Messages

1

Graphics / Re: Efficient way to work ontop of a static image each frame

« on: April 23, 2012, 02:10:35 pm »

Thanks for the suggestion! I'll experiment with this.

2

Graphics / Efficient way to work ontop of a static image each frame

« on: April 23, 2012, 01:02:19 pm »

Hi,

I'm doing something which is similar, but not quite the same as, rendering a movie.

Every frame, I need to override pixel data with CPU operations on top of a read-only sf::Image "staticImage" (this never changes over all frames) - e.g. i'm looking to copy, change, and send the data of the static image extremely efficiently.

Currently i'm just doing:
   

Code: [Select]

dynamicImage.create(staticImage.getWidth(), staticImage.getHeight(), sf::Color(0,0,0,0));
dynamicImage.copy(staticImage, 0, 0);

dynamicImage.setPixel(...);

dynamicTexture.loadFromImage(dynamicImage);

- Could anyone help explain to me to a faster way to do this: I looked at sf::RenderTexture but wasn't sure whether or how I could use this in my situation. I'd like to support larger images, e.g. 8192 x 8192, but currently copying in the way above is about 1 FPS for such sizes.

Thanks

3

Graphics / Porting "Fullscreen" Shaders

« on: March 17, 2012, 11:37:56 am »

Haha yeah namespace SF = sf; to be consistent with the previous work - thankfully that's the only major change in style though now that your using lowercase camelCase

4

Graphics / Porting "Fullscreen" Shaders

« on: March 16, 2012, 06:28:34 pm »

Thanks for the quick reply!

1. Sorry, there's no problem with this at all - I just thought their may have been "GetScreenVertices()" or something - but setting vertices each frame works well:

Code: [Select]

const SF::Vertex vertices[] = { SF::Vertex(SF::Vector2f(0.0f, app.getSize().y)), SF::Vertex(SF::Vector2f(app.getSize().x, app.getSize().y)), SF::Vertex(SF::Vector2f(app.getSize().x, 0.0f)), SF::Vertex(SF::Vector2f(0.0f, 0.0f)) };

Simple is good for this library and its intentions

Thanks.

5

Graphics / Porting "Fullscreen" Shaders

« on: March 16, 2012, 02:25:15 pm »

Hi,

I'm porting some work to SFML - I need to apply fullscreen shaders to the entire window render area, instead of just a texture rectangle. I see you can draw shaders by specifying vertices, but after trying:

Code: [Select]


const SF::Vertex vertices[] = { SF::Vertex(SF::Vector2f(-1.0f, 1.0f)), SF::Vertex(SF::Vector2f(1.0f, 1.0f)), SF::Vertex(SF::Vector2f(1.0f, -1.0f)), SF::Vertex(SF::Vector2f(-1.0f, -1.0f)) };
target.draw(&vertices[0], 4, SF::PrimitiveType::Quads, states);

1. I found it doesn't use normalized coords - how can I therefore draw to the entire window render area efficiently (do I really have to recreate 4 vertices from the app size each frame)?

2. I need to use "uniform mat4" for several of my shaders. Do I just do it manually with glUniformMatrix4fv(..., 1, GL_FALSE, &viewMatrix[0][0]); ?

3. I was looking at the code, I noticed setParameter is called in the update loop, which implicitly does: GLint location = glGetUniformLocationARB(m_shaderProgram, name.c_str()); - isn't this slow when you have quite a lot of parameters? What about providing a int getFastParamID(...)  to call in init, then instead of having strings force users who don't care about optimizations to do setParameter(getID("name"), ...) - but this allows users who do care about performance to just send the private GLint from the init - do you think this'll make much difference? (I don't know much about the framework/principles/conventions in SFML yet though so may have missed something)

4. (Offtopic) Is there an easy way to use subpixel font rendering/expose it from freetype? - I haven't looked into this yet.

- P.S. Thanks so much for keeping SFML lightweight! :wink:  I was very happy to see the output size, with static linking, so small.