Instead of re-emitting some dynamic states when a new shader is bound,
only re-emit the user SGPR states. This is slightly more optimal.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31160>
Instead of recreating the packets in the DGC prepare shader, the best
solution is to emit them to a temporary CS object at pipeline creation
time. Then in the DGC prepare shader, the driver just needs to copy
the packets.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31101>
Instead of storing the stride/rsrc_word3 as part of the VBO descriptors,
pass them as parameters. This is cleaner and this will allow us
to simplify VBO in DGC.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30918>
Dynamic rendering local read allows the application to use subpass input
attachments with feedback loops. But unless legacy RPs where it's
possible to determine feedback look at creation time, with dynamic
rendering it's not possible.
To fix that, the driver needs to determine at draw time if a feedback
loop is present, and it needs to decompress DCC/HTILE if necessary.
See https://gitlab.khronos.org/vulkan/vulkan/-/issues/3928 for more
information.
Note that VKCTS is still missing coverage but this has been reported.
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/11127
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30124>
vkd3d-proton always sets the acceleration structure size to be the
whole buffer size. Because of that, allocating read back buffers
for all acceleration structures causes a system with a finite amount
of RAM to OOM.
This is solved by allocating read back buffers on build where the
required size is known.
Reviewed-by: Friedrich Vock <friedrich.vock@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29537>
This used to work by luck because the current DGC prepare shader
is using one descriptor set and it was the currently bound compute
shader... Using two descriptor sets or starting from 1 would just fail.
For indirect compute pipelines, descriptors must be emitted from the
DGC shader because there is no bound compute pipeline at all. This
solution is using indirect descriptor sets because it's much shorter
and easier to implement. This could be improved but nothing uses
indirect compute pipelines and this is like experimental stuff.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29700>
We already track a couple of registers per cmdbuf and this introduces
a generic mechanism, instead of having a bunch of last_xxx fields.
Loosely based on RadeonSI.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28644>
SPI_SHADER_COL_FORMAT/CB_SHADER_MASK are used slightly differently
for PS epilogs, shader objects and monolithic graphics pipelines.
This introduces a new state that will allow us to emit these two
registers in only place. The main motivation is for depth-only RB+
support and for tracking context registers in the cmdbuf.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28976>
This is always the non-compacted format because it's compacted right
before it's emitted. This looks much cleaner to me.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28976>
