to upload
* small piece each frame. There is ONE case where bUpload is forced to be true inside the method: if the texture
* must be converted to RGBA. \see bAllUploaded
* \param bAllUploaded true if any upload arise (texture invalid, must convert texture etc...).
* \param bMustRecreateSharedTexture if true and if the texture supportSharing, then the texture is recreated
* (and uploaded if bUpload==true) into the shared DrvInfo (if found). Default setup (false) imply that the DrvInfo
* is only bound to tex (thus creating and uploading nothing)
* NB: the texture must be at least touch()-ed for the recreate to work.
*/
virtual bool setupTextureEx (ITexture& tex, bool bUpload, bool& bAllUploaded,
bool bMustRecreateSharedTexture= false) = 0;
/** The texture must be created or uploadTexture do nothing.
* These function can be used to upload piece by piece a texture. Use it in conjunction with setupTextureEx(..., false);
* For compressed textures, the rect must aligned on pixel block. (a block of pixel size is 4x4 pixels).
*/
virtual bool uploadTexture (ITexture& tex, NLMISC::CRect& rect, uint8 nNumMipMap) = 0;
virtual bool uploadTextureCube (ITexture& tex, NLMISC::CRect& rect, uint8 nNumMipMap, uint8 nNumFace) = 0;
/**
* Invalidate shared texture
*/
bool invalidateShareTexture (ITexture &);
/**
* Get the driver share texture name
*/
static void getTextureShareName (const ITexture& tex, std::string &output);
/** if true force all the uncompressed RGBA 32 bits and RGBA 24 bits texture to be DXTC5 compressed.
* Do this only during upload if ITexture::allowDegradation() is true and if ITexture::UploadFormat is "Automatic"
* and if bitmap format is RGBA.
*/
virtual void forceDXTCCompression(bool dxtcComp)=0;
/** if !=1, force mostly all the textures (but TextureFonts lightmaps, interfaces etc..)
* to be divided by Divisor (2, 4, 8...)
* Default is 1.
* NB: this is done only on TextureFile
*/
virtual void forceTextureResize(uint divisor)=0;
/** Sets enforcement of native fragment programs. This is by default enabled.
*
* \param nativeOnly If set to false, fragment programs don't need to be native to stay loaded,
* otherwise (aka if true) they will be purged.
*/
virtual void forceNativeFragmentPrograms(bool nativeOnly) = 0;
virtual bool setupMaterial(CMaterial& mat)=0;
/**
* Activate a shader, NULL to disable the current shader.
*/
virtual bool activeShader(CShader *shd)=0;
/** Special for Faster Specular Setup. Call this between lot of primitives rendered with Specular Materials.
* Visual Errors may arise if you don't correctly call endSpecularBatch().
*/
virtual void startSpecularBatch()=0;
virtual void endSpecularBatch()=0;
/// \name Material multipass.
/** NB: setupMaterial() must be called before those methods.
* NB: This is intended to be use with the rendering of simple primitives.
* NB: Other render calls performs the needed setup automatically
*/
// @{
/// init multipass for _CurrentMaterial. return number of pass required to render this material.
virtual sint beginMaterialMultiPass() = 0;
/// active the ith pass of this material.
virtual void setupMaterialPass(uint pass) = 0;
/// end multipass for this material.
virtual void endMaterialMultiPass() = 0;
// @}
// Setup the camera mode as a perspective/ortho camera. NB: znear and zfar must be >0 (if perspective).
virtual void setFrustum(float left, float right, float bottom, float top, float znear, float zfar, bool perspective=true)=0;
virtual void setFrustumMatrix(CMatrix &frust)=0;
virtual CMatrix getFrustumMatrix()=0;
virtual float getClipSpaceZMin() const = 0;
/** setup the view matrix (inverse of camera matrix).
*
* NB: you must setupViewMatrix() BEFORE setupModelMatrix(), or else undefined results.
*/
virtual void setupViewMatrix(const CMatrix& mtx)=0;
/** setup the view matrix (inverse of camera matrix).
* Extended: give a cameraPos (mtx.Pos() is not taken into account but for getViewMatrix()),
* so the driver use it to remove translation from all ModelMatrixes (and lights pos).
* This approach improves greatly ZBuffer precision.
*
* This is transparent to user, and getViewMatrix() return mtx (as in setupViewMatrix()).
*
* NB: you must setupViewMatrixEx() BEFORE setupModelMatrix(), or else undefined results.
*
* \param mtx the same view matrix (still with correct "inversed" camera position) as if passed in setupViewMatrix()
* \param cameraPos position of the camera (before inversion, ie mtx.getPos()!=cameraPos ).
*/
virtual void setupViewMatrixEx(const CMatrix& mtx, const CVector &cameraPos)=0;
/** setup the model matrix.
*
* NB: you must setupModelMatrix() AFTER setupViewMatrix(), or else undefined results.
