On MRT-capable systems, the framebuffer format is encoded as a 64-bit
word in the render target descriptor. Previously, the two 32-bit
words were exposed as opaque hex values. This commit identifies a 12-bit
Mali swizzle and a 2-bit channel counter, removing some of the magic. It
also adds decoding support for the AFBC and MSAA enable bits, which were
already known but otherwise ignored in pandecode.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
These ops were discovered by invoking the correspondingly names GLSL
functions. The rounding ops here behave exact as expected and are mapped
to their corresponding NIR ops where applicable. The ffma behaves as a
LUT instruction and requires some special argument packing (since
Midgard normally only allows for 2 arguments); this quirk will be
addressed in the future, but for now FMA is still lowered.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
This flag corresponds to what was MEM_COHERENT_LOCAL in the vendor
driver, which seems to influence the cache policy, necessary for the
varying temporary storage but nothing else.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Potentially, the kernel could optimize these allocations, or perhaps we
can save on mapping costs.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
For reasons that are still unclear (speculation included in the comment
added in this patch), the tiler? metadata has a fast path that we were
not enabling; there looks to be a possible time/memory tradeoff, but the
details remain unclear.
Regardless, this patch improves performance dramatically. Particular
wins are for geometry-heavy scenes. For instance, glmark2-es2's
Phong-shaded bunny, rendering at fullscreen (2400x1600) via GBM, jumped
from ~20fps to hitting vsync cap at 60fps. Gains are even more obvious
when vsync is disabled, as in glmark2-es2-wayland.
With this patch, on GLES 2.0 samples not involving FBOs, it appears
performance is converging with (and sometimes surpassing) the blob.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
The nir_swizzle helper is used some on it's own but it's also called by
nir_channel and nir_channels which are used everywhere. It's pretty
quick to check while we're walking the swizzle anyway whether or not
it's an identity swizzle. If it is, we now don't bother emitting the
instruction. Sure, copy-prop will clean it up for us but there's no
sense making more work for the optimizer than we have to.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
This header has been unused since f8f2520e88 ("st/mesa: Remove
unnecessary headers"). And in the more than 8 years since, this
hasn't been useful. So let's just get rid of it.
Signed-off-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>
From the bash manual:
string1 == string2
string1 = string2
True if the strings are equal. = should be used with the test
command for POSIX conformance.
Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Test using array deref on vectors in loads and stores. These are
marked DISABLED_ as this optimization is currently not done.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Replace find_next_intrinsic(intrinsic, after) with
get_intrinsic(intrinsic, index). This makes slightly more convenient
to check the resulting loads/stores/copies, since in most tests we
know which one we care about. The cost is to perform more traversals,
but for such tests this is not a problem.
Added the ASSERT_EQ() on count to some tests missing it, so the
indices queried are always expected to find something.
Also, drop two nir_print_shader leftover calls in a test.
v2: Remove redundant assertions. nir_src_comp_as_uint already
assert what we need. (Jason)
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
When a copy_entry is SSA, store not only the nir_ssa_def* for each
component, but also the source component they come from. At the
moment this is always a match (i.e. 'component[i] == i'), because all
the operations for a copy_entry happen using definitions with the same
size. This prepares the code for array_derefs of vectors, in which
'component[i] != i'.
Also, extract setting all SSA components into a function of its own.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Disabled by default, to be used during development. Adding those
so I don't rewrite some ad-hoc version of them everytime I'm working
with this pass.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
For now these derefs are not handled, so don't let these get into the
copies list -- which would cause wrong propagations. For load_derefs,
do nothing. For store_derefs, invalidate whatever the store is
writing to. For copy_derefs, invalidate whatever the copy is writing
to.
These cases will happen once derefs to SSBOs/UBOs are kept around long
enough to get optimized by copy_prop_vars.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Rather than only lowering if all srcs are scalarizable we instead
check that at least one src is scalarizable.
