I'm not exactly sure if the results I'm getting are correct, can someone tell me if this is how it should look like?

This is what I do in my deferred renderer:

- Render GBuffer (nothing happening here)

- Render light to accumulation buffer (read specular albedo as linear, meaning pow(color, 2.2f))

- Compose Pass (read albedo from gbuffer as linear and combine it with the lighting but don't convert back)

- Do post processing

- Do tonemapping and write back to the backbuffer (which is R8G8B8A8_UNORM_SRGB)

Is that the correct way of doing it ?

And here's some screens:

http://d.pr/i/mVEW

http://d.pr/i/jeyR

http://d.pr/i/LwSJ

Sorry, can't see your images.

You should do all calculation in linear space and only convert it while reading it from the texture and writing it back to the framebuffer, and the current hardware is able to handle sRGB textures automatically (no need to transform it yourself), further infos are found here.

For testing purpose I would render three bars (greyscale 0...100%):

1. bar: linear texture (e.g. tga)

2. bar: sRGB texture (use sRGB texture format)

3. bar: shader which generate linear ramp

All bars should look the same and should be comparable to an according color ramp image seen in a browser/image processing tool.