As explained in the previous commit, GFX9+ has issues with addressing
mipmaps in block-compressed images. In the case of copy commands, we fix
this by doing an extra copy for the missing blocks.
For GFX10, the mipmap layout in memory allows us to do better than that. We
can change the base level of the descriptor to one level bigger than the
requested level and adjust the extent and address to match. This is done by
ComputeNonBlockCompressedView in addrlib. Thus on GFX10 we can skip the
fixup copy workaround, and this will also fix cases outside of explicit
copy commands.
Signed-off-by: John Brooks <john@fastquake.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Acked-by: Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17970>
GFX9+ hardware has an issue where mipmap degradations are calculated
incorrectly due to using divide-by-two integer math and certain mipmap
sizes lose blocks.
This issue has been documented before, and we ported a workaround from
AMDVLK to increase the extent that is programmed into the descriptor, so
that the hardware arrives at the correct result. However, this is
insufficient as we cannot safely increase the extent beyond the physical
extent of the image in memory. If we can't increase it enough, the image
will still be missing blocks.
But there is still hope. In cases where RADV is responsible for copying to
or from an image (such as vkCmdCopyBufferToImage/vkCmdCopyImageToBuffer),
we can perform a second copy of the blocks that the hardware excluded so
that the resulting image is complete. This is another workaround from
AMDVLK.
This fixes corrupted textures in Halo: The Master Chief Collection.
v2: Add RADV_CMD_FLAG_INV_L2 | RADV_CMD_FLAG_INV_VCACHE to flush_bits
just in case (Samuel Pitoiset)
Closes: #3347
Signed-off-by: John Brooks <john@fastquake.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Acked-by: Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17970>
This workaround need not apply to subsampled formats.
Signed-off-by: John Brooks <john@fastquake.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Acked-by: Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17970>
This assumption is probably valid with render passes but it won't for
dynamic rendering where we may not have a layout for depth at all.
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18084>
It's entirely redundant with radv_can_fast_clear_depth except that it's
missing a few checks around view masks and the layer range in the actual
clear rect and assumes you're always clearing the whole image view.
This isn't necessarily true, even for classic Vulkan render passes.
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18084>
This is only used in the case of resolve image clears which are going to
go away entirely when we start doing dynamic rendering only.
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18084>
As soon as the if ends (or possibly before?) the VkSampleLocationsInfoEXT
goes out of scope. We need it to stay valid until after the
vkCmdPipelineBarrier completes.
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18084>
This commit fixes
dEQP-VK.rasterization.rasterization_order_attachment_access.format on
GFX9 because changing the layout for Vulkan feedback loops will trigger
a fast-clear eliminate. Though, the root cause is unrelated to that and
it's because the CMASK/FMASK initialization on GFX9 is currently broken
for TC-compatible images (there is a TODO somewhere).
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18084>
Even with dynamic rendering, you have to bind both aspects of the image
if the image contains both depth and stencil. One day, we may see this
restriction lifted but that will require deeper driver surgery into the
way we handle depth/stencil layouts.
Fixes: 42db590006 ("radv: convert the meta blit 2d path to dynamic rendering")
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18084>
This is a link step because it needs to know if the previous stage is
a mesh shader.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18138>
This also moves merging the tess info.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18138>
Replace these ugly loops with something much cleaner and easier to
maintain.
There shouldn't be any functional changes.
No fossils-db changes (tested on NAVI21, VEGA10 and POLARIS10).
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18138>
It can be accessed from the pipeline key instead.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18138>
For consistency.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18138>
This looks cleaner.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18138>
We do something similar for shader args and this allows to remove this
ugly hack when postprocessing NIR shaders.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18138>
To copy info to merged stage at the same place.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18138>
If the pipeline type is graphics, VK_PIPELINE_CREATE_LIBRARY_KHR
shouldn't never be set, so this was redundant.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18138>
When SCC is clobbered between s_cmp and its operand's writer,
the current optimization that eliminates s_cmp won't kick in.
However, when s_cmp is the only user of its operand temporary,
it is possible to "pull down" the instruction that wrote the operand.
Fossil DB stats on Navi 21:
Totals from 63302 (46.92% of 134906) affected shaders:
CodeSize: 176689272 -> 176418332 (-0.15%)
Instrs: 33552237 -> 33484502 (-0.20%)
Latency: 205847485 -> 205816205 (-0.02%); split: -0.02%, +0.00%
InvThroughput: 34321285 -> 34319908 (-0.00%); split: -0.00%, +0.00%
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16266>
This was simply left out by accident when I wrote this.
Fossil DB stats on Navi 21:
Totals from 70165 (52.01% of 134906) affected shaders:
CodeSize: 246375656 -> 245814396 (-0.23%)
Instrs: 46519773 -> 46379458 (-0.30%)
Latency: 385159303 -> 385089261 (-0.02%); split: -0.02%, +0.00%
InvThroughput: 66490172 -> 66487867 (-0.00%); split: -0.00%, +0.00%
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16266>
Also delete them when they are already dead in process_instruction().
No Fossil DB changes.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16266>
Thos are only included once, so just move the code into the
corresponding .comp files.
Signed-off-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17963>
get_ssa_temp's and NIR's bit size can differ for scalar sources.
This causes broken packing of the MIMG operands with A16/G16.
Fixes: f5f73db846 ("aco: Support 16bit sources for texture ops.")
Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18008>
Separately track descriptor set layouts for pools with
FREE_DESCRIPTOR_SET disabled, instead of reusing
`radv_descriptor_pool::entries`. This saves 8 bytes for each entry and
also avoids some indirection.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16541>
DCC shouldn't be compressed for Vulkan feedback loops (ie. render pass
with input attachments). It looks like it has always been broken...
Note that GFX9 and earlier chips aren't affected because they don't
compress DCC on GENERAL.
This fixes
dEQP-VK.rasterization.rasterization_order_attachment_access.format*.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Joshua Ashton <joshua@froggi.es>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18104>
Section 11.2.6 of the Vulkan spec states:
"Importing memory from a particular host pointer may not be possible due to
additional platform-specific restrictions beyond the scope of this
specification in which case the implementation must fail the memory
import operation with the error code VK_ERROR_INVALID_EXTERNAL_HANDLE_KHR."
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17995>
This experimental and very suboptimal implementation of graphics
pipeline library shouldn't be used by anyone, except for development
purposes. It's still under active development from my side. This
extension is only exposed via RADV_PERFTEST=gpl anyways.
It's very suboptimal because RADV doesn't yet support PS epilogs and
VS prologs need to be improved/reworked because GPL changed the logic
(eg. can't know the next stage when compiling a prolog). So, currently
the driver ALWAYS retains NIR shaders during libs creation and it links
and compiles in the final pipeline.
I have WIP branches for approximately everything but it looks simpler
to start merging an initial implementation in order to improve steps
by steps from main. This commit might look simple and short but I
already merged TON of preliminary work. RADV was definitely not ready
for GPL and it's still not completely fine. Expect refactoring again.
This implementation currently pass
dEQP-VK.pipeline.pipeline_library.* on NAVI21, except few feedback
creation tests due to CTS bugs.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17542>
For gfx pipeline libraries we would have to keep track of the gfx
state, so change how it's allocated.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17542>
It currently inherits from radv_graphics_pipeline because it's simpler
but I think it could be improved because most of fields won't be useful
for libs.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17542>
It also initializes dynamic states and active stages. It will also be
used when creating a graphics pipeline library.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17542>