We change undef type to return false otherwise it will cause
regressions when it is the only scalarizable src.
total instructions in shared programs: 13219105 -> 13024547 (-1.47%)
instructions in affected programs: 1153797 -> 959239 (-16.86%)
helped: 581
HURT: 74
total cycles in shared programs: 333968972 -> 324807922 (-2.74%)
cycles in affected programs: 129809402 -> 120648352 (-7.06%)
helped: 571
HURT: 131
total spills in shared programs: 57947 -> 29130 (-49.73%)
spills in affected programs: 53364 -> 24547 (-54.00%)
helped: 351
HURT: 0
total fills in shared programs: 51310 -> 25468 (-50.36%)
fills in affected programs: 44882 -> 19040 (-57.58%)
helped: 351
HURT: 0
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
This fixes a bug where SWR will fail to render in cases with large
buffer allocations, e.g. very large meshes whose vertex buffers exceed
2GB
CC: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>
This might have decreased performance for radeonsi/tgsi, because most
most shaders claimed they used bindless.
Cc: 18.3 19.0 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Note that emit_intrinsic_load_image() already swaps a .3d flag with an
.a flag. I tried doing things the other way around (going back to .3d)
but that didn't work. And treating cube images as 2d array is also what
blob does, so let's just go with that.
Fixes dEQP-GLES31.functional.image_load_store.cube.load_store.*
Signed-off-by: Rob Clark <robdclark@gmail.com>
Fixes nearly all of dEQP-GLES31.functional.texture.border_clamp.* when
run after a test that binds textures used in vertex shader.
Signed-off-by: Rob Clark <robdclark@gmail.com>
Fixes dEQP-GLES31.functional.shaders.opaque_type_indexing.sampler.const_literal.vertex.samplercubeshadow
and few other similar tests that do multiple texture fetches into
individual components of a packet output. Mostly works around the
issue mentioned in ra_block_find_definers().
Signed-off-by: Rob Clark <robdclark@gmail.com>
vulkan_core.h defines non-dispatchable handles as (struct object *)
on 64-bit systems, but uint64_t on 32-bit systems. The former can be
implicitly cast to void *, but the latter requires an explicit cast.
While here, %lu is the wrong format specifier for uint64_t on 32-bit
systems, so use PRIu64, fixing a warning.
Reported-by: Mike Lothian <mike@fireburn.co.uk>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
On one side, when emitting 3DSTATE_SF, VertexSubPixelPrecisionSelect is
used to select between 8 bit subpixel precision (value 0) or 4 bit
subpixel precision (value 1). As this value is not set, means it is
taking the value 0, so 8 bit are used.
On the other side, in the Vulkan CTS tests, if the reference rasterizer,
which uses 8 bit precision, as it is used to check what should be the
expected value for the tests, is changed to use 4 bit as ANV was
advertising so far, some of the tests will fail.
So it seems ANV is actually using 8 bits.
v2: explicitly set 3DSTATE_SF::VertexSubPixelPrecisionSelect (Jason)
v3: use _8Bit definition as value (Jason)
v4: (by Jason)
anv: Explicitly set 3DSTATE_CLIP::VertexSubPixelPrecisionSelect
This field was added on gen8 even though there's an identically defined
one in 3DSTATE_SF.
CC: Jason Ekstrand <jason@jlekstrand.net>
CC: Kenneth Graunke <kenneth@whitecape.org>
CC: 18.3 19.0 <mesa-stable@lists.freedesktop.org>
Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Fill out "Vertex Sub Pixel Precision Select" possible values.
CC: 18.3 19.0 <mesa-stable@lists.freedesktop.org>
Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
float16 types can have non-flat interpolation so set up the HW
correctly for that.
