Extract VkDeviceQueueShaderCoreControlCreateInfoARM from the queue
create info pNext chain and use shaderCoreCount to limit
max_compute_cores and max_fragment_cores in the panthor group create
ioctl. The core masks remain unchanged, letting the kernel pick which
cores to schedule on.
Signed-off-by: Christian Gmeiner <cgmeiner@igalia.com>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40063>
These don't really fit in pan_texture anymore, so move them to a
separate pan_buffer-file.
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39294>
The function pan_emit_bview_surface_with_stride is only used by
pan_buffer_texture_emit, which cannot use ASTC compressed formats
(already disabled in the format list).
As well as dropping this functionality, also inline the static function
in its only caller.
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39294>
Originally, I thought I could get away with not doing this because we
were sizing up the framebuffer boundary to include the superblocks.
However, that's not actually good enough because it's not just about
when we do a clear. It's actually required for AFBC writes to work
properly with small tiles.
Fixes: 6a56d35705 ("pan/fb: Fill out our own descriptors")
Reviewed-by: Christoph Pillmayer <christoph.pillmayer@arm.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39324>
When the the tile size is not a multiple of a superblock for the given
render target, there's a potential race condition between the write at
the end of one tile and reads in some other tile where the reads may
access a superblock in a partially updated state. If the tile size is a
multiple of a superblock, this shouldn't happen because no superblock is
ever used by more than one tile. But for small tiles, this can happen.
The hardware provides us an escape hatch: the reverse_issue_order bit.
This bit comes with a bit of a performance cost but it's necessary to
ensure rendering correctness.
Fixes: 6a56d35705 ("pan/fb: Fill out our own descriptors")
Reviewed-by: Christoph Pillmayer <christoph.pillmayer@arm.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39324>
Every time we need to adjust what goes to these helpers, it's a giant
pain because of all the function prototypes and the mode table. Just
put everything in a struct and pass that around.
Reviewed-by: Christoph Pillmayer <christoph.pillmayer@arm.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39324>
These changes to the nightly results were missed. We're now passing one
more test, and the trace renders ever so slightly differently. The
differences doesn't seem to suggest anything wrong, so let's just accept
the change.
Fixes: 399c0d22f3 ("panfrost: blending fixes for Midgard")
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/40090>
Wire up the existing Panfrost dithering infrastructure to the Vulkan
extension. The library already supports dithered formats via
pan_dithered_format_from_pipe_format(), so this plumbs the
VK_RENDERING_ENABLE_LEGACY_DITHERING_BIT_EXT flag through to the blend
descriptor emission and color attachment internal conversion paths.
Dithering is only applied to color attachments, not depth or framebuffer
preloads.
Signed-off-by: Christian Gmeiner <cgmeiner@igalia.com>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39781>
I don't know why but these seem to fail only intermittantly now.
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
As a side-effect, we're now also respecting storeOp for separate Z/S.
For interleaved, we still always store because interleaved is annoyingly
complicated.
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
The new framebuffer code properly handles attachments without a store
and the incremental rendering code doesn't rely on the store, either, so
we don't need to always store.
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
This is mostly a matter of reworking the frame shader code to pull image
views from a temp struct instead of always reaching for the pan_fb_load.
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
We already have this check for the JM back-end but the CSF back-end
based the choice on whether or not the shader is writing color which
isn't quite what we want.
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
This makes cmd_emit_dcd() about actually emitting the DCD and moves the
bit that's operation-specific to the operation-specific function.
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
Also rename some stuff internally to stop using the word "preload".
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
Instead of having the pan_fb code have to `max(rt_count, 1)` add then
constantly be doing the same in the driver, just make the condition an
assert and do the `max(colorAttachmentCount, 1)` up-front. This way
we're not constantly worried if everyone has the same count or not. The
pan_fb code will handle having an empty render target just fine.
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
The resolve sits between the render pass and the write at the end. This
adds the resolve ops themselves, support in the shader builder for
building the post-frame shaders to handle them, and an optimization pass
that attempts to fold the resolve op into the store and potentially
eliminate the post-frame shader.
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
This commit both adds a couple more MSAA modes and adds support for them
in shaders. While resolving an image into a framebuffer isn't that
interesting, most of the code to do it will be reused for resolving
render targets, which we do want. The refactor is also nice.
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
This generalizes the shader building code at the cost of an additional
level of abstraction through the pan_fb_shader_op enum. This will let
us share code between load and resolve shaders while keeping the client
interfaces to both of those clean and idiomatic.
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
Right now, we have this concept of sections where one packet can be
broken up into multiple sub-packets. Currently, we handle ths by every
packet's `_packed` struct is just an array of `uint32_t` and we do
pointer math and cast to get at the subset of the `uint32_t` array
belonging to the given section. This works okay if we're operating on
`void *` but it blows up on strict aliasing rules if we're ever trying
to pack sections into a `struct foo_packed` on the stack.
To work around this, make it so that anything with sections emits
something like this:
struct foo_packed {
union {
uint32_t packed[N];
struct {
struct foo_sec1_packed sec1;
struct foo_sec2_packed sec2;
};
};
};
This is entirely safe because C allows unions and structs to overlap as
long as the members that have the same base type end up on top of each
other. Since everything is a `uint32_t` internally, this is safe
according to the C pointer rules. And now that everything is a
substruct, everything is nicely named and we can easily get to sections
without pointer magic.
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
Now that we're doing a merge to fill out the various enable bits, we
don't actually need the pan_fb_info in the attachment helpers.
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
Previously, when we had an unaligned render area, we would flush to
force a full FB load and then clear the render area manually after
loading. This means we're doing a full FB load and then throwing most
of that data away.
The new FB load shaders, however, are capable of doing partial loads,
with the inside of the render area loaded or cleared as we wish and the
border pixels loaded. This way we can handle partial clears directly in
the load shaders instead of having to do this dance. This should
improve perf whenever a non-trivial render area is used.
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
We already had this for LOAD_OP_DONT_CARE but we also need it for
LOAD_OP_NONE.
Cc: mesa-stable
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
The current code looks a lot like how you might do this in GL. We start
off with the load, which might be loads or clears. Then, after the
first flush, we replace that with the spill load. This doesn't really
fit the load/spill/store that the new pan_fb code wants.
Instead, this commit add a couple of separate tracker bits to keep track
of whether or not we've loaded yet and whether or not we're in the final
EndRendering and want to do the "real" store. This allows us to track
where we are in the render pass and do the right load/spill/store
operations.
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
The old code was all out of order and made no sense. There's a reason
it made no sense. It was wrong. Cleaning this up fixes a solid 1/3 of
the remaining Bifrost CTS fails in CI.
Cc: mesa-stable
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39759>
