On Bifrost, fragment shaders might output either FP16 or FP32. The blend
shader will access the output as-is within the register, so depending on
the precision of the blend shader's logic, it may need to insert a
f2f16 or f2f32 conversion. This requires expanding the blend shader key.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10393>
We only want it to trigger if MRT is actually in use. This is a cheap
key (only require multiple variants for an obscure edge case) and avoids
the perf regression of using this pass which is needed for conformance.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10393>
Creating Android swapchain image from gralloc buffer requires to use
VkImageDrmFormatModifierExplicitCreateInfoEXT. To fill the struct info,
we need to query extended resource info from gralloc.
With the queried modifier from gralloc, we can ask the driver for the
plane count of the given format and modifier pair.
Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Reviewed-by: Chia-I Wu <olvaffe@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10553>
We allocate them all align16, so mark the unions (and their container
structs) that way so the compiler can do aligned SSE load/stores.
glmark2 -b loop FPS +0.197265% +/- 0.117633% (n=1906)
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10604>
When a TSY barrier is hit, the entire supergroup will be synchronized.
If the supergoup is large and uses all available QPU threads it would
mean that we would sychronize and stall all running threads until all
of them reach the barrier, which may be inefficient.
This patch makes it so that if the compute shader has any such barriers
we limit the supergroup size so each supergroup only takes half of the
QPU threads available at most, so that if one supergroup hits a
barrier we have at least one other supergroup we can run, reducing
idle QPU time.
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10541>
Each supergroup executes a number batches. Each batch has 16 elements
(one per QPU lane), except possibly the last batch which might be
incomplete. Until now, we packed a single workgroup in each supergroup,
which can lead to more incomplete batches and less efficient use
of the QPUs depending on the configuration of workgroups being dispatched.
This patch computes a number of workgroups per supergroup so that
we reduce or completely eliminate incomplete batches if possible.
It should be noted however, that TSY barriers act on supergroups,
so larger supergroups lead to larger syncpoints on barriers too.
A follow-up patch will try to find a good balance for compute shaders
that use such barriers.
This improves performance of the Sascha Willem's computecloth demo
by ~13%.
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10541>
The V3D hardware allows us to pack multiple workgroups together to avoid
wasting execution lanes in shader cores.
For example, if we dispatch 16 workgroups with a local size of 1 element, we
can pack all 16 workgroups in a single 16-wide dispatch where each lane
executes a different workgroup, instead of 16 1-wide dispatches.
When we do this, we don't have a uniform workgroup id any more.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10541>
Now that all drivers are using brw_cs_get_dispatch_info() we can
remove one function (which is now unused) and reduce the scope of the
other.
Reviewed-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10504>
And since right_mask is already provided as part of dispatch_info,
just use that instead of storing it.
Reviewed-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10504>
We have this small calculations repeated in each Intel driver, so move
them to a single place to be reused. Also includes "right_mask" since
is always used in the same context and depends on the dispatch info
values.
Reviewed-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10504>
The mode one was used before 0bc5a829dd ("nir: Remove shared support from
lower_io").
The others were used before 5f7c7c9a7f ("nir: add src and dest types
to all IO loads and stores for mediump").
All conditions now are always true, so drop them.
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10533>
We recently added 5 more VIM3s to the lavalab, this should be more than
enough to run the full GLES 3.0 testsuite on G52.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10614>
For unsigned comparisons with zero these ops can be eliminated.
v2: Add comparison optimizations with -1 (Rhys Perry)
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Eric Anholt <eric@anholt.net> (v1)
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10583>
NIR provides two helper macros to run transformation passes correctly,
NIR_PASS() and NIR_PASS_V(). So far we've seemingly been a bit haphazard
about when to use them.
Let's correct that, and consistently use the NIR helpers here. This
helps us in two ways:
1. We now run nir_validate_shader after each pass, ensuring we didn't
break the shader
2. We now respect the NIR_PRINT environment variable for all NIR passes,
making debugging much less surprising.
In addition, we had an OPT()-macro that doesn't seem to provide much
help other than to hiding some trivial details. But they make our code
different to other users of NIR, which doesn't seem ideal. So let's drop
that macro while we're at it.
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10585>