*/
virtual void setupModelMatrix(const CMatrix& mtx)=0;
virtual CMatrix getViewMatrix(void)const=0;
/** Force input normal to be normalized by the driver. default is false.
* NB: driver force the normalization himself if:
* - current Model matrix has a scale.
*/
virtual void forceNormalize(bool normalize)=0;
/** return the forceNormalize() state.
*/
virtual bool isForceNormalize() const =0;
/** Get max number of per stage constant that can be used simultaneously.
* This will usually match the number of texture stages, but with a D3D driver, this feature is not available most of the time
* so it is emulated. If pixel shaders are available this will be fully supported.
*/
virtual void getNumPerStageConstant(uint &lightedMaterial, uint &unlightedMaterial) const = 0;
/** return true if driver support VertexBufferHard.
*/
virtual bool supportVertexBufferHard() const =0;
/** return true if volatile vertex buffer are supported. (e.g a vertex buffer which can be created with the flag CVertexBuffer::AGPVolatile or CVertexBuffer::RAMVolatile)
* If these are not supported, a RAM vb is created instead (transparent to user)
*/
virtual bool supportVolatileVertexBuffer() const = 0;
/** return true if driver support indices offset. That is, allow to specify a constant value that is added to each
* index in current active active index buffer when rendering indexed primitives
*/
virtual bool supportIndexOffset() const = 0;
/** return true if driver support VertexBufferHard, but vbHard->unlock() are slow (ATI-openGL).
*/
virtual bool slowUnlockVertexBufferHard() const =0;
/* Returns true if static vertex and index buffers must by allocated in VRAM, false in AGP.
* Default is false.
*/
bool getStaticMemoryToVRAM() const { return _StaticMemoryToVRAM; }
/* Set to true if static vertex and index buffers must by allocated in VRAM, false in AGP.
* Default is false.
*/
void setStaticMemoryToVRAM(bool staticMemoryToVRAM);
/** Return the driver reset counter.
* The reset counter is incremented at each driver reset.
*/
uint getResetCounter () const { return _ResetCounter; }
/** return How many vertices VertexBufferHard support
*/
virtual uint getMaxVerticesByVertexBufferHard() const =0;
/** Allocate the initial VertexArray Memory. (no-op if !supportVertexBufferHard()).
* VertexArrayRange is first reseted, so any VBhard created before will be deleted.
* NB: call it after setDisplay(). But setDisplay() by default call initVertexBufferHard(16Mo, 0);
* so this is not necessary.
* NB: If allocation fails, mem/=2, and retry, until mem < 500K.
* \param agpMem amount of AGP Memory required. if 0, reseted.
* \param vramMem amount of VRAM Memory required. if 0, reseted.
* \return false if one the Buffer has not been allocated (at least at 500K).
*/
virtual bool initVertexBufferHard(uint agpMem, uint vramMem=0) =0;
/** Return the amount of AGP memory allocated by initVertexBufferHard() to store vertices.
*/
virtual uint32 getAvailableVertexAGPMemory () =0;
/** Return the amount of video memory allocated by initVertexBufferHard() to store vertices.
*/
virtual uint32 getAvailableVertexVRAMMemory () =0;
/** active a current VB, for future render().
* This method suppose that all vertices in the VB will be used.
*
* NB: please make sure you have setuped / unsetuped the current vertex program BEFORE activate the vertex buffer.
* Don't change the vertex buffer format/size after having activated it.
* Don't lock the vertex buffer after having activated it.
*
* \see activeVertexProgram
*/
virtual bool activeVertexBuffer(CVertexBuffer& VB)=0;
/** active a current IB, for future render().
*
* Don't change the index buffer format/size after having activated it.
* Don't lock the index buffer after having activated it.
*/
virtual bool activeIndexBuffer(CIndexBuffer& IB)=0;
/** Render a list of indexed lines with previously setuped VertexBuffer / IndexBuffer / Matrixes.
* \param mat is the material to use during this rendering
* \param firstIndex is the first index in the index buffer to use as first line.
* \param nlines is the number of line to render.
*/
virtual bool renderLines(CMaterial& mat, uint32 firstIndex, uint32 nlines)=0;
/** Render a list of indexed triangles with previously setuped VertexBuffer / IndexBuffer / Matrixes.
* \param mat is the material to use during this rendering
* \param firstIndex is the first index in the index buffer to use as first triangle.
* \param ntris is the number of triangle to render.
*/
virtual bool renderTriangles(CMaterial& mat, uint32 firstIndex, uint32 ntris)=0;
/** Render a list of triangles with previously setuped VertexBuffer / IndexBuffer / Matrixes, AND previously setuped MATERIAL!!
* This use the last material setuped. It should be a "Normal shader" material, because no multi-pass is allowed
* with this method.
* Actually, it is like a straight drawTriangles() in OpenGL.
