Finally, a DirectX 10 video driver for Irrlicht

[Granyte]

the loop

Code: Select all

 
for (u32 i=0; i<DepthBuffers.size(); ++i)
    {
        if ((DepthBuffers[i]->Size.Width>=optSize.Width) &&
            (DepthBuffers[i]->Size.Height>=optSize.Height))
        {
            if ((DepthBuffers[i]->Size.Width<destSize.Width) &&
                (DepthBuffers[i]->Size.Height<destSize.Height))
            {
                depth = DepthBuffers[i];
                destSize=DepthBuffers[i]->Size;
            }
        }
    }
 
 

in void CD3D11Driver::checkDepthBuffer(ITexture* tex)

at line 3017


should be

Code: Select all

for (u32 i=0; i<DepthBuffers.size(); ++i)
    {
        if ((DepthBuffers[i]->Size.Width==optSize.Width) &&
            (DepthBuffers[i]->Size.Height==optSize.Height))
        {
            if ((DepthBuffers[i]->Size.Width<destSize.Width) &&
                (DepthBuffers[i]->Size.Height<destSize.Height))
            {
                depth = DepthBuffers[i];
                destSize=DepthBuffers[i]->Size;
            }
        }
    }

otherwise dx11 will refuse bind the depth texture as it will often be larger then the current texture



EDIT after more test this fix is barely better when using more then 1 rtt of diferent sizes the driver always end up crashing

[Darktib]

@Nadro: That's a perfectly valid point, but yes I think naming it set_used_safe would be better, as it would not break existing code will giving the user the choice.

[Nadro]

@Darktib
I fixed set_used method.

@Granyte
Sorry for a delay with commits related to instancing, but I have some problems with my new dev platform. I'll try to fix it ASAP.

[Granyte]

it's alright i'm fighting some oddities inside irrlicht at the moment

many of the things I had working in the 1.8 branch are broken in this one some of those are delaying my work on instancing

[Darktib]

Thank you for the commit !

[The_Glitch]

I'd also like to add I've been trying to use spark with shader-pipeline branch. If you make the recommended change that Darktib suggest, it breaks hardware skinning for some reason? I did a test with and without change and confirmed that models no longer render if you make that change in Cvertexbuffer file.

I'd like to also say I haven't updated beyond the orginal pipeline file's beyond the first HWS example. You may have fixed this already let me know if you have so I can update.

[Nadro]

After changes HW skinning still works properly, because IMeshManipulator::copyVertices at now uses 'fill' method instead of 'set_used'.

[The_Glitch]

Okay thanks, Darktib has been helping me with the spark library, he suspected some of the issues I was having was due to my Irrlicht copy of pipeline branch not being fully up to date. I'll re download and compile you shader pipeline version again to reflect your latest changes thanks.

[robmar]

Are cube maps supported on the driver at this time?

[Granyte]

not that I know but i'm working on them for dx9

[robmar]

Thought the focus was on a dx11 driver. Is there some complexity with dx11?

[Granyte]

Render target are broken on dx 11 so i cannot port my main project to dx11 when i finally fix them i will start pushing new feature equally across dx9 and 11

Also we don' even have a dx11 driver and dx12 is comming

[Nadro]

@Granyte
DirectX 12 will be not available too fast IMHO probably late 2015 with Windows 9.

[robmar]

Yeah I imagine its a lot of work. I could spare a couple of hours a week if you need someone to do anything specific, code testing, bug finding..

[Granyte]

@Granyte
DirectX 12 will be not available too fast IMHO probably late 2015 with Windows 9.

It depend the rumor is they wanna kill mantle with it and since mantle is out i doubt they will wait a year and an half to give mantle a head start

Yeah I imagine its a lot of work. I could spare a couple of hours a week if you need someone to do anything specific, code testing, bug finding..

So far transparent materials and rtt are broken and i don't understand why especially as rtt used to work before the port to the fvf