Software renderer beating Radeon/GeForce

[Panajev2001a]

This would be quite INTERESTING for Cell

32 * 4 = 128 FP SIMD units, but the back-end re-write would require pro-tools by Sony to do it...

If I were Nick I would approach Sony... they might like the idea of having also available games easily ported from PC to steal all the games Xbox 2 would get... take that DX advantage...

You would still have power for physics and A.I. if you targeted 30-60 fps and a moderate resolution and who cares if you only do 1.5 MVertices/frame... yeah it is not like more modern engines for PlayStation 3 ( a custom made engine for PlayStation 3 would push more Vertices/frame than what an engine using this software Shader would, but it would not be that much of an issue as the graphics would still be beyond what current generation can do [by a small or a big amount] ), but it is light years beyond PlayStation 2, Xbox and GCN...

 

[Nick]

If I were Nick I would approach Sony... they might like the idea of having also available games easily ported from PC to steal all the games Xbox 2 would get... take that DX advantage...

I'm a student and I put all my free time into working on my projects, but I can't do a full-time job although I'd love to...

 

[Saem]

Take some time off, make some fat cash. Might be interesting to see what Sony says and possibly/hopefully offers. Don't hesitate in exploring that option, remember, YOU can always say no. But finding out doesn't hurt.

 

[darkblu]

If I were Nick I would approach Sony... they might like the idea of having also available games easily ported from PC to steal all the games Xbox 2 would get... take that DX advantage...

I'm a student and I put all my free time into working on my projects, but I can't do a full-time job although I'd love to...

i'd second Panajev's advice: get the project onto the next level - depending on your preferences and plans you can contact some CGI tools devhouses (they tend to be interested in software technologies such as yours) or game middleware devhouses. Or if you'd like to keep the project open you can approach some of the larger open 3d API/engine projects out there like Mesa and CrystalSpace. Finally, i bet those guys at realtech would be interested in your work, too. i believe if you choose any of the above routes you'd stay focused on the project, and gain additional momentum. and don't worry about full-time - there are alternatives to that.

 

[Panajev2001a]

I understand you are a student, but I heard that in a nice Corvette going to classes is particularly sweeter



If I were Sony I'd hop onto your project or one similar... you have a good technology and the sooner you could offer one of these guys ( you could contact Criterion: they'd love it for Renderware ) this opportunity the less probable is that they have an alternative in their R&D garage and the more fat cash as SA put it they would offer for access to this technology...

Pixel Shaders only had the MS made Reference Rasterizer which is ridicously slow: there are tons of opportunities for a fast software Pixel Shader especially on an architecture that is fast enough...

You are playing with SSE that do not even have a fast dot product operation (not as useful as a 1 cycle pipelined FP MADD )...

This could be interesting ( in a more limited fashion as the EE is not as fast as Cell ) also for PlayStation 2...

You have 2 VU with a total of 9 FP MADD/cycle ( two Vector FP MADD + 1 FP MADD [EFU] ) and then you have the FPU with another FP MADD/cycle.

Basically that is two 128 bits SIMD units and two FPUs and you have two 64 bits ALUs ( which can work as a single 128 bits SIMD Integer Unit ).

Even if it is 300 MHz that doesn't sound too bad compared to MMX + SSE of the Pentium 3

I think that PlayStation 3 will offer more HW features in its GPU, but the programmability of the GPU is related to the fact that it uses Cell technology ( APUs ).

Each APU contains 4 FP Units and 4 Integer Units and each group of 4 units ( the 4 FP Units, the 4 Integer Units ) can work as a 128 bits SIMD unit.

You would basically find APUs on the CPU and the GPU of PlayStation 3 ( if they keep their plans similar to what the land grabbing patent specified or 32 APUs on the CPU and 8 APUs on the GPU ).

You know better than me that the kind of architecture they seem to be following is open to MANY possible rendering approaches.

Yours could be one of them.

You could have also a software Vertex Shader, I think it would not be impossible to match VS 2.0+ to the APUs ( they are standalone processors with branching capability [no branch prediction IIRC] and they support SIMD instructions ) as they could be used for both Vertex Shaders and Pixel Shaders.

Then you could have programmers writing code for DirectX 9 and with some small fixes and using your tools they could port their code on PlayStation 3.

Sure, some developers like Konami and Square might prefer to have their in-house engine and Sony is not going to force them not to use the regular PlayStation 3 APIs or Renderware, but what matters is that your solution would be even cheaper and could allow even more developers to code for PlayStation 3 and Sony would like that ( if their games are crap they will not pass the Q&A testing and won't get Sony's approval, but the ones that are good will enrich their library ).

Even if your solution would push less Vertices/frame than a very optimized engine it would still manage more than what PlayStation 2 could, quite a bit more, and using more effects and costing much less than a very customized and optimized PlayStation 3 engine allowing small developers to concentrate on art and gameplay.

I think your work could be integrated in a complete low cost SDK which would be priced competitvely with Renderware ( cheaper, Renderware doesn't seem THAT cheap ) and would be the joy of small developers.

 

[Nick]

Wow that's a long post!

Thanks for the advice, I'll think about it!

 

[Panajev2001a]

You are welcome, I hope you do succeed...

And if you decide to also have a freeware version for PS2Linux users... well that buys me a PS2Linux kit... I cannot take it out my mind...

 

[Nick]

I've made a small update. I fixed a bug with source swizzling, speeded up compilation of large shaders a lot and added a check for MMX/SSE compatibility.

http://sourceforge.net/project/showfiles.php?group_id=76723&release_id=155845

 

[Nick]

Another update: it now supports all depth compare modes, depth write enable/disable and cull modes. You can have over 10^14 combinations of render modes. Anyone want to test them all? ;-)

http://sourceforge.net/project/showfiles.php?group_id=76723&release_id=155845

 

[Sage]

cell, cell, cell. im tired of people still talking about cell as though its near because its not. best bet is 2007. PS3 will be supped up PS2, so if you can make it run on a PS2 then they should be VERY excited about it.

 

[chavvdarrr]

any info how to use modify used shader?
there is compiled version of the shader(shader.asm), but no pre-compiled source or info how to "compile" ps source...

 

[Nick]

any info how to use modify used shader?
there is compiled version of the shader(shader.asm), but no pre-compiled source or info how to "compile" ps source...

It's in Shader/Shader.txt. I also suggest to check Application/Application.cpp and Renderer/Renderer.hpp.

I'll write some documentation when I've got the time, but right now I have to study for my exams...