* NB: nlassert() if ntris is 0!!!! this is unlike other render() call methods. For optimisation concern.
* NB: this is useful for landscape....
* \param firstIndex is the first index in the index buffer to use as first triangle.
* \param ntris is the number of triangle to render.
*/
virtual bool renderSimpleTriangles(uint32 firstIndex, uint32 ntris)=0;
/** Render points with previously setuped VertexBuffer / Matrixes.
* Points are stored as a sequence in the vertex buffer.
* \param mat is the material to use during this rendering
* \param startVertex is the first vertex to use during this rendering.
* \param numPoints is the number of point to render.
*/
virtual bool renderRawPoints(CMaterial& mat, uint32 startVertex, uint32 numPoints)=0;
/** Render lines with previously setuped VertexBuffer / Matrixes.
* Lines are stored as a sequence in the vertex buffer.
* \param mat is the material to use during this rendering
* \param startVertex is the first vertex to use during this rendering.
* \param numLine is the number of line to render.
*/
virtual bool renderRawLines(CMaterial& mat, uint32 startVertex, uint32 numTri)=0;
/** Render triangles with previously setuped VertexBuffer / Matrixes.
* Triangles are stored as a sequence in the vertex buffer.
* \param mat is the material to use during this rendering
* \param startVertex is the first vertex to use during this rendering.
* \param numTri is the number of triangle to render.
*/
virtual bool renderRawTriangles(CMaterial& mat, uint32 startVertex, uint32 numTri)=0;
/** If the driver support it, primitive can be rendered with an offset added to each index
* These are the offseted version of the 'render' functions
* \see supportIndexOffset
*/
virtual bool renderLinesWithIndexOffset(CMaterial& mat, uint32 firstIndex, uint32 nlines, uint indexOffset)=0;
virtual bool renderTrianglesWithIndexOffset(CMaterial& mat, uint32 firstIndex, uint32 ntris, uint indexOffset)=0;
virtual bool renderSimpleTrianglesWithIndexOffset(uint32 firstIndex, uint32 ntris, uint indexOffset)=0;
/** render quads with previously setuped VertexBuffer / Matrixes.
* Quads are stored as a sequence in the vertex buffer.
* There's a guaranty for the orientation of its diagonal, which is drawn as follow :
*
* 3----2
* | / |
* | / |
* |/ |
* 0----1
*
* \param mat is the material to use during this rendering
* \param startVertex is the first vertex to use during this rendering.
* \param numQuad is the number of quad to render.
*/
virtual bool renderRawQuads(CMaterial& mat, uint32 startVertex, uint32 numQuads)=0;
/** Say what Texture Stage use what UV coord.
* by default activeVertexBuffer*() methods map all stage i to UV i. You can change this behavior,
* after calling activeVertexBuffer*(), by using this method.
*
* eg: mapTextureStageToUV(0,2) will force the 0th texture stage to use the 2th UV.
*
* Warning! This DOESN'T work with VertexProgram enabled!! (assert)
*
* Warning!: some CMaterial Shader may change automatically this behavior too when setupMaterial()
* (and so render*()) is called. But Normal shader doesn't do it.
*/
virtual void mapTextureStageToUV(uint stage, uint uv)=0;
/// Swap the back and front buffers.
virtual bool swapBuffers(void)=0;
/// Copy a string to system clipboard.
virtual bool copyTextToClipboard(const ucstring &text) =0;
/// Paste a string from system clipboard.
virtual bool pasteTextFromClipboard(ucstring &text) =0;
/** set the number of VBL wait when a swapBuffers() is issued. 0 means no synchronisation to the VBL
* Default is 1. Values >1 may be clamped to 1 by the driver.
*/
virtual void setSwapVBLInterval(uint interval)=0;
/// get the number of VBL wait when a swapBuffers() is issued. 0 means no synchronisation to the VBL
virtual uint getSwapVBLInterval()=0;
/// \name Profiling.
// @{
/** Get the number of primitives rendered from the last swapBuffers() call.
* \param pIn the number of requested rendered primitive.
* \param pOut the number of effective rendered primitive. pOut==pIn if no multi-pass material is used
* (Lightmap, Specular ...).
*/
virtual void profileRenderedPrimitives(CPrimitiveProfile &pIn, CPrimitiveProfile &pOut) =0;
/** Return the amount of Texture memory requested. taking mipmap, compression, texture format, etc... into account.
* NB: because of GeForce*, RGB888 is considered to be 32 bits. So it may be false for others cards :).
*/
virtual uint32 profileAllocatedTextureMemory() =0;
/** Get the number of material setuped from the last swapBuffers() call.
*/
virtual uint32 profileSetupedMaterials() const =0;
/** Get the number of matrix setuped from the last swapBuffers() call.
