Add support for reading depth buffers to the PackFramebufferSync function #15924
Conversation
…tion. Also adds D16 conversion code (though, could probably get away with just creating a 16-bit depth buffer on the GPU and blit to it, where depth blit is available). Not yet used, just getting it in separately.
hrydgard
added
GE emulation
| } else { | ||
| for (uint32_t y = 0; y < height; ++y) { | ||
| for (uint32_t x = 0; x < width; ++x) { | ||
| dst16[x] = (uint16_t)(src32[x] * 65535.0f); |
There was a problem hiding this comment.
This won't be a correct conversion for our purposes, though? I thought I wrote some logic in... I guess it was #11676?
Under some versions of our depth support, we intentionally use 1/4 of the range (in the middle) for the actual depth values, where the rest is meant to be clamped. This reduces the bugs from not having depth clamp. If we have depth clamp, I think we use the full range, though.
We need to account for that somewhere (maybe not in thin3d, not sure) for 24 or 32 bit depth buffers.
-[Unknown]
There was a problem hiding this comment.
Yeah, that's right. I'll leave a TODO for now, since I might indeed try to address this elsewhere. This'll work when we don't reduce the range which is most of the time on Vulkan, for example.
Also adds D16 conversion code (though, could probably get away with just creating a 16-bit depth buffer on the GPU and blit to it, where depth blit is available).
Not yet used, just getting it in separately.
Part of work on #15923 .