Jump to content

  • Log In with Google      Sign In   
  • Create Account


Directional light and matrix issues


Old topic!
Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.

  • You cannot reply to this topic
5 replies to this topic

#1 KaiserJohan   Members   -  Reputation: 1004

Like
0Likes
Like

Posted 20 March 2013 - 02:57 PM

I'm trying to implement basic directional lightning in OpenGL 3.3 by emulating the logic shown in this guide: http://www.arcsynthesis.org/gltut/Illumination/Tutorial%2009.html

 

I do not understand what matrix to send to the vertex shader to use for calculating lightning effects. The guide says modelView matrix, but to me, and when I'm running the game, it dosn't make sense; when I orient myself in the world with mouse and keyboard, the lightning direction gets altered aswell!

 

 

Here's the items to which I fill my 'render queue', which contains all data needed to draw the models:

 

/* RenderItem definition */
    struct RenderItem
    {
        VertexBufferPtr mVertexBuffer;
        const Vec4 mColor;
        const Mat4 mMVP;
        const Mat3 mMVMatrix;    // I used model-view matrix here, but I dont know really which one to use

        const Vec4 mLightIntensity;
        const Vec3 mDirToLight;

       // simply extracts all the struct data into one buffer to be uploaded as UniformBuffer to shader
        void CopyUniformData(std::vector<float>& buffer) const
        {
            std::vector<float> colorValues = GetTypeValues(mColor);
            std::vector<float> mvpMatrixValues = GetTypeValues(mMVP);
            std::vector<float> mvMatrixValues = GetTypeValues(mMVMatrix);
            std::vector<float> lightIntensityMatrixValues = GetTypeValues(mLightIntensity);
            std::vector<float> dirToLightMatrixValues = GetTypeValues(mDirToLight);

            buffer.insert(buffer.end(), colorValues.begin(), colorValues.end());
            buffer.insert(buffer.end(), mvpMatrixValues.begin(), mvpMatrixValues.end());
            buffer.insert(buffer.end(), mvMatrixValues.begin(), mvMatrixValues.end());
            buffer.insert(buffer.end(), lightIntensityMatrixValues.begin(), lightIntensityMatrixValues.end());
            buffer.insert(buffer.end(), dirToLightMatrixValues.begin(), dirToLightMatrixValues.end());
        }

        RenderItem(VertexBufferPtr vertexBuffer, const Vec4& color, const Mat4& mvp, const Mat3& mv, const Vec4& lightIntensity, const Vec3& dirToLight);
    };

 

 

Here's where I create all the RenderItems for my renderqueue:

 


void Engine::CreateModelRenderables(const Model* model, const Mat4& viewMatrix, const Mat4& perspectiveMatrix, const Mat4& nodeTransform, std::vector<RenderItem>& renderQueue, const std::vector<LightPtr>& activeLights)
    {
        const Mat4 modelMatrix = nodeTransform * model->mTransform;       // 'nodeTransform' is the SceneNodes transform, while model->mTransform is the models transform... so this gives the model-world matrix
        const Mat4 modelViewMatrix = viewMatrix * modelMatrix;
        const Mat4 modelViewProjMatrix = perspectiveMatrix * modelViewMatrix;

        // 'activelights' is a vector, but for now only one light is supported, so I just pick the first element

        // TODO: handle multiple lights
        BOOST_FOREACH(const Mesh& mesh, model->mMeshes)
            if (activeLights.size() >= 1)
                renderQueue.push_back(RenderItem(mesh.mVertexBuffer, Vec4(1.0f, 0.3f, 0.3f, 1.0f), modelViewProjMatrix, Mat3(....???....), activeLights.front()->mLightIntensity, activeLights.front()->mLightDirection));
            else
                renderQueue.push_back(RenderItem(mesh.mVertexBuffer, Vec4(1.0f, 0.3f, 0.3f, 1.0f), modelViewProjMatrix, Mat3(....???....), Vec4(0.0f), Vec3(0.0f)));

        BOOST_FOREACH(const Model& childModel, model->mChildren)
            CreateModelRenderables(&childModel, viewMatrix, perspectiveMatrix, modelMatrix, renderQueue, activeLights);
    }

 

 

 

 

for drawing, I just loop over the renderitems, update the UniformBuffer object and glDrawElements (vertexbuffer->render):

 

void OpenGLRenderer::DrawRenderables(const std::vector<RenderItem>& renderQueue)
    {
        BOOST_FOREACH(const RenderItem& renderItem, renderQueue)
        {
            mUniBuffer.SetData(renderItem);

            renderItem.mVertexBuffer->Render();
        }
    }

 

 

 

this is my vertex shader:

 

 const std::string gVertexShader =	"#version 330								                \n	\
                                                                                                    \n  \
                                        layout(std140) uniform DefaultUniform                       \n  \
                                        {                                                           \n  \
                                            vec4 Color;                                             \n  \
                                            mat4 MVPMatrix;                                         \n  \
                                            mat3 MVMatrix;                                          \n  \
                                                                                                    \n  \
                                            vec4 LightIntensity;                                    \n  \
                                            vec3 DirToLight;                                        \n  \
                                        };                                                          \n  \
                                                                                                    \n	\
                                        layout(location = 0) in vec3 vert_position;				    \n	\
                                        layout(location = 1) in vec3 vert_normal;					\n	\
                                        out vec4 frag_color;                                        \n	\
                                                                                                    \n	\
                                        void main()									                \n	\
                                        {											                \n	\
                                            gl_Position = MVPMatrix * vec4(vert_position, 1.0f);    \n  \
                                                                                                    \n  \
                                            vec3 normCamSpace = normalize(MVMatrix * vert_normal);  \n  \
                                            float angIncidence = dot(normCamSpace, DirToLight);     \n  \
                                            angIncidence = clamp(angIncidence, 0, 1);               \n  \
                                                                                                    \n  \
                                            frag_color = LightIntensity * Color * angIncidence;     \n  \
                                        }											                \n";

 

 

the fragment shader simply passes-through the frag_color..