*/
virtual uint32 profileSetupedModelMatrix() const =0;
/** Enable the sum of texture memory used since last swapBuffers() call. To retrieve the memory used call getUsedTextureMemory().
*/
virtual void enableUsedTextureMemorySum (bool enable=true) =0;
/** Return the amount of texture video memory used since last swapBuffers() call. Before use this method, you should enable
* the sum with enableUsedTextureMemorySum().
*/
virtual uint32 getUsedTextureMemory() const =0;
/** If the driver support it, enable profile VBHard locks.
* No-Op if already profiling
*/
virtual void startProfileVBHardLock() = 0;
/** If the driver support it, stop profile VBHard locks, and "print" result
* No-Op if already profiling
* NB: The results are the Locks in Chronological time (since last swapBuffers).
* Since multiple frame are summed, an "*" is marked against the VBHard name to show if it was not
* always this one (ptr test and not name test) in the chronological order.
* NB: if the driver does not support VBHard or VBHard profiling (like ATI VBHard), result is empty.
* NB: ???? string is displayed if the VBHard has no name or if was just deleted.
*/
virtual void endProfileVBHardLock(std::vector<std::string> &result) = 0;
/** display VBhards allocated
*/
virtual void profileVBHardAllocation(std::vector<std::string> &result) = 0;
// Index buffer profiling, same use than with vertex buffers
virtual void startProfileIBLock() = 0;
virtual void endProfileIBLock(std::vector<std::string> &result) = 0;
virtual void profileIBAllocation(std::vector<std::string> &result) = 0;
/** For each texture setuped in the driver, "print" result: type, shareName, format and size (mipmap included)
*/
void profileTextureUsage(std::vector<std::string> &result);
// @}
/// \name Fog support.
// @{
virtual bool fogEnabled()=0;
virtual void enableFog(bool enable)=0;
/// setup fog parameters. fog must enabled to see result. start and end are distance values.
virtual void setupFog(float start, float end, CRGBA color)=0;
/// Get.
virtual float getFogStart() const =0;
virtual float getFogEnd() const =0;
virtual CRGBA getFogColor() const =0;
// @}
/// Deriver should calls IDriver::release() first, to destroy all driver components (textures, shaders, VBuffers).
virtual bool release(void);
/// Return true if driver is still active. Return false else. If he user close the window, must return false.
virtual bool isActive()=0;
/// Return the depth of the driver after init().
virtual uint8 getBitPerPixel ()=0;
/** Output a system message box and print a message with an icon. This method can be call even if the driver is not initialized.
* This method is used to return internal driver problem when string can't be displayed in the driver window.
* If the driver can't open a messageBox, it should not override this method and let the IDriver class manage it with the ASCII console.
*
* \param message This is the message to display in the message box.
* \param title This is the title of the message box.
* \param type This is the type of the message box, ie number of button and label of buttons.
* \param icon This is the icon of the message box should use like warning, error etc...
*/
virtual TMessageBoxId systemMessageBox (const char* message, const char* title, TMessageBoxType type=okType, TMessageBoxIcon icon=noIcon);
/** Set the current viewport
*
* \param viewport is a viewport to setup as current viewport.
*/
virtual void setupViewport (const class CViewport& viewport)=0;
/** Get the current viewport
*/
virtual void getViewport(CViewport &viewport) = 0;
/** Set the current Scissor.
* \param scissor is a scissor to setup the current Scissor, in Window relative coordinate (0,1).
*/
virtual void setupScissor (const class CScissor& scissor)=0;
/**
* Get the driver version. Not the same than interface version. Incremented at each implementation change.
*
* \see InterfaceVersion
*/
virtual uint32 getImplementationVersion () const = 0;
/**
* Get driver information.
* get the nel name of the driver (ex: "Opengl 1.2 NeL Driver")
*/
virtual const char* getDriverInformation () = 0;
/**
* Get videocard information.
* get the official name of the driver
*/
virtual const char* getVideocardInformation () = 0;
/// \name Mouse / Keyboard / Game devices
// @{
/// show cursor if b is true, or hide it if b is false
virtual void showCursor (bool b) = 0;
/// x and y must be between 0.0 and 1.0
virtual void setMousePos (float x, float y) = 0;
/** Enable / disable low level mouse. This allow to take advantage of some options (speed of the mouse, automatic wrapping)
* It returns a interface to these parameters when it is supported, or NULL otherwise
* The interface pointer is valid as long as the low level mouse is enabled.
* A call to disable the mouse returns NULL, and restore the default mouse behavior
* NB : - In this mode the mouse cursor isn't drawn.
* - Calls to showCursor have no effects
* - Calls to setCapture have no effects
*/
virtual NLMISC::IMouseDevice *enableLowLevelMouse(bool enable, bool exclusive) = 0;
/** Enable / disable a low level keyboard.
* Such a keyboard can only send KeyDown and KeyUp event. It just consider the keyboard as a
* gamepad with lots of buttons...
