dany lepage
- Thread starter Anonymous
- Start date
Dany Lepage
Newcomer
AA support could be fixed in DX9a (assuming the copy function from a multisample render target to a texture is fast enough) but SC (1) won't be "supporting" DX9A. You need to have access to Direct3D9 objects but the code is written with "Direct3DDevice8" calls and changing all the calls to use DX9 won't happen with this game.
It could have been nice to support AA in normal vision mode (there is a noise effect in night vision anyway) but it won't happen with SC1.
Sorry about that.
It could have been nice to support AA in normal vision mode (there is a noise effect in night vision anyway) but it won't happen with SC1.
Sorry about that.
I wanted to ask the same question, especially with regard to 3DMark03 - where the change to DX9a was made - but then I realized that it must have been quite easy for the onions with their code obviously already based on DX9. Changing the code from DX8 to DX9 framework is really not worth the effort at this point, I guess.
DirectX Shadow Buffering
Greetings Dany,
Elsewhere in this forum, there was some discussion regarding general shadow buffer support for non-nVidia graphics cards (e.g. Radoen 9700). Would the use of DirectX 9.0 have allowed the shadow buffer functionality of cards such as the Radeon 9700 to be exposed with enough precision to be useful in the PC version of Splinter Cell?
Greetings Dany,
Elsewhere in this forum, there was some discussion regarding general shadow buffer support for non-nVidia graphics cards (e.g. Radoen 9700). Would the use of DirectX 9.0 have allowed the shadow buffer functionality of cards such as the Radeon 9700 to be exposed with enough precision to be useful in the PC version of Splinter Cell?
Dany Lepage
Newcomer
AA
I stand by a comment I made a while ago in the www.splintercell.com forum. If DirectX would be supporting multisampled texture render targets, that would solve the problem right away without any need for us to do anything. Obviously, that won't happen with DX8. However, I heard Nvidia is looking at providing the option in OpenGL.
They could create a special case and detect SC and make the operation we are doing "possible" (mixing a multisampled depth/stencil buffer with a texture render target (non multisampled)).
I implemented a "normal vision" AA patch but the performance quite simply suck. SC is already very fill limited and rendering the scene twice (once to the mainbuffer and the other to the texture to create the mask for glow) takes a huge toll on performance. I would rather crank resolution by 2 notchs before turning on AA under those circonstances (and it will never work in night vision and you spend 50% of the game (or more) with that post-filter).
I stand by a comment I made a while ago in the www.splintercell.com forum. If DirectX would be supporting multisampled texture render targets, that would solve the problem right away without any need for us to do anything. Obviously, that won't happen with DX8. However, I heard Nvidia is looking at providing the option in OpenGL.
They could create a special case and detect SC and make the operation we are doing "possible" (mixing a multisampled depth/stencil buffer with a texture render target (non multisampled)).
I implemented a "normal vision" AA patch but the performance quite simply suck. SC is already very fill limited and rendering the scene twice (once to the mainbuffer and the other to the texture to create the mask for glow) takes a huge toll on performance. I would rather crank resolution by 2 notchs before turning on AA under those circonstances (and it will never work in night vision and you spend 50% of the game (or more) with that post-filter).
Dany Lepage
Newcomer
Shadow Buffer precision
Even if I didn't try it. I'm almost 100% sure the ATI DX9 implementation would have been fine. In fact, during the porting of the xbox version, we were confronted with which algorithm we would use for a replacement to the depth texture extension used on NV hardware.
The reason we went with shadow projectors is that we wanted an algorithm that could work on DX7 hardware too (ATI can have shadow buffer running on their R200 with PS1.4). The GF2/4MX market is still huge out there.
Considering that I wasn't going to move the SC code base to DX9, the only available option to have shadow maps on ATI cards was to use PS1.4. I almost did it but couldn't find the time.
Even if I didn't try it. I'm almost 100% sure the ATI DX9 implementation would have been fine. In fact, during the porting of the xbox version, we were confronted with which algorithm we would use for a replacement to the depth texture extension used on NV hardware.
The reason we went with shadow projectors is that we wanted an algorithm that could work on DX7 hardware too (ATI can have shadow buffer running on their R200 with PS1.4). The GF2/4MX market is still huge out there.
Considering that I wasn't going to move the SC code base to DX9, the only available option to have shadow maps on ATI cards was to use PS1.4. I almost did it but couldn't find the time.
Luminescent
Veteran
If you can teach me, I'll do it
.It's just a pity us Ati users can't have the shadow buffer support. Check out the difference between projected shadows and shadow buffers:
Guess which is which.
Luminescent said:If you can teach me, I'll do it
How its done (clever use of all 12 bits and a big texture) from ATI themself.
http://www.ati.com/developer/samples/shadowmap.html
Actually works quite well considering how few bits you've got.
Similar threads
