For reads, we use the LD_PKA (AKA LD_BUFFER) so we can directly
pass the buffer index. For writes, we still convert the SSBO index
into a global address before doing a global load/store/atomic
operation, but we do that with an LEA_PKA instruction that takes
care of bounds checking.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31164>
On Mali(Valhall), we have a way to load SSBO data without going through
an SSBO index -> global address translation, so let's provide a way
to tell nir_lower_ssbo() when it shouldn't lower loads.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31164>
We don't merge subpasses, so we can't turn subpass attachment
loads into tile buffer reads yet. Let's just treat those as
regular 2D textures for now (as we do on Bifrost).
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/11875
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Tested-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31166>
On valhall, the sample index should go in the R component
of the image load/store/lea instruction. This provides a
straightforward way to implement image2DMS and
image2DMSArray image load and store for valhall.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30521>
The nir_lower_image_atomics_to_global pass can create some image
load/stores, so we need to do the multisample image load/store
lowering after this.
Also, the pass only actually works on bifrost and below, so skip it
for valhall.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30521>
The pan_arch function is useful elsewhere, and doesn't rely on
anything else within genxml/gen_macros.h.
It's useful, for example, to find the architecture from the
GPU id in bifrost_compile.c, where before we were using ad-hoc
shifting.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30521>
Previous changes of the helper invocation pass fixed missing conditional
control flow tracking but this is not enough.
Propagation of the dependency chain also need to handle value outside of
direct predecessors.
This fix "dEQP-VK.graphicsfuzz.cov-nested-loops-sample-opposite-corners"
for real this time.
Fixes: 33fef27356 ("bi: Do not mark tex ops as skip when dest is used by control flow")
Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30754>
On Valhall, we can store the layer index in PositionFIFO attributes and
have the primitives dispatched to the appropriate list in the tiler
context, which means we no longer have to issue N IDVS jobs when doing
layered rendering.
On the fragment shader side, we can pass the layer index through the
frame_argument field, which can be preloaded in r62-r63, so do that to
save a push uniform slot.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30695>
Make pseudo instructions for the IR separate from real Bifrost and
Valhall instructions, which are kept in their own ISA.xml files.
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30179>
Make the valhall ISA parser valhall.py have a functional interface
returning a tuple, rather than making users directly access variables
within it.
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30179>
Updates the ISA.xml parser to be able to handle some of the constructs
from the Valhall ISA.xml (which differs in significant ways from the
Bifrost ISA.xml). The eventual intent is to avoid duplicating instructions
in the two files, although that isn't enabled in this patch.
The new features aren't used yet, that will be in a future commit.
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30179>
We were using the first character of names to indicate the execution unit
('+' for add, '*' for fma). Change the ISA.xml file to have an explicit
`unit` attribute for instructions; this makes the XML more flexible
for future architectures and matches what the valhall ISA.xml does.
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30179>
Apply the same optimisation as ACO and AGX.
Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com>
Reviewed-by: Eric R. Smith <eric.smith@collabora.com>
Acked-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30484>
This trim vector srcs to the appropriate component count
based on the write mask.
This also should help with image store as the vector srcs
will be trimed according to the format if its known.
Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30484>
Ensure we vectorize load/store when possible.
Also move lower pack after loop optimization.
This drastically reduce the shader size of
"dEQP-VK.graphicsfuzz.spv-stable-maze-flatten-copy-composite" and allow
it to pass instead of timing out but it might greatly help others.
Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30484>
Previously, it was possible to have a texture operation marked as SKIP
while one of the dests was in use in conditional control flow.
If an helper thread was to execute that instruction, it would result
in an undefined value being used.
This fix
"dEQP-VK.graphicsfuzz.cov-nested-loops-sample-opposite-corners" where
helper threads would get stuck inside a loop depending on the result of
a TEXS_2D invocation.
Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30484>
Will be used for DCE and helper invocations pass changes.
Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30484>
rewrote most of the impl but shrug.
regresses code gen for mediump but I'm not too bothered given the lackluster
perf of fp16 on bifrost :(
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30567>
ir3's lowering of variables to scratch memory has to treat 8-bit values as
16-bit ones when comparing such value's size against the given threshold
since those values are handled through 16-bit half-registers. But those
values can still use natural 8-bit size and alignment for storing inside
scratch memory.
nir_lower_vars_to_scratch now accepts two size-and-alignment functions,
one used for calculating the variable size and the other for calculating
the size and alignment needed for storing inside scratch memory. Non-ir3
uses of this pass can just duplicate the currently-used function. ir3
provides a separate variable-size function that special-cases 8-bit types.
Signed-off-by: Zan Dobersek <zdobersek@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29875>
Not the most efficient approach but functional.
Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com>
Acked-by: Rebecca Mckeever <rebecca.mckeever@collabora.com>
Reviewed by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30088>
Also ensure that we never emit vector with more than 4 components.
Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com>
Reviewed-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30088>
The semantics of discard differ between GLSL and HLSL and
their various implementations. Subsequently, numerous application
bugs occurred and SPV_EXT_demote_to_helper_invocation was written
in order to clarify the behavior. In NIR, we now have 3 different
intrinsics for 2 things, and while demote and terminate have clear
semantics, discard still doesn't and can mean either of the two.
This patch entirely removes nir_intrinsic_discard and
nir_intrinsic_discard_if and replaces all occurences either with
nir_intrinsic_terminate{_if} or nir_intrinsic_demote{_if} in the
case that the NIR option 'discard_is_demote' is being set.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/27617>
VK_KHR_buffer_device_address was enabled but actually depends on
VK_KHR_device_group.
This trivialy implement device group extensions like other drivers.
Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29410>
This move most of the disassembler detail to disasm.py.
Signed-off-by: Mary Guillemard <mary.guillemard@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29161>
Base offset of the push constant access shouldn't be taken into
account when selecting the push constant words to load. We should
instead assume the first word in the range is the base of the
dynamic indexing, which also simplifies the code.
Fixes: d53e848936 ("pan/bi: Lower load_push_constant with dynamic indexing")
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29670>
This is the granularity of global loads/stores. Since all other type
of memory accesses have the same or even smaller constraints, we don't
even bother testing the intrinsic type.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29670>
The code for checking flow control did not realize that
`LD_TEX` and `LD_TEX_IMM` were memory accesses, and hence was
not inserting waits where these were necessary. This showed up
as flakes in KHR-GLES31.core.shader_image_load_store.basic-glsl-misc-fs
Cc: mesa-stable
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29363>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28899>
We were erroneously specifying Y for 1D arrays
Cc: mesa-stable
Suggested-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28899>
This removes the need for drivers to handle both versions. The base will
get added once in nir_lower_system_values when converting from deref to
intrinsic and will be replaced by a zero for users not supporting it.
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/26800>
The current code was doing some assumptions on the maximum
texture/sampler sizes which no longer stand with panvk. Add explicit
range checks on the direct texture/sampler_index.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
Fragment color outputs can be defined as a vec4 array. As long as the
indexing is not dynamic, we can add the constant offset passed to
the store_output() intrinsic to the RT index we extracted from the
location.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28417>
Push constants are exposed as special registers on Bifrost/Valhall,
this means we can't index the push constant region with a dynamic
index. In order to support dynamic indexing, we need iterative CSELs
to select the right value from the access range.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28175>
On Bifrost, push constants are exposed as 64-bit registers which can
be accessed at a 32-bit granularity. Make sure push constant accesses
are lowered to guarantee a 32-bit alignment.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28175>
The gallium and vulkan drivers deal with vertex attribute emission
differently. The gallium driver re-emits the VS attributes on each
draw, while the vulkan driver uses explicit attribute/image
descriptor dirtiness tracking, and could keep the attribute array
around if a new pipeline using a different number of attribute is
bound. If we want to be able to do that, we need to assign a fixed
offset for image attributes, such that the Vulkan descriptor
lowering pass knows where the images are in the attribute table.
We could teach the Bifrost backend how to deal with a custom offset
but it doing that in a lowering pass also simplifies the Midgard
code.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/28200>