* This returns a interface to some parameters when it is supported, or NULL otherwise.
* The interface pointer is valid as long as the low level keyboard is enabled.
* A call to disable the keyboard returns NULL, and restore the default keyboard behavior
*/
virtual NLMISC::IKeyboardDevice *enableLowLevelKeyboard(bool enable) = 0;
/** Get the delay in ms for mouse double clicks.
*/
virtual uint getDoubleClickDelay(bool hardwareMouse) = 0;
/** If true, capture the mouse to force it to stay under the window.
* NB : this has no effects if a low level mouse is used
*/
virtual void setCapture (bool b) = 0;
// see if system cursor is currently captured
virtual bool isSystemCursorCaptured() = 0;
// Add a new cursor (name is case unsensitive)
virtual void addCursor(const std::string &name, const NLMISC::CBitmap &bitmap) = 0;
// Display a cursor from its name (case unsensitive)
virtual void setCursor(const std::string &name, NLMISC::CRGBA col, uint8 rot, sint hotSpotX, sint hotSpotY, bool forceRebuild = false) = 0;
// Change default scale for all cursors
virtual void setCursorScale(float scale) = 0;
/** Check whether there is a low level device manager available, and get its interface. Return NULL if not available
* From this interface you can deal with mouse and keyboard as above, but you can also manage game device (joysticks, joypads ...)
*/
virtual NLMISC::IInputDeviceManager *getLowLevelInputDeviceManager() = 0;
// @}
/// Get the width and the height of the window
virtual void getWindowSize (uint32 &width, uint32 &height) = 0;
/// Get the position of the window always (0,0) in fullscreen
virtual void getWindowPos (sint32 &x, sint32 &y) = 0;
/** get the RGBA back buffer. After swapBuffers(), the content of the back buffer is undefined.
*
* \param bitmap the buffer will be written in this bitmap
*/
virtual void getBuffer (CBitmap &bitmap) = 0;
/** get the ZBuffer (back buffer).
*
* \param zbuffer the returned array of Z. size of getWindowSize() .
*/
virtual void getZBuffer (std::vector<float> &zbuffer) = 0;
/** get a part of the RGBA back buffer. After swapBuffers(), the content of the back buffer is undefined.
* NB: 0,0 is the bottom left corner of the screen.
*
* \param bitmap the buffer will be written in this bitmap
* \param rect the in/out (wanted/clipped) part of Color buffer to retrieve.
*/
virtual void getBufferPart (CBitmap &bitmap, NLMISC::CRect &rect) = 0;
// copy the first texture in a second one of different dimensions
virtual bool stretchRect (ITexture * srcText, NLMISC::CRect &srcRect, ITexture * destText, NLMISC::CRect &destRect) = 0;
// is this texture a rectangle texture ?
virtual bool isTextureRectangle(ITexture * tex) const = 0;
// return true if driver support Bloom effect.
virtual bool supportBloomEffect() const =0;
// return true if driver support non-power of two textures
virtual bool supportNonPowerOfTwoTextures() const =0;
/** get a part of the ZBuffer (back buffer).
* NB: 0,0 is the bottom left corner of the screen.
*
* \param zbuffer the returned array of Z. size of rec.Width*rec.Height.
* \param rect the in/out (wanted/clipped) part of ZBuffer to retrieve.
*/
virtual void getZBufferPart (std::vector<float> &zbuffer, NLMISC::CRect &rect) = 0;
/** Set the current render target.
*
* The render target can be a texture (tex pointer) or the back buffer (tex = NULL).
* The texture must have been right sized before the call.
* This mark the texture as valid, but doesn't copy data to system memory.
* This also mean that regenerating texture data will erase what
* has been copied before in the device memory.
* This doesn't work with compressed textures.
* Ideally, the FrameBuffer should have the same format than the texture.
*
* When direct render to texture is not available (openGl), it uses the frame buffer for the rendering and copy the frame buffer
* content into the texture when setRenderTarget(NULL) is called.
*
* The x, y, width and height parameters are only used in this case to optimize the copy from the framebuffer
* to the texture.
*
* If a texture is set as target, the viewport and the scissor are now relative to the texture sizes,
* and not to the x, y, width and height parameters.
*
* The texture content can be lost after the first setRenderTarget().
*
* The texture must have the render target abilities enabled. (ITexture::setRenderTarget ())
*
* \param tex the texture to render into.
* \param x x position within the destination texture of the renderable area.
* \param y y position within the destination texture of the renderable area.
* \param width width of the renderable area, if 0, use the whole size.
* \param height height of the renderable area, if 0, use the whole size.
* \param mipmapLevel the mipmap to copy texture to.
* \param cubaFace the face of the cube to copy texture to.
