If the optional flag is set, compaction groups TES inputs based on which
outputs they are used for:
- inputs generating only POS/CLIP outputs are first
- inputs generating both POS/CLIP and VAR outputs are next
- inputs generating only VAR outputs are last
shader-db with ACO:
143 shaders have -1.44% average decrease in code size.
There are fewer input loads and more of them are vec4 instead of vec1-3.
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32262>
The first pass computes which shader instructions contribute to each
output. It can be used to query how data flows within shaders towards
outputs.
The second pass computes which shader input components and which types of
memory loads are used to compute shader outputs.
The third pass uses the second pass to gather which input components are
used to compute pos and clip dist outputs, which input components are used
to compute all other outputs, and which input components are used to
compute both. This will be used by compaction in nir_opt_varyings for
drivers that split TES into a separate position cull shader and varying
shader to make it less likely that the same vec4 inputs are needed in both.
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32262>
With a 64 bit pointer model, instead of doing -1 the pass ended up doing
+4294967295. The reason here was some implicit integer conversion going
horribly wrong, so just do the offset math in 64 bit to get a nice result.
Cc: mesa-stable
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/13023
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34669>
the only dynamic allocation left for geometry shaders is all done in the setup
indirect kernel. so just pass the heap to that kernel directly, so we don't
reserve a heap for direct draws with GS (including pure-VS XFB). this should
reduce our memory footprint a lot in certain apps.
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/34661>
use 8-bit index buffer instead of 32-bit to significantly decrease the size of
serialized geometry shaders (agx_gs_info is not dynamic).
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/34661>
rather than a bunch of subtle booleans telling the driver how to invoke the GS
rast shader, collect everything into a common enum, and provide (CL safe)
helpers to do the appropriate calculations rather than duplicating across
GL/VK/indirects.
this fixes suboptimal handling of instancing with list topologies.
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/34661>
For h264/h265/av1/vp9, give warning when application is
sending more slices than allowed by limit, and stop copying
remaining slices to avoid unwanted behaviour.
Cc: mesa-stable
Signed-off-by: Yinjie Yao <yinjie.yao@amd.com>
Reviewed-by: Boyuan Zhang <Boyuan.Zhang@amd.com>
Reviewed-by: Ruijing Dong <ruijing.dong@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34633>
They were both missing subtle cases. While we're here, fix a bunch of
SrcTypes. They shouldn't matter in practice since it's just used to
determine how many GPRs to allocate but we may as well get them right.
Fixes: 078ffb860b ("nak/sm20: Add initial SM20 encoding")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34675>
Since an FAbs or FNegAbs modifier is going to force the source out to a
register no matter what, we should do this first. That way we avoid the
unnecessary source swaps or other evictions when we have to evict the
source anyway because it has a modifier.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34675>
After the previous changes to varyings, this test no longer crashes.
Update the expected result from Crash to Fail to reflect that.
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34074>
This logic to enable LD_VAR_BUF[_IMM] is on the conservative side.
For fixed varyings, we would need to know what the VS outputs to correctly
compute the indices the FS has to load from. For general varyings, the
locations are aligned either by the linker or by the application in case
of separable shaders.
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34074>
This is no longer used after the previous commit and should therefore
be removed.
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34074>
This change removes the limit of 16 varyings caused by the 8-bit offset
value used in LD_VAR_BUF[_IMM]. LD_VAR[_IMM] is used instead and the
necessary ADs are emitted at draw time.
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34074>
This is already supported, all we have to do is advertise it.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34648>
A8_UNORM on v9+ is using RGBA8_UNORM as a pixel format with the
A8_UNORM clump format to dealing with the diffences between
RGBA8 and the actual A8 in-memory layout.
The problem is, LEA_TEX only loads the InternalConversionDescriptor
which contains only the pixel format, and that's what ST_CVT uses
to do the conversion, so we'll actually store 4 components instead
of one.
This shows up with
dEQP-VK.image.load_store.without_any_format.buffer.a8_unorm* after
enabling maintenance5.
For now I've turned off the image storage capability for A8_UNORM
on all gens, but I'd be fine disabling it only on v9+ if you think
that's preferable.
Fixes: d95423686f ("pan/format: Add PAN_BIND_STORAGE_IMAGE flag")
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34648>
It doesn't do much, but let's call it, just in case this changes at
some point.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34648>
It's not used, and we could retrieve the device from
image->vk.base.device if we had to anyway.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34648>
The image is not supposed to be modified there.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34648>
This is needed for maintenance5.
While at it, move the buffer offseting opertaion to CmdBindVertexBuffers2()
instead of applying the offset at draw time.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34648>
This is already supported, all we have to do is advertise it.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Reviewed-by: Lars-Ivar Hesselberg Simonsen <lars-ivar.simonsen@arm.com>
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34648>
si_destroy_context needs to call context->set_debug_callback(...) to
avoid the debug logs to access the destroyed context.
Adding this change introduced a different problem: when an aux context
is destroyed from si_destroy_screen, parts of the screen have been
freed already: the shader_compiler_queue_*.
c467a87e06 ("radeonsi: Destroy queues before the aux contexts") moved
the util_queue_destroy calls above the context destruction, but with
the 59a3f38ff6 change, it's not needed anymore: si_destroy_context
will finish the screen shader queues before proceeding with releasing,
so use-after-free isn't possible.
Fixes: 59a3f38ff6 ("radeonsi: clear the debug callback on ctx destroy")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/12035
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34574>
The execution_times structure was filled the same way in all the cases of an
if. As a side effect, the retry evaluation doesn't include the last result and
can lead, as a race condition, into triggering one extra time than the
indication in the stress argument.
Signed-off-by: Sergi Blanch Torne <sergi.blanch.torne@collabora.com>
Reviewed-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33750>
When one launches a stress test on a single job, the script behaves like the
stress number is not set. After this wrong end, relaunch the command works
only if stress is bigger than 2. In case 2, it can confuse the number of
executions.
When in stress mode, don't exit the monitor_pipeline method as if there were
only one job run. One job run, prints in std the job trace, but in stress
mode, there are more than one job execution.
The stress_status_counter structure lost the information about job IDs, and
the bug happens when it counts twice the same job.
Reported-by: Martin Roukala (né Peres) <martin.roukala@mupuf.org>
Signed-off-by: Sergi Blanch Torne <sergi.blanch.torne@collabora.com>
Reviewed-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33750>
The former version was problematic because:
- time.perf_counter() returns seconds relative to an arbitrary point in
time (monotonic clock)
- time.mktime() converts to epoch time (seconds since 1970)
Signed-off-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Reviewed-by: Sergi Blanch Torne <sergi.blanch.torne@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33750>
Modify the pretty_wait function to use a two-digit width for seconds
display, ensuring consistent and aligned output when showing the
countdown timer.
Signed-off-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Reviewed-by: Sergi Blanch Torne <sergi.blanch.torne@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33750>