Fixes: 62024fa775 "radv: enable VK_KHR_16bit_storage extension / 16bit storage features"
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
It causes more trouble than it's worth. Now vl tries to create compute
shaders without all the proper checking. Since there's really no
(current) way to use compute on nv50, just mark it disabled.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=109742
Fixes: f6ac0b5d71 ("gallium/auxiliary/vl: Add compute shader to support video compositor render")
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Fixes the following building error:
including ./external/mesa/Android.mk ...
build/core/base_rules.mk:183: *** external/mesa/src/intel:
MODULE.TARGET.STATIC_LIBRARIES.libmesa_isl_tiled_memcpy already defined by external/mesa/src/intel.
make: *** [build/core/ninja.mk:164: out/build-android_x86_64.ninja] Error 1
ISL_TILED_MEMCPY_FILES is isl/isl_tiled_memcpy_normal.c
and that source file includes isl_tiled_memcpy.c source
Fixes: 96bb328 ("iris: add Android build")
Signed-off-by: Mauro Rossi <issor.oruam@gmail.com>
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
This is needed for gen_clflush.h intrinsics to work on 32-bit builds.
i965 and anv both set these, and iris needs to as well.
Tested-by: Mark Janes <mark.a.janes@intel.com>
Even though the hardware spec claims that any "integer DWord multiply"
operation is affected by the regioning restrictions of CHV/BXT/GLK,
this is inconsistent with the behavior of the simulator and with
empirical evidence -- Return false from has_dst_aligned_region_restriction()
for such instructions as a micro-optimization.
Tested-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Strides up to 32B can be implemented for the source regions of most
instructions by leveraging either the vertical or the horizontal
stride of the hardware Align1 region. The main motivation for this is
that currently the lower_integer_multiplication() pass will happily
double the stride of one of the 32-bit sources, which can blow up if
the stride of the original source was already the maximum value
allowed by the hardware.
An alternative would be to use the regioning legalization pass in
order to lower such strides into the composition of multiple legal
strides, but that would be somewhat less efficient.
This showed up as a regression from my commit cbea91eb57
in Vulkan 1.1 CTS tests on CHV/BXT platforms, however it was really a
pre-existing problem that had affected conformance on other platforms
without native support for integer multiplication. CHV/BXT were
getting around it because the code I removed in that commit had the
"fortunate" side effect of emitting narrower regions that didn't hit
the hardware stride limit after lowering. Beyond fixing the
regression this fixes ~90 additional Vulkan 1.1 subgroup CTS tests on
ICL (that's why this patch is marked for inclusion in mesa-stable even
though the original regressing patch was not).
According to Jason, a nearly equivalent change had been committed
previously as e8c9e65185 and then (mistakenly?) reverted as
a31d038208.
Cc: mesa-stable@lists.freedesktop.org
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=109328
Reported-by: Mark Janes <mark.a.janes@intel.com>
Tested-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
This is required in combination with the following commit, because
otherwise if a source region with an extended 8+ stride is present in
the instruction (which we're about to declare legal) we'll end up
emitting code that attempts to write to such a region, even though
strides greater than four are still illegal for the destination.
Tested-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Because the "low" temporary needs to be accessed with word type and
twice the original stride, attempting to preserve the alignment of the
original destination can potentially lead to instructions with illegal
destination stride greater than four. Because the CHV/BXT alignment
restrictions are now being enforced by the regioning lowering pass run
after lower_integer_multiplication(), there is no real need to
preserve the original strides anymore.
Note that this bug can be reproduced on stable branches, but
back-porting would be non-trivial, because the fix relies on the
regioning lowering pass recently introduced.
Tested-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Currently the execution type calculation will return a bogus value in
cases like:
mov_indirect(8) vgrf0:w, vgrf1:w, vgrf2:ud, 32u
Which will be considered to have a 32-bit integer execution type even
though the actual indirect move operation will be carried out with
16-bit precision.
Similarly there's no need to apply the CHV/BXT double-precision region
alignment restrictions to such control sources, since they aren't
directly involved in the double-precision arithmetic operations
emitted by these virtual instructions. Applying the CHV/BXT
restrictions to control sources was expected to be harmless if mildly
inefficient, but unfortunately it exposed problems at codegen level
for virtual instructions (namely the SHUFFLE instruction used for the
Vulkan 1.1 subgroup feature) that weren't prepared to accept control
sources with an arbitrary strided region.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=109328
Reported-by: Mark Janes <mark.a.janes@intel.com>
Fixes: efa4e4bc5f "intel/fs: Introduce regioning lowering pass."
Tested-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