* \return true if the render target has been changed
*/
virtual bool setRenderTarget (ITexture *tex,
uint32 x = 0,
uint32 y = 0,
uint32 width = 0,
uint32 height = 0,
uint32 mipmapLevel = 0,
uint32 cubeFace = 0
) = 0 ;
/** Trick method : copy the current texture target into another texture without updating the current texture.
*
* This method copies the current texture into another texture.
* WARNING : at the next setRenderTarget () call, the current texture target WILL NOT BE UPDATED.
*
* When direct render to texture is not available, this method can save a texture copy :
*
* Use this method to copy a temporary texture target into a destination texture.
* Then, resets the rendering target with setRenderTarget().
*
* The temporary texture is copied into the final texture direct from the frame buffer. The temporary texture is not filled in VRAM when
* the framebuffer is set back as render target.
*
* Works only if a texture is used as render target.
*
* This method invalidates the vertex buffer, the view and model matrices, the viewport and the frustum.
*
* \param tex the texture to render into.
* \param offsetx x position within the destination texture.
* \param y y position within the destination texture.
* \param x x position within the current texture target.
* \param y y position within the current texture target.
* \param width width of the renderable area to copy, if 0, use the whole size.
* \param height height of the renderable area to copy, if 0, use the whole size.
* \param mipmapLevel the mipmap to copy texture to.
*/
virtual bool copyTargetToTexture (ITexture *tex,
uint32 offsetx = 0,
uint32 offsety = 0,
uint32 x = 0,
uint32 y = 0,
uint32 width = 0,
uint32 height = 0,
uint32 mipmapLevel = 0
) = 0;
/** Retrieve the render target size.
* If the render target is the frame buffer, it returns the size of the frame buffer.
* It the render target is a texture, it returns the size of the texture mipmap selected as render target.
*/
virtual bool getRenderTargetSize (uint32 &width, uint32 &height) = 0;
/** fill the RGBA back buffer
*
* \param bitmap will be written in the buffer. no-op if bad size.
* \return true if success
*/
virtual bool fillBuffer (CBitmap &bitmap) = 0;
/** Set the global polygon mode. Can be filled, line or point. The implementation driver must
* call IDriver::setPolygonMode and active this mode.
*
* \param polygon mode choose in this driver.
* \see getPolygonMode(), TPolygonMode
*/
virtual void setPolygonMode (TPolygonMode mode)
{
_PolygonMode=mode;
}
/**
* return the number of light supported by driver. typically 8.
*
* \see enableLight() setLight()
*/
virtual uint getMaxLight () const = 0;
/**
* Setup a light.
*
* You must call enableLight() to active the light.
*
* \param num is the number of the light to set.
* \param light is a light to set in this slot.
* \see enableLight()
*/
virtual void setLight (uint8 num, const CLight& light) = 0;
/**
* Enable / disable light.
*
* You must call setLight() if you active the light.
*
* \param num is the number of the light to enable / disable.
* \param enable is true to enable the light, false to disable it.
* \see setLight()
*/
virtual void enableLight (uint8 num, bool enable=true) = 0;
/**
* Set ambient.
*
* \param color is the new global ambient color for the scene.
* \see setLight(), enableLight()
*/
virtual void setAmbientColor (CRGBA color) = 0;
/** Setup the light used for per pixel lighting. The given values should have been modulated by the material diffuse and specular.
* This is only useful for material that have their shader set as 'PerPixelLighting'
* \param the light used for per pixel lighting
*/
virtual void setPerPixelLightingLight(CRGBA diffuse, CRGBA specular, float shininess) = 0;
/** Setup the unique light used for Lightmap Shader.
* Lightmaped primitives are lit per vertex with this light (should be local attenuated for maximum efficiency)
* This is only useful for material that have their shader set as 'LightMap'
* \param the light used for per pixel lighting
*/
virtual void setLightMapDynamicLight (bool enable, const CLight& light) = 0;
/** Get the global polygon mode.
*
* \param polygon mode choose in this driver.
* \see setPolygonMode(), TPolygonMode
*/
TPolygonMode getPolygonMode ()
{
return _PolygonMode;
}
/// \name Vertex program interface
// @{
enum TMatrix
{
ModelView= 0,
Projection,
ModelViewProjection,
NumMatrix
};
enum TTransform
{
Identity=0,
Inverse,
Transpose,
InverseTranspose,
NumTransform
};
/**
* Does the driver supports vertex programs ?
*/
virtual bool isVertexProgramSupported () const =0;
/**
* Does the driver supports vertex program, but emulated by CPU ?
*/
virtual bool isVertexProgramEmulated () const =0;
/**
* Activate / disactivate a vertex program
*
* \param program is a pointer on a vertex program. Can be NULL to disable the current vertex program.
*
* \return true if setup/unsetup succeeded, false else.
*/
virtual bool activeVertexProgram (CVertexProgram *program) =0;
/**
* Setup constant values.