 

 

SO, my question is, what matrix should the MVMatrix be in the vertex shader? The guide says the 'modelToCameraMatrix', which I suppose is my modelViewMatrix in Engine::CreateModelRenderables(), but I don't want the lights direction to change when I re-orient my camera!

 

 

EDIT: Here is my camera class used to generate the viewMatrix through getCameraTransform(), in case there might be something wrong there? 

I don't think so though, as I can move around in the world with WASD + mouse just as expected, its just the lightning that is wierd...

 

Camera::Camera() : mTranslation(0.0f), mHorizontalAngle(0.0f), mVerticalAngle(0.0f)
    {
    }
        
    Camera::~Camera()
    {
    }


    void Camera::SetPosition(const Vec3& position)
    {
        mTranslation = position;
    }

    void Camera::TranslateCamera(const Vec3& translateVec)
    {
        mTranslation += translateVec;
    }
        
    void Camera::RotateCamera(const float offsetHorizontalAngle, const float offsetVerticalAngle)
    {
        mHorizontalAngle += offsetHorizontalAngle;
        mVerticalAngle   += offsetVerticalAngle;
    }


    Vec3 Camera::Forward() const
    {
        Vec4 forward = Inverse(Orientation()) * Vec4(0, 0, 1, 1);

        return Vec3(forward);
    }

    Vec3 Camera::Right() const
    {
        Vec4 right = Inverse(Orientation()) * Vec4(-1, 0, 0, 1);

        return Vec3(right);
    }

    Mat4 Camera::Orientation() const
    {
        Quaternion rotation;
        rotation = AngleAxisToQuaternion(mVerticalAngle, Vec3(1.0f, 0.0f, 0.0f));
        rotation = rotation * AngleAxisToQuaternion(mHorizontalAngle, Vec3(0.0f, 1.0f, 0.0f));

        return QuaternionToMat4(rotation);
    }

       
    Mat4 Camera::GetCameraTransform() const
    {
        Mat4 viewMatrix(1.0f);
        viewMatrix = Orientation() * Translate(viewMatrix, mTranslation);

        return viewMatrix;
    }

Edited by KaiserJohan, 20 March 2013 - 03:37 PM.


Sponsor:

#2 TheChubu   Crossbones+   -  Reputation: 3631

Like
0Likes
Like

Posted 20 March 2013 - 09:04 PM

Edit: Nope. Wasn't that. No idea really.


Edited by TheChubu, 20 March 2013 - 09:17 PM.

"I AM ZE EMPRAH OPENGL 3.3 THE CORE, I DEMAND FROM THEE ZE SHADERZ AND MATRIXEZ"

 

My journals: dustArtemis ECS framework and Making a Terrain Generator


#3 KaiserJohan   Members   -  Reputation: 1004

Like
0Likes
Like

Posted 21 March 2013 - 07:52 AM

I added the camera class which is used to get the viewMatrix aswell, but I dont think that is the problem since I can navigate with keyboard and mouse naturally as in any FPS



#4 TheChubu   Crossbones+   -  Reputation: 3631

Like
0Likes
Like

Posted 22 March 2013 - 05:51 AM

Is your "dirToLight" vec3 in camera space?

 

I think you might have a dirToLight (an unit vector I guess) specified in word space, and since you're doing your lighting in camera space, you get weird lighting from it.

 

Since its an uniform, its better if you transform your dirToLight to camera space (MVMatrix * dirToLight) in the CPU and then pass it as an uniform to the GPU.


"I AM ZE EMPRAH OPENGL 3.3 THE CORE, I DEMAND FROM THEE ZE SHADERZ AND MATRIXEZ"

 

My journals: dustArtemis ECS framework and Making a Terrain Generator


#5 KaiserJohan   Members   -  Reputation: 1004

Like
0Likes
Like

Posted 22 March 2013 - 07:55 AM

I don't want the light direction to change when I orient/move my camera around; that happens when I try to do lighting in camera space right?

 

EDIT: So I tried doing the lighting normalization in world space. So, the LightDir is unchanged (its specified in world space) and then as 'MVMatrix' I actually pass the 'modelMatrix', and otherwise the code is unchanged.

As a result, I see only light being applied when I increase the X-values of LightDir; changing Y and Z values dosn't cause any lightning whatsoever. Why could this be?


Edited by KaiserJohan, 22 March 2013 - 11:04 AM.


#6 KaiserJohan   Members   -  Reputation: 1004

Like
0Likes
Like

Posted 22 March 2013 - 05:10 PM

Alright, I solved it.

 

For those interested, I can do the lighting calculations in world space just fine if I supply a mat4 model-to-world matrix rather than a mat3 model-to-world matrix to the vertexshader. I believe it has something to do with translation, since the objects in question were translated away abit from the camera, but if anyone has a complete answer I would be happy to hear it. 

 

 

EDIT: One thing itches me though, why do people/guides do all lighting transformations in camera space rather than world space?


Edited by KaiserJohan, 22 March 2013 - 05:14 PM.





Old topic!
Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.



PARTNERS