cl images and samplers come through untyped, so they need to be typed
before they can be used
samplers are also not combined into the descriptor, so track a mask which
can be used later for emission
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19327>
this simplifies all the different sampler metadata tracking fields to be
more comprehensible
it also increases some array sizes in case future work increases them outside
the compiler areas
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19327>
This specialized version prints out the liveness count as well as the
maximum liveness count. It was eye opening when seeing the max
liveness jump after lowering of packing instructions which should not
have changed the count.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18657>
When lowering a single instruction with a destination VGRF to 2 or
more, the VGRF is now considered partially written by each generated
instruction and that increases its liveness especially in loops. Thus
potentially increasing the number of spills/fills due to register
allocation.
Putting an UNDEF instruction in front of the lowered instructions
allows the IR to limit the liveness of the VGRF, reducing register
pressure.
This has a pretty dramatic effect on spills/fills for RT shaders. Here
the stats on Q2RTX shaders on DG2 (wipping out any spills/fills due to
register allocation) :
Instructions in all programs: 26150 -> 24955 (-4.6%)
SENDs in all programs: 1148 -> 1148 (+0.0%)
Loops in all programs: 4 -> 4 (+0.0%)
Cycles in all programs: 392179 -> 332787 (-15.1%)
Spills in all programs: 132 -> 116 (-12.1%)
Fills in all programs: 262 -> 154 (-41.2%)
Shader-db results on TGL :
total instructions in shared programs: 21158140 -> 21158377 (<.01%)
instructions in affected programs: 76629 -> 76866 (0.31%)
helped: 18
HURT: 20
helped stats (abs) min: 1 max: 60 x̄: 18.89 x̃: 12
helped stats (rel) min: 0.21% max: 3.61% x̄: 1.02% x̃: 0.77%
HURT stats (abs) min: 1 max: 79 x̄: 28.85 x̃: 18
HURT stats (rel) min: 0.04% max: 2.81% x̄: 1.13% x̃: 0.79%
95% mean confidence interval for instructions value: -4.82 17.30
95% mean confidence interval for instructions %-change: -0.34% 0.57%
Inconclusive result (value mean confidence interval includes 0).
total loops in shared programs: 5753 -> 5753 (0.00%)
loops in affected programs: 0 -> 0
helped: 0
HURT: 0
total cycles in shared programs: 798856834 -> 798870688 (<.01%)
cycles in affected programs: 6208395 -> 6222249 (0.22%)
helped: 22
HURT: 17
helped stats (abs) min: 2 max: 8794 x̄: 1438.18 x̃: 782
helped stats (rel) min: 0.05% max: 2.28% x̄: 0.63% x̃: 0.44%
HURT stats (abs) min: 2 max: 19178 x̄: 2676.12 x̃: 1358
HURT stats (rel) min: 0.04% max: 23.49% x̄: 2.25% x̃: 0.71%
95% mean confidence interval for cycles value: -952.19 1662.65
95% mean confidence interval for cycles %-change: -0.64% 1.90%
Inconclusive result (value mean confidence interval includes 0).
total spills in shared programs: 4078 -> 4066 (-0.29%)
spills in affected programs: 40 -> 28 (-30.00%)
helped: 2
HURT: 0
total fills in shared programs: 2856 -> 2832 (-0.84%)
fills in affected programs: 127 -> 103 (-18.90%)
helped: 2
HURT: 0
total sends in shared programs: 998554 -> 998554 (0.00%)
sends in affected programs: 0 -> 0
helped: 0
HURT: 0
LOST: 0
GAINED: 0
Total CPU time (seconds): 2346.06 -> 2304.80 (-1.76%)
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18657>
To match PIPE_COMPUTE_CAP_MAX_THREADS_PER_BLOCK, having it be higher in any
dimension is nonsensical and can confuse apps. Fixes tests in
KHR-GLES31.core.texture_buffer.* on Mali-T860.