*/
virtual void setConstant (uint index, float, float, float, float) =0;
virtual void setConstant (uint index, double, double, double, double) =0;
virtual void setConstant (uint index, const NLMISC::CVector& value) =0;
virtual void setConstant (uint index, const NLMISC::CVectorD& value) =0;
/// setup several 4 float csts taken from the given tab
virtual void setConstant (uint index, uint num, const float *src) =0;
/// setup several 4 double csts taken from the given tab
virtual void setConstant (uint index, uint num, const double *src) =0;
/**
* Setup constants with a current matrix.
*
* This call must be done after setFrustum(), setupViewMatrix() or setupModelMatrix() to get correct
* results.
*
* \param index is the base constant index where to store the matrix. This index must be a multiple of 4.
* \param matrix is the matrix id to store in the constants
* \param transform is the transformation to apply to the matrix before store it in the constants.
*
*/
virtual void setConstantMatrix (uint index, TMatrix matrix, TTransform transform) =0;
/**
* Setup the constant with the fog vector. This vector must be used to get the final fog value in a vertex shader.
* You must use it like this:
* DP4 o[FOGC].x, c[4], R4;
* With c[4] the constant used for the fog vector and R4 the vertex local position.
*
* This call must be done after setFrustum(), setupViewMatrix(), setupModelMatrix() and setupFog() to get correct
* results.
*
* \param index is the index where to store the vector.
*
*/
virtual void setConstantFog (uint index) =0;
/// Check if the driver support double sided colors vertex programs
virtual bool supportVertexProgramDoubleSidedColor() const = 0;
// test if support for cloud render in a single pass
virtual bool supportCloudRenderSinglePass() const = 0;
/**
* Activate VertexProgram 2Sided Color mode. In 2Sided mode, the BackFace (if material 2Sided enabled) read the
* result from o[BFC0], and not o[COL0].
* default is false. you should reset to false after use.
* NB: no-op if not supported by driver
*/
virtual void enableVertexProgramDoubleSidedColor(bool doubleSided) =0;
// @}
/// \name Texture addressing modes aka textures/pixels shaders
// @{
/// test whether the device supports some form of texture shader. (could be limited to DX6 EMBM for example)
virtual bool supportTextureShaders() const = 0;
// Is the shader water supported ? If not, the driver caller should implement its own version
virtual bool isWaterShaderSupported() const = 0;
//
/// test whether a texture addressing mode is supported
virtual bool isTextureAddrModeSupported(CMaterial::TTexAddressingMode mode) const = 0;
/** setup the 2D matrix for the OffsetTexture, OffsetTextureScale and OffsetTexture addressing mode
* It should be stored as the following
* [a0 a1]
* [a2 a3]
*/
virtual void setMatrix2DForTextureOffsetAddrMode(const uint stage, const float mat[4]) = 0;
//@}
/** \name EMBM support. If texture shaders are present, this is not available, must use them instead.
* EMBM is a color op of CMaterial.
* NB : EMBM is the equivalent of the CMaterial::OffsetTexture addressing mode. However, it is both a texture
* addressing mode and a color op.
* NB : EMBM may not be supported by all stages.
*
* if embm unit is at last last stage, it operates on texture at first stage
* otherwise it operates on texture at next stage
*/
// @{
// Test if EMBM is supported.
virtual bool supportEMBM() const = 0;
// Test if EMBM is supported for the given stage
virtual bool isEMBMSupportedAtStage(uint stage) const = 0;
// set the matrix used for EMBM addressing
virtual void setEMBMMatrix(const uint stage, const float mat[4]) = 0;
// @}
// Does the driver support the per-pixel lighting shader ?
virtual bool supportPerPixelLighting(bool specular) const = 0;
/// \name Misc
// @{
/** Does the driver support Blend Constant Color ??? If yes CMaterial::blendConstant* enum can be used
* for blend Src ord Dst factor. If no, using these enum will have undefined results.
*/
virtual bool supportBlendConstantColor() const =0;
/** see supportBlendConstantColor(). Set the current Blend Constant Color.
*/
virtual void setBlendConstantColor(NLMISC::CRGBA col)=0;
/** see supportBlendConstantColor(). Get the current Blend Constant Color.
*/
virtual NLMISC::CRGBA getBlendConstantColor() const =0;
/** force the driver to flush all command. glFinish() in opengl.
* Interesting only for debug and profiling purpose.
*/
virtual void finish() =0;
// Flush command queue an immediately returns
virtual void flush() = 0;
/** Use AntiAliasing For polygons (GL_POLYGON_SMOOTH like, not the FSAA).
* See GL_POLYGON_SMOOTH help, and GL_SRC_ALPHA_SATURATE OpenGL doc (not yet implemented now since
* used only for alpha part in ShadowMap gen)
*/
virtual void enablePolygonSmoothing(bool smooth) =0;
/// see enablePolygonSmoothing()
virtual bool isPolygonSmoothingEnabled() const =0;
// @}
/** Special method to internally swap the Driver handle of 2 textures.
