Build Testing With DX12 Reserved For Windows 10

I have the depth buffer rendered in the bottom right hand corner too, I'm not sure why since it should be set by the scale/offset value.

I enabled mrt read/write. Similar to depth_read, when context_dma_color_a/b/c/d changes, or when a semaphore is released, the content of the color buffer is written to.
The cube_mrt.ppu.elf test should work properly now.

when the dx12 branch is merged, could you do it so that you dont need mvs140(something) to run the emu? because i dont know how the emu is gonna work on linux for example...

I don't understand what you meant. Why is msvc140 problematic for Linux build ?

i dont know, but for some reason, it does not let me run rpcs3 on my windows 7. is there a newer version of Visual C++ Redistributable Packages? i have them for year 2013. So i dont know why the emu does not even run... even if the emu supports dx12, it should be at least able to run and choose the opengl if i want to.

Last Build 30/05/2015

With Last commit, look Here

Other Video

Last Build 30/05/2015
Snip
With Last commit, look Here

Amazing.

Nice progress on the DX12 renderer.

[raven02]

I tested the DX12 renderer with Arkedo swap game, with current code it crashes but with some dirty hack I got it running.
I'm getting almost twice the fps as with GL renderer in game (ie 17 fps with DX12 vs 9-10 with GL).

There is some room for improvement for DX12 (per frame vertex buffer caching is one of them) and GL (again per frame vbo + texture caching) but I think we can't reach 60 fps without tracking texture data modification in local memory (to avoid reuploading texture each frame).

Unfortunalty I don't know where in rpcs3 RSX local memory is "stored" so that I can try to record memory write to region of data.

I tested the DX12 renderer with Arkedo swap game, with current code it crashes but with some dirty hack I got it running.
I'm getting almost twice the fps as with GL renderer in game (ie 17 fps with DX12 vs 9-10 with GL).

There is some room for improvement for DX12 (per frame vertex buffer caching is one of them) and GL (again per frame vbo + texture caching) but I think we can't reach 60 fps without tracking texture data modification in local memory (to avoid reuploading texture each frame).

Unfortunalty I don't know where in rpcs3 RSX local memory is "stored" so that I can try to record memory write to region of data.

This might work. Best idea i've heard. Doing this can help so much more on games.

With last commit Arkedo 2 swap should work (I tested arcade mode)