SVN2445 : Software rendering

[jacky400]

What would be the benefit of the upcoming implementation of software rendering ? (Accuracy of graphic emulation ?)

[serio]

i'm guessing the emulator will work for people with poor graphic cards like all of the intel chipsets thanks to it.

[bugmenot]

If it's anything like other emulator software rendering :
- psp-like graphic
- use enormous cpu cycle relative to hw rendereing
- less buggy and more accurate than hw rendering in its complete form

[Ghost Ripper]

Are shader options on AMD fixable? Cause that would be a huge benefit (

[bugmenot]

Do you even understand what "software rendering" is?

[Ghost Ripper]

Do you even understand what "software rendering" is?

Software rendering will circumvent current glitches with shaders (because it is software assisted rendering, ie not using graphical and other hardware assisted accelerations, amirite?), which exist specificaly on AMD videocards. So what are you referring to when you ask about understanding?

[bugmenot]

It's not software assisted, it's fully software, there's no point in doing one if it would still condone to specific manufacturer limitation.

[Ghost Ripper]

It's not software assisted, it's fully software, there's no point in doing one if it would still condone to specific manufacturer limitation.

I meant that, sorry for my broken english.

[gid15]

What would be the benefit of the upcoming implementation of software rendering ? (Accuracy of graphic emulation ?)

The software rendering is a CPU implementation of the PSP rendering engine. The current Jpcsp implementation is using OpenGL to implement the PSP rendering engine, i.e. is using the GPU (the processors on the graphic card).
The software rendering is not using the GPU at all (well, just a little bit at the end of the rendering to display the final result).

My goal with this implementation is to try to workaround the different rendering problems reported on some hardware (ATI/AMD, Intel...). It also allows the implementation of some PSP features that were up to now quite difficult to provide using OpenGL.

The main drawback of the software rendering is its performance. I currently put my focus on compatibility but I have several ideas to improve the performance in the future. I still do not know if it will be possible to reach a normal FPS rate in 3D games. We will see...

I also want to mix both renderings: render a part of a frame with OpenGL when it can do its job correctly and render the rest of the frame in software when the OpenGL implementation fails. This would mitigate the performance problem. But such an approach has still to be investigated.

It will probably take a few months to get a decent software rendering, so do not put too high expectations on the first draft.

I'm currently working on the last bits of the lighting engine and testing against the demo applications. I will provide a setting option to enable the software rendering when these basic tests are successful.

[bugmenot]

Would love to see hybrid rendering as full software mode usually looks, eh, ugly. Thanks for your hard work as always grid, thought it's also probably an interesting project to code.