Copy color to depth, defer depth buffer copies #15858
Conversation
hrydgard
force-pushed
the
copy-color-to-depth
branch
from
35366ab to
58a3da6
Compare
|
EDIT: This turned into #15859 (chose the depal path for now), though there are issues. I think this is ready to go in as-is now. |
…th_render to GPUCommon.
hrydgard
force-pushed
the
copy-color-to-depth
branch
from
f8ab610 to
12db0e5
Compare
|
I'm gonna get this in soon, it seems to work well now, with the lone exception of Colin McRae. I do plan to fix that one again, when I go look at the other problems it has (like its sun effect). |
| @@ -1685,6 +1685,16 @@ void GPUCommon::Execute_Prim(u32 op, u32 diff) { | |||
| return; | |||
| } | |||
|
|
|||
| if (!gstate_c.usingDepth) { | |||
| bool isClearingDepth = gstate.isModeClear() && gstate.isClearModeDepthMask();; | |||
There was a problem hiding this comment.
nit: ;;.
-[Unknown]
| } else if (source.channel == RASTER_COLOR) { | ||
| VirtualFramebuffer *src = source.vfb; | ||
| // Copying color to depth. | ||
| BlitUsingRaster(src->fbo, 0.0f, 0.0f, src->renderWidth, src->renderHeight, dest->fbo, 0.0f, 0.0f, dest->renderWidth, dest->renderHeight, false, DRAW2D_565_TO_DEPTH, "565_to_depth"); |
There was a problem hiding this comment.
nit: Maybe log drawnFormat if not 565? Maybe related to the Colin thing?
-[Unknown]
There was a problem hiding this comment.
I'm planning to remove drawnFormat soon, and instead have duplicate VirtualFramebuffer, and automatic reinterpret copies between them on-demand. But yeah, can certainly add such logging in the meantime.
| @@ -1957,7 +1959,7 @@ void FramebufferManagerCommon::UpdateFramebufUsage(VirtualFramebuffer *vfb) { | |||
|
|
|||
| checkFlag(FB_USAGE_DISPLAYED_FRAMEBUFFER, vfb->last_frame_displayed); | |||
| checkFlag(FB_USAGE_TEXTURE, vfb->last_frame_used); | |||
| checkFlag(FB_USAGE_RENDERTARGET, vfb->last_frame_render); | |||
| checkFlag(FB_USAGE_RENDER_COLOR, vfb->last_frame_render); | |||
There was a problem hiding this comment.
Hm, what if it's only rendering depth? I could see it being useful to know that... hm. Maybe for render-to-texture selection too. Can't remember if I've seen games render a scene with RGB and stencil all masked out, just to establish depth... feel like I have...
-[Unknown]
There was a problem hiding this comment.
Yeah, this tracking isn't perfect yet. I don't think I have seen such a case though, but I also don't think it'll hurt very much if we get that wrong..
Goes on top of #15857 (currently includes all of it, but will be rebased once that's merged).This supersedes #15777 , and includes all of it. Anyway, it:
This has one known regression, which we can live with temporarily (the fix might shake out of future changes):