Fixes: 9b19104a30 ("pan/mdg: Lower PIPE_COMPUTE_CAP_MAX_THREADS_PER_BLOCK on Midgard")
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19238>
Now we're back to a single XFB implementation for all gens. Fixes:
KHR-GLES31.core.draw_indirect.advanced-twoPasses-transformFeedback-arrays
KHR-GLES31.core.draw_indirect.advanced-twoPasses-transformFeedback-elements
Cc: mesa-stable
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19238>
So we can use the common XFB lowering, and try to reduce the differences between
Midgard and Bifrost (given that Bifrost is conformant and actively maintained,
and Midgard is neither, getting Midgard as close to Bifrost is in the best
interests of Midgard's long term prospects upstream). Piles of KHR-GLES31 tests
from Fail -> Skip.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19238>
Even with hierarchical tiling. Otherwise if there's garbage leftover (due to BO
caching), a fragment-only batch can raise DATA_INVALID_FAULT. Fixes many tests
in KHR-GLES31.core.compute_shader.* on Mali-T860, including
KHR-GLES31.core.compute_shader.build-separable
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19238>
Now that we only have one convert_instr path, this is simpler.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19303>
NIR has two implementations of lower_idiv, keyed on the
imprecise_32bit_lowering flag. This flag is misleading: the results when
setting this flag "imprecise", they're completely wrong for some values.
If a backend has a native implementation of umul_high, the correct path
isn't that much more expensive. If it doesn't, it's substantially slower
for highp integer divison... but in practice, non-constant highp integer
division is pretty rare.
After a painful migration of the tree, this code path has no more users.
Remove it so nobody else gets the bright idea of using it again.
Closes: #6555
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19303>
imprecise_32bit_lowering produces wrong outputs for some inputs. Do not
use it, it is wrong. The correct lowering generates umul_high
instructions, which need to be lowered via lower_alu. That lowering in
turn produces uadd_carry instructions which also need lowering.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19303>
As a memory object, the MemConstant check should be the same as
MemGlobal.
Signed-off-by: Italo Nicola <italonicola@collabora.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19352>
This breaks MSAA, even when compression is not activated. The issue is
mostly theoretical, as we always enable color compression with MSAA, but
I ran into some GPU hangs when I disabled compression to run some tests.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19066>
Fixes MSAA corruptions on GC3000. 0x4 seems to be enough to fix the
misrendering, but blob seems to always emit 0x5.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19066>
On GPUs with the SMALL_MSAA feature, color tiles of the MSAA resource
are 256B, even if the GPU doesn't have the CACHE128B256BPERLINE feature.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19066>
For MSAA downsampling to work correctly, the RS engine needs to know
the exact format of the blit source/target. The compatible formats are
fine as long as the RS is only used as a tiler without doing any
conversion.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19066>
The RS window alignment restrictions apply to the downsampled side of the
blit, so we must increase the blit size alignment by the MSAA scale to
avoid RS hangs. If a multi-pipe resolve is used (when the GPU doesn't have
the singlebuffer feature) then the Y alignment needs to be increased by
the number of pixel pipes.
The resource allocation has already been fixed to take this additional
alignment requirement into account.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19066>
The RS window aligment restrictions apply to the downsampled size of a
MSAA resource, so in order to be able to do the downsample blit for all
possible sizes, we must make sure to increase the alignment of the
multisampled resource by the MSAA scale.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19066>
Instead of special casing compressed formats in the caller, handle
them properly when working out the resource alignment. For good measure
add an assert that the layout is linear, as expected.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19066>
Instead of adjusting the Y alignment in the callers, just do the
right (and consistent) thing in etna_layout_multiple.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19066>
This function assigns proper values to the padding and halign parameters
for all valid layouts. Using unreachable() in the default path of the
switch statement asserts that we got a valid layout, so we can get rid of
some superfluous variable initializations and asserts in the call sites.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19066>
The call sites of this function make a number of adjustments to the
padding/alignment returned by this function, which are inconsistent
and still don't cover all necessary cases. To be able to extend this
function move it out of the header and make the parameters passed
more useful by providing all necessary information at once.
No functional change, just a preparation for the following changes.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19066>
Having only-recently hooked monitoring up for zink-on-radv, I am
starting to capture some flakes. So, I am unsure about the reproduction
rate of this failure, but I can say it is not extremely common.
Signed-off-by: Martin Roukala (né Peres) <martin.roukala@mupuf.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19335>