* USE IT WITH CARE (eg: may have Size problems, mipmap problems, format problems ...)
* Actually, it is used only by CAsyncTextureManager, to manage Lods of DXTC CTextureFile.
* NB: internally, all textures slots are disabled.
*/
virtual void swapTextureHandle(ITexture &tex0, ITexture &tex1) =0;
/** Advanced usage. Get the texture Handle.Useful for texture sorting for instance
* NB: if the texture is not setuped in the driver, 0 is returned.
* NB: if implementation does not support it, 0 may be returned. OpenGL ones return the Texture ID.
* NB: unlike isTextureExist(), this method is not thread safe.
*/
virtual uint getTextureHandle(const ITexture&tex)=0;
// see if the Multiply-Add Tex Env operator is supported (see CMaterial::Mad)
virtual bool supportMADOperator() const = 0;
// Adapter class
class CAdapter
{
public:
std::string Driver;
std::string Description;
std::string DeviceName;
std::string Vendor;
sint64 DriverVersion;
uint32 VendorId;
uint32 DeviceId;
uint32 SubSysId;
uint32 Revision;
};
// Get the number of hardware renderer available on the client platform.
virtual uint getNumAdapter() const=0;
// Get a hardware renderer description.
virtual bool getAdapter(uint adapter, CAdapter &desc) const=0;
/** Choose the hardware renderer.
* Call it before the setDisplay and enumModes methods
* Choose adapter = 0xffffffff for the default one.
*/
virtual bool setAdapter(uint adapter)=0;
/** Tell if the vertex color memory format is RGBA (openGL) or BGRA (directx)
* BGRA :
* *****************************************************************
* Offset: * 0 * 1 * 2 * 3 *
* *****************************************************************
* RGBA * red * green * blue * alpha *
* *****************************************************************
* BGRA * blue * green * red * alpha *
* *****************************************************************
*/
virtual CVertexBuffer::TVertexColorType getVertexColorFormat() const =0;
/// \name Bench
// @{
// Start the bench. See CHTimer::startBench();
virtual void startBench (bool wantStandardDeviation = false, bool quick = false, bool reset = true) =0;
// End the bench. See CHTimer::endBench();
virtual void endBench () =0;
// Display the bench result
virtual void displayBench (class NLMISC::CLog *log) =0;
// @}
/// \name Occlusion query mechanism
// @{
// Test whether this device supports the occlusion query mechanism
virtual bool supportOcclusionQuery() const = 0;
/** Create an occlusion query object.
* \return NULL is not enough resources or if not supported
*/
virtual IOcclusionQuery *createOcclusionQuery() = 0;
// Delete an occlusion query object previously obtained by a call to createOcclusionQuery
virtual void deleteOcclusionQuery(IOcclusionQuery *oq) = 0;
// @}
// get the number of call to swapBuffer since the driver was created
virtual uint64 getSwapBufferCounter() const = 0;
/** Set cull mode
* Useful for mirrors / cube map rendering or when the scene must be rendered upside down
*/
virtual void setCullMode(TCullMode cullMode) = 0;
virtual TCullMode getCullMode() const = 0;
/** Set stencil support
*/
virtual void enableStencilTest(bool enable) = 0;
virtual bool isStencilTestEnabled() const = 0;
virtual void stencilFunc(TStencilFunc stencilFunc, int ref, uint mask) = 0;
virtual void stencilOp(TStencilOp fail, TStencilOp zfail, TStencilOp zpass) = 0;
virtual void stencilMask(uint mask) = 0;
protected:
friend class IVBDrvInfos;
friend class IIBDrvInfos;
friend class CTextureDrvShare;
friend class ITextureDrvInfos;
friend class IMaterialDrvInfos;
friend class IVertexProgramDrvInfos;
friend class IShaderDrvInfos;
/// remove ptr from the lists in the driver.
void removeVBDrvInfoPtr(ItVBDrvInfoPtrList vbDrvInfoIt);
void removeIBDrvInfoPtr(ItIBDrvInfoPtrList ibDrvInfoIt);
void removeTextureDrvInfoPtr(ItTexDrvInfoPtrMap texDrvInfoIt);
void removeTextureDrvSharePtr(ItTexDrvSharePtrList texDrvShareIt);
void removeMatDrvInfoPtr(ItMatDrvInfoPtrList shaderIt);
void removeShaderDrvInfoPtr(ItShaderDrvInfoPtrList shaderIt);
void removeVtxPrgDrvInfoPtr(ItVtxPrgDrvInfoPtrList vtxPrgDrvInfoIt);
private:
bool _StaticMemoryToVRAM;
};
// --------------------------------------------------
}
#endif // NL_DRV_H