- 19 Mar, 2021 3 commits
-
-
Mike Blumenkrantz authored
-
Mike Blumenkrantz authored
these don't do anything without the device extension
-
Mike Blumenkrantz authored
this should work
-
- 18 Mar, 2021 37 commits
-
-
Mike Blumenkrantz authored
at some point this will get fixed to work for format conversions
-
Mike Blumenkrantz authored
-
Mike Blumenkrantz authored
this reworks all handling for ubos and ssbos such that they get dumped in as a single array<buffer> descriptor which can then use dynamic indexing in shaders I don't know if this is good, but it works as a proof of concept, potentially at the cost of incurring overhead by sucking up extra descriptors which won't actually be used
-
Mike Blumenkrantz authored
this is allowed for fb attachments, so we can use it to avoid needing to change layouts for zs textures if we know that it isn't going to be written to during a given subpass
-
Mike Blumenkrantz authored
since surfaces are cached and deduplicated, we no longer have to do deep comparisons to determine whether two surfaces are equal and can just compare the pointers
-
Mike Blumenkrantz authored
if we get one of these, there's two possibilities: * fbfetch * haven't yet gotten a set_framebuffer_state the first case isn't implemented yet so who cares, but in the latter case this resource may get set to a layout to apply pending clears before it's actually used as a descriptor, so the descriptor layout needs to be deferred to draw time to ensure that this is taken into account
-
Mike Blumenkrantz authored
this will be useful for applying layouts
-
Mike Blumenkrantz authored
this is more consistent with how gallium applies the states
-
Mike Blumenkrantz authored
-
Mike Blumenkrantz authored
-
Mike Blumenkrantz authored
this lets us avoid the spec-mandated barrier and use flags based on what we'll actually be doing in the pass
-
Mike Blumenkrantz authored
only the earliest stage should be applied for these barriers, so we can skip ones where the existing access is earlier
-
Mike Blumenkrantz authored
this is implicit from queue submit
-
Mike Blumenkrantz authored
we don't need a separate function for this
-
Mike Blumenkrantz authored
this doesn't actually do anything other than test the codepath since the hashed pipeline state will still change on every frontface change, but it's a start
-
Mike Blumenkrantz authored
this is pretty straightforward, just adding some extra calls
-
Mike Blumenkrantz authored
-
Mike Blumenkrantz authored
the contents of these may change frequently, so we should skip all the flush/invalidate
-
Mike Blumenkrantz authored
nvidia claims format feature support for various flags but then doesn't actually support them for a 3D image, only 2D, meaning that an extra step is necessary before creating a linear image
-
Mike Blumenkrantz authored
VK_EXT_4444_formats strikes again
-
Mike Blumenkrantz authored
just to be pedantic
-
Mike Blumenkrantz authored
nvidia is a total liar about what it can do with these in linear mode
-
I hate everything.
-
Mike Blumenkrantz authored
is this really a thing?
-
Mike Blumenkrantz authored
technically shouldn't matter but whatever
-
Mike Blumenkrantz authored
-
Mike Blumenkrantz authored
enforce linear bind more logically and mark with more accurate usage
-
Mike Blumenkrantz authored
by using the bind counts, each rebind can be evaluated to ensure that every descriptor was rebound
-
Mike Blumenkrantz authored
we want to avoid updating these values when possible, which means that if a descriptor is being replaced, we only should update if the replacement is not the same resource
-
Mike Blumenkrantz authored
-
Mike Blumenkrantz authored
with tc enabled, we get genuine multidraws with valid drawids, so we can generally avoid this
-
Mike Blumenkrantz authored
temp while this gets fixed to stop hanging
-
Mike Blumenkrantz authored
-
Mike Blumenkrantz authored
-
Mike Blumenkrantz authored
-
Mike Blumenkrantz authored
this is unnecessary since the locations will match
-
Mike Blumenkrantz authored